The Power of PBO: Unleashing Raw Performance on AMD Ryzen 9000 Series CPUs

When it comes to CPU overclocking, AMD's Ryzen 9000 series has been making waves with its impressive power and performance capabilities. In this article, we'll delve into the world of PBO (Performance Boosting Options) enabled on these chips and explore the benefits and limitations of this feature.

One of the most notable aspects of the Ryzen 9000 series is its ability to reach extremely high clock speeds. In our testing, we saw that even with the chiplet design and tightly packed substrate, temperatures could climb to as high as 90 degrees Celsius. However, when we compared these results to Intel CPUs, it became clear that AMD's Ryzen 9000 series has a different approach to heat management. While Intel CPUs may push their voltage limits higher, AMD's chips tend to stay within a more limited range, making them more appealing to those who prioritize stability and reliability.

In our testing, we found that PBO enabled on the Ryzen 9000 series CPU pushed its frequency even further than expected, with some tests reaching speeds of over 4.5 GHz. This is particularly impressive when compared to other high-end CPUs, which often struggle to reach such high frequencies without significant power consumption. However, this increased performance came at a cost: our voltage readings spiked significantly during these tests, indicating that the chip was operating at its maximum capacity.

Interestingly, when we compared our results to Intel's PBO enabled chips, it became clear that AMD's approach has some unique advantages and disadvantages. While Intel's chips tend to push their voltage limits higher, resulting in faster frequencies, they also come with a risk of overheating and reduced stability. In contrast, AMD's Ryzen 9000 series CPUs seem to strike a balance between performance and efficiency, making them more appealing to those who prioritize raw power without compromising on stability.

Despite these advantages, some critics argue that AMD has leaned too heavily on the efficiency discussion when promoting their Ryzen 9000 series CPUs. While it's true that these chips offer impressive performance per dollar, they also come with a higher price tag than their Intel counterparts. For those who value raw power above all else, this may be less of an issue.

One area where AMD could improve is in providing more detailed information about their PBO enabled CPUs. In our testing, we found that the voltage limits for these chips were surprisingly tight, with only a small margin for error before the CPU began to overheat. This highlights the importance of proper cooling and heat management when overclocking any CPU.

To address this issue, AMD could provide more detailed information about their PBO enabled CPUs, including recommended cooling solutions and safety guidelines for overclocking. By doing so, they can empower users to make informed decisions about their system builds and help prevent overheating and stability issues.

In conclusion, the Ryzen 9000 series CPUs offer impressive performance and efficiency, making them an attractive option for those looking to upgrade their systems. However, it's essential to carefully consider the pros and cons of PBO enabled chips, including their voltage limits and potential for overheating. By understanding these limitations and taking steps to mitigate them, users can unlock even more performance from these powerful CPUs.

Our next step is to test two more 9,000 Series CPUs and provide additional insights into their performance capabilities. We'll also be exploring the results of other reviewers and comparing our findings to establish a comprehensive picture of what to expect from these chips. In the meantime, we encourage you to share your thoughts on PBO enabled CPUs and what information is most important to you when making system builds.

"WEBVTTKind: captionsLanguage: enso my uh 9700 9600 x video just went live and one of the main comments was actually you should have enabled PBO so that's what we're going to do today and see what kind of curve optimization we can get and whether or not overclocking one of these CPUs is even going to yield any yield any Real Results it actually probably will for those looking for a high-end custom gaming experience look no further than Falcon Northwest Falcon Northwest has been building PCS made for gamers for over 30 years with a focus on a true high-end gaming experience custom cases available only through Falcon Northwest feature state-of-the-art testing and design to ensure that every component is performing at their best through thermal imaging and rigorous lab testing designed and overseen by the Falcon Northwest founder himself with a complete lineup of systems ranging from small to large every Falcon Northwest system includes a three-year warranty policy and a year of two-way overnight shipping coverage providing the ultimate peace of mind to see all that Falcon Northwest has to offer follow the sponsored Link in the description below this is our AMD am5 test bench it consists of an x670 ta Chi the board from ASRock it's got 32 GB of 6,000 MHz crucial Pro which we are running at 5200 MHz Expo 2 profile for our tests um it's got a 360 EK AIO in there I just haven't replaced it yet with something different um probably should have done that I guess when we had all the CPUs coming in and out of there anyway and this is a Founders Edition Nvidia uh RTX 4090 which we use for our bottleneck tests so and then the drive is a I think it's a crucial um nbme Drive doesn't really matter cuz we're doing super CPU evaluations anyway let's talk about PBO real quick that stands for precision boost overdrive it's amd's built-in overclocking utility that you can use on their CPU because pretty much every CPU of theirs is unlocked to a degree it's like the black edition stuff with FX they were all unlocked um Precision boost overdrive works very similar to how like a GPU boost does where as long as there's temperature Headroom and there is a power limit Headroom and we've got voltage Headroom it will push the clocks up as high as it possibly can to get you the best performance now you have to enable that feature the different the reason why we don't enable that feature in our benchmarks is we test out of the box experience the same way on Intel CPUs uh we run them at the Intel default extreme profile the default profile like what actually is listed as defaults on Intel it's not really defaults at all to be honest it's like a super it's so far below what they advertise it's like criminal at that point they call it Baseline yes Baseline which means this is what it this is what it will turn it on basically uh anyway the extreme profile is really the normal profile which is like the 253 Watts for like the N the I9 stuff um or like 180 Watts for like the i7 stuff so we also made sure that none of the auto overclocking Auto features that are built into motherboards which have been destroying Intel CPUs part of Intel's whole ecosystem problem wasn't present but we don't enable PBO because it's not a guaranteed performance uplift not all AMD CPUs are going to uplift the same not all of them are going to get the same types of scores so we test out of the box box then we start playing around with our particular Asic samples or Asic qualities to see what kind of performance that we get so you guys have already seen the charts what we're going to do today is I'm going to enable PBO we're going to see where it goes to U in terms of frequency then we're going to do the auto curve Optimizer now curve Optimizer it's a more granular type of PBO where Precision boost overdrive will lift all of the limits all these guys up here all our power limits and stuff but what it will do is it will also do its own stability test and stress testing on a per core basis if you do per core or all core which is what I'll be optimizing for and it will test each core individually to see where its Max offset is for a frequency boost and then each core can actually get a different boost than each other that's actually the best way to do it but I do all core one because it's faster it takes a while like on this CPU the 9700x it's going to take like an hour to do an all core it would take even longer to do per core cuz then it has to go through and test all those cores individually where per core will see how they all do with the same workload evenly so it kind of tunes for the worst core technically and then they all get limited to where the worst core is so per core curve Optimizer would be the best result but it takes hours to complete and if you have a big CPU like a thread Ripper you will be there for probably a day or more getting curve Optimizer anyway moving on we're on all default settings right here uh let me double check my memory is where it was on our previous runs yep 5200 right there it will run 6,000 if we do a um a a 1:1 ratio for the F clock or the fabric but this is just I just chose Expo 2 5200 it'll run it'll guarantee it's not going to crash mostly we might catch a blue screen or two on here