The brand new Nvidia RTX 4080 is one other speedy graphics card—it bloody needs to be for $1,200—and whenever you take DLSS 3 into consideration you’re getting on for twice the efficiency of the equally priced RTX 3080 Ti (opens in new tab) from the final technology. Critically, Body Era is black magic.
However reviewing the RTX 4080 is harder than being Jen-Hsun’s spatula wrangler. Although it is much more easy now there’s solely a 16GB model and it would not include some extra 12GB half breed trailing it round. However whereas I really feel that I’ve obtained all the information essential to decide on how good a GPU the brand new Ada Lovelace card is true on this second, I am ultra-aware that second is not going to final lengthy.
You might all the time argue the non permanent nature of a GPU’s preliminary success (or failure) is a daily problem with graphics card critiques; they will solely ever exist as a snapshot in time in mild of the inevitable modifications that availability, pricing, and competitors will wreak following any launch.
However this technology feels totally different to some other, and the RTX 4080 launch extra so. The battle I am having, you see, is that AMD (doubtless intentionally) introduced its Radeon RX 7900 XTX (opens in new tab) simply forward of this GeForce launch. And that’s AMD’s high GPU from this new technology, which is particularly concentrating on the RTX 4080, and never the Nvidia RTX 4090 (opens in new tab).
Realizing that, it nearly seems like I am reviewing this card with simply half the information I actually want to have the ability to make a totally knowledgeable choice. As it’s, I can in all probability extrapolate simply sufficient from what little AMD has let slip formally to make a name. Nevertheless it’s nonetheless a little bit of a stretch.
What’s additionally a stretch is any type of justification for the pricing of contemporary graphics playing cards.
Throughout a time of maximum financial problem and uncertainty throughout the globe, it is not an awesome search for each the primary GPU makers to disclose their subsequent technology graphics playing cards with the beginning worth being $900 at greatest. There will probably be arguments the $1,200 RTX 4080’s efficiency over and above the RTX 3080 Ti renders it an unqualified success. However I’ve ideas on that, too.
Nvidia RTX 4080 specs
What’s contained in the RTX 4080?
I assume the spotlight right here is the actual fact that is the RTX 4080 16GB card, not the ‘unlaunched’ RTX 4080 12GB (opens in new tab). However the important thing to why Nvidia ended up having to kill (nonetheless quickly) the 12GB card lies not simply within the quantity of reminiscence it sported, however the precise GPU make-up of the chip itself.
This RTX 4080 16GB card is utilizing the second-tier Ada Lovelace GPU, the AD103 chip. The 12GB card was set to make use of the AD104, which is a third-tier GPU, and realistically solely appropriate to enter an RTX 4070-series card, not some bastardised RTX 4080. Which I assume is the place we will see it subsequent pop up, both with or with out the ‘Ti’ suffix.
Nevertheless, the AD104 stuff shakes out within the ultimate reckoning, the RTX 4080’s AD103 GPU accommodates all of the architectural goodness that made the RTX 4090 such a robust new chip. Meaning you get the fourth gen Tensor Core and third-gen RT Core, in addition to the body producing magic of DLSS 3. Inevitably this far smaller slice of silicon can’t hope to comprise the identical stage of particular person componentry the highest AD102 accommodates.
Picture 1 of 3
The RTX 4080’s GPU is round 62% the scale of the RTX 4090, accommodates round 60% the numbers of transistors, and is dwelling to round 59% of the CUDA cores the highest client Ada Lovelace GPU homes. Which could have you ever questioning why it prices simply 25% lower than the large boi card.
The chip used within the RTX 4080 is not the total AD103 config, nonetheless. That has 80 SMs and 10,240 CUDA cores, versus the 76 SMs and 9,728 CUDA cores the RTX 4080 design accommodates.
Nevertheless it’s nonetheless mighty spectacular whenever you examine the brand new TSMC 4N AD103 towards the highest Samsung 8nm GA102 of the earlier technology. That was a 628.4mm² GPU with 28.3 billion transistors inside it. This can be a 378.6mm² chip with 45.9 billion transistors inside. For those who needed a fast numerical snapshot of the brand new nominally 4nm GPU lithography then it is proper there within the disparity between the discount in die measurement and the huge increase in transistor rely.
|GeForce RTX 4080||GeForce RTX 3080 Ti|
|Lithography||TSMC 4N||Samsung 8N|
|Reminiscence||16GB GDDR6X||12GB GDDR6X|
|L1 | L2 cache||9,728KB | 65,536KB||10,240KB | 6,144KB|
|Transistors||45.9 billion||28.3 billion|
|Worth||$1,200 | £1,269||$1,200 | £1,200|
Regardless of having a GPU that is smaller than the Ampere chip on the coronary heart of the RTX 3060 Ti, nonetheless, Nvidia has nonetheless opted to make use of the very same MASSIVE cooler that it picked for the RTX 4090 Founders Version. The add in board companions (AIBs) have adopted go well with, too—I’ve obtained a PNY RTX 4080 that is the scale of a small moon.
