**Nvidia's 30 Series: A Powerhouse with a Price to Pay**

The Nvidia GeForce RTX 3090 Ti, also known as the "Titan," is a powerful graphics card that promises exceptional performance and features. In this review, we'll take a closer look at the card's design, power consumption, and pricing.

**A Well-Built Card at an Atrocious Price**

The EVGA variant of the RTX 3090 Ti is well-built and sturdy, with a solid aluminum heatsink and a robust power delivery system. However, its price tag of $2200 is, indeed, atrocious. This is not surprising, considering the card's power consumption of up to 500 watts. The high power draw makes it a challenging product to develop and manufacture.

**Transient Load Spikes: A Growing Concern**

One of the significant concerns with the 30 series is transient load spikes, which refer to brief periods of high power consumption that can be detrimental to power supplies. These spikes occur due to the card's ability to handle extreme performance levels, but they also put a strain on power supplies. We've seen this issue in the past with other Nvidia cards and hope that the 3090 Ti will not exhibit similar problems.

**The Importance of Power Supplies**

To mitigate the risk of transient load spikes, it's essential to choose a power supply that can handle the card's power requirements. Even high-end power supplies like those from EVGA or Corsair may struggle to keep up with the card's demands. We recommend using a 1600-watt power supply or higher to ensure stable operation.

**Content Creation and Gaming**

While the RTX 3090 Ti is primarily designed for content creation, it can also handle demanding gaming workloads. The card features a robust cooling system, which helps maintain performance levels even during intense gaming sessions. However, its high price tag may deter some buyers, especially those on a budget.

**Pricing: A Barrier to Entry**

The RTX 3090 Ti's pricing is a significant barrier to entry for many potential buyers. The card's cost of $2200 is equivalent to the price of an EVGA FTW3 card, which is already a premium product. This raises questions about the value proposition of the 3090 Ti and whether it's worth the additional expense.

**A Market Reality**

The GPU market is characterized by intense competition and limited supply, leading to high prices for popular models like the RTX 3090 Ti. Buyers are often forced to compromise on their budget or wait for more affordable alternatives. This reality is particularly challenging for content creators who require powerful GPUs to produce high-quality content.

**A Final Verdict**

In conclusion, the Nvidia GeForce RTX 3090 Ti is a powerful graphics card that offers exceptional performance and features. However, its pricing is a significant concern, making it inaccessible to many buyers. While we appreciate the card's capabilities, our hard pass on this product is due to its steep price tag.

**The Market Will Decide**

Ultimately, the Nvidia GeForce RTX 3090 Ti will be decided by the market, and its popularity will depend on several factors, including pricing, performance, and availability. For now, we'll continue to monitor the situation and provide updates as more information becomes available.

