The Power of AMD's Ryzen 7 2700 X: A Performance and Value Review

In recent years, the CPU market has seen significant advancements with the introduction of new technologies and architectures. One of the key players in this space is AMD's Ryzen series, which has consistently delivered high-performance processors at competitive prices. In this article, we'll take a closer look at AMD's Ryzen 7 2700 X, a flagship processor that offers exceptional performance and value.

The Ryzen 7 2700 X is a behemoth of a processor, boasting eight cores and a 3.4 GHz base clock speed. However, with the inclusion of AMD's SenseMI technology, the processor can automatically adjust its clock speed to match the workload, ensuring optimal performance in games and content creation applications. This level of flexibility and overclocking headroom is unparalleled in the market, making it an attractive option for enthusiasts who want to push their processors to the limit.

One of the standout features of the Ryzen 7 2700 X is its compatibility with AMD's Wraith Prism CPU cooler. In our testing, we found that this cooler provides excellent cooling performance without sacrificing much in terms of noise levels. The fan selector feature allows users to customize the fan speed to suit their needs, making it an ideal choice for those who want to balance performance and noise.

Our testing revealed that the Ryzen 7 2700 X is capable of delivering outstanding performance in a wide range of workloads. In our gaming benchmarks, we saw the processor deliver results comparable to Intel's i-5 processors, with some titles even outperforming them. The single-threaded performance was impressive, but it was the multi-core capabilities that truly shone in our testing.

In terms of value, AMD has consistently delivered on its promise of delivering exceptional performance at competitive prices. When compared to Intel's i-5 processors, the Ryzen 7 2700 X offers significantly better value, especially when overclocked. In fact, we found that the processor was able to deliver similar or even superior results to Intel's flagship processors in certain workloads.

However, it's worth noting that AMD's marketing message of a 5-10% performance increase is somewhat modest. While this may not be entirely unfounded, our testing revealed that the processor delivered significantly better performance than expected, especially in multi-core workloads. In fact, we saw some instances where the Ryzen 7 2700 X was able to deliver results up to 40-50 megahertz faster than Intel's i-5 processors.

In conclusion, AMD's Ryzen 7 2700 X is an exceptional processor that delivers outstanding performance and value. With its eight cores, 3.4 GHz base clock speed, and SenseMI technology, it's an attractive option for enthusiasts who want to push their processors to the limit. The Wraith Prism CPU cooler provides excellent cooling performance without sacrificing much in terms of noise levels. While AMD's marketing message may be somewhat modest, our testing revealed that the processor delivers significantly better performance than expected.

A Word from the Nerd: Thanks and Feedback

I'd like to extend my gratitude to Brad Morris for helping me crunch the numbers and do some analysis on this article. His expertise was invaluable in understanding the performance capabilities of the Ryzen 7 2700 X. I'd also like to thank Mike CFM for his help in setting up our test systems, which were used to generate the data in this article.

To all our readers out there, we appreciate your interaction and feedback. If you have any questions or comments about this article, please don't hesitate to reach out. We're always looking for ways to improve and provide more valuable content to our audience. Thanks for watching, and we'll catch you in the forums at Level One!

"WEBVTTKind: captionsLanguage: enRison mm has launched and we've been hard at work poking and prodding it and seeing what it has to offer unlocking its secrets it doesn't make the zelda knows when you unlock it though but we've got an interesting story and it's gonna be a long one at least it's interesting to me Rison mm sure it's an incremental update you know these second generation processors are what's called Zen plus and so really this is just a refinement of the first generation rising CPUs that came out last year you know this it was a big deal I mean it's new chipset and you CPU to microarchitecture it's crazy to do all that at once and then was a very very ambitious product well a project in general I mean not just a product and we've got new motherboards and new chipsets this time around you can still mix and match CPUs on a m4 just like AMD promised but you also get some perks with the new x4 70 boards perks like seriously fast memory that's a huge improvement over the last generation of boards especially because Rosen's performance it can improve pretty significantly with any uplifts that you can do to memory speed or decreases in latency by clocking memory faster and that sort of thing so to be clear yes you can totally use a 2000 series rise in CPU in a first generation board the you know the rise in 720 700 X so though I would caution you you know that's it's 105 watt CPU we've seen it use power up to 140 watts I would caution you not to put that in a be 350 board or you know a first gen X 370 board that might have a questionable vrm implementation if you get a high-end X 370 board you can you can totally do that and that'll totally be fine for the other CPUs it doesn't really matter as much because those CPUs don't use near as much power but if you want the flagship see for you you know it's probably worth it to get the perks that you get with the extra board it's not just the memory space there's a lot of other things like store in my we did a separate video on storm I should check that out there are other little perks to well if I could if I could interject there for a second I mean if you got a great first chitin board like the X 370 gaming pro carbon Mossad or the Ezreal Itachi here thinking about gaming five or another you know a higher-end X 370 board you know on those first invoice of first gen CPUs we were seeing Dior for 3200 basically at the top in maybe three thousand twenty-nine thirty threes a little bit more stable and mixing and matching the second gen CPUs with the first gen boards it doesn't really seem like the situations improved for anything maybe it's regressed a little bit this is anecdotal testing it's not formal testing but it does seem like 29 to 33 with those second gen chips it's a little bit more stable than the faster speeds so it's also true that it's kind of diminishing returns past 29 33 past 3200 I mean if you can tighten the timings often you're better off than the higher clock speed anyway because when you increase the memory speed you're pushing the Infinity fabric and and that whole thing plays into it so you can mix and match CPUs but it does think we could get a better memory speed on the second generation boards but of course diminishing returns past 29 33 if 3200 so just you know way that for what it's worth all right I'll let's get back to it so we're seeing one big improvement one obvious one and one not so obvious one one that's a little bit spooky and one that Amy hasn't really mentioned but that we've experienced personally and that's why I say this videos gonna be a little longer a little more difficult for me to explain but first for the quick and dirty yeah yeah I give it to me straight I don't