The day is finally here and we can now share our official review of the NVIDIA 3080 cards. Right here we have an ASUS 3080 TUF Gaming OC GPU, this is a factory overclocked variant and is approximately 75 MHz faster than the standard. In this article we will focus on gaming performance including Ray Tracing and also some productivity tests.
Jumping straight into it – in the reveal presentation Jensen referred to 3080 as the new flagship GPU with 10GB of GDDR6X VRAM which is the world’s fastest memory. It also boasts 3rd gen Tensor cores and 2nd gen RT cores which are supposed to be up to 2 times faster. With all of this in mind, 3080 should be up to 2 times more powerful than 2080 at the same price.
So we have set out to check how much of that is engineer Jensen talking and how much is his marketing alter ego. Since we don’t have 2080, we will be doing all the testing with 2080 super instead.
We are running all the tests using our intel 10700k bench. We chose this because at higher clock speeds it is still a faster platform when it comes to gaming and we want to minimise any potential bottlenecks.
GPU 1 – ASUS RTX 3080 TUF Gaming OC GPU 2 – ASUS ROG-STRIX-RTX2080Super GAMING Intel i7-10700k ASUS Z490 Maximus XII Extreme motherboard KLEVV CRAS X RGB (8GB x2 KLEVV CRAS C700 RGB NVME SSD Fractal SFX 650w power supply Fractal Design Celsius+ S24 Dynamic Cooler Streacom BC1 Open Benchtable
With that out of the way, let’s jump into the gaming benchmarks, and spoiler alert – in every single test we did the 3080 was faster so we will focus on the actual improvement amount and more specifically on the 1 percentiles as these are the ones that ensure smooth gameplay. Starting with Shadow of the Tomb Raider at 1080p there is 22% improvement over 2080 super. At 1440p the difference jumps to 50% and at 4k the difference is 64%. When we look at the raw frame rates, we essentially moved from 1440p to a comfortable 4k gaming.
In Ghost Recon Breakpoint while using DirectX 11, improvements to both 1080p and 1440p are about 27% and then at 4k it is 46%. In this title it looks like 4K gaming with 60 FPS and above is becoming possible using RTX3080 but it is a little close to the line.
When looking at the same game but using Vulcan API – the story is very different. Here at 1080p the older card was doing so well that we have only 7% improvement from RTX3080 and only 15% at 1440p. On