"WEBVTTKind: captionsLanguage: enthank you today reviewing the Nvidia RTX 3090 TI 2200 gets you this and it also gets you one of these we're at that at that stage now we got the octopus cables going on three eight pins for this one feeding into a single 12 pin we're gonna have a full tear down obviously but in this video we're focusing on the performance to some extent the value and we'll also be going over extensively The Thermals and the design characteristics how it performs overall but the prices what you need to know right away the cheapest ones of these we're aware of start at two thousand dollars let's get started before that this video is brought to you by Kale mods pcie 4.0 Riser cables available in 90 degree and straight connections the pcie Gen 4 Riser cables run up to 30 centimeters long and pair well with cable mods vertical GPU brackets designed to maintain a more optimal distance between the video card fans and the case side panel kill mod markets that these have full speed signal transmission because they use a Pure copper tinning process to achieve better signal quality learn more at the link in the description below the money angle is pretty known at this point for these cars we've done we've done the gag before now it's more of a power thing though the 3090 TI is the most power hungry card we've tested before overclocking so we'll go over the numbers in a little bit in detail but you're looking at 500 watts plus or minus maybe 20 here doesn't help you a lot uh for the 3090 TI and that is drastically increased over the original 30 90. what that extra power gets you we're going to save everyone the time here and just give you the quick version up front is somewhere between about eight and twelve percent on average uplift over an RTX 3090 non-ti which we have been calling the business casual edition of the 3090 or the filthy pores edition if you will because the 390s high does the business person's 30 90. so the ti is a 10 752 Cuda core card versus the 3090 original card that's at ten thousand four hundred and Ninety Six Cuda core is not a big difference there a couple percent uh it's a non-linear increase in performance for Cuda core increases because there are other factors that play too like bass and boost clocks uh rops tmus uh the memory bus the memory controllers the amount of memory all that stuff factors in but that's the change just for the GPU core uh and the Cuda core account within it so anyway the pricing was mostly hidden up until the last minute for this one the whole launch was actually pretty pretty last minute from Nvidia where the card showed up on Friday for us we started testing immediately the review goes up now and so that means we had basically all of Saturday Sunday to produce it and Monday to produce it turn it around pretty well came out as we like it but our Focus here is going to be on thermal and power first and foremost because the one thing this has going for it within the pricing spectrum is that it's a flagship Ultra high-end card so everyone expects that the value is going to be bad that's just how it works from that angle there's still some fun to be had for this being an Enthusiast card which is why we're doing all the overclocking with it and the focus on things like power consumption uh additional things you need to know 24 gigabytes of dddr6x memory it's a 384-bit bus and then for the die it's ga-102-350 for the Silicon with the TDP F450 I don't know I put this many in there they kept coming out for the rest nvidia's marketing is largely focused on 8K well largely it's about it's about 30 8K gaming so they've backed off a little bit from the mockery of the 3090 or Nvidia torpedoed its own launch on that one by focusing on 8K gaming as if that's a thing anyone gives a about but it's also talking about being a Creator card we actually don't disagree with this aspect because the 24 gigabytes of memory is extremely helpful for our workloads and working on what now everyone calls content creation so we use Unreal Engine 5 for example we use it for visualizing our office we use it for building out animations these days largely replaced blenders rendering with real time rendering an Unreal Engine and the memory does help there we use 3090s for that so we can see the value we get it for most people of course these flagships should be in a way considered to be Titans they should probably rename so that Nvidia doesn't bear the expectation of being a gaming card that is granted by an on Titan card but they should be thought of the same way now ultimately this value is atrocious it's terrible value versus a 30 90. there's the same memory capacity you don't gain much 8 to 12 like you said and really the takeaway is that the memory moves all to one side of the PCB that's the value here so there's not a lot of value outside of that it's easier to cool and that's kind of it anyway let's get started with the benchmarks we have a lot to talk about for testing and then we'll go through a little bit more of the conclusion but you've basically gotten it here already power consumption testing is up first we're front loading this because the card is so impressively power hungry that this test has actually become really interesting we have no doubt that transient load survival is going to be one of the hardest tasks for power supply to keep up with with a 39 DTI but that's harder to test in a short review window in the very least we can start with raw power consumption in a torture workload the total card power draw of the RTX 3090 TI ftw3 was a shocking 500 watts running two of these with anything else on the computer at all would put you into dual power supply territory on a 115 volt wall drop so we're truly backed in the era of ultra high power draw components the GTX 480s would be proud of this one it's like true to form for older style Nvidia now the big difference now versus then is that cooling has become far more advanced but mostly at the cost of form factor that's why this card is so fat the ocv BIOS pulled the same 504 Watts as it only adjusts for fan speed the power consumption versus the stock 3090 non-ti also ftw3 previously at 389 watts is 29 higher with a 30 90 TI ftw3 that's going to set the tone for this entire review because it sets the tone for efficiency or the percent Improvement for the percent increase in power drop overclocking the 3090 TI got it up to 533 watts in line with the seven percent power Target offset but take a look at this chart the per rail power consumption looks like this during the overclocked testing the total power consumption maximally spiked to 560 Watts but it may have spiked even higher if they're in a microbursted transients which typically are up to about 100 microseconds in length we can only measure 100 milliseconds with this particular equipment although we're working on that but the sudden drops and the total power consumption seen in this chart are from stability of the overclock it's actually somewhat difficult to stabilize this card under some conditions like torture workloads but more on that later in our overclocking section the pcie cables fairly universally ran at about 108 watts each we'll plot them one by one here that makes sense because they all feed into the same single 12 pin connector ultimately the pcie slot is capped at 66 Watts at 12 volts and 75 Watts total for the spec and we're well under those numbers here the card is primarily leaning on the cables which is sensible when there are three available I'll plug it into an octopus but the slot draw was 10 watts on 12 volts and about zero Watts on 3 volts our thermal testing is extensive on this one but we'll split it into two charts and we'll start with the simplest one equilibrium temperatures for the GPU core GPU hotspot and memory tested across all 3v bios options is what we're looking at first the V bios is on this card only adjust fan speed not the power limit which is capped at 480 Watts so we'll look at that next in the slower RPM normal V bios the card ran well within acceptable tolerance is at 70 degrees Celsius core and 79 degrees on the hot spot with ambient temperature at 20 degrees