Today the review embargo for GeForce RTX 4080 has lifted and we have a total of eight reviews for you:
NVIDIA Founders Edition,
ASUS STRIX OC,
Colorful Ultra White OC,
Gainward Phantom GS,
MSI Gaming X Trio,
MSI Suprim X,
PNY Verto OC,
Zotac AMP Extreme AIRO.
With the GeForce RTX 4080, NVIDIA is introducing their second Ada Lovelace generation card, positioned at $1200, offering performance higher than the GeForce RTX 3090 Ti, besides generational improvements such as DLSS 3 and faster RT cores. While the GeForce RTX 3080 used the same GA102 GPU as the RTX 3090/3090 Ti, NVIDIA has changed that approach with the GeForce 40 series. We now get a new GPU chip called "AD103" that sits between AD104 and AD102, so the company is better prepared to combat AMD's upcoming Navi 31 offerings. AD103 comes with a much smaller die size than AD102, 379 mm² vs 608 mm² which makes it cheaper to produce—an important capability, because AMD is betting on chiplets with Navi 31, considerably strengthening their position.
On the RTX 4080, NVIDIA has enabled 9,728 out of 10,752 GPU cores available in the silicon. The number of ROPs is 112, there's 304 TMUs and 76 ray tracing cores. The memory capacity has been increased to 16 GB, up from 10 GB on the RTX 3080, to better handle today's complex workloads. Memory bus width is now only 256-bit, whereas the RTX 3080 has 320-bit. Technically this is a step backwards, because memory bandwidth is now 717 GB/s (760 GB/s on the RTX 3080), despite the higher VRAM clock frequencies, but this difference is probably compensated for by the MUCH bigger L2 cache (64 MB vs 5 MB).
Gainward's RTX 4080 Phantom GS comes overclocked out of the box, to a rated boost clock of 2640 MHz, which is a 135 MHz increase over the NVIDIA FE speeds. Averaged over all the 25 games in our GPU test suite at 4K resolution we find the Gainward RTX 4080 to be 19% faster than the GeForce RTX 3090 Ti—last generation's $2000 flagship card—very impressive. Compared to the GeForce RTX 3080, the gen-over-gen performance uplift is a a massive 54%. Compared to the RTX 2080 from two generations ago, the RTX 4080 is 2.5x as fast! The recently-launched RTX 4090 is another 23% faster yet. AMD's most high-performance offering, the Radeon RX 6950 XT is 25% behind the RTX 4080, the RX 6900 XT is 30% slower. AMD has announced the Radeon RX 7900 XT and XTX a few weeks ago. While we don't have our own performance numbers yet, it is expected that the RX 7900 XTX's rasterization performance ends up somewhere between the RTX 4080 and the RTX 4090—so things will stay interesting in that segment.
RTX 4080 excels at gaming in 4K. Pretty much all the titles in our test suite ran at more than 60 FPS—at max settings. While RTX 4090 was really designed just for 4K, I feel like RTX 4080 will be a pretty amazing option for 1440p gaming, too. If you can afford it, it will achieve astonishing framerates at that resolution, so you can drive a 144 Hz+ monitor with ease—without having to sacrifice any details. For 1080p gaming the RTX 4080 really is too fast and ends up CPU-limited fairly often, so I wouldn't recommend it for that, because it's just too expensive, rather, buy a decent monitor first and then look at GPU choices.
NVIDIA is betting big on ray tracing with GeForce RTX 4080. Their new Ada architecture comes with several improvements to run RT faster and more efficiently, and adoption rates in games are getting better and better. In their announcement AMD showed Radeon RX 7900 XTX performance numbers with ray tracing enabled, and it looks like they will lose against the RTX 4080, when RT is enabled. Older Navi 2x cards have even lower RT capability, so for the best ray tracing performance it seems that NVIDIA will keep the performance crown, even after AMD releases their own cards.
Gainward's GeForce RTX 4080 Phantom GS looks great and features adjustable RGB lighting. Under the hood we found a powerful cooler that ensures temperatures are low, and Gainward paired this with great fan settings. Under full load the card runs at a quiet 31.5 dBA. If you prefer even lower temperatures, then you can activate the "quiet" BIOS, which runs the fans a bit slower, at 30.4 dBA—almost whisper quiet. I do wish that Gainward had set fan speeds even lower. Currently the difference between both BIOSes is rather small. Temperatures are excellent anyway: the difference between default BIOS and quiet BIOS is only +1°C. Imagine what noise levels we could get had they allowed +5°C or +7°C, which are both still very conservative. Nevertheless, the card is definitely a good choice if you like a low-noise cooling solution.