really care about all your egghead mercury I just want to know you know give me the bottom line what's up with this alright the 2700 X versus the six core Intel i7 maybe 700 K the 2700 X holds its own it's we're basically at parity with Intel there are certain scenarios where the 2700 X will outperform the Intel i7 87 RK yeah I know shocker and there are certain scenarios where the 87th K will be marginally faster we're talking 1080p gaming and 1440p gaming mainly you know 1440p and 4k gaming tend to be more as the GPU bound than CPU bound so there's honestly not much difference between these CPUs for gaming type workloads and the AMD fanboys right now we're like free yes we you know AMD masterrace everybody wins AMD better than Intel whoo that's not come on nobody needs a fanboy these both CPUs are seriously impressive pieces of engineering they really are and they're different critters they approach problems in different ways AMD's approach is very pragmatic I mean you know honestly there's you know the ragtag underdog of Awesomeness is able to do you know unseat the giant behemoth I mean you know it's it's any kind of David and Goliath story no matter how you want to cut the cheese but it's actually a good product like you know just set all that aside I don't have time for that I don't care about the backstory or anything I just want something that works it's good product does really well and it does even better than Intel and things that aren't gaming which we'll talk about in a second so where did the gaming performance improvement come from I mean the gaming performance on 1st gen Rison was was there except in high frame rate games and it was really hard for first gen Rosen to deliver you know that 150 FPS plus gaming experience so this is thanks in large part due to improvements in the latency on the platform and so what are we talking about what do you mean latency so we've got the level 1 the level 2 in the level 3 cache memory current level 1 get it there's memory that's on the CPU it's really high speed memory it's really close to the the computational mechanism of the CPU and the data that the CPU is working on is stored there because system memory is too slow because literally everything else in the entire system is too slow and level 1 is insanely way faster than level 2 is insanely way faster than level 3 cache memory and then your your memory like the system memory the ddr4 memory is just glacial in comparison you know the level 1 in the level 2 cache memory are operating on the order of nanoseconds I mean the level 2 cache memory is typically like 12 nanoseconds or 12 cycles or well 12 cycles which is actually less than 12 nanoseconds because the clock speeds faster than 1 gigahertz so anyway anyway it is just a massive massive engineering accomplishment that AMD has been able to do that I cannot overstate that it is just it is mind-blowing and it's really awesome and it's really awesome that they've been able to improve that between Zen and Zen plus I thought sure that it was gonna be Zen too but that was really improved his we were seeing you know whom the benchmarks for the first-generation rosin it's really awesome to see that that is fixed because holy cow those fixes basically fixed a lot of other stuff as we'll see in just in just a minute so story time its benchmarks you know it's all about the benchmarks right well not so fast you know we had a mystery on our hands when we were benchmarking the 1000 series rise in CPUs it was there was some stuff happening that didn't make sense from a computer engineering for a computer science standpoint and it's not really an easier cleaner way to explain that that's one of the reasons why this video is coming gonna drag on hopefully the stories is interesting so there are a ton of benchmarks around the web and we're not really super good at benchmarking yet I mean we try but we the people not we level one have got to get an automated and repeatable benchmarking platform something to automate games something to be sure that it's like I'm gonna run this game benchmark with exactly the same settings and all the same stuff as you know a reviewer or as somebody else just one to be sure that my machine is performing the way that it should be in case I haven't missed something obvious it's like whoops my memory's running at 2133 it's like I've had a brick tied to my head for all these years and all I had to do is just turn on one thing and it was gonna be a way better computing experience there are people out there right now that are experiencing that it doesn't have to be that way we have the tools to make that better I think so you know I've also seen a lot of squabbling between and among the tech press you know about performance anomalies and game settings and all this kind of stuff and you know life's too short for that we don't it doesn't make sense to get all worked up about that and synthetic benchmarks are also kind of annoying and not very good well we really just need are repeatable benchmarks from actual like real world using an application this is gonna help buyers verify their systems are working you're performing like they should and not just with games I mean you know there's things like the FIR Onix test suite which is really good support on Windows now I've used it on on Linux for a lot of our videos it's pretty well so let's think about something like that but on steroids and lots of well-maintained tests now the spooky thing I mentioned the spooky thing what's this what's the spooky things like well the memory did latency discussion kind of leads into it so you might recall we did a lot of GTA 5 testing last year like just a crazy amount because we're gluttons for punishment I guess and when we tested the first gender Aizen we noticed something weird the experience when playing the game was better but the frames per second was lower and so we had a lower fps but playing the game was actually a better experience that's kind of weird we did a lot of testing and we did a lot of high-speed footage and so what we figured out at least our guess we're pretty sure was that the game engine spazzes out past 140 FPS so like from 140 FPS to 180 7.5 ft s the gta5 engine is exponentially more likely to spaz out because the gta5 engine is basically garbage hypothesis why do we say that well when we were testing on the first gen Rison CPUs it just couldn't push above about 140 FPS and the result of that was that it was a smoother experience than when we were testing on a 7700 K which would routinely hit 187 point five eight FPS it was like it hit a 187 point 5 FPS and there was a trip over his own shoelaces it was pretty much guaranteed but there was a chance that it would just randomly trip over its own shoelaces even not hitting 180 7.5 FPS it was really easy to see in the fraps crafts it was really easy to see pretty much everywhere and we did not get anywhere near that kind of behavior on 1st gen horizon and that drove was crazy we really wanted to sort of figure that out so in other games too like fallout the extra CPU horsepower who really helped a lot it really helps the game chug through some of the scenes and some of the situations where it needed a lot of CPU horsepower I'm talking about first gen rise in here versus the relatively peeny for core 77 third K of the area so it's like why am i why am i rehashing all of this I don't I'm gonna really understand it's like well the fallout 4 engine is also pretty broken had had a frame locked by default that you had to turn