Celsius but this is not delta T over m is just the temperature of the card at 20c for ambient the memory thermal has averaged at 71 degrees for this V bios the performance V bios is the default EVGA choice though and this one had us at 63 degrees 71 degrees for a hot spot and 64 degrees for memory the OCB bios has no meaningful impact on anything in our testing and doesn't even look like it's changed anything here the chart we're showing now is for an AMD CPU actually but it's helpful to illustrate this next Point as a reminder GPU thermals directly impacts boosting frequency every couple degrees Celsius can be the difference of maybe 12 or 15 megahertz on the clock for most cars but these really fat high-end cards can normally Flatline the frequency the Nvidia behavior is more nuanced than what we saw on that AMD charge but modern AMD CPUs behave in a similar fashion by leveraging additional thermal Headroom to boost higher here's the frequency chart for the 39 DTI the normal vbios has us at about 20 25 megahertz dead for most of the test although it peaked at 20 40 megahertz at the start proving the point on how a couple degrees results in that 15 megahertz reduction we saw the performance of BIOS runs a slightly higher clock at 2040 for the full test that's not for any reason though other than the more aggressive fan curve programmed into V bios for the ocv BIOS without any overclock actually applied frequency starts at about 20 55 megahertz but settles at the same 2040 as performance the flat line is somewhat remarkable here Nvidia Fe cards typically have much more chaotic frequency plots as do lower end board versions for almost any GPU that's because they're more power limited for the Target here EVGA has boosted its power Target enough that the common limiter is V Rel or voltage not power if you were to boost voltage externally or through a mod the calf would become power or thermal instead so v-rel or voltage is the reason we're stuck at this flat line but the flat Line's a good thing fan speed is the next most important part seen as it's the only thing that changes with these V buyers options the normal V bios has us at zero RPM and boosts to about 1657 RPM over sustained load hysteresis has us at about 383 seconds required to stabilize the fan RPM against the V bios temperature Target the performance fee bios jumps nearly immediately to 1980 to 2000 RPM under the same test conditions hysteresis or fan ramp lag is closer to just 200 seconds to ramp but the jump from 0 to 1800 takes place in about two seconds that's very noticeable to the end user and will be more audible than the curved ramp of the normal V bios which only instantaneously jumps to about 1200 RPM the ocv BIOS doesn't seem to do anything different than performance V bios but remains useful as a switch for flashing haxby bios is onto later like for xoc so the main takeaway here is that you can run this card sized such that it is at a much lower RPM of about 200 to 400 RPM reduced off of the automatic speeds we're seeing and still be within an acceptable thermal tolerance and what that allows you to do obviously is just have sort of a better user experience without really giving up very much it's still well with an acceptable range the frequency doesn't change much more than 15 25 megahertz Max and that's kind of what our recommendation would be if you were to buy one of these cards uh or really if you buy anything that is sized like this on a lower end card as well the other takeaway is the opposite end there's room to ramp the fan speed and there's room to leverage the thermal capabilities of the cooler for a bigger overclock but if you're going to do that really you need an unlocked to be bios where the power Target is uncapped or at least capped higher right now it's capped at uh well about 480 Watts stock and you do seven percent offset and that kind of limits you so we're seeing Max 500 530 Watts somewhere in there for total board power and that's just not enough to get really much more out of it so you would need a mod for that to go further and finally this is at least showing that this cooler is attached two of the components and cooling them properly we've seen other massive cards in the past like the Zotac 1080 TI where it has this huge cooler but it's not really connected to anything except the GPU sometimes so it is at least being leveraged here now for the memory and voltage regulator module thermals this is where things have historically Fallen apart for the non-ti 3090 cards due to low heat sinking capability of the back of the PCB for this one using the hottest V bios as a measuring stick the tested memory modules that never exceeded 67 degrees Celsius on the NTC thermistor logging the data this is well within spec and as a relief as compared to the original RTX 3090s although the price is Extreme the vrm components never exceeded 65 degrees Celsius during testing which is also well within spec these could scale up to 100 degrees Celsius and still technically would be within spec so there's a lot of Headroom here EVGA has built an insanely Overkill cooler for an insanely Overkill rights okay next we're going to talk about overclock performance and stepping the really important thing we want to lay out here is that as you're overclocking you have to do validation so it's not good enough anymore to just run it with the core offset the memory offset and do that until it crashes now once you think it's running stable you also need to do a performance check with something like Port Royal or a game that's repeatable anything that gives you a repeatable score because if that's not done then what can happen is silent memory errors in the background that would actually produce a lower result than stock we've seen that it's possible to think you've overclocked your card but actually you've just heard performance worse than stock because it's not crashing so it doesn't present the catastrophic failure but it also is silently failing each application will behave differently here so you want to tune for sort of the most fussy of all of them and then for frequency offset and power percent remember that these are always relative to what the the card is doing anyway so if one card has 115 power Target the other one's 107 it doesn't mean that 115 is higher than 107 other than that has true numerical value because if the power of one is 480 watts and the power of the other one is 450 Watts obviously the percentage means a different thing here's the step in when using Port Royal to validate we had already determined that 125 megahertz offset was the most reliable for wide reaching stability in various games so we settled there although it sometimes could hold 160 megahertz offset and Times by extreme that would not hold globally and especially not in games so it's not really valid because it wasn't stable first off we noticed that the EVGA Precision voltage slider actually does something here The Slider helps stabilize the maximum frequency without any other changes and you can see that when our limiter switches from v-rail to power the frequency jumped up by 15 megahertz just by moving that slider and changing nothing else which also allowed an immediate jump in score for about 15 480 or 15 490 maximally to about 15 500 20. that's an increase of about 0.25 percent and if you've seen our extreme overclocking streams that's actually a pretty big jump for just doing one slider from zero to Max this slider only tells the voltage to run its peak available option consistently so it can be thought of more as a binary switch than an actual voltage offset is not offsetting voltage by 100 units of anything it's just kind of fully offset to the peak or not from the stock 14 880 results on normal V bios up to the max 15 549 score we got with our heavier OC we got about a four and a half percent uplift overall there's about two hours of work to get this stable and operational we'd say that this generally isn't worth overclocking today but we've seen cards in the past like the 1080 TI or the 780 Ti or 980 TI where overclocking becomes disproportionately valuable as the card agents we