NVIDIA is building AD103 on the same 5 nanometer TSMC process as AD102 (RTX 4090), so efficiency should be good. It's actually shockingly good, beating even the RTX 4090 in efficiency, which makes the RTX 4080 the most energy-efficient graphics card ever made. Energy efficiency not only helps with your power bill, but it's also the magic bullet to reduce GPU temperature and noise levels. High efficiency means that less complex cooling solutions can be used, and ultimately that the card can clock higher and run faster overall. For engineers, energy efficiency is the limiting factor for GPU performance today. We're pretty much at the limit of what cooling solutions can handle, and users will not accept sitting in a hot room during long gaming sessions, and power draw can't go much higher either. It seems that the RGB or some changes to the VRM circuitry increased non-gaming power levels quite a bit. Gaming power draw is increased, too, but by a very reasonable 10%.
In this review we've tested NVIDIA's new DLSS 3 frame-generation capability, and I have to say I'm impressed. At first I was highly sceptical and thought it would be like the soap opera interpolation effect on TVs, but no, it works REALLY well. The algorithm takes two frames, measures how things have moved in those two frames and calculates an intermediate frame in which these things moved only half the distance. While this approach is definitely not problem-free, especially when pixel-peeping at stills or slowed down video, in real-time it's nearly impossible to notice any difference. As you run at higher FPS and resolution it becomes even more difficult because the deltas between each frame are getting smaller and smaller. I also feel like we're only seeing the beginning of this technology, and there will be numerous improvements in the future. Adoption rates should be good, because implementing DLSS 3 frame generation is very easy if you already have DLSS 2 support in your game. Another interesting NVIDIA Tech is "Reflex," which reduces the total gaming latency, so you see things earlier on your screen and can react faster, to get more kills or survive for longer.
Gainward is the only company that hasn't given us any information on pricing yet, no idea why, they definitely know it. For this review we assumed a price point of $1300, which is in-line with similar offerings. Even a $100 price increase isn't easy to justify, certainly not with +2% out of the box performance from a factory OC. The better, quieter cooler could be something, but $100, I'm not convinced, for $50 I would definitely think about it. Even at $1200, the RTX 4080 Founders Edition is expensive. There's no doubt that performance numbers are impressive, and that a lot of tech has been integrated in the product, but in these times I have to wonder "isn't this a bit much?" Looking at the price/performance ratio based on the numbers from our review, we find the RTX 4080 at $1200 matching the $950 RTX 3080 Ti almost exactly in price/performance, just like the $900 RTX 3090. It conclusively beats the RTX 3090 Ti ($1400) in both price and performance. The RTX 4090 is scalped into oblivion right now, going for an insane $2400 on Newegg—so not an option if you care about value in any way. GeForce RTX 3080 is heavily discounted at the moment, at sub-$700 and that's a very interesting value proposition. The RTX 3080 comes with 10 GB VRAM though, and is from the last generation, which might scare away many buyers—it's still a fantastic card. Probably the most interesting alternative is Radeon RX 6900 XT, which is listed for $655 at the moment. that's almost half the price of RTX 4080, with 70% the performance offered. RX 6900 XT has lower RT performance, too, and it doesn't offer the DLSS 3.0 frame generation capability, but if you're just looking for an affordable high-end graphics card, that's a compromise you will have to make.
AMD's upcoming Radeons certainly have the ability to disrupt NVIDIA's plans for the RTX 4080, and I'm sure NVIDIA is aware of that. To me it looks like they are pricing the RTX 4080 in a way that ensures they can sell off their existing GeForce 30 inventory, and that still gives them enough breathing room to react to AMD's new offerings with price cuts, if needed. Also, they will justify the higher price with additional ray tracing performance and features like DLSS 3.0. At the end of the day, with miners no longer snatching up all cards, it comes down to you the customer: if you feel any product is too expensive, or doesn't meet your expectations in any other way, just don't buy it, vote with your wallet. However, given the numbers in this review I am confident that RTX 4080 will be a hit and that it will sell in numbers—it's the second-fastest graphics card available—and has all the new features and technologies—available and working today.