Following last month’s launch of the $1,599 Nvidia GeForce RTX 4090, a graphics card meant largely for professional use, today the new generation of GPUs has finally arrived for the slightly less wealthy crowd of PC gamers in the form of the Nvidia GeForce RTX 4080 16GB Founders Edition, which launches tomorrow for $1,199. That means the entry point for the RTX 40-series’ lofty performance claims and frame-generating DLSS 3 has come down at least a bit. We’re still at the beginning of this new GPU generation, but so far the RTX 4080 is a strong showing.
What’s in a Name
Before we get to the spec sheet and performance numbers, we should cover the RTX 4080 naming kerfuffle and talk a bit about how Nvidia’s GPUs are typically positioned – and why this generation is a bit different than years previous. Dating back more than a decade, Nvidia GPUs bearing the “-80” moniker are considered the flagship mainstream cards, and have been priced in the $500 to $700 range. Looking more specifically at recent trends, the GTX 1080 launched at $599, and both the RTX 2080 and 3080 launched at $699.
You might be wondering, then, why the RTX 4080 starts at nearly double that range. The answer is in that aforementioned naming fiasco. Originally, Nvidia planned two variants of the RTX 4080 – a 16GB for $1,199 (the version we’re reviewing here) and a 12GB, priced at $899. This wouldn’t have been the first time Nvidia has launched VRAM-variant cards, but typically the amount of VRAM was the only difference, whereas in this case the two cards also carried different core counts and clockspeeds – differences that previously would have warranted a bump down to another tier (in this case, the RTX 4070).
People rightfully complained about the confusion this was already starting to cause, and to Nvidia’s credit it responded, opting to “unlaunch” the RTX 4080 12GB; it’s now rumored that those cards will be re-announced with the “RTX 4070 Ti” name, though as of yet there’s nothing official.
That’s all well and good, but it still leaves us with a “mainstream” card bearing the enthusiast-tier pricing that would previously have been reserved for a card bearing the “Ti” label – representing a mid-generation step up. In other words, typical generational comparisons are a bit skewed this time around, so we’ll mostly be comparing the RTX 4080 against the RTX 3080 Ti, which also launched at $1,199 in June 2021, as opposed to the RTX 3080.
Nvidia GeForce RTX 4080 – Design and Features
If you read my review of the RTX 4090, you’ll remember that it’s an absolutely massive card, in both size and performance terms. The RTX 4080, meanwhile, is…not any smaller. It bears the same triple-slot designation, measuring 11.9 inches (304mm) long, 5.4 inches (137mm) wide, and 2.4 inches (61mm) thick – the exact same dimensions as the RTX 4090. This is a big card. For comparison, the RTX 3080 measured 11.2 inches (285mm) long, 4.4 inches (112mm) wide, and 1.5 inches (40mm) thick, while the RTX 2080 and GTX 1080 were even smaller.
Most of that heft comes from the large, dual-axial flowthrough cooling solution required to keep temperatures in check. The cooler design is mostly similar to that of the RTX 3090, but with larger fans and taller fins in order to achieve what Nvidia says is 15% more airflow at the same acoustic level. In practice, the RTX 4080 remained whisper-quiet while keeping temperatures hovering around 53-55C, with a peak of 57C, during a long stretch of benchmarking.
Compared to the RTX 3080 Ti, the RTX 4080 has 9,728 CUDA cores (down from 10,240), 304 fourth-gen Tensor cores (vs 320 third-gen), and 76 third-gen RT cores (vs 80 second-gen). In other words: it has newer cores, but slightly fewer of them overall. The decrease in count shouldn’t be alarming though, since the 4080 includes a boost clock speed of 2,505MHz compared to the RTX 3080 Ti’s 1,665MHz clock, not to mention the 16GB of GDDR6X VRAM, compared to 12GB on its 30-series “predecessor.”
Like the RTX 4090, the 4080 uses the somewhat contentious 16-pin 12VHPWR power connector that has recently been in the news due to reports of it overheating and melting. We haven’t had any issues with it in any of our testing, but we’ll definitely be monitoring the situation as this generation of graphics cards matures.
Speaking of power, the RTX 4080 has a TDP of 320W, down from 350W on the RTX 3080 Ti. Nvidia recommends using a 750W power supply at minimum. There’s also a 3x 8-pin adapter in the box for people whose power supplies don’t have the new connector.