off and all this other kind of stuff so so unrelated but kind of related when we look at the benchmark data and artificial benchmarks like fire strike we noticed that the combined score was lower than we would have expected given the relatively high scores of the physics and graphics components of that test now so that's just like GTA 5 where GTA if I was a pretty CPU intensive game so when we got the combined CPU and GPU stuff going on you know rosin just couldn't the first gen rosin just couldn't push past 140 FPS so it seemed like something was bottlenecking so that bottleneck as it turns out was likely related to that extra latency that we were saying so just a few extra cycles just a few extra nanoseconds on every load and store operation was sort of limiting us I guess at least that's that's the current guess so shuffling things between CPU and memory but also i/o sending packets down the PCI Express bus the latency is higher than the system designer expected I think so that can reduce the effective available bandwidth the Rison platform was new you know the Intel platform had a certain amount of latency in it the rasen platform had a little bit more latency peripherals not maybe not designed for that programs might be not designed for that so maybe some of what we were seeing was the extra latency in the platform sort of manifest as well GTA just can't go past her in 40 FPS and you get a smoother experience in GTA as a result because the game engine designers never figured anybody would have a system that can do you know that Phi I mean Rockstar for god sakes was probably designing for a console that's like ah 30 FPS you're doing pretty good yeah I'm a rider am i right yeah a PC master-race I don't know that was for the uploads so so we couldn't 100% confirm it though it's like how do you confirm that it's like well there's their obscure Nvidia driver options to limit GTA 5 to 140 FPS and so at that point the smoothness experience on both systems was identical but you know we we couldn't push we didn't close the loop they're closing the loop there would be push the rising system to 180 7.5 ft s or high well you know some above 140 fps and see if you could get the glitching and the stuttering and all this kind of stuff and whoo we did boy howdy we did we had for a day or two we had the world record for the fastest single GPU 1800 X system in fire strike yeah pretty nuts and that part of the reason that we did that was to try to see if we could push you know push through you know faster memory clocks and whatever we did all kinds of crazy stuff you just don't even know and we could not get GTA to perform really well so long story short why does any of that matter why why is that woody were you to rise in 2000 so what we're talking about in this video rise in 2000 so with rise of 2000 this magically insanely way better CPU can we get those crazy high FPS GTA 5 yes we can we are in fact getting those crazy high FPS in GTA 5 and does the engine break oh yes yes absolutely the engine just but it just just doesn't eat like it just trips over its own shoelaces it is pretty much at parity with Intel so all of the game breaking weirdnesses and bugs is like you're flying with the jet and the benchmark and the whole game just stutters yep that happens on rise in 2002 just as well is it you know does on on Intel whoo I mean I think that's actually a great result there's no like these latency improvements have translated into real-world performance improvements in pretty much everything we've also been working on a project called looking glass now Looking Glass is really exciting this technology lets you copy the frame buffer from Windows virtual machine to a Linux host so it's very interesting for us in terms of how long does it take you to copy a 4k 32 megabytes frame buffer from that windows host through shared memory from the windows guest I'm sorry through shared memory to the Linux host how long does it take to copy 32 megabytes if we can do it in a millisecond that means we can do a thousand FPS at 4k assuming the system doesn't have anything else to do but really we're shooting for like one 60 FPS so you know 60 out of a thousand gives the CPU because the rest of the system a lot of time to do a whole bunch of stuff and so we always look how long does it actually take to do that and to just give you a baseline on our test systems all of which have thirty-two hundred megahertz memory the same g.skill memory kit thanks g.skill that's really awesome t school memories like worked out super amazing do you skill memory with both of these check it out 3400 3600 it's gonna be all part of our stuff but I digress so yeah 32 megabytes on the thread Ripper system which is the fastest system that we've tested this on to date 0.8 milliseconds so yeah thread refer has got the memory bandwidth there's no problem at all but one point five milliseconds on an i-9 on a quad channel I nine that seems glacial in comparison why is the I nine so slow even the 3200 it makes no sense so then we've also got first gen rising systems also about 1.5 milliseconds for the dual channel configuration you know wha-wha not not very good so memory speed beyond 2400 on the Intel 87 okay doesn't really seem to make much of a difference and it's at about one point three milliseconds give or take the rise in 2000 time one point one to one point two milliseconds in our best-case scenario and that is running 29 33 with really really tight timings on the memory so we've tuned the memory a little bit we can get a much better or much worse time depending on what we're running our memory at so again memory bandwidth you know this memory speed affects the performance analyzing a lot more than it does on Intel now the really interesting thing about this 32 megabytes is kind of a lot of memory to copy and because the Rison system can do it faster than the Intel system that suggests that a lot of the benchmarks do a lot of a lot of small operations really rapidly and so when you do a lot of small operations really rapidly maybe the benchmark program waits for that memory operation to complete and then it does the next one and then it waits for that one to complete minute that's the next one and so it's again it's a little different critter than the way that it works on the rosin platform because well the rosin platform has a measurably higher maury latency when you you know when you I want to nail to your the same when you talk about layer three cache and cross CCX cores and then maybe it might have to go do something somewhere else but the different memory controller and then the memory actually getting stuff from memory and we can measure those times with benchmarking programs like a 264 and we know that they're worse but for us for lookingglass because 32 megabytes is kind of a lot of data it sort of makes up for it I guess because rosin is better able to take advantage of the higher memory bandwidth so even though the latency is a little bit higher we can we can get stuff there faster so think think about like you've got to move your entire existence from New York to San Francisco and you've got you know the car from Knight Rider which you can do 600 miles an hour and you get a straight shot from you know Elon Musk he's dug you a tunnel and so you can load up all your furniture in the Knight Rider car and you know do a lot of back and forth is 600 miles an hour from New York to San Francisco or you've got a box truck and the box truck can only do a hundred miles an hour you know