It is obtained a mighty acquainted PCB as nicely, utilizing a modified PG136 just like the RTX 4090 Founder Version. Meaning the identical cutout and closely energy phased board to deal with the 320W Complete Graphics Energy (TGP) the RTX 4080 generates.
It is also obtained a well-known 16-pin energy connector, too, with a triple-headed adapter versus the quad-head one equipped with the RTX 4090. We nonetheless do not know fairly why the 16-pin adapters out within the wild have been melting—not a very good search for a brilliant costly, high-powered graphics card—however Nvidia is persevering with it. As are the AIBs.
It might sound unusual to say, however I wager there will probably be a bunch of individuals making an attempt to get an RTX 4080 adapter to soften as quickly as they’re launched.
Then we come to the reminiscence. As has been broadly reported, this RTX 4080 comes with 16GB of GDDR6X reminiscence, although perhaps not as nicely reported that it comes with a 256-bit aggregated reminiscence interface. In contrast with the 384-bit bus the identically priced RTX 3080 Ti shipped with that feels a bit restricted. Once more, we’re again to referencing the RTX 3060 Ti relating to that specific spec within the Ampere technology.
What doubtless mitigates that and helps the RTX 4080—with its fewer CUDA cores, fewer texture models, and weaker reminiscence bus in contrast with the RTX 3080 Ti—is the truth that the Ada structure follows present chip development process and jams a ton extra cache into the die. We’re speaking ten occasions the quantity of L2 cache within the RTX 4080 as with the RTX 3080 Ti.
It is also utilizing sooner 22.4 Gbps reminiscence, courtesy of a better clock velocity. And once we’re speaking about clocks, the precise GPU clock itself is totally price referencing right here—Nvidia charges the RTX 4080 with a 2,505MHz increase clock, whereas the RTX 3080 Ti comes with a 1,665MHz increase.
Conservative is not the phrase right here, although, as a result of in my testing I used to be getting a median clock velocity of over 2,700MHz in contrast with 1,700MHz on my RTX 3080 Ti Founders Version. Once more, Ada is providing a 1GHz clock velocity bump over the earlier technology.
And relating to uncooked rasterised gaming efficiency, I would say it is leaning closely on each that clock velocity bump and the additional L2 cache to ship its lead over the RTX 3080 Ti.
Nvidia RTX 4080 benchmarks and efficiency
How does the RTX 4080 carry out?
For all of the technical nuances of a GPU that is round 60% the scale of the AD102 chip of Nvidia’s RTX 4090 flagship in most respects that rely, the RTX 4080 is barely on common 33% slower than its barely older sibling throughout our benchmarking suite. And that is within the considerably rarified air of 4K ray traced gaming at extremely presets.
Flip it round and that simply highlights how a lot further energy and silicon is required to squeeze rather more efficiency out of a GPU structure. It is definitely not a linear graph relating to what’s required to nail excessive body charges in trendy video games.
4K gaming efficiency
Picture 1 of 10
Does that make the RTX 4080 extra of a candy spot for non-4K gaming than the RTX 4090? Possibly, however when you’re shopping for both RTX 40-series card to play video games at 1080p then I would perhaps recommend one thing’s not fairly proper relating to your notion of propriety or worth. Nonetheless, the RTX 4080 is virtually nearly as good because the flagship Ada card relating to sub-4K resolutions the place you are arguably nonetheless sure by the boundaries of your processor.
It is attention-grabbing then that Nvidia’s newest drivers have truly improved the GPU efficiency of its playing cards in contrast with the RTX 4090 launch software program. For those who’re doing PC gaming proper, and operating the RTX 4090 at 4K, you’ll not discover the distinction, however the brand new drivers utilized in testing the RTX 4080 have proven an outlined enhance on the decrease resolutions, the place the highly effective GPUs would have been CPU sure earlier than.