For ports, the RTX 4080 has 3x DisplayPort 1.4a and 1x HDMI 2.1a. This is the typical layout for current-gen graphics cards, though AMD’s recently-announced RX 7900 XT and XTX use the newer DisplayPort 2.1, which has more than triple the bandwidth and enables 4K resolution at up to 480Hz, or 8K up to 165Hz, versus 240Hz at 4K and 60Hz at 8K for DisplayPort 1.4. Most games and monitors won’t be able to take advantage of that bandwidth, so it’s sort of a moot point, but AMD does technically have the advantage.
Nvidia Geforce RTX 4080 – Performance
Starting off with our synthetic benchmarks, the RTX 4080 comes out swinging in 3D Mark Fire Strike Ultra with a 17% improvement over the RTX 3090 Ti and 28% over AMD’s RX 3950 XT – the two best GPUs of the previous generation – and a 35% boost over its generational price-equivalent predecessor, the RTX 3080 Ti. As you’d expect it falls considerably short of the RTX 4090, however, with a score of 16,255 compared to the RTX 4090’s 21,872, which makes total sense given that card costs $400 more.
Moving on to Unigine Heaven, the RTX 4080 edges out the RTX 3090 Ti and RX 6950 XT at 1080p and 1440p, but actually falls short of both cards at 4K. Against the RTX 3080 Ti, though, it consistently wins out with a 13% lead at 1080p, 14% at 1440p, and a slight 4% at 4K.
The ray tracing synthetics are more dramatic. The RTX 4080 offers an average uplift of 28% compared to the RTX 3090 Ti across our three tests, and of course absolutely demolishes the RX 6950 XT, which lacks the ray tracing chops of Nvidia’s hardware. Comparing it to the 3080 Ti offers even more impressive results, with an average improvement of 45% compared to that card.
Moving on to our gaming benchmarks, the RTX 4080 again has a strong showing across all three resolutions tested. At this point our benchmark tests are basically CPU bound at 1080p, with the RTX 4080 pinging the meter alongside the more powerful RTX 4090. 1440p is relatively similar, with the card showing large gains over the last generation in tests that aren’t CPU bound, and of course matching the best in tests that are.
Given the high-end nature of this hardware, however, the real story is at 4K. (If you’re not playing at 4K or greater resolutions, you shouldn’t spend this much money on a GPU.) Expanding our test suite slightly, you can see that the RTX 4080 offers considerable gains over the previous generation, with an average improvement of 27% over the RTX 3090 Ti and 45% versus the RTX 3080 Ti. Keep in mind that the latter of those cards launched at the same $1,199 price, while the former carried a $2,000 MSRP when it launched earlier this year (though prices have now fallen to around what you should expect to pay for an RTX 4080 fresh off the shelf).
Those are impressive gains, but not really out of the ordinary when you consider this is a new graphics generation. Looking back to our RTX 3080 review, that card offered 50% to 70% improvements over its generational predecessor, the RTX 2080 Super. That’s not to discount the RTX 4080 – 4K framerates well above 60fps in the most demanding games will raise eyebrows for a few years to come – I just feel it’s important to remember that we’re talking about high-end, if not enthusiast-level pricing here, so my expectations are sky high.
Finally, I want to touch on Nvidia’s new DLSS 3 frame generation technology. Check out my RTX 4090 review for a more in-depth explanation, but in short the GPU looks at two sequential frames, calculates the difference between them, and then uses AI to generate a frame in between them. As with the RTX 4090, I tested DLSS 3 and frame generation in Cyberpunk 2077.
DLSS again offered a stunning uplift, bringing the RTX 4080’s framerate up to 73 without frame generation, and 108 with it. Those are awesome numbers for one of the most technically demanding games available on PC today – and remember, this benchmark is run at 4K with max settings and ray tracing enabled. The 30-series RTX cards, meanwhile, both receive less of an uplift from DLSS, and don’t have access to frame generation all together.
Of course, DLSS 3 is still a new technology, and game support is limited for the time being. That said, it’s steadily rolling out to more games, including Microsoft Flight Simulator, A Plague Tale: Requiem, and Marvel’s Spider-Man Remastered. If support continues to grow as expected and the performance uplift remains as formidable, DLSS 3 will be the killer feature that truly makes upgrading to a 40-series card worth it for high-resolution, high-framerate gaming.
from IGN Reviews https://ift.tt/KoZeMJa
This could be a real lead forward for personal gaming... Revolutionise gaming
No comments:
Post a Comment