as we go um but anyways you can see right here our temperature limits 85 which is interesting because that's actually a 10° drop over previous generation 7,000 series was up to 95c so this is very similar to how the x3d stuff was so the x3d stuff had an 85 or 89 C limit depending on the CPU so that's kind of interesting but you can see we only have 88 Watts available to us although It's a 65 watt part it will allow itself to push as high as 88 Watts for more bursty workloads that might suddenly Spike what I'm looking for right here is at least in cinebench in this instruction what is our Max clock going to do for all core so you can see we're getting 4.65 GHz all core it's actually a little bit higher than what some of the other guys were seeing I think some of the other guys were seeing like 4.4 4.45 maybe 4.5 ours is doing pretty Prett good actually at 4.65 that's still disappointing for an all core boost clock considering like we're used to seeing over 5 gz now boost clock so when it comes to Intel so Intel's kind of indoctrinate um and then I'll just might as well let the test finish here but it stayed steady at 56c and it was at 4.65 the entire time for a score of 20524 it's going to be up or down a few percent run to run to run because of things happening so anyway that's where we wanted to see where we went oh let's look at our power once again let me just start the test one more time so we see where our power and amps peek out at so we are hitting the 88 Watts right there well that's total PPT but our CPU power is at 70 point or 71 Watts technically a little bit above the 65 watt that they show but again this is software monitoring right so this is what's what's being reported and we're at 89 % of our 75 amp TDC but again the temperatures you know 56 to 57c it's got a 360 AIO on there with Kingpin extreme thermal past so it's performing pretty damn good when it comes to cooling it should at that TDP honestly now here's what we're going to do I'm going to show you something that's kind of broken and this is one thing to consider when it comes to using bios features that the motherboards Implement which I think a lot of people have finally learned their lesson with hopefully Intel has created change for the industry as a whole including AMD because most motherboards especially x670 x670 motherboards have built in PBO profiles all right so here I am in the Bios memory aside it's set to 5200 Expo 2 profile if you're curious you can see it right there see a gasa default 5200 MHz I can run the 5600 um but we were getting crashing when we did the 5600 so that's why we're running 52 it'll probably run 6,000 to be honest just fine with um if I was to do the F clock at 2000 but as you can see this Ram doesn't show an expo profile up to 6,000 even though it's 6,000 MHz memory so it is showing XMP equivalence for Expo but it won't show us the full 6,000 which I think is interesting but that's fine um also too for those wondering on my system with the 78 or 7950 x3d I did get 6,000 running just fine I didn't realize the F clock was defaulting itself to 1733 so it wasn't a one to1 multiplier so it should have been set to 2,000 soon as I did that it's been Rock Solid ever since all my lagging and stuff I was getting in Photoshop all went away so remember f- clock matters anyway if we go over here to Performance preset they do have builtin like profiles that you can use you can just enable PBO which will probably give us an uplift no matter what because remember we were hitting the 4.46 GHz all core um TJ Max of 85 this is where I think there's still some bios updates that need to happen regarding the new series of processors I don't understand why 85c is the new limit when 95c was the previous limit on 7,000 and what you'll notice with these limits right here the reason why 85c exists on here is because of the 7,000 Ser series having a 95c so these were put in for people that were concerned at seeing that high wattage although that's the way it was intended to actually perform so you could drop it by 10 degrees at a time so 85 75 65 so I still feel like it should be 95c probably but I don't know I guess we'll see um so that also has a built-in curve Optimizer so curve Optimizer right here also has a minus as you can see molt in there which is almost like a um uh an undervolting if you will again probably for temperature reasons so I'm just going to do the curve Optimizer Min curve Optimizer minus 40 and it just rebooted on me right now got into windows and now you can see right here if you look here it is spazzing out there okay we're finally I think we're finally getting ready to get an image here this could be a stability thing too with that that- 40 molt so I could also try it at- 20 I think prove the point why we you shouldn't do this are we going to get into Windows I just want to run it once since okay so we're going to let ryzen Master do it so I highly recommend don't even enable PBO here let ryzen Master do it leave your bios at Auto we'll do it live if I'm going to do overclocking or whatever I'll just go to like profile one and save one there so all want to do here is precision boost overdrive not even Auto overclocking I don't want to do auto overclocking I want curve off to to do all that sort of stuff for me cuz as we curve optimize we're also going to optimize voltage for temperatures as well because if temperature limit is our Headroom is our is our limit we run out of that if we can get the voltage down stability up frequency up we can get more frequency before we hit that temperature limit we can lift the max temperature dare I I'll go 90 I think 90c will be fine what could go wrong so when you reboot it should automatically launch ryzen Master again yes okay should normally also too if you do a setting in there that's unstable and it detects a crash when you restart it won't apply the settings that's why ryzen M that's why ryzen Master is preferred over bios because if you apply it in the Bios it's just going to keep trying to apply it and it can keep crashing so let ryzen Master do it uh hey check it out 90c there it goes right there okay and all our other limits are where we want them to be 360 AIO do your freaking job that 5.23 5.24 all core at 87c 86 140 wats so double the power draw suddenly but let's see what happens to our score remember last time we were at 20,5 37 I think somewhere around there from 4.6 to 5.3 allore this is this is now like Intel territory by the way on all core which is kind of nice the only it's lacking is core count to really keep up you see the max it could go as 5.5 according to that and a score went up to a 23, 325 so nearly 3,000 Point increase by literally clicking a button and then I upped it to 90 so as you can see we would have hit throttle a little bit on temperature prior to hitting the the 5.4 or what was it 5.34 so it probably would have been like 5 point or was it 5.24 it probably would hit like 5.2 somewhere around there okay what I'm going to do right now is I'm going to stop C I'm going to run through all the tests again with these settings to compare how they were with our launch results and then I'm going to show you guys Precision boost uh and curve Optimizer show you what it what it does like how it does it run the test again with curve Optimizer to see if there's any Improvement or benefit to waiting all that time to curve optimized to see what happens and then we'll put all those scores against each other to see how like this basically one button overclocking does for AMD this is nothing new this has been around since ryzen has existed only I think it's getting better we're actually seeing things happen now all right so I'm in the middle of my testing I want to point something else or something out that just happened that's really weird um we talk about these AMD weird problems all the time that AMD Fanboys are always like there's no weird you're just dumb when it comes to AMD but no this thank God Steve had his problems otherwise you guys would never believe me that the platform is not perfect anyway moving on I'm getting my results and stuff through the PBO test and I had a crash when launching time spy Phil noticed the crash cuz the fans which were running full speed uh slowed down he's like what happened I'm like it it crashed it did the thing like I saw earlier where it wasn't a blue screen it was just a restart so the random restarts with AMD has been a thing for years and no one can figure out why the random restarts just happened anyway on booting back up I noticed now my limit is 95c you guys saw me set this to 90 and I wanted to sity check myself and say did I screw up if I go back to profile one it still set a 90 so on the restart it still applied PBO but now with a 95c limit which is what I was saying I thought was weird because 7,000 series had a 95c limit I don't know if this is a 4 nanometer thing that's just some new limmit in terms of temp's not going as high but now if I hit profile one and apply now my 90c is back so I don't have any clue as to why it shows 95c or showed 95c cuz it has never showed 95c with a 9000 Series