In video games the place beforehand the 1080p and 1440p scores had been kind of equivalent the brand new drivers have given it that little bit further, to the purpose the place I used to be initially shocked to see the RTX 4080 truly taking the lead over the RTX 4090 in a bunch of CPU sure video games. I’ve since retested the Ada monster on these titles with this newest software program, and the rising tide has positively lifted all of the silicon boats within the water. In Far Cry 6 and F1 22 I’ve seen a 12% and 16% enhance in 1080p efficiency.
1440p gaming efficiency
Picture 1 of 8
That is all towards the flagship RTX 40-series card, however how in regards to the $1,200 GPU it is ostensibly changing, the RTX 3080 Ti? Averaged throughout our benchmarking suite, the RTX 4080 is round 38% faster than the Ampere card. That is a hefty efficiency hike for a similar worth.
Throw the $100 cheaper RX 6950 XT (opens in new tab) to the Ada wolves, nonetheless, and AMD’s strongest graphics card will get savaged. The ray tracing titles in our suite are those with blood on their arms right here, however you are trying on the RTX 4080 being 68% sooner on common. Within the video games the place rasterised efficiency is all that is required, nonetheless, the efficiency delta is way nearer, nearer to the mark of the RTX 3080 Ti.
The advantages of Nvidia’s frankly unbelievable (up to now) DLSS 3 with Body Era are plain to see whenever you take a look at what it may possibly do within the likes of F1 22 and Cyberpunk 2077. That is the place Nvidia’s claims of a far higher efficiency lead over the RTX 3080 Ti start to make sense, with the AI upscaling tech delivering an efficient doubling of efficiency on the High quality upscaling stage at 4K.
Picture 1 of 7
However actually, it is not the comparisons with different playing cards that I discover essentially the most startling with DLSS 3, it is the uplift over native efficiency and for less than a extremely hard-to-see distinction in constancy. In Cyberpunk 2077, utilizing the 4K extremely ray tracing preset, and the High quality DLSS setting, you are getting a 144% enhance in body charges. And in F1 22, with the identical stage of settings, it is 131% faster. In Flight Sim although it is simply 110% sooner with DLSS 3 with Body Era.
I do know AMD has claimed to have its personal body technology tech coming to FSR 3, however the inexperienced crew’s body interpolating characteristic is already beginning to filter into video games and is clearly making a tangible distinction right this moment.
There may be, nonetheless, nonetheless a big a part of me actually battling the vastly minimize down AD103 GPU being introduced as this $1,200 chip. However the efficiency numbers are there, and whenever you combine them with the ability figures you are a much more environment friendly gaming structure in contrast with Ampere.
Picture 1 of 4
Once more, it is that vs. RTX 3080 Ti and RX 6950 XT comparability that basically hits dwelling right here. It makes use of round 10% much less energy than the earlier technology Nvidia card and is over 50% extra environment friendly by way of efficiency per watt.
However towards the Radeon competitors, which often wins out within the effectivity stakes, the distinction is much more stark. On common the RTX 4080 is utilizing round 12% much less power in contrast with AMD’s strongest GPU, and has 20% decrease peaks. And relating to the efficiency per watt metrics the newest Ada card is sort of twice as environment friendly.
Nvidia RTX 4080 evaluation
How does the RTX 4080 stack up?
What’s the RTX 4080 combating towards? Sooner or later will probably be the RX 7900 XTX, however proper now you are both the RTX 3080 Ti or the RX 6950 XT. And, given the efficiency of this newest Ada GPU, you would not spend $1,000 on the Radeon card or $1,200 on the GeForce card from these benchmarks above.
However that is not likely the place issues stand as of late. Granted, the RTX 3080 Ti remains to be costly, however yow will discover it for across the $1,000 mark when you dig round and ignore the $1,400 choices. On the opposite aspect, the RX 6950 XT has, like just about each single AMD GPU of the final technology, seen a hefty worth minimize. You will discover that card for $800 right this moment. The RX 6900 XT for even much less, round $650, practically half the value of the RTX 4080.
However even that less expensive than their preliminary costs they signify an enormous outlay, and I’d nonetheless battle to advocate both card within the face of the brand new RTX 40-series right this moment. The AMD card has extra of a case, as a result of its rasterised efficiency is nearer to that of the RTX 4080 compared with the additional cash you’d need to drop on the brand new Ada card. However ray tracing, nonetheless you are feeling about it, is nicely on its solution to being a ubiquitous lighting characteristic of contemporary video games, and the older RDNA 2 structure actually suffers beneath the mirrored glare.