don't know if that'll be true here though in other words 4.5 today in today's applications may turn into 7 or 10 in the future either way the overclocking aspect is clearly more fun from an Enthusiast tinkering standpoint than it is from a practical one at four and a half percent in Shadow of the Tomb Raider at 4K the RTX 3090 TI ftw3 stock card ran at 119 FPS average leading the original 3090 ftw3 by about 12 percent that's with brand new data too we we ran another 3090 and got about the same averages as we did previously the end result is two and a half percent more power consumption for about every one percent more performance showing that we're beyond the operating point for efficiency overclocking the 390 TI certainly doesn't help that argument but it did boost the results by a five percent average FPS while maintaining good low FPS consistency the 6900 XT it runs at about 98 FPS average here that's too bad because the human eye can only see above 100 FPS so this card is best left to the pores and The Peasants the lead over the 6900 XT is about 21 at 4K okay in this game though this would reduce as we approach lower resolutions but there's obviously a limit to how low of a resolution you'll go with with these cards at 1440p the results have the RTX 3090 TI ftw3 at 176 FPS average with lows still in good shape for overall frame time consistency that has the 3090 tie leading the 3090 ftw3 non-ti by 3.7 percent in exchange for 15 more money comparing ftw3 to ftw3 the 6900 XT encroaches here but we're predictably limited by the CPU in this set of data as well clearly in terms of value you really shouldn't be buying either of these cards for gaming that's not news we've been saying that they don't make any sense for the vast majority of use cases the 3080 or the 6800 XT achieve most of the performance for a fraction of the price even factoring in scalp's prices this isn't news to Nvidia either they're trying to distance themselves from the gaming marketing for these cards because it worked out very poorly last time when they tried to make it a whole big 8K gaming thing anyway these are more worthwhile if you need the vram like we do for some of our blender and Unreal Engine workflows but because Nvidia shipped these out with sort of two days of notice there wasn't a whole lot of time to factor that in Total War 3 Kingdoms is run at ultra settings in our test Suite which makes it one of the most abusive Benchmark scenarios that we run the RTX 3090 Thai stock card plotted at about 60 FPS average at 4K which gives it a 10 lead over the RTX 3090 business casual ftw3 the overclock pushes another three FPS average not noticeable to a human at this level but probably noticeable to whatever this is the lead over the RTX 3080 is about 26 percent with a 6900xc roughly in the same spot ultra settings at 4K are ridiculous though you could easily make this playable and enjoyable at 4K with these cards by dropping settings one run we leave them high for the value of creating a strain and for showing scaling though at 1440p the 3090 TI ftw3 ran at 116 FPS average still constrained by the game more than the CPU and the ram the lead remains about eight percent not as much as before but not bound either the lead over the 3080 is about 13 so not particularly worthwhile going much above that at 1080p we still see scaling links to our aggressive settings for this one the 3090 TI fw3 ran at 161 FPS Average stock lowest paced as we'd expect and well with the OC giving us seven percent that's one of the best gains from overclock and we'll see on this card for the whole review it's not particularly impressive but uh it's better than what we saw elsewhere the 3090 filthy casual Edition wasn't much behind here either in the division 2 at 4K we measured the RTX 3090 TI ftw3 at 133 FPS average with lows paced in accordance with other frame Diamond consistency pacing on this charge the overclock boosted Us by five percent consistent with previous result results up to 140 FPS average compared to the RTX 3090 master that we just re-ran tonight the lead is about 16.5 percent the 3080 TI still sits behind that with the 3080 Eagle at 105 FPS average about tied with the 6900 XT for reference the 39 DTI leads to 30 80 and 6900 XT by about 27 in exchange for the massive price increase again uh even at those scalp prices gaming is best out on cards at or below the level of a 3080 or 6800 XT which we're only really saying just in case some of you are newer here and haven't been in the market for this cycle to know that at 1440p the RTX 39 TI fcw3 and the overclocked results align with no difference as they bounce off of other limitations the leadership over the 3090 is reduced to nine percent due to this limitation so there's not much to learn here other than as we already knew running at lower resolutions and settings won't be benefited by Extreme Graphics processing capabilities at all now AMD has maintained an advantage as the resolution goes down to 1080 or 1440 price to price in the past and that's true here as well if you remove the CPU Bond the 6800 XT and the 3080 are in the 184 FPS average range and are more than capable here with the 3090 TI leading by 19 as capped by other devices in the system in Rainbow Six Siege the RTX 390 TI ftw3 ran about 239 FPS average and we see there's still plenty of scaling Room by The OC landed at 249 FPS average that's a four percent uplift which still doesn't feel particularly worth it for the effort but considering the level performance this car is already at 530 Watts there's not much more room without just committing to competitive overclocking the 3090 ftw3 allows the 3090 TI a lead of 12.3 percent about in line with our previous results for perspective the 3080 manages 80 percent of the 3090 TI's performance in case that's helpful in making your choices later at 1440p the results boosted to 417 FPS average One stock or 430 over clocked there's still sufficient room to see scaling at 1440p with these graphic settings as well the 39 ETI ended up leading the 1600 XT by about 18 percent with AMD gaining some ground as resolution drops the 3090 Masters led by about 10 percent in this one for one last Benchmark we'll look at GTA 5 at 4K really just for fun with our test settings this old game remains challenging for gpus to run at 4K the 39 DTI ftw3 capped out at 126 FPS average again with good load timing while the OC pushed 5.4 percent ahead the lead over the slavin Lee 3090 non-ti poor people Edition is about 10 percent once again showing that the average Gap is consistent across games okay so as noted we were trying to sort of respect everyone's time and just put it all up in the front of the video you get the idea uh the increase versus the 30 90 is about 10 plus or minus a little bit increase versus the 30 80 is sometimes 19 to 27 depending on the application of the game and 3080 6800 XT really just were just re-emphasizing this point for people who might be new to Building Systems new to the market haven't been in the market for a few years the maximum card you need to be looking at for a gaming computer really stops at a 30 80 or 6800 XT and our original reviews for those largely stand the numbers have shuffled a little bit with drivers but the concepts are the same and when you would buy each one Remains the Same so that's where you need to stop looking if you're buying for gaming for 30 90 type cards I mean the 30 90 what it came out was an absolute mockery of a card it made a buffoon of its Nvidia made a buffoon of itself by doing the whole 8K gaming push doing 8K sponsored content before the launch of the card and trying to present it as if that's even remotely realistic as a wide sweeping thing it's just not don't expect that that's stupid uh and we're not there yet and that's fine not many people have an 8K monitor or TV or whatever so it's just a silly or relevant marketing point but it's still there on the slides as if Nvidia couldn't be like reducing the focus to zero percent isn't enough we need we need to still look like we're committed to this idea so let's do 30 focus on it the rest was on content creation there's some