processor in there until this moment and then I decided to check one more thing and the ram reset itself to 5600 MHz it's that times two right double data rate so so yeah I mean check it out here 5600 so on that restart it put the ram back to 5600 which like I keep it won't save what I put when it does that stupid restart it's incredibly irritating okay so after several restarts and crashes and freezes um every time we just let it resume it finally came back with a minus 33 offset so minus 40 obviously was too much of an undervolt - 33 not much difference though we're talking 7 molts here I guess that could be the difference all right so curve Optimizer is done interestingly enough it drops the is that 95 now it was at 85 a second ago curve Optimizer all course Precision boost overdrive it did a minus 33 offset I don't think it's going to be very um and by the way there was a lot of crashes during it doing its optimizing normally it shouldn't crash restarts should happen but crashes like hard locks Mouse won't move and all that is a is a problem anyway it raise the limit back up to 95c um which is going to be too high I think okay Rising Master doesn't well there we go can't make this up I'm sorry I know Fanboys out there don't want to admit this but there's there's problems with some of the software anyway um will it run though on the co allore minus 33 run multicore move my mouse no and it's frozen so we're going to throw the curve Optimizer out entirely um just going to go ahead and now kind of do uh cinebench R23 10-minute runs just kind of with PBO on PBO off take a look at the frequencies over time look at the voltages over time and look at the temperatures over time to see what kind of uplift we're seeing in power draw and all that sort of stuff just by enabling PBO as you guys already saw it doubles the wattage of the CPU which is pretty significant but if we get a good performance uplift I think a lot of people would be willing to say I don't care what the power is in fact just read the comments of the videos a lot of people are like efficienc is cool but at the end of the day I want the performance I don't give a flying F about the efficiency most people kind of feel that way to be honest and so I guess we'll get it up and running and do that chart now all right so it is actually a couple days later we ended up doing that micro code video in between when we started this one and went on to this one so that's why you're going to notice some 14900 k ZX 0x1 129 and blah blah blah on the chart so disregard that you probably already seen it but moving on a lot of you Keen eagle-eyed viewers out there we know there's plenty of you uh notice the PBO numbers on those charts so forgive me if you've already seen this but we need to finish this video here so PB it's interesting I showed you guys already how we tried to do a couple of different things like I enabled so first of all the curve Optimizer just did not work well with this generation I don't know why I don't know if it's something to do with the fact that the the 9000 Series right now just seems to be incredibly unstable on our motherboard with the BIOS provided by AMD even reinstalled the BIOS to see if whether or not that would fix some of the stability problems it does all point to being memory related and I don't get why for instance memory management is the most recent like blue screen code we continue to get with this and we've tried different Ram we've tried different Ram speed it just continues to happen but it doesn't really happen on our 7,000 series CPUs and my 7950 x3d which I'm using as my game /wor system here at work that I did a few weeks back has been Rock Solid without a single restart a single um blue screen and no issues whatsoever on 6,000 MHz memory at a 2,000 MHz F clock so we'll have to play around with that more and I've been in communication with AMD regarding some of the stability problems unfortunately right now they're kind of like quiet on getting me any information back so we're just kind of left our own devices to figure that one out moving on uh we only did the 9700 X right here but you can see we also had the 90c limit on there um and the weird weird thing that's incredibly inconsistent is that when it crashes and comes back online as I've already showed it'll lift the 90c to 95c which is super odd and I don't understand why it's doing that but whatever 90c was where the profile I set um left it curve Optimizer with PBO on was actually giving us less result than just enabling PBO so I did away with the Curve op Optimizer altogether as I actually showed was that minus 33 it came up with was not stable at all so my recommendation is still to just enable PBO and leave it let PBO do the work don't let curve Optimizer do anything I think these are such low voltage low wattage Parts as it is reducing any molts already increases instability because of the fact that they're such low wattage to begin with so here we are with blender CPU test uh classroom you can see we actually went up from a 63.9 h samples um higher is better to 71.8 3 so if you move to junk shop again a pretty good jump from 88.2 to 98.8 4 um again not catching the anywhere near the 14700 K which this is much more similarly priced to actually more expensive than the 14700 K when it comes to CPU tasks but uh blender test for Monster as you can see we jumped up from 12542 to a 13984 moving on to cinebench which is a mult R1 core one thread test this is where um it doesn't really improve if anything we we lost 10 points went from a 2197 to a 2187 but that's because PBO doesn't really push single core the single core boost is already allowed to go pretty high it really does do a multi-core uplift when it comes to the multicore uh total frequency jump and I'll show that in the uh frequency over time chart when we get there but upper down 10 points could be margin of error I mean we do have things like uh updates disabled we don't have any background tasks going we try and keep the OS as clean and simple as possible so that no program doing something in the background searching for updates or whatever could potentially uh impact results this is just an up and down margin of error between run to run but as you can see on single core thread though or single core it's holding its own pretty well against the 14 uh 14th gen and 13th gen obviously respectively the 9600 X down there at 2159 looking real good that shows the IPC Improvement um versus Intel and versus 7000 series unfortunately that IPC it still has a clock disadvantage versus Intel where Intel has a higher clock even though it has less IPC it has more Cycles um which is getting it the the better performance cinebench R23 multicore you can see right here we jumped from a 20352 to a 23, 325 so that particular AVX instruction set um with that CPO and CPU and PBO definitely gave us a pretty big jump unfortunately as you can see it doesn't get us anywhere near its cheaper priced Intel Brethren which is unfortunate this is why so many people are saying the cost to Performance right now with 9,000 Series at least at these CPU levels the 9600 X and 9700x is just not there unfortunately and the reason for that and again we'll talk about this when we get to the the frequency over time is although on a core per core IPC basis AMD right now is superior unfortunately all of these CPUs have a thread count disadvantage which is why in multi-threaded tests like this they fall behind because they simply do not have as many cores we're talking 16 threads eight core eight full fat cores uh and might say full fat I mean they're not talking EC cor P core like Intel has but full-fledged cores with simultaneous multi-threading for 16 threads versus 20 threads on the 14 uh or the 146 and 28 threads on the 147 and the ecores we've seen definitely pick up a lot of the slack when it comes to the multi-core workload so realistically these CPUs just have a thread count disadvantage which is why I would have loved to have seen them come up with some more of a hybrid type of 10 core2 thread or something like that unfortunately the chiplet design just doesn't allow for it so even though we're uplifting the frequencies it's not enough to make up for the thread count difference uh the single thread though as you can see no uplift at all uh 132 to 132 and the reason for that like I said is PBO doesn't really do much for the max single core frequency if anything at all it's all about the all core frequency but the all the cinebench r24 multi-thread as you can see here once again a severe multi-thread disadvantage in fact these charts are sorted price left to right and you can see we got a very linear curve going up unfortunately when you get to the price of the 9700x there's a severe drop in performance because it's unfortunately its value proposition in multi-threaded workloads is just simply not there a lot of you right now going be going but the efficiency the efficiency wait till you see what happens to the efficiency when we enable PBO we're definitely going to show that now we look at geek bench single thread uh it goes from a 