flat plank running down the highway but it's a box truck and so it's not really like really it's like a small box truck versus a large box truck there's not really that much difference here in in terms of speed it's a little exaggerated but for us for looking glass the rise in memory performance because it's good it seems to be better able to take advantage of the higher memory bandwidth you get from the higher clock memory more so than Intel it sort of makes up for the latency deficiency a little bit and in that one particular use case now know you know it's gonna be different from game to game it's gonna be different from from this that any other but if you look at all the games benchmarks around the internet they've basically closed the gap and even for like oddball things like our Looking Glass project the gap was closed and then some so that's really exciting to see from an engineering perspective as well so IMD with those bundled coolers this is the rosin 5 bundled cooler it's basically the non RGB version of the same cooler that we saw with the risin 7 1700 last year you also got the wraith prism you know RGB that comes with the 2700 X flagship CPU both of those coolers are perfectly fine and as long as your case has adequate ventilation you will totally see the on the Box performance for both CPUs you you don't need to get an aftermarket cooler so we say and the stilt agrees with us that out of the box the 2700 X stock cooler can already performed exceptionally well at manual overclocking will often hurt single core performance so you really shouldn't do manual overclocking on the 2700 X and we can agree with that I'll go a step further and say that if you do want to overclock you can forget overclocking in the traditional sense and you're just gonna have to let the CPU do its thing and you're probably gonna need something crazy like an all-in-one cooler like a beefy all-in-one cooler or something even better like that now in our testing with a good cooler we say 3.9 gigahertz to 40 50 megahertz on all core clock speeds but we could manually overclocked to 4.1 to 4.8 with a good all-in-one cooler but we lost performance in one to four threads just as we predicted so Andy I think has anticipated this situation and given us more options for still letting us push past that limit so you know the 43 50 megahertz xfr to limit which is you know out of the box out of the box on your on your isin 743 50 megahertz you really want to try to hang on to that and you can do that on X 470 motherboards for sure maybe on X 370 motherboards it's really up to the board vendor to add support for that full support I mean partial support something like that so you understand what's happening under the hood the CPU is monitoring its thermal limits its power wattage limits and its reliability so it's like three points of a triangle it's basically keeping track of you know am I am I about to lose my marbles like am I gonna have like a stress freak out like did you know and just lose everything what's my temperature like and how much power am i drawing now how it exactly precisely behaves it's gonna vary from workload to workload if you're doing something like running pron 95 you're gonna see a much lower all core boost speed at least we did in our testing but even for doing things like running a rise of the Tomb Raider you know we were seeing 40 200 megahertz 4140 200 megahertz across all cores without really well I don't wanna say all course it's like you know four cores I'm seeing it on like 4 cores but when we did our you know customized super fancy overclocking we were still able to get the 4.2 gigahertz all core when the CPU really as long as we can keep it cool and we still got to keep our 43 50 megahertz boost clogged so in the comparison with the 8700 K you know can you overclock the 87 RK and get back some more of the performance edge it's like me you know it's like level one you're telling me that they're basically a parody but I can overclock the 8700 K it's like well yes and no when you overclock the 87 hard K you're really just over talking the core most of the time you should also overclock the cache and other parts of the CPU they all sort of clock independently so when you add you know the 4.5 to 5 gigahertz if you don't overclock everything else it's really kind of diminishing returns you can get a little bit of a lead but in terms of stability heat production after market cooler that you definitely definitely will need because they're run away thermals on the 8700 K I mean I don't its you can yes that's not wrong but would you rather just get a CPU that just you know you just plug it in and use the stock cooler and it already works great so overall you know the 2700 X which is mostly what we've talked about both for the rosin 5 2600 X Verizon 5 2600 X I think we're gonna have some separate coverage on that because it's an interesting CPU in and of itself 6 cores 12 threads short version is the 2600 X is the value king it's just it's an almost unbeatable value because you know right in the box you've got a good cooler that is more than adequate to keep that CPU cool but this thing's gonna get a turbo up to 4.2 gigahertz or you know xfr to up to 4.2 gigahertz with this cooler basically no problems so great performance in games great performance compiling and core clothes productivity you know content creation whatever that you might want to do it's it's really you can't you can't beat the value you just can't I mean even if you got an i-5 and you overclock the Jesus out of it it's still the 27 the 2600 X is gonna give the i-5 a run for its money and that's just what a time to be alive really it's just it's great with the gaming benchmarks we've got GTA 5 rise of the Tomb Raider and Civ 6 you know as I mentioned the perform essentially at parity the Intel test system is an Intel 8700 K with multi-core enhancement on so that's 4.7 gigahertz all core turbo so that's a little bit of an overclock in the Intel CPU in our comparisons we're using a data ddr4 2400 dual channel with the founders edition 1080i and all our test systems now for our workloads that can actually take advantage of you know eight cores things really skew in favor of Rison so I'm talking about you know development compiling code working with databases multi-threaded applications things like handbrake rising just you know the risin seven twenty seven hundred x with its eight cores just it wins hands down we typically saw 40 50 megahertz only in things like prime95 that it dip down to like 37 five 3800 something like that which is the advertised base clock on the risin 7 2,700 X but we almost even in eight core workloads we almost never saw it actually operating at the base clock now we didn't have the fan selector on our you know Wraith prism CPU cooler set to high so the CPU cooler does get quite loud well not quite loud it's not as loud as a blower sometimes as loud as the founders Edition 1080 TI does get pretty loud so alright based on all this you know kept the nerdery it's like I don't miss that in my life based on my experience with the 2700 X AMD's claim of five to ten percent important a performance improvement is true there's some Asterix there like if you bought say the risin 1700 and overclocked the bejesus out of it you know 3.9 gigahertz 4 gigahertz maybe 4.1 gigahertz on the 1700 it was the deal of the century it was about $300 I think AMD has delivered on that value I mean it's a year later and so we expect at 5 to 10 percent performance bump but if