gaming discussion so it Stacks up like this very simply the card as it is from EVGA is well built it is at atrocious price at 2200 but this is not surprising to anybody the thing that is a little more surprising is the power consumption at 500 watts this changes the landscape for the class of power supply you have to buy at the high end and more importantly transient load spikes are a massive consideration for the 30 series and they will be for the 40 series as well from what we're hearing at least right now from the power supply manufacturers and transient load spikes is something we're actually working on doing some testing with this particular card but what you need to know is that these are things that sustain from actually something about 100 microseconds and it sounds really short and it is except with a lot of power supplies including some of C Sonic's higher end units a lot of 850 watt gold units thousand watt gold units a transient Spike of a couple X the expected power consumption uh or even a couple hundred Watts over not necessarily full 2X or something that's enough to where if the power supply doesn't have enough capacitance to soak the change you'd you lose power and the system will shut down from functionally an ocp or over current protection so that's what you need to watch out for we did not run into that issue with this test setup we were using the EVGA T2 1600 watt though which is absurd I mean that's a lot of capability from a power supply so we're working on a follow-up content piece to look at that this has been an issue with the 30 series in the past we haven't seen it yet on the 3090 TI because we've only had it for two full days of testing and we didn't spend any of that on transients so anyway something for you to be aware of where this is a gap in our testing coverage that we are working on to fill in over the next few days or a week or so uh so we we can't include that in a conclusion here ultimately then obviously like the Titans have been for us in the past this is largely I probably don't buy this we do see the value in content creation because we've used it that way as the price hurts a lot um if you're making money on it then it's kind of like business class internet where they say you could have consumer internet actually you can't because you're a business so you have to pay business class rates get basically uh the same thing that used to be available for consumers but you pay a lot more that's kind of where the 390 TI slots in where if you're a professional especially if you're making money it starts to get easier to justify but it still hurts in the pricing so for us a hard pass on this 8 to 12 percent for twenty two hundred dollars is insane the only reason that cash grabs like this and the 30 80 12 gigabyte can work is because if you're new to the market and they know this if you're coming to the market and you need a video card and you see the 3090 is out of stock but the out of stock pricing says whatever two thousand dollars which is um the the current price Nineteen Hundred twenty dollars is the price of the evj FTW three to Thirty ninety you see that and you see the new one for 2200 or something 300 difference you've talked yourself into already two grand so now you're like huh what's another couple hundred dollars to just get something and be done with this and that's kind of the market we're in right now everyone knows that we've said this for many months at this point uh it is getting boring from a Content perspective but it Remains the reality so we do want to put that out there but the power is really we think the main story here at the incredible 500 plus watch so that's it for this one check back for follow-ups as always again two days with this full days um we are exhausted and tired but we got through the review I'd like to do a lot more depth to it but that's the limit of what we can do and still hit embargo which we have a responsibility to do to make sure people are informed as soon as this thing goes up so that's it for this one thanks for watching and we'll see you all next timethank you today reviewing the Nvidia RTX 3090 TI 2200 gets you this and it also gets you one of these we're at that at that stage now we got the octopus cables going on three eight pins for this one feeding into a single 12 pin we're gonna have a full tear down obviously but in this video we're focusing on the performance to some extent the value and we'll also be going over extensively The Thermals and the design characteristics how it performs overall but the prices what you need to know right away the cheapest ones of these we're aware of start at two thousand dollars let's get started before that this video is brought to you by Kale mods pcie 4.0 Riser cables available in 90 degree and straight connections the pcie Gen 4 Riser cables run up to 30 centimeters long and pair well with cable mods vertical GPU brackets designed to maintain a more optimal distance between the video card fans and the case side panel kill mod markets that these have full speed signal transmission because they use a Pure copper tinning process to achieve better signal quality learn more at the link in the description below the money angle is pretty known at this point for these cars we've done we've done the gag before now it's more of a power thing though the 3090 TI is the most power hungry card we've tested before overclocking so we'll go over the numbers in a little bit in detail but you're looking at 500 watts plus or minus maybe 20 here doesn't help you a lot uh for the 3090 TI and that is drastically increased over the original 30 90. what that extra power gets you we're going to save everyone the time here and just give you the quick version up front is somewhere between about eight and twelve percent on average uplift over an RTX 3090 non-ti which we have been calling the business casual edition of the 3090 or the filthy pores edition if you will because the 390s high does the business person's 30 90. so the ti is a 10 752 Cuda core card versus the 3090 original card that's at ten thousand four hundred and Ninety Six Cuda core is not a big difference there a couple percent uh it's a non-linear increase in performance for Cuda core increases because there are other factors that play too like bass and boost clocks uh rops tmus uh the memory bus the memory controllers the amount of memory all that stuff factors in but that's the change just for the GPU core uh and the Cuda core account within it so anyway the pricing was mostly hidden up until the last minute for this one the whole launch was actually pretty pretty last minute from Nvidia where the card showed up on Friday for us we started testing immediately the review goes up now and so that means we had basically all of Saturday Sunday to produce it and Monday to produce it turn it around pretty well came out as we like it but our Focus here is going to be on thermal and power first and foremost because the one thing this has going for it within the pricing spectrum is that it's a flagship Ultra high-end card so everyone expects that the value is going to be bad that's just how it works from that angle there's still some fun to be had for this being an Enthusiast card which is why we're doing all the overclocking with it and the focus on things like power consumption uh additional things you need to know 24 gigabytes of dddr6x memory it's a 384-bit bus and then for the die it's ga-102-350 for the Silicon with the TDP F450 I don't know I put this many in there they kept coming out for the rest nvidia's marketing is largely focused on 8K well largely it's about it's about 30 8K gaming so they've backed off a little bit from the mockery of the 3090 or Nvidia torpedoed its own launch on that one by focusing on 8K gaming as if that's a thing anyone gives a about but it's also talking about being a Creator card we actually don't disagree with this aspect because the 24 gigabytes of memory is extremely helpful for our workloads and working on what now everyone calls content creation so we use Unreal Engine 5 for example we use it for visualizing our office we use it for building