3303 to a 3314 again fairly margin of error the multi-thread though once again we go from a 16,063 to a 16 881 so we got 800 points over 800 points of uplift there but not enough to close that Gap with the 14700 K which is unfortunately cheaper than the AMD CPU um moving on to handbrake we do drop significantly we go from a 93 second transcode on our three uh over 3 minute 4K video down to 1080p U it's important to note that we are not using Quicks sync or or any of that stuff when it comes to Intel this is just raw CPU x264 encoder but we dropped from 93 to 85 un fortunately the 137 and the 147 are significantly faster at 68 and 63 so we do see a performance increase for a one cl one click button but it's not enough to make up that performance Gap that I was really hoping to see that Gap close because the all core frequencies are pretty gimped in my opinion out of the box at about 4.4 to 4.6 G all core it just wasn't enough to make up for it um gaming shadow of the Tomb Raider we see a moderate uplift from a 49 average fps to 258 average FPS again medium setting settings 1080p no scaling no RTX 4090 just pushing all the frames we can to see how the CPU can handle it uh but you can see the 137 and the 147 that are flanking it um outperforming it by more than 20 FPS and up to 30 FPS in some instances there so a lot left to be desired we did see a pretty good jump though in Times by extreme we went from a 6589 to a 7634 so it shows that there are instances where we can get a pretty good jump in performance but unfortunately tests like this are designed to really have the chart emphasize an improvement so that type of percentage Improvement doesn't translate to that type of percentage Improvement in games unfortunately and then if we look at Cyber Punk a weird title that just doesn't seem to favor AMD I think this is something a lot of people have realized already it is definitely a heavily Intel optimized title as far as we can tell cuz you can see the 7600 X versus 9600 X versus 7700x versus 9700x and 9700x with PBO enabled are all like pretty much the same which is a weird almost like an artificial cap that's happening there and their minimum FPS is also nearly the same across the board it's very odd it's almost telling like this is an intentional type of bottleneck or something for AMD CPUs but when we look at the Intel CPUs we can see the same sort of thing the 13600 K and the 14900 K with the latest micro code is 218 versus 229 so Intel is seeing the same type of like across the board sort of cap regardless of the tier of of the skew it's just a higher cap and then their their minimum FPS is also a very similar story but it's much higher in the 180 190 range versus the 130 Range found on AMD so cyber Punk's one of those titles which I feel like heavily heavily favors Intel in this instance let's talk about the charts over time because this is where the the real differences uh show let's start with core clocks here so you'll notice the green chart right here which is core clock average megahertz across you know obviously all the cores this is a 10-minute cinebench R23 multicore run you can see there's some frequency spikes as the test is loaded which is normal we've showed this as the test gets loaded it's more of a single core type of a workload to load the test so we see a frequency Spike well over 5,000 MHz up to almost 5.5 MHz single core but then as soon as all cores go under load it drops down significantly our particular run was running about 465 GHz on average but you can see that that line is very wavy it's all over the place um temperatures were pretty solid so that's why we're not seeing a curve downward or anything because we're not seeing thermals really come into play when it comes to the frequencies when we enable PBO you can see we jump up to 5.3 to 5.4 GHz all core and it's definitely locked like it is a more Flat Line because that's as far as the frequency scaler will allow PBO to push it so it definitely hits that 53 54 Mark kind of fluctuates between like those 50 MHz there between like 535 54 and then you can see the individual spikes uh as it's loading the test those jump up to about 56 on the single core now it shows here in this test that the single core does jump up but on the other tests as we showed you um which do a longer test under single core don't really seem to benefit um the single core bump because it doesn't stay up there it drops back down to like this 5554 range on single core uh and stays there so that's why the single core threads on like geekbench and r24 and R23 don't really show any sort of an uplift over time so you can see by adjusting the PBO we get a nearly 1 GHz bump in performance it's something I wish that AMD had sort of allowed out of the box and maybe not heavily emphasize the efficiency and maybe push these clocks a little bit more out of the box and maybe made these like an 95 watt part 85 watt part something like that versus 65 but when we look at the Watts it takes to do that this isn't this is a this is a scary chart I've seen some people comment saying any CPU that draws over 150 watts is scary to me well this should be terrifying to you because if you look at our PPT Watts which is um the actual power package power for the CPU itself that's memory controller the whole deal not just the core but the whole package of the CPU you can see it goes from about 88 Watts solid a straight line no fluctuation at all to over 160 Watts averaging about 158 Watts so double the power draw for the small amount of performance uplift that you saw that's the part that's a little disappointing here is the fact that PBO was able to push the frequencies but it pushes the power to double for up to 10% performance gain that's a bad trade-off that's where the efficiency goes out the window with these particular CPUs but what about the temperature it took to do that well if we look at the T the T diey temperature here um this is a more accurate measurement of amd's like they have core package temp you have package Temp and then you have T diey T die is what we're looking at here because this is the more sensitive number that actually handles a lot of PBO scaling and stuff you can see we go from the mid-50s spiking up to about 57 58 while we do the single cord load and that's that single core that jumps up that it's measuring that that die to up to 90c so it actually pushes it up to 90c cuz that's what I set because if you enable PBO it's still 85c I want to see where it would if we push beyond that 85 how high will that actually go the numbers went red in Hardware info 64 the numbers are red cuz anything over 85 is considered Beyond spec so it pushed itself to 90c makes sense considering the fact that we're talking double the wattage but you can see right now with with the chiplet design and it's so tightly packed in the way that the the whole substrate is designed it becomes very focused heat so we go from 55c to 90 which is nuts but double the power to do it makes sense that we would see that type of temperature climb and the voltage the voltage to do it the the one thing that we were super excited about with this because of the super efficient frequencies and the power limit it only took about 1.05 volts to be able to run the CPU at itss out of the box specs with spikes though leading up to about 1.35 volt but the spikes in the chart as you already know is every time it reloads R23 uh cin bench R23 it spikes but check out the orange line which are which is our PBO enabled what I'd like to see though is you can see that the green spikes and the orange spikes are the same not like the Intel thing where it pushes it way up higher so AMD does have a pretty hard-coded voltage limit in there that says regardless of PBO on or off don't exceed this voltage and it's very obvious when you look at the green line versus the orange line but look at the Baseline when it comes to the average frequencies here to maintain in allcore overclock we're talking about 1.28 to 1.32 volts is what we were averaging according to Hardware info so a significant voltage bump which as you saw led to a significant PPT increase which also led to a significant temperature increase taking all of the efficiency benefits of this chip out of the window and giving us about 10% more Improvement for double the power it is a one button overclock it does allow itself to be able to give you more performance and if efficiency is not that big of a deal to you I saw a lot of comments saying the efficiency is great but we wish AMD hadn't leaned so heavily on the efficiency discussion and maybe lean more into raw power this is what happens when you do that I feel like AMD could have split the difference here somewhere in the middle they are very appealing chips to people that are not interested in any of the Intel drama right now and don't want to even consider going down that route with a CPU that you can't trust whether or not it's already degraded or the one that you're buying doesn't already have potential oxidation or whatever you have a plethora of drama you can pick from right now to say I have buyer uh