That ought to all make this a fairly straightforward advice, then. The RTX 4080 comfortably outperforms each the equally priced playing cards from the earlier technology, most notably the $1,200 RTX 3080 Ti, and due to this fact is admittedly hitting that gen-on-gen efficiency uplift we crave.
However neither GeForce card ought to ever have been a $1,200 GPU.
I really feel just like the RTX 3080 Ti’s pricing was a response to Nvidia lacking out on cashing in on the unprecedented worth hikes the RTX 3080 was seeing over and above the $699 MSRP the corporate launched it with. It appears to me that Nvidia needed its piece of the silicon scarcity/crypto growth motion, and so the RTX 3080 Ti got here out with a $1,200 price ticket.
And that has made all of it too straightforward for Nvidia to match this nearly synthetic worth level with its newest Ada card due to the actual fact it outperforms the last-gen GPU on all counts. Although no less than the RTX 3080 Ti was nonetheless utilizing an enormous GA102 chip at its coronary heart. As I’ve mentioned earlier, Nvidia has pared the silicon again an entire lot to create the AD103 GPU compared to the AD102 chip of the RTX 4090.
Generalising, it is 60% the scale, has 60% of the transistors, and 60% of the CUDA cores, and but is 75% of the value of the RTX 4090. For those who needed to do some easy maths the RTX 4080 actually should be round $960.
However due to the abortive choice to launch with a pair of RTX 4080 playing cards at $1,200 and $899 respectively—with totally different ranges of reminiscence and wholly totally different GPUs—Nvidia was locked into this worth even as soon as it unlaunched the cheaper 12GB card.
I get the TSMC 4N course of node goes to be a dear one at the start of a brand new technology, however by way of GPU yield alone, I would be shocked if the manufacturing margins on the AD103 chip did not begin outstripping the GA102 comparatively shortly. That is nice for Nvidia, and wonderful for its shareholders, however has little profit for us PC players if it will keep an nearly artificially excessive pricing technique.
After which we come full circle to the purpose made at the start of this assessment; how will it stack up towards the upcoming Radeon RX 7900 XTX? I am unable to assist however really feel we is perhaps a really close-run factor. I am anticipating the brand new AMD card to carry out at the same stage to the RTX 3090, and for a $999 card that may make it tempting in comparison with a barely faster $1,200 RTX 4080, particularly with its improved ray tracing capabilities.
I believe AMD will lose on efficiency for essentially the most half however will sit comfortably decrease by way of pricing.
However therein lies the rub. Although the $999 MSRP will include the reference version playing cards from AMD, and sure just a few Sapphire and ASRock playing cards at launch, too, there will probably be myriad $1,000+ overclocked variations doing the rounds. Historically these pricier playing cards change into the norm as preliminary launch inventory runs dry. And when the costs of the RTX 4080 and RX 7900 XTX get shut the choice between them will doubtless change into something however.
However let’s lob within the inevitable premium priced merchandise Nvidia’s board companions will unleash post-launch, too. As all the time, I’d advocate the Founders Version over any overclocked card, and any MSRP model too. Manufacturing facility overclocked GPUs merely aren’t definitely worth the ridiculous worth premium connected to them. Even forward of launch you possibly can see AIB playing cards pushing in direction of the value of the RTX 4090. In equity that is extra prevalent within the UK (opens in new tab) than within the US, the place there’s extra cheap pricing (opens in new tab).
Nvidia RTX 4080 verdict
Do you have to purchase an RTX 4080?
Taking all the pieces else out of the equation, the truth that Nvidia can lop a lot of the GPU good things from its AD103 silicon and nonetheless carry out at a stage that is only a third decrease than the far larger AD102 core means it delivers what you’d need from the RTX 4080. It is quick, the GPU is impressively small (although the shroud is unfeasibly huge), and I am deeply in love with DLSS with Body Era.
It is simply that $1,200 price ticket that stings.
Nvidia has a historical past of pricing playing cards purely in response to efficiency relatively than the precise underlying silicon. Wanting again so far as the Kepler structure and the GTX 680 (opens in new tab), the corporate has made efficiency and effectivity leaps and priced up nominally decrease spec GPUs accordingly.
As a enterprise it is the best factor to do in a capitalist market. However whenever you’re trying on the comparative specs of the RTX 4080, RTX 3080 Ti, and RTX 4090, it is robust to not suppose this can be a GPU that should not be priced at this prohibitive stage. If the RTX 4080 got here out at $999 that may be a much more cheap worth given the cut-down silicon it is utilizing and the relative 40-series efficiency. Throw within the spectre of a brutal macro financial local weather and that feeling’s solely going to be compounded.