out animations these days largely replaced blenders rendering with real time rendering an Unreal Engine and the memory does help there we use 3090s for that so we can see the value we get it for most people of course these flagships should be in a way considered to be Titans they should probably rename so that Nvidia doesn't bear the expectation of being a gaming card that is granted by an on Titan card but they should be thought of the same way now ultimately this value is atrocious it's terrible value versus a 30 90. there's the same memory capacity you don't gain much 8 to 12 like you said and really the takeaway is that the memory moves all to one side of the PCB that's the value here so there's not a lot of value outside of that it's easier to cool and that's kind of it anyway let's get started with the benchmarks we have a lot to talk about for testing and then we'll go through a little bit more of the conclusion but you've basically gotten it here already power consumption testing is up first we're front loading this because the card is so impressively power hungry that this test has actually become really interesting we have no doubt that transient load survival is going to be one of the hardest tasks for power supply to keep up with with a 39 DTI but that's harder to test in a short review window in the very least we can start with raw power consumption in a torture workload the total card power draw of the RTX 3090 TI ftw3 was a shocking 500 watts running two of these with anything else on the computer at all would put you into dual power supply territory on a 115 volt wall drop so we're truly backed in the era of ultra high power draw components the GTX 480s would be proud of this one it's like true to form for older style Nvidia now the big difference now versus then is that cooling has become far more advanced but mostly at the cost of form factor that's why this card is so fat the ocv BIOS pulled the same 504 Watts as it only adjusts for fan speed the power consumption versus the stock 3090 non-ti also ftw3 previously at 389 watts is 29 higher with a 30 90 TI ftw3 that's going to set the tone for this entire review because it sets the tone for efficiency or the percent Improvement for the percent increase in power drop overclocking the 3090 TI got it up to 533 watts in line with the seven percent power Target offset but take a look at this chart the per rail power consumption looks like this during the overclocked testing the total power consumption maximally spiked to 560 Watts but it may have spiked even higher if they're in a microbursted transients which typically are up to about 100 microseconds in length we can only measure 100 milliseconds with this particular equipment although we're working on that but the sudden drops and the total power consumption seen in this chart are from stability of the overclock it's actually somewhat difficult to stabilize this card under some conditions like torture workloads but more on that later in our overclocking section the pcie cables fairly universally ran at about 108 watts each we'll plot them one by one here that makes sense because they all feed into the same single 12 pin connector ultimately the pcie slot is capped at 66 Watts at 12 volts and 75 Watts total for the spec and we're well under those numbers here the card is primarily leaning on the cables which is sensible when there are three available I'll plug it into an octopus but the slot draw was 10 watts on 12 volts and about zero Watts on 3 volts our thermal testing is extensive on this one but we'll split it into two charts and we'll start with the simplest one equilibrium temperatures for the GPU core GPU hotspot and memory tested across all 3v bios options is what we're looking at first the V bios is on this card only adjust fan speed not the power limit which is capped at 480 Watts so we'll look at that next in the slower RPM normal V bios the card ran well within acceptable tolerance is at 70 degrees Celsius core and 79 degrees on the hot spot with ambient temperature at 20 degrees Celsius but this is not delta T over m is just the temperature of the card at 20c for ambient the memory thermal has averaged at 71 degrees for this V bios the performance V bios is the default EVGA choice though and this one had us at 63 degrees 71 degrees for a hot spot and 64 degrees for memory the OCB bios has no meaningful impact on anything in our testing and doesn't even look like it's changed anything here the chart we're showing now is for an AMD CPU actually but it's helpful to illustrate this next Point as a reminder GPU thermals directly impacts boosting frequency every couple degrees Celsius can be the difference of maybe 12 or 15 megahertz on the clock for most cars but these really fat high-end cards can normally Flatline the frequency the Nvidia behavior is more nuanced than what we saw on that AMD charge but modern AMD CPUs behave in a similar fashion by leveraging additional thermal Headroom to boost higher here's the frequency chart for the 39 DTI the normal vbios has us at about 20 25 megahertz dead for most of the test although it peaked at 20 40 megahertz at the start proving the point on how a couple degrees results in that 15 megahertz reduction we saw the performance of BIOS runs a slightly higher clock at 2040 for the full test that's not for any reason though other than the more aggressive fan curve programmed into V bios for the ocv BIOS without any overclock actually applied frequency starts at about 20 55 megahertz but settles at the same 2040 as performance the flat line is somewhat remarkable here Nvidia Fe cards typically have much more chaotic frequency plots as do lower end board versions for almost any GPU that's because they're more power limited for the Target here EVGA has boosted its power Target enough that the common limiter is V Rel or voltage not power if you were to boost voltage externally or through a mod the calf would become power or thermal instead so v-rel or voltage is the reason we're stuck at this flat line but the flat Line's a good thing fan speed is the next most important part seen as it's the only thing that changes with these V buyers options the normal V bios has us at zero RPM and boosts to about 1657 RPM over sustained load hysteresis has us at about 383 seconds required to stabilize the fan RPM against the V bios temperature Target the performance fee bios jumps nearly immediately to 1980 to 2000 RPM under the same test conditions hysteresis or fan ramp lag is closer to just 200 seconds to ramp but the jump from 0 to 1800 takes place in about two seconds that's very noticeable to the end user and will be more audible than the curved ramp of the normal V bios which only instantaneously jumps to about 1200 RPM the ocv BIOS doesn't seem to do anything different than performance V bios but remains useful as a switch for flashing haxby bios is onto later like for xoc so the main takeaway here is that you can run this card sized such that it is at a much lower RPM of about 200 to 400 RPM reduced off of the automatic speeds we're seeing and still be within an acceptable thermal tolerance and what that allows you to do obviously is just have sort of a better user experience without really giving up very much it's still well with an acceptable range the frequency doesn't change much more than 15 25 megahertz Max and that's kind of what our recommendation would be if you were to buy one of these cards uh or really if you buy anything that is sized like this on a lower end card as well the other takeaway is the opposite end there's room to ramp the fan speed and there's room to leverage the thermal capabilities of the cooler for a bigger overclock but if you're going to do that really you need an unlocked to be bios where the power Target is uncapped or at least capped higher right now it's capped at uh well about 480 Watts stock and you do seven percent offset and that kind of limits you so we're seeing Max 500 530 Watts somewhere in there for total board power and that's just not enough to get really much more out of it so you would need a mod for that to go further and finally this