anxiety and it's all valid AMD right now you just have to decide whether or not their price to performance on these chips whether PBO enabled or not are worth it to you and as a buyer that's 100% your decision I can tell you people like me don't give a rat's ass about what you buy it's your money it's your decision all we do is present the data here as it presents itself the numbers are what they are so PBO enabled on the 20 or the 9700x did obviously lead to an increase a bigger increase than we typically see on the higher core count chips with the 16 core 32 thread chips because of the fact that there is PPT and temperature Headroom in those the more dense core count chips tend to not benefit from PBO quite as much because they're already near that Headroom out of the box so there you go guys PBO numbers you guys said why aren't you showing PBO well cuz I always do this as a separate video cuz I want to play around with them some more anyway we've got two more 9,000 Series CPUs that need to be tested so we got to get on out of here and uh I may or may not include PBO at the same time with those videos we'll see we've got a lot of regression testing here our charts are lacking a lot of data and I'm trying to fill these charts in with data to make it more relevant for you it's taking time to catch up with what other people have established so do yourself a favor I know you already do check out as many reviewers as you can to get their takes on these CPUs because everyone kind of comes at these tests a little differently thanks for watching guys thanks for supporting us and as always we'll see you in the next one oh and please give me the data points that you think are important to you I want to know what information you think is relevant and I'm trying to get it all in there and sometimes what I think is relevant you don't and vice versa so sound off in the comments below or hit me up on Twitter and as always I'll see you guys in the next oneso my uh 9700 9600 x video just went live and one of the main comments was actually you should have enabled PBO so that's what we're going to do today and see what kind of curve optimization we can get and whether or not overclocking one of these CPUs is even going to yield any yield any Real Results it actually probably will for those looking for a high-end custom gaming experience look no further than Falcon Northwest Falcon Northwest has been building PCS made for gamers for over 30 years with a focus on a true high-end gaming experience custom cases available only through Falcon Northwest feature state-of-the-art testing and design to ensure that every component is performing at their best through thermal imaging and rigorous lab testing designed and overseen by the Falcon Northwest founder himself with a complete lineup of systems ranging from small to large every Falcon Northwest system includes a three-year warranty policy and a year of two-way overnight shipping coverage providing the ultimate peace of mind to see all that Falcon Northwest has to offer follow the sponsored Link in the description below this is our AMD am5 test bench it consists of an x670 ta Chi the board from ASRock it's got 32 GB of 6,000 MHz crucial Pro which we are running at 5200 MHz Expo 2 profile for our tests um it's got a 360 EK AIO in there I just haven't replaced it yet with something different um probably should have done that I guess when we had all the CPUs coming in and out of there anyway and this is a Founders Edition Nvidia uh RTX 4090 which we use for our bottleneck tests so and then the drive is a I think it's a crucial um nbme Drive doesn't really matter cuz we're doing super CPU evaluations anyway let's talk about PBO real quick that stands for precision boost overdrive it's amd's built-in overclocking utility that you can use on their CPU because pretty much every CPU of theirs is unlocked to a degree it's like the black edition stuff with FX they were all unlocked um Precision boost overdrive works very similar to how like a GPU boost does where as long as there's temperature Headroom and there is a power limit Headroom and we've got voltage Headroom it will push the clocks up as high as it possibly can to get you the best performance now you have to enable that feature the different the reason why we don't enable that feature in our benchmarks is we test out of the box experience the same way on Intel CPUs uh we run them at the Intel default extreme profile the default profile like what actually is listed as defaults on Intel it's not really defaults at all to be honest it's like a super it's so far below what they advertise it's like criminal at that point they call it Baseline yes Baseline which means this is what it this is what it will turn it on basically uh anyway the extreme profile is really the normal profile which is like the 253 Watts for like the N the I9 stuff um or like 180 Watts for like the i7 stuff so we also made sure that none of the auto overclocking Auto features that are built into motherboards which have been destroying Intel CPUs part of Intel's whole ecosystem problem wasn't present but we don't enable PBO because it's not a guaranteed performance uplift not all AMD CPUs are going to uplift the same not all of them are going to get the same types of scores so we test out of the box box then we start playing around with our particular Asic samples or Asic qualities to see what kind of performance that we get so you guys have already seen the charts what we're going to do today is I'm going to enable PBO we're going to see where it goes to U in terms of frequency then we're going to do the auto curve Optimizer now curve Optimizer it's a more granular type of PBO where Precision boost overdrive will lift all of the limits all these guys up here all our power limits and stuff but what it will do is it will also do its own stability test and stress testing on a per core basis if you do per core or all core which is what I'll be optimizing for and it will test each core individually to see where its Max offset is for a frequency boost and then each core can actually get a different boost than each other that's actually the best way to do it but I do all core one because it's faster it takes a while like on this CPU the 9700x it's going to take like an hour to do an all core it would take even longer to do per core cuz then it has to go through and test all those cores individually where per core will see how they all do with the same workload evenly so it kind of tunes for the worst core technically and then they all get limited to where the worst core is so per core curve Optimizer would be the best result but it takes hours to complete and if you have a big CPU like a thread Ripper you will be there for probably a day or more getting curve Optimizer anyway moving on we're on all default settings right here uh let me double check my memory is where it was on our previous runs yep 5200 right there it will run 6,000 if we do a um a a 1:1 ratio for the F clock or the fabric but this is just I just chose Expo 2 5200 it'll run it'll guarantee it's not going to crash mostly we might catch a blue screen or two on here as we go um but anyways you can see right here our temperature limits 85 which is interesting because that's actually a 10° drop over previous generation 7,000 series was up to 95c so this is very similar to how the x3d stuff was so the x3d stuff had an 85 or 89 C limit depending on the CPU so that's kind of interesting but you can see we only have 88 Watts available to us although It's a 65 watt part it will allow itself to push as high as 88 Watts for more bursty workloads that might suddenly Spike what I'm looking for right here is at least in cinebench in this instruction what is our Max clock going to do for all core so you can see we're getting 4.65 GHz all core it's actually a little bit higher than what some of the other guys were seeing I think some of the other guys were seeing like 4.4 4.45 maybe 4.5 ours is doing pretty Prett good actually at 4.65 that's still disappointing for an all core boost clock considering like we're used to seeing over 5 gz now boost clock so when it comes to Intel so Intel's kind of indoctrinate um and then I'll just might as well let the test finish here but it stayed steady at 56c and it was at 4.65 the entire time for a score of 20524 it's going to be up or down a few percent run to run to run because of things happening so anyway that's where we wanted to see where we went oh let's look at our power once again let me just start the test one more time so we see where our power and amps peek out at so we are hitting the 88 Watts right there well that's total PPT but our CPU power is at 70 point or 71 Watts technically a little bit above the 65 watt that they show but again this is software monitoring right so this is what's what's being reported and we're at 89 % of our 75 amp TDC but again the temperatures you know 56 to 57c it's got a 360 AIO on there with Kingpin extreme thermal past so it's performing pretty damn good when it comes to cooling it should at that TDP honestly now here's what we're going