you look at we've come from the 1700 to the 2700 x better cooler 4.3 gigahertz out of the box for you know single around 4.3 gigahertz dual thread workloads gaming performance is basically up there because they really have split the difference between single thread performance the multi-core performance now Canada overclocked as much as Intel no will it deliver an amazing experience nonetheless yes it absolutely will so based on my experience with the 2700 X I could say that Amy's marketing message of about a 5 to 10 percent performance increases is actually pretty modest I think most users in most scenarios would definitely see more than a 10 percent performance uplift especially more than a 10 percent performance uplift per dollar the value of these things really is just I mean it's great I mean it well I guess without saying I mean competition in the cpu market the consumer wins so this is no matter which CPU you're leaning toward this only benefits you the consumer because we've got real actual competition in the marketplace this is a great product it does exactly what it says it does and it does it well that's all you can ask of a product I think it's pretty good I'd also like to give a shout out to Brad Morris for helping me crunch the numbers and do some analysis here and also to Mike CFM for helping me put the test systems together that we use in our testing for I'll just this video but also a couple other videos some of which are probably not out yet and some of which probably already came out with the envy me rate video so kudos thanks I thought this video would come out first but editing was tricky and the performance numbers and we had to do some analysis so hey I'm Windell this is level one should like and subscribe and hit the bell or thumbs down if you want I don't care just some sort of interaction to feed the algorithm or at least that's what I'm told so I'll see you the forums at level oneRison mm has launched and we've been hard at work poking and prodding it and seeing what it has to offer unlocking its secrets it doesn't make the zelda knows when you unlock it though but we've got an interesting story and it's gonna be a long one at least it's interesting to me Rison mm sure it's an incremental update you know these second generation processors are what's called Zen plus and so really this is just a refinement of the first generation rising CPUs that came out last year you know this it was a big deal I mean it's new chipset and you CPU to microarchitecture it's crazy to do all that at once and then was a very very ambitious product well a project in general I mean not just a product and we've got new motherboards and new chipsets this time around you can still mix and match CPUs on a m4 just like AMD promised but you also get some perks with the new x4 70 boards perks like seriously fast memory that's a huge improvement over the last generation of boards especially because Rosen's performance it can improve pretty significantly with any uplifts that you can do to memory speed or decreases in latency by clocking memory faster and that sort of thing so to be clear yes you can totally use a 2000 series rise in CPU in a first generation board the you know the rise in 720 700 X so though I would caution you you know that's it's 105 watt CPU we've seen it use power up to 140 watts I would caution you not to put that in a be 350 board or you know a first gen X 370 board that might have a questionable vrm implementation if you get a high-end X 370 board you can you can totally do that and that'll totally be fine for the other CPUs it doesn't really matter as much because those CPUs don't use near as much power but if you want the flagship see for you you know it's probably worth it to get the perks that you get with the extra board it's not just the memory space there's a lot of other things like store in my we did a separate video on storm I should check that out there are other little perks to well if I could if I could interject there for a second I mean if you got a great first chitin board like the X 370 gaming pro carbon Mossad or the Ezreal Itachi here thinking about gaming five or another you know a higher-end X 370 board you know on those first invoice of first gen CPUs we were seeing Dior for 3200 basically at the top in maybe three thousand twenty-nine thirty threes a little bit more stable and mixing and matching the second gen CPUs with the first gen boards it doesn't really seem like the situations improved for anything maybe it's regressed a little bit this is anecdotal testing it's not formal testing but it does seem like 29 to 33 with those second gen chips it's a little bit more stable than the faster speeds so it's also true that it's kind of diminishing returns past 29 33 past 3200 I mean if you can tighten the timings often you're better off than the higher clock speed anyway because when you increase the memory speed you're pushing the Infinity fabric and and that whole thing plays into it so you can mix and match CPUs but it does think we could get a better memory speed on the second generation boards but of course diminishing returns past 29 33 if 3200 so just you know way that for what it's worth all right I'll let's get back to it so we're seeing one big improvement one obvious one and one not so obvious one one that's a little bit spooky and one that Amy hasn't really mentioned but that we've experienced personally and that's why I say this videos gonna be a little longer a little more difficult for me to explain but first for the quick and dirty yeah yeah I give it to me straight I don't really care about all your egghead mercury I just want to know you know give me the bottom line what's up with this alright the 2700 X versus the six core Intel i7 maybe 700 K the 2700 X holds its own it's we're basically at parity with Intel there are certain scenarios where the 2700 X will outperform the Intel i7 87 RK yeah I know shocker and there are certain scenarios where the 87th K will be marginally faster we're talking 1080p gaming and 1440p gaming mainly you know 1440p and 4k gaming tend to be more as the GPU bound than CPU bound so there's honestly not much difference between these CPUs for gaming type workloads and the AMD fanboys right now we're like free yes we you know AMD masterrace everybody wins AMD better than Intel whoo that's not come on nobody needs a fanboy these both CPUs are seriously impressive pieces of engineering they really are and they're different critters they approach problems in different ways AMD's approach is very pragmatic I mean you know honestly there's you know the ragtag underdog of Awesomeness is able to do you know unseat the giant behemoth I mean you know it's it's any kind of David and Goliath story no matter how you want to cut the cheese but it's actually a good product like you know just set all that aside I don't have time for that I don't care about the backstory or anything I just want something that works it's good product does really well and it does even better than Intel and things that aren't gaming which we'll talk about in a second so where did the gaming performance improvement come from I mean the gaming performance on 1st gen Rison was was there except in high frame rate games and it was really hard for first gen Rosen to deliver you know that 150 FPS plus gaming experience so this is thanks in large part due to improvements in the latency on the platform and so what are we talking about what do you mean latency so we've got the level 1 the level 2 