is at least showing that this cooler is attached two of the components and cooling them properly we've seen other massive cards in the past like the Zotac 1080 TI where it has this huge cooler but it's not really connected to anything except the GPU sometimes so it is at least being leveraged here now for the memory and voltage regulator module thermals this is where things have historically Fallen apart for the non-ti 3090 cards due to low heat sinking capability of the back of the PCB for this one using the hottest V bios as a measuring stick the tested memory modules that never exceeded 67 degrees Celsius on the NTC thermistor logging the data this is well within spec and as a relief as compared to the original RTX 3090s although the price is Extreme the vrm components never exceeded 65 degrees Celsius during testing which is also well within spec these could scale up to 100 degrees Celsius and still technically would be within spec so there's a lot of Headroom here EVGA has built an insanely Overkill cooler for an insanely Overkill rights okay next we're going to talk about overclock performance and stepping the really important thing we want to lay out here is that as you're overclocking you have to do validation so it's not good enough anymore to just run it with the core offset the memory offset and do that until it crashes now once you think it's running stable you also need to do a performance check with something like Port Royal or a game that's repeatable anything that gives you a repeatable score because if that's not done then what can happen is silent memory errors in the background that would actually produce a lower result than stock we've seen that it's possible to think you've overclocked your card but actually you've just heard performance worse than stock because it's not crashing so it doesn't present the catastrophic failure but it also is silently failing each application will behave differently here so you want to tune for sort of the most fussy of all of them and then for frequency offset and power percent remember that these are always relative to what the the card is doing anyway so if one card has 115 power Target the other one's 107 it doesn't mean that 115 is higher than 107 other than that has true numerical value because if the power of one is 480 watts and the power of the other one is 450 Watts obviously the percentage means a different thing here's the step in when using Port Royal to validate we had already determined that 125 megahertz offset was the most reliable for wide reaching stability in various games so we settled there although it sometimes could hold 160 megahertz offset and Times by extreme that would not hold globally and especially not in games so it's not really valid because it wasn't stable first off we noticed that the EVGA Precision voltage slider actually does something here The Slider helps stabilize the maximum frequency without any other changes and you can see that when our limiter switches from v-rail to power the frequency jumped up by 15 megahertz just by moving that slider and changing nothing else which also allowed an immediate jump in score for about 15 480 or 15 490 maximally to about 15 500 20. that's an increase of about 0.25 percent and if you've seen our extreme overclocking streams that's actually a pretty big jump for just doing one slider from zero to Max this slider only tells the voltage to run its peak available option consistently so it can be thought of more as a binary switch than an actual voltage offset is not offsetting voltage by 100 units of anything it's just kind of fully offset to the peak or not from the stock 14 880 results on normal V bios up to the max 15 549 score we got with our heavier OC we got about a four and a half percent uplift overall there's about two hours of work to get this stable and operational we'd say that this generally isn't worth overclocking today but we've seen cards in the past like the 1080 TI or the 780 Ti or 980 TI where overclocking becomes disproportionately valuable as the card agents we don't know if that'll be true here though in other words 4.5 today in today's applications may turn into 7 or 10 in the future either way the overclocking aspect is clearly more fun from an Enthusiast tinkering standpoint than it is from a practical one at four and a half percent in Shadow of the Tomb Raider at 4K the RTX 3090 TI ftw3 stock card ran at 119 FPS average leading the original 3090 ftw3 by about 12 percent that's with brand new data too we we ran another 3090 and got about the same averages as we did previously the end result is two and a half percent more power consumption for about every one percent more performance showing that we're beyond the operating point for efficiency overclocking the 390 TI certainly doesn't help that argument but it did boost the results by a five percent average FPS while maintaining good low FPS consistency the 6900 XT it runs at about 98 FPS average here that's too bad because the human eye can only see above 100 FPS so this card is best left to the pores and The Peasants the lead over the 6900 XT is about 21 at 4K okay in this game though this would reduce as we approach lower resolutions but there's obviously a limit to how low of a resolution you'll go with with these cards at 1440p the results have the RTX 3090 TI ftw3 at 176 FPS average with lows still in good shape for overall frame time consistency that has the 3090 tie leading the 3090 ftw3 non-ti by 3.7 percent in exchange for 15 more money comparing ftw3 to ftw3 the 6900 XT encroaches here but we're predictably limited by the CPU in this set of data as well clearly in terms of value you really shouldn't be buying either of these cards for gaming that's not news we've been saying that they don't make any sense for the vast majority of use cases the 3080 or the 6800 XT achieve most of the performance for a fraction of the price even factoring in scalp's prices this isn't news to Nvidia either they're trying to distance themselves from the gaming marketing for these cards because it worked out very poorly last time when they tried to make it a whole big 8K gaming thing anyway these are more worthwhile if you need the vram like we do for some of our blender and Unreal Engine workflows but because Nvidia shipped these out with sort of two days of notice there wasn't a whole lot of time to factor that in Total War 3 Kingdoms is run at ultra settings in our test Suite which makes it one of the most abusive Benchmark scenarios that we run the RTX 3090 Thai stock card plotted at about 60 FPS average at 4K which gives it a 10 lead over the RTX 3090 business casual ftw3 the overclock pushes another three FPS average not noticeable to a human at this level but probably noticeable to whatever this is the lead over the RTX 3080 is about 26 percent with a 6900xc roughly in the same spot ultra settings at 4K are ridiculous though you could easily make this playable and enjoyable at 4K with these cards by dropping settings one run we leave them high for the value of creating a strain and for showing scaling though at 1440p the 3090 TI ftw3 ran at 116 FPS average still constrained by the game more than the CPU and the ram the lead remains about eight percent not as much as before but not bound either the lead over the 3080 is about 13 so not particularly worthwhile going much above that at 1080p we still see scaling links to our aggressive settings for this one the 3090 TI fw3 ran at 161 FPS Average stock lowest paced as we'd expect and well with the OC giving us seven percent that's one of the best gains from overclock and we'll see on this card for the whole review it's not particularly impressive but uh it's better than what we saw elsewhere the 3090 filthy casual Edition wasn't much behind here either in the division 2 at 4K we measured the RTX 3090 TI ftw3 at 133 FPS average with lows paced in accordance with other frame Diamond consistency pacing on this charge the overclock boosted Us by five percent consistent with previous result results up to 140 FPS average compared to the RTX 3090 master that we just re-ran tonight the lead is about 16.5 percent