to do I'm going to show you something that's kind of broken and this is one thing to consider when it comes to using bios features that the motherboards Implement which I think a lot of people have finally learned their lesson with hopefully Intel has created change for the industry as a whole including AMD because most motherboards especially x670 x670 motherboards have built in PBO profiles all right so here I am in the Bios memory aside it's set to 5200 Expo 2 profile if you're curious you can see it right there see a gasa default 5200 MHz I can run the 5600 um but we were getting crashing when we did the 5600 so that's why we're running 52 it'll probably run 6,000 to be honest just fine with um if I was to do the F clock at 2000 but as you can see this Ram doesn't show an expo profile up to 6,000 even though it's 6,000 MHz memory so it is showing XMP equivalence for Expo but it won't show us the full 6,000 which I think is interesting but that's fine um also too for those wondering on my system with the 78 or 7950 x3d I did get 6,000 running just fine I didn't realize the F clock was defaulting itself to 1733 so it wasn't a one to1 multiplier so it should have been set to 2,000 soon as I did that it's been Rock Solid ever since all my lagging and stuff I was getting in Photoshop all went away so remember f- clock matters anyway if we go over here to Performance preset they do have builtin like profiles that you can use you can just enable PBO which will probably give us an uplift no matter what because remember we were hitting the 4.46 GHz all core um TJ Max of 85 this is where I think there's still some bios updates that need to happen regarding the new series of processors I don't understand why 85c is the new limit when 95c was the previous limit on 7,000 and what you'll notice with these limits right here the reason why 85c exists on here is because of the 7,000 Ser series having a 95c so these were put in for people that were concerned at seeing that high wattage although that's the way it was intended to actually perform so you could drop it by 10 degrees at a time so 85 75 65 so I still feel like it should be 95c probably but I don't know I guess we'll see um so that also has a built-in curve Optimizer so curve Optimizer right here also has a minus as you can see molt in there which is almost like a um uh an undervolting if you will again probably for temperature reasons so I'm just going to do the curve Optimizer Min curve Optimizer minus 40 and it just rebooted on me right now got into windows and now you can see right here if you look here it is spazzing out there okay we're finally I think we're finally getting ready to get an image here this could be a stability thing too with that that- 40 molt so I could also try it at- 20 I think prove the point why we you shouldn't do this are we going to get into Windows I just want to run it once since okay so we're going to let ryzen Master do it so I highly recommend don't even enable PBO here let ryzen Master do it leave your bios at Auto we'll do it live if I'm going to do overclocking or whatever I'll just go to like profile one and save one there so all want to do here is precision boost overdrive not even Auto overclocking I don't want to do auto overclocking I want curve off to to do all that sort of stuff for me cuz as we curve optimize we're also going to optimize voltage for temperatures as well because if temperature limit is our Headroom is our is our limit we run out of that if we can get the voltage down stability up frequency up we can get more frequency before we hit that temperature limit we can lift the max temperature dare I I'll go 90 I think 90c will be fine what could go wrong so when you reboot it should automatically launch ryzen Master again yes okay should normally also too if you do a setting in there that's unstable and it detects a crash when you restart it won't apply the settings that's why ryzen M that's why ryzen Master is preferred over bios because if you apply it in the Bios it's just going to keep trying to apply it and it can keep crashing so let ryzen Master do it uh hey check it out 90c there it goes right there okay and all our other limits are where we want them to be 360 AIO do your freaking job that 5.23 5.24 all core at 87c 86 140 wats so double the power draw suddenly but let's see what happens to our score remember last time we were at 20,5 37 I think somewhere around there from 4.6 to 5.3 allore this is this is now like Intel territory by the way on all core which is kind of nice the only it's lacking is core count to really keep up you see the max it could go as 5.5 according to that and a score went up to a 23, 325 so nearly 3,000 Point increase by literally clicking a button and then I upped it to 90 so as you can see we would have hit throttle a little bit on temperature prior to hitting the the 5.4 or what was it 5.34 so it probably would have been like 5 point or was it 5.24 it probably would hit like 5.2 somewhere around there okay what I'm going to do right now is I'm going to stop C I'm going to run through all the tests again with these settings to compare how they were with our launch results and then I'm going to show you guys Precision boost uh and curve Optimizer show you what it what it does like how it does it run the test again with curve Optimizer to see if there's any Improvement or benefit to waiting all that time to curve optimized to see what happens and then we'll put all those scores against each other to see how like this basically one button overclocking does for AMD this is nothing new this has been around since ryzen has existed only I think it's getting better we're actually seeing things happen now all right so I'm in the middle of my testing I want to point something else or something out that just happened that's really weird um we talk about these AMD weird problems all the time that AMD Fanboys are always like there's no weird you're just dumb when it comes to AMD but no this thank God Steve had his problems otherwise you guys would never believe me that the platform is not perfect anyway moving on I'm getting my results and stuff through the PBO test and I had a crash when launching time spy Phil noticed the crash cuz the fans which were running full speed uh slowed down he's like what happened I'm like it it crashed it did the thing like I saw earlier where it wasn't a blue screen it was just a restart so the random restarts with AMD has been a thing for years and no one can figure out why the random restarts just happened anyway on booting back up I noticed now my limit is 95c you guys saw me set this to 90 and I wanted to sity check myself and say did I screw up if I go back to profile one it still set a 90 so on the restart it still applied PBO but now with a 95c limit which is what I was saying I thought was weird because 7,000 series had a 95c limit I don't know if this is a 4 nanometer thing that's just some new limmit in terms of temp's not going as high but now if I hit profile one and apply now my 90c is back so I don't have any clue as to why it shows 95c or showed 95c cuz it has never showed 95c with a 9000 Series processor in there until this moment and then I decided to check one more thing and the ram reset itself to 5600 MHz it's that times two right double data rate so so yeah I mean check it out here 5600 so on that restart it put the ram back to 5600 which like I keep it won't save what I put when it does that stupid restart it's incredibly irritating okay so after several restarts and crashes and freezes um every time we just let it resume it finally came back with a minus 33 offset so minus 40 obviously was too much of an undervolt - 33 not much difference though we're talking 7 molts here I guess that could be the difference all right so curve Optimizer is done interestingly enough it drops the is that 95 now it was at 85 a second ago curve Optimizer all course Precision boost overdrive it did a minus 33 offset I don't think it's going to be very um and by the way there was a lot of crashes during it doing its optimizing normally it shouldn't crash restarts should happen but crashes like hard locks Mouse won't move and all that is a is a problem anyway it raise the limit back up to 95c um which is going to be too high I think okay Rising Master doesn't well there we go can't make this up I'm sorry I know Fanboys out there don't want to admit this but there's there's problems with some of the software anyway um will it run though on the co allore minus 33 run multicore move my mouse no and it's frozen so we're going to throw the curve Optimizer out entirely um just going to go ahead and now kind of do uh cinebench R23 10-minute runs just kind of with PBO on PBO off take a look at the frequencies over time look at the voltages over time and look at the temperatures over time to see what kind of uplift we're seeing in power draw and all that sort of stuff just by enabling PBO as you guys already saw it doubles the wattage of the CPU which is pretty significant but if we get a good performance uplift I think a lot of people would be willing to