in the level 3 cache memory current level 1 get it there's memory that's on the CPU it's really high speed memory it's really close to the the computational mechanism of the CPU and the data that the CPU is working on is stored there because system memory is too slow because literally everything else in the entire system is too slow and level 1 is insanely way faster than level 2 is insanely way faster than level 3 cache memory and then your your memory like the system memory the ddr4 memory is just glacial in comparison you know the level 1 in the level 2 cache memory are operating on the order of nanoseconds I mean the level 2 cache memory is typically like 12 nanoseconds or 12 cycles or well 12 cycles which is actually less than 12 nanoseconds because the clock speeds faster than 1 gigahertz so anyway anyway it is just a massive massive engineering accomplishment that AMD has been able to do that I cannot overstate that it is just it is mind-blowing and it's really awesome and it's really awesome that they've been able to improve that between Zen and Zen plus I thought sure that it was gonna be Zen too but that was really improved his we were seeing you know whom the benchmarks for the first-generation rosin it's really awesome to see that that is fixed because holy cow those fixes basically fixed a lot of other stuff as we'll see in just in just a minute so story time its benchmarks you know it's all about the benchmarks right well not so fast you know we had a mystery on our hands when we were benchmarking the 1000 series rise in CPUs it was there was some stuff happening that didn't make sense from a computer engineering for a computer science standpoint and it's not really an easier cleaner way to explain that that's one of the reasons why this video is coming gonna drag on hopefully the stories is interesting so there are a ton of benchmarks around the web and we're not really super good at benchmarking yet I mean we try but we the people not we level one have got to get an automated and repeatable benchmarking platform something to automate games something to be sure that it's like I'm gonna run this game benchmark with exactly the same settings and all the same stuff as you know a reviewer or as somebody else just one to be sure that my machine is performing the way that it should be in case I haven't missed something obvious it's like whoops my memory's running at 2133 it's like I've had a brick tied to my head for all these years and all I had to do is just turn on one thing and it was gonna be a way better computing experience there are people out there right now that are experiencing that it doesn't have to be that way we have the tools to make that better I think so you know I've also seen a lot of squabbling between and among the tech press you know about performance anomalies and game settings and all this kind of stuff and you know life's too short for that we don't it doesn't make sense to get all worked up about that and synthetic benchmarks are also kind of annoying and not very good well we really just need are repeatable benchmarks from actual like real world using an application this is gonna help buyers verify their systems are working you're performing like they should and not just with games I mean you know there's things like the FIR Onix test suite which is really good support on Windows now I've used it on on Linux for a lot of our videos it's pretty well so let's think about something like that but on steroids and lots of well-maintained tests now the spooky thing I mentioned the spooky thing what's this what's the spooky things like well the memory did latency discussion kind of leads into it so you might recall we did a lot of GTA 5 testing last year like just a crazy amount because we're gluttons for punishment I guess and when we tested the first gender Aizen we noticed something weird the experience when playing the game was better but the frames per second was lower and so we had a lower fps but playing the game was actually a better experience that's kind of weird we did a lot of testing and we did a lot of high-speed footage and so what we figured out at least our guess we're pretty sure was that the game engine spazzes out past 140 FPS so like from 140 FPS to 180 7.5 ft s the gta5 engine is exponentially more likely to spaz out because the gta5 engine is basically garbage hypothesis why do we say that well when we were testing on the first gen Rison CPUs it just couldn't push above about 140 FPS and the result of that was that it was a smoother experience than when we were testing on a 7700 K which would routinely hit 187 point five eight FPS it was like it hit a 187 point 5 FPS and there was a trip over his own shoelaces it was pretty much guaranteed but there was a chance that it would just randomly trip over its own shoelaces even not hitting 180 7.5 FPS it was really easy to see in the fraps crafts it was really easy to see pretty much everywhere and we did not get anywhere near that kind of behavior on 1st gen horizon and that drove was crazy we really wanted to sort of figure that out so in other games too like fallout the extra CPU horsepower who really helped a lot it really helps the game chug through some of the scenes and some of the situations where it needed a lot of CPU horsepower I'm talking about first gen rise in here versus the relatively peeny for core 77 third K of the area so it's like why am i why am i rehashing all of this I don't I'm gonna really understand it's like well the fallout 4 engine is also pretty broken had had a frame locked by default that you had to turn off and all this other kind of stuff so so unrelated but kind of related when we look at the benchmark data and artificial benchmarks like fire strike we noticed that the combined score was lower than we would have expected given the relatively high scores of the physics and graphics components of that test now so that's just like GTA 5 where GTA if I was a pretty CPU intensive game so when we got the combined CPU and GPU stuff going on you know rosin just couldn't the first gen rosin just couldn't push past 140 FPS so it seemed like something was bottlenecking so that bottleneck as it turns out was likely related to that extra latency that we were saying so just a few extra cycles just a few extra nanoseconds on every load and store operation was sort of limiting us I guess at least that's that's the current guess so shuffling things between CPU and memory but also i/o sending packets down the PCI Express bus the latency is higher than the system designer expected I think so that can reduce the effective available bandwidth the Rison platform was new you know the Intel platform had a certain amount of latency in it the rasen platform had a little bit more latency peripherals not maybe not designed for that programs might be not designed for that so maybe some of what we were seeing was the extra latency in the platform sort of manifest as well GTA just can't go past her in 40 FPS and you get a smoother experience in GTA as a result because the game engine designers never figured anybody would have a system that can do you know that Phi I mean Rockstar for god sakes was probably designing for a console that's like ah 30 FPS you're doing pretty good yeah I'm a rider am i right yeah a PC master-race I don't know that was for the uploads so so we couldn't 100% confirm it though it's like how do you confirm that it's like well there's their obscure Nvidia driver options to limit GTA 5 to 140 FPS and so at that