the 3080 TI still sits behind that with the 3080 Eagle at 105 FPS average about tied with the 6900 XT for reference the 39 DTI leads to 30 80 and 6900 XT by about 27 in exchange for the massive price increase again uh even at those scalp prices gaming is best out on cards at or below the level of a 3080 or 6800 XT which we're only really saying just in case some of you are newer here and haven't been in the market for this cycle to know that at 1440p the RTX 39 TI fcw3 and the overclocked results align with no difference as they bounce off of other limitations the leadership over the 3090 is reduced to nine percent due to this limitation so there's not much to learn here other than as we already knew running at lower resolutions and settings won't be benefited by Extreme Graphics processing capabilities at all now AMD has maintained an advantage as the resolution goes down to 1080 or 1440 price to price in the past and that's true here as well if you remove the CPU Bond the 6800 XT and the 3080 are in the 184 FPS average range and are more than capable here with the 3090 TI leading by 19 as capped by other devices in the system in Rainbow Six Siege the RTX 390 TI ftw3 ran about 239 FPS average and we see there's still plenty of scaling Room by The OC landed at 249 FPS average that's a four percent uplift which still doesn't feel particularly worth it for the effort but considering the level performance this car is already at 530 Watts there's not much more room without just committing to competitive overclocking the 3090 ftw3 allows the 3090 TI a lead of 12.3 percent about in line with our previous results for perspective the 3080 manages 80 percent of the 3090 TI's performance in case that's helpful in making your choices later at 1440p the results boosted to 417 FPS average One stock or 430 over clocked there's still sufficient room to see scaling at 1440p with these graphic settings as well the 39 ETI ended up leading the 1600 XT by about 18 percent with AMD gaining some ground as resolution drops the 3090 Masters led by about 10 percent in this one for one last Benchmark we'll look at GTA 5 at 4K really just for fun with our test settings this old game remains challenging for gpus to run at 4K the 39 DTI ftw3 capped out at 126 FPS average again with good load timing while the OC pushed 5.4 percent ahead the lead over the slavin Lee 3090 non-ti poor people Edition is about 10 percent once again showing that the average Gap is consistent across games okay so as noted we were trying to sort of respect everyone's time and just put it all up in the front of the video you get the idea uh the increase versus the 30 90 is about 10 plus or minus a little bit increase versus the 30 80 is sometimes 19 to 27 depending on the application of the game and 3080 6800 XT really just were just re-emphasizing this point for people who might be new to Building Systems new to the market haven't been in the market for a few years the maximum card you need to be looking at for a gaming computer really stops at a 30 80 or 6800 XT and our original reviews for those largely stand the numbers have shuffled a little bit with drivers but the concepts are the same and when you would buy each one Remains the Same so that's where you need to stop looking if you're buying for gaming for 30 90 type cards I mean the 30 90 what it came out was an absolute mockery of a card it made a buffoon of its Nvidia made a buffoon of itself by doing the whole 8K gaming push doing 8K sponsored content before the launch of the card and trying to present it as if that's even remotely realistic as a wide sweeping thing it's just not don't expect that that's stupid uh and we're not there yet and that's fine not many people have an 8K monitor or TV or whatever so it's just a silly or relevant marketing point but it's still there on the slides as if Nvidia couldn't be like reducing the focus to zero percent isn't enough we need we need to still look like we're committed to this idea so let's do 30 focus on it the rest was on content creation there's some gaming discussion so it Stacks up like this very simply the card as it is from EVGA is well built it is at atrocious price at 2200 but this is not surprising to anybody the thing that is a little more surprising is the power consumption at 500 watts this changes the landscape for the class of power supply you have to buy at the high end and more importantly transient load spikes are a massive consideration for the 30 series and they will be for the 40 series as well from what we're hearing at least right now from the power supply manufacturers and transient load spikes is something we're actually working on doing some testing with this particular card but what you need to know is that these are things that sustain from actually something about 100 microseconds and it sounds really short and it is except with a lot of power supplies including some of C Sonic's higher end units a lot of 850 watt gold units thousand watt gold units a transient Spike of a couple X the expected power consumption uh or even a couple hundred Watts over not necessarily full 2X or something that's enough to where if the power supply doesn't have enough capacitance to soak the change you'd you lose power and the system will shut down from functionally an ocp or over current protection so that's what you need to watch out for we did not run into that issue with this test setup we were using the EVGA T2 1600 watt though which is absurd I mean that's a lot of capability from a power supply so we're working on a follow-up content piece to look at that this has been an issue with the 30 series in the past we haven't seen it yet on the 3090 TI because we've only had it for two full days of testing and we didn't spend any of that on transients so anyway something for you to be aware of where this is a gap in our testing coverage that we are working on to fill in over the next few days or a week or so uh so we we can't include that in a conclusion here ultimately then obviously like the Titans have been for us in the past this is largely I probably don't buy this we do see the value in content creation because we've used it that way as the price hurts a lot um if you're making money on it then it's kind of like business class internet where they say you could have consumer internet actually you can't because you're a business so you have to pay business class rates get basically uh the same thing that used to be available for consumers but you pay a lot more that's kind of where the 390 TI slots in where if you're a professional especially if you're making money it starts to get easier to justify but it still hurts in the pricing so for us a hard pass on this 8 to 12 percent for twenty two hundred dollars is insane the only reason that cash grabs like this and the 30 80 12 gigabyte can work is because if you're new to the market and they know this if you're coming to the market and you need a video card and you see the 3090 is out of stock but the out of stock pricing says whatever two thousand dollars which is um the the current price Nineteen Hundred twenty dollars is the price of the evj FTW three to Thirty ninety you see that and you see the new one for 2200 or something 300 difference you've talked yourself into already two grand so now you're like huh what's another couple hundred dollars to just get something and be done with this and that's kind of the market we're in right now everyone knows that we've said this for many months at this point uh it is getting boring from a Content perspective but it Remains the reality so we do want to put that out there but the power is really we think the main story here at the incredible 500 plus watch so that's it for this one check back for follow-ups as always again two days with this full days um we are exhausted and tired but we got through the review I'd like to do a lot more depth to it but that's the limit of what we can do and still hit embargo which we have a responsibility to do to make sure people are informed as soon as this thing goes up so that's it for this one thanks for watching and we'll see you all next time\n"