say I don't care what the power is in fact just read the comments of the videos a lot of people are like efficienc is cool but at the end of the day I want the performance I don't give a flying F about the efficiency most people kind of feel that way to be honest and so I guess we'll get it up and running and do that chart now all right so it is actually a couple days later we ended up doing that micro code video in between when we started this one and went on to this one so that's why you're going to notice some 14900 k ZX 0x1 129 and blah blah blah on the chart so disregard that you probably already seen it but moving on a lot of you Keen eagle-eyed viewers out there we know there's plenty of you uh notice the PBO numbers on those charts so forgive me if you've already seen this but we need to finish this video here so PB it's interesting I showed you guys already how we tried to do a couple of different things like I enabled so first of all the curve Optimizer just did not work well with this generation I don't know why I don't know if it's something to do with the fact that the the 9000 Series right now just seems to be incredibly unstable on our motherboard with the BIOS provided by AMD even reinstalled the BIOS to see if whether or not that would fix some of the stability problems it does all point to being memory related and I don't get why for instance memory management is the most recent like blue screen code we continue to get with this and we've tried different Ram we've tried different Ram speed it just continues to happen but it doesn't really happen on our 7,000 series CPUs and my 7950 x3d which I'm using as my game /wor system here at work that I did a few weeks back has been Rock Solid without a single restart a single um blue screen and no issues whatsoever on 6,000 MHz memory at a 2,000 MHz F clock so we'll have to play around with that more and I've been in communication with AMD regarding some of the stability problems unfortunately right now they're kind of like quiet on getting me any information back so we're just kind of left our own devices to figure that one out moving on uh we only did the 9700 X right here but you can see we also had the 90c limit on there um and the weird weird thing that's incredibly inconsistent is that when it crashes and comes back online as I've already showed it'll lift the 90c to 95c which is super odd and I don't understand why it's doing that but whatever 90c was where the profile I set um left it curve Optimizer with PBO on was actually giving us less result than just enabling PBO so I did away with the Curve op Optimizer altogether as I actually showed was that minus 33 it came up with was not stable at all so my recommendation is still to just enable PBO and leave it let PBO do the work don't let curve Optimizer do anything I think these are such low voltage low wattage Parts as it is reducing any molts already increases instability because of the fact that they're such low wattage to begin with so here we are with blender CPU test uh classroom you can see we actually went up from a 63.9 h samples um higher is better to 71.8 3 so if you move to junk shop again a pretty good jump from 88.2 to 98.8 4 um again not catching the anywhere near the 14700 K which this is much more similarly priced to actually more expensive than the 14700 K when it comes to CPU tasks but uh blender test for Monster as you can see we jumped up from 12542 to a 13984 moving on to cinebench which is a mult R1 core one thread test this is where um it doesn't really improve if anything we we lost 10 points went from a 2197 to a 2187 but that's because PBO doesn't really push single core the single core boost is already allowed to go pretty high it really does do a multi-core uplift when it comes to the multicore uh total frequency jump and I'll show that in the uh frequency over time chart when we get there but upper down 10 points could be margin of error I mean we do have things like uh updates disabled we don't have any background tasks going we try and keep the OS as clean and simple as possible so that no program doing something in the background searching for updates or whatever could potentially uh impact results this is just an up and down margin of error between run to run but as you can see on single core thread though or single core it's holding its own pretty well against the 14 uh 14th gen and 13th gen obviously respectively the 9600 X down there at 2159 looking real good that shows the IPC Improvement um versus Intel and versus 7000 series unfortunately that IPC it still has a clock disadvantage versus Intel where Intel has a higher clock even though it has less IPC it has more Cycles um which is getting it the the better performance cinebench R23 multicore you can see right here we jumped from a 20352 to a 23, 325 so that particular AVX instruction set um with that CPO and CPU and PBO definitely gave us a pretty big jump unfortunately as you can see it doesn't get us anywhere near its cheaper priced Intel Brethren which is unfortunate this is why so many people are saying the cost to Performance right now with 9,000 Series at least at these CPU levels the 9600 X and 9700x is just not there unfortunately and the reason for that and again we'll talk about this when we get to the the frequency over time is although on a core per core IPC basis AMD right now is superior unfortunately all of these CPUs have a thread count disadvantage which is why in multi-threaded tests like this they fall behind because they simply do not have as many cores we're talking 16 threads eight core eight full fat cores uh and might say full fat I mean they're not talking EC cor P core like Intel has but full-fledged cores with simultaneous multi-threading for 16 threads versus 20 threads on the 14 uh or the 146 and 28 threads on the 147 and the ecores we've seen definitely pick up a lot of the slack when it comes to the multi-core workload so realistically these CPUs just have a thread count disadvantage which is why I would have loved to have seen them come up with some more of a hybrid type of 10 core2 thread or something like that unfortunately the chiplet design just doesn't allow for it so even though we're uplifting the frequencies it's not enough to make up for the thread count difference uh the single thread though as you can see no uplift at all uh 132 to 132 and the reason for that like I said is PBO doesn't really do much for the max single core frequency if anything at all it's all about the all core frequency but the all the cinebench r24 multi-thread as you can see here once again a severe multi-thread disadvantage in fact these charts are sorted price left to right and you can see we got a very linear curve going up unfortunately when you get to the price of the 9700x there's a severe drop in performance because it's unfortunately its value proposition in multi-threaded workloads is just simply not there a lot of you right now going be going but the efficiency the efficiency wait till you see what happens to the efficiency when we enable PBO we're definitely going to show that now we look at geek bench single thread uh it goes from a 3303 to a 3314 again fairly margin of error the multi-thread though once again we go from a 16,063 to a 16 881 so we got 800 points over 800 points of uplift there but not enough to close that Gap with the 14700 K which is unfortunately cheaper than the AMD CPU um moving on to handbrake we do drop significantly we go from a 93 second transcode on our three uh over 3 minute 4K video down to 1080p U it's important to note that we are not using Quicks sync or or any of that stuff when it comes to Intel this is just raw CPU x264 encoder but we dropped from 93 to 85 un fortunately the 137 and the 147 are significantly faster at 68 and 63 so we do see a performance increase for a one cl one click button but it's not enough to make up that performance Gap that I was really hoping to see that Gap close because the all core frequencies are pretty gimped in my opinion out of the box at about 4.4 to 4.6 G all core it just wasn't enough to make up for it um gaming shadow of the Tomb Raider we see a moderate uplift from a 49 average fps to 258 average FPS again medium setting settings 1080p no scaling no RTX 4090 just pushing all the frames we can to see how the CPU can handle it uh but you can see the 137 and the 147 that are flanking it um outperforming it by more than 20 FPS and up to 30 FPS in some instances there so a lot left to be desired we did see a pretty good jump though in Times by extreme we went from a 6589 to a 7634 so it shows that there are instances where we can get a pretty good jump in performance but unfortunately tests like this are designed to really have the chart emphasize an improvement so that type of percentage Improvement doesn't translate to that type of percentage Improvement in games unfortunately and then if we look at Cyber Punk a weird title that just doesn't seem to favor AMD I think this is something a lot of people have realized already it is definitely a heavily Intel optimized title as far as we can tell cuz you can