point the smoothness experience on both systems was identical but you know we we couldn't push we didn't close the loop they're closing the loop there would be push the rising system to 180 7.5 ft s or high well you know some above 140 fps and see if you could get the glitching and the stuttering and all this kind of stuff and whoo we did boy howdy we did we had for a day or two we had the world record for the fastest single GPU 1800 X system in fire strike yeah pretty nuts and that part of the reason that we did that was to try to see if we could push you know push through you know faster memory clocks and whatever we did all kinds of crazy stuff you just don't even know and we could not get GTA to perform really well so long story short why does any of that matter why why is that woody were you to rise in 2000 so what we're talking about in this video rise in 2000 so with rise of 2000 this magically insanely way better CPU can we get those crazy high FPS GTA 5 yes we can we are in fact getting those crazy high FPS in GTA 5 and does the engine break oh yes yes absolutely the engine just but it just just doesn't eat like it just trips over its own shoelaces it is pretty much at parity with Intel so all of the game breaking weirdnesses and bugs is like you're flying with the jet and the benchmark and the whole game just stutters yep that happens on rise in 2002 just as well is it you know does on on Intel whoo I mean I think that's actually a great result there's no like these latency improvements have translated into real-world performance improvements in pretty much everything we've also been working on a project called looking glass now Looking Glass is really exciting this technology lets you copy the frame buffer from Windows virtual machine to a Linux host so it's very interesting for us in terms of how long does it take you to copy a 4k 32 megabytes frame buffer from that windows host through shared memory from the windows guest I'm sorry through shared memory to the Linux host how long does it take to copy 32 megabytes if we can do it in a millisecond that means we can do a thousand FPS at 4k assuming the system doesn't have anything else to do but really we're shooting for like one 60 FPS so you know 60 out of a thousand gives the CPU because the rest of the system a lot of time to do a whole bunch of stuff and so we always look how long does it actually take to do that and to just give you a baseline on our test systems all of which have thirty-two hundred megahertz memory the same g.skill memory kit thanks g.skill that's really awesome t school memories like worked out super amazing do you skill memory with both of these check it out 3400 3600 it's gonna be all part of our stuff but I digress so yeah 32 megabytes on the thread Ripper system which is the fastest system that we've tested this on to date 0.8 milliseconds so yeah thread refer has got the memory bandwidth there's no problem at all but one point five milliseconds on an i-9 on a quad channel I nine that seems glacial in comparison why is the I nine so slow even the 3200 it makes no sense so then we've also got first gen rising systems also about 1.5 milliseconds for the dual channel configuration you know wha-wha not not very good so memory speed beyond 2400 on the Intel 87 okay doesn't really seem to make much of a difference and it's at about one point three milliseconds give or take the rise in 2000 time one point one to one point two milliseconds in our best-case scenario and that is running 29 33 with really really tight timings on the memory so we've tuned the memory a little bit we can get a much better or much worse time depending on what we're running our memory at so again memory bandwidth you know this memory speed affects the performance analyzing a lot more than it does on Intel now the really interesting thing about this 32 megabytes is kind of a lot of memory to copy and because the Rison system can do it faster than the Intel system that suggests that a lot of the benchmarks do a lot of a lot of small operations really rapidly and so when you do a lot of small operations really rapidly maybe the benchmark program waits for that memory operation to complete and then it does the next one and then it waits for that one to complete minute that's the next one and so it's again it's a little different critter than the way that it works on the rosin platform because well the rosin platform has a measurably higher maury latency when you you know when you I want to nail to your the same when you talk about layer three cache and cross CCX cores and then maybe it might have to go do something somewhere else but the different memory controller and then the memory actually getting stuff from memory and we can measure those times with benchmarking programs like a 264 and we know that they're worse but for us for lookingglass because 32 megabytes is kind of a lot of data it sort of makes up for it I guess because rosin is better able to take advantage of the higher memory bandwidth so even though the latency is a little bit higher we can we can get stuff there faster so think think about like you've got to move your entire existence from New York to San Francisco and you've got you know the car from Knight Rider which you can do 600 miles an hour and you get a straight shot from you know Elon Musk he's dug you a tunnel and so you can load up all your furniture in the Knight Rider car and you know do a lot of back and forth is 600 miles an hour from New York to San Francisco or you've got a box truck and the box truck can only do a hundred miles an hour you know flat plank running down the highway but it's a box truck and so it's not really like really it's like a small box truck versus a large box truck there's not really that much difference here in in terms of speed it's a little exaggerated but for us for looking glass the rise in memory performance because it's good it seems to be better able to take advantage of the higher memory bandwidth you get from the higher clock memory more so than Intel it sort of makes up for the latency deficiency a little bit and in that one particular use case now know you know it's gonna be different from game to game it's gonna be different from from this that any other but if you look at all the games benchmarks around the internet they've basically closed the gap and even for like oddball things like our Looking Glass project the gap was closed and then some so that's really exciting to see from an engineering perspective as well so IMD with those bundled coolers this is the rosin 5 bundled cooler it's basically the non RGB version of the same cooler that we saw with the risin 7 1700 last year you also got the wraith prism you know RGB that comes with the 2700 X flagship CPU both of those coolers are perfectly fine and as long as your case has adequate ventilation you will totally see the on the Box performance for both CPUs you you don't need to get an aftermarket cooler so we say and the stilt agrees with us that out of the box the 2700 X stock cooler can already performed exceptionally well at manual overclocking will often hurt single core performance so you really shouldn't do manual overclocking on the 2700 X and we can agree with that I'll go a step further and say that if you do want to overclock you can forget overclocking in the traditional sense and you're just gonna have to let the CPU do its thing and you're probably gonna need something crazy like an all-in-one cooler like a beefy all-in-one cooler or something even better like that now in our testing with a good cooler we say 3.9 gigahertz to 40 50 megahertz on all core clock speeds but we could manually overclocked to 4.1 to 4.8 with a good all-in-one cooler but we lost performance in one to four threads just as we predicted so Andy I think has anticipated this situation and given us more options for still letting us push past that limit so you know the 43 50 megahertz xfr to limit which is you know out of the box out of the box on your on your isin 743 50 megahertz you really want to try to hang on to that and you can do that on X 470 motherboards for sure maybe on X 370 motherboards it's really up to the board vendor to add support for that full support I mean partial support something like that so you understand what's happening under the hood the CPU is monitoring its thermal limits its power wattage limits and its reliability so it's like three points of a triangle it's basically keeping track of you know am I am I about to lose my marbles like am I gonna have like a stress freak out like did you know and just lose everything what's my temperature like and how much power am i drawing now how it exactly precisely behaves it's gonna vary from workload to workload if you're doing something like running pron 95 you're gonna see a much lower all core boost speed at least we did in our testing but even for doing things like running a rise of the Tomb Raider you know we were seeing 40 200 megahertz 4140 200 megahertz across all cores without really well I don't wanna say all course it's like you know four cores I'm seeing it on like 4 cores but when we did our you know customized super fancy overclocking we were still able to get the 4.2 gigahertz all core when the CPU really as long as we can keep it cool and we still got to keep our 43 50 megahertz boost clogged so in the comparison with the 8700 K you know can you overclock the 87 RK and get back some more of the performance edge it's like me you know it's like level one you're telling me that they're basically a parody but I can overclock the 8700 K it's like well yes and no when you overclock the 87 hard K you're really just over talking the core most of the time you should also overclock the cache and other parts of the CPU they all sort of clock independently so when you add you know the 4.5 to 5 gigahertz if you don't overclock everything else it's really kind of diminishing returns you can get a little bit of a lead but in terms of stability heat production after market cooler that you definitely definitely will need because they're run away thermals on the 8700 K I mean I don't its you can yes that's not wrong but would you rather just get a CPU that just you know you just plug it in and use the stock cooler and it already works great so overall you know the 2700 X which is mostly what we've talked about both for the rosin 5 2600 X Verizon 5 2600 X I think we're gonna have some separate coverage on that because it's an interesting CPU in and of itself 6 cores 12 threads short version is the 2600 X is the value king it's just it's an almost unbeatable value because you know right in the box you've got a good cooler that is more than adequate to keep that CPU cool but this thing's gonna get a turbo up to 4.2 gigahertz or you know xfr to up to 4.2 gigahertz with this cooler basically no problems so great performance in games great performance compiling and core clothes productivity you know content creation whatever that you might want to do it's it's really you can't you can't beat the value you just can't I mean even if you got an i-5 and you overclock the Jesus out of it it's still the 27 the 2600 X is gonna give the i-5 a run for its money and that's just what a time to be alive really it's just it's great with the gaming benchmarks we've got GTA 5 rise of the Tomb Raider and Civ 6 you know as I mentioned the perform essentially at parity the Intel test system is an Intel 8700 K with multi-core enhancement on so that's 4.7 gigahertz all core turbo so that's a little bit of an overclock in the Intel CPU in our comparisons we're using a data ddr4 2400 dual channel with the founders edition 1080i and all our test systems now for our workloads that can actually take advantage of you know eight cores things really skew in favor of Rison so I'm talking about you know development compiling code working with databases multi-threaded applications things like handbrake rising just you know the risin seven twenty seven hundred x with its eight cores just it wins hands down we typically saw 40 50 megahertz only in things like prime95 that it dip down to like 37 five 3800 something like that which is the advertised base clock on the risin 7 2,700 X but we almost even in eight core workloads we almost never saw it actually operating at the base clock now we didn't have the fan selector on our you know Wraith prism CPU cooler set to high so the CPU cooler does get quite loud well not quite loud it's not as loud as a blower sometimes as loud as the founders Edition 1080 TI does get pretty loud so alright based on all this you know kept the nerdery it's like I don't miss that in my life based on my experience with the 2700 X AMD's claim of five to ten percent important a performance improvement is true there's some Asterix there like if you bought say the risin 1700 and overclocked the bejesus out of it you know 3.9 gigahertz 4 gigahertz maybe 4.1 gigahertz on the 1700 it was the deal of the century it was about $300 I think AMD has delivered on that value I mean it's a year later and so we expect at 5 to 10 percent performance bump but if you look at we've come from the 1700 to the 2700 x better cooler 4.3 gigahertz out of the box for you know single around 4.3 gigahertz dual thread workloads gaming performance is basically up there because they really have split the difference between single thread performance the multi-core performance now Canada overclocked as much as Intel no will it deliver an amazing experience nonetheless yes it absolutely will so based on my experience with the 2700 X I could say that Amy's marketing message of about a 5 to 10 percent performance increases is actually pretty modest I think most users in most scenarios would definitely see more than a 10 percent performance uplift especially more than a 10 percent performance uplift per dollar the value of these things really is just I mean it's great I mean it well I guess without saying I mean competition in the cpu market the consumer wins so this is no matter which CPU you're leaning toward this only benefits you the consumer because we've got real actual competition in the marketplace this is a great product it does exactly what it says it does and it does it well that's all you can ask of a product I think it's pretty good I'd also like to give a shout out to Brad Morris for helping me crunch the numbers and do some analysis here and also to Mike CFM for helping me put the test systems together that we use in our testing for I'll just this video but also a couple other videos some of which are probably not out yet and some of which probably already came out with the envy me rate video so kudos thanks I thought this video would come out first but editing was tricky and the performance numbers and we had to do some analysis so hey I'm Windell this is level one should like and subscribe and hit the bell or thumbs down if you want I don't care just some sort of interaction to feed the algorithm or at least that's what I'm told so I'll see you the forums at level one\n"