see the 7600 X versus 9600 X versus 7700x versus 9700x and 9700x with PBO enabled are all like pretty much the same which is a weird almost like an artificial cap that's happening there and their minimum FPS is also nearly the same across the board it's very odd it's almost telling like this is an intentional type of bottleneck or something for AMD CPUs but when we look at the Intel CPUs we can see the same sort of thing the 13600 K and the 14900 K with the latest micro code is 218 versus 229 so Intel is seeing the same type of like across the board sort of cap regardless of the tier of of the skew it's just a higher cap and then their their minimum FPS is also a very similar story but it's much higher in the 180 190 range versus the 130 Range found on AMD so cyber Punk's one of those titles which I feel like heavily heavily favors Intel in this instance let's talk about the charts over time because this is where the the real differences uh show let's start with core clocks here so you'll notice the green chart right here which is core clock average megahertz across you know obviously all the cores this is a 10-minute cinebench R23 multicore run you can see there's some frequency spikes as the test is loaded which is normal we've showed this as the test gets loaded it's more of a single core type of a workload to load the test so we see a frequency Spike well over 5,000 MHz up to almost 5.5 MHz single core but then as soon as all cores go under load it drops down significantly our particular run was running about 465 GHz on average but you can see that that line is very wavy it's all over the place um temperatures were pretty solid so that's why we're not seeing a curve downward or anything because we're not seeing thermals really come into play when it comes to the frequencies when we enable PBO you can see we jump up to 5.3 to 5.4 GHz all core and it's definitely locked like it is a more Flat Line because that's as far as the frequency scaler will allow PBO to push it so it definitely hits that 53 54 Mark kind of fluctuates between like those 50 MHz there between like 535 54 and then you can see the individual spikes uh as it's loading the test those jump up to about 56 on the single core now it shows here in this test that the single core does jump up but on the other tests as we showed you um which do a longer test under single core don't really seem to benefit um the single core bump because it doesn't stay up there it drops back down to like this 5554 range on single core uh and stays there so that's why the single core threads on like geekbench and r24 and R23 don't really show any sort of an uplift over time so you can see by adjusting the PBO we get a nearly 1 GHz bump in performance it's something I wish that AMD had sort of allowed out of the box and maybe not heavily emphasize the efficiency and maybe push these clocks a little bit more out of the box and maybe made these like an 95 watt part 85 watt part something like that versus 65 but when we look at the Watts it takes to do that this isn't this is a this is a scary chart I've seen some people comment saying any CPU that draws over 150 watts is scary to me well this should be terrifying to you because if you look at our PPT Watts which is um the actual power package power for the CPU itself that's memory controller the whole deal not just the core but the whole package of the CPU you can see it goes from about 88 Watts solid a straight line no fluctuation at all to over 160 Watts averaging about 158 Watts so double the power draw for the small amount of performance uplift that you saw that's the part that's a little disappointing here is the fact that PBO was able to push the frequencies but it pushes the power to double for up to 10% performance gain that's a bad trade-off that's where the efficiency goes out the window with these particular CPUs but what about the temperature it took to do that well if we look at the T the T diey temperature here um this is a more accurate measurement of amd's like they have core package temp you have package Temp and then you have T diey T die is what we're looking at here because this is the more sensitive number that actually handles a lot of PBO scaling and stuff you can see we go from the mid-50s spiking up to about 57 58 while we do the single cord load and that's that single core that jumps up that it's measuring that that die to up to 90c so it actually pushes it up to 90c cuz that's what I set because if you enable PBO it's still 85c I want to see where it would if we push beyond that 85 how high will that actually go the numbers went red in Hardware info 64 the numbers are red cuz anything over 85 is considered Beyond spec so it pushed itself to 90c makes sense considering the fact that we're talking double the wattage but you can see right now with with the chiplet design and it's so tightly packed in the way that the the whole substrate is designed it becomes very focused heat so we go from 55c to 90 which is nuts but double the power to do it makes sense that we would see that type of temperature climb and the voltage the voltage to do it the the one thing that we were super excited about with this because of the super efficient frequencies and the power limit it only took about 1.05 volts to be able to run the CPU at itss out of the box specs with spikes though leading up to about 1.35 volt but the spikes in the chart as you already know is every time it reloads R23 uh cin bench R23 it spikes but check out the orange line which are which is our PBO enabled what I'd like to see though is you can see that the green spikes and the orange spikes are the same not like the Intel thing where it pushes it way up higher so AMD does have a pretty hard-coded voltage limit in there that says regardless of PBO on or off don't exceed this voltage and it's very obvious when you look at the green line versus the orange line but look at the Baseline when it comes to the average frequencies here to maintain in allcore overclock we're talking about 1.28 to 1.32 volts is what we were averaging according to Hardware info so a significant voltage bump which as you saw led to a significant PPT increase which also led to a significant temperature increase taking all of the efficiency benefits of this chip out of the window and giving us about 10% more Improvement for double the power it is a one button overclock it does allow itself to be able to give you more performance and if efficiency is not that big of a deal to you I saw a lot of comments saying the efficiency is great but we wish AMD hadn't leaned so heavily on the efficiency discussion and maybe lean more into raw power this is what happens when you do that I feel like AMD could have split the difference here somewhere in the middle they are very appealing chips to people that are not interested in any of the Intel drama right now and don't want to even consider going down that route with a CPU that you can't trust whether or not it's already degraded or the one that you're buying doesn't already have potential oxidation or whatever you have a plethora of drama you can pick from right now to say I have buyer uh anxiety and it's all valid AMD right now you just have to decide whether or not their price to performance on these chips whether PBO enabled or not are worth it to you and as a buyer that's 100% your decision I can tell you people like me don't give a rat's ass about what you buy it's your money it's your decision all we do is present the data here as it presents itself the numbers are what they are so PBO enabled on the 20 or the 9700x did obviously lead to an increase a bigger increase than we typically see on the higher core count chips with the 16 core 32 thread chips because of the fact that there is PPT and temperature Headroom in those the more dense core count chips tend to not benefit from PBO quite as much because they're already near that Headroom out of the box so there you go guys PBO numbers you guys said why aren't you showing PBO well cuz I always do this as a separate video cuz I want to play around with them some more anyway we've got two more 9,000 Series CPUs that need to be tested so we got to get on out of here and uh I may or may not include PBO at the same time with those videos we'll see we've got a lot of regression testing here our charts are lacking a lot of data and I'm trying to fill these charts in with data to make it more relevant for you it's taking time to catch up with what other people have established so do yourself a favor I know you already do check out as many reviewers as you can to get their takes on these CPUs because everyone kind of comes at these tests a little differently thanks for watching guys thanks for supporting us and as always we'll see you in the next one oh and please give me the data points that you think are important to you I want to know what information you think is relevant and I'm trying to get it all in there and sometimes what I think is relevant you don't and vice versa so sound off in the comments below or hit me up on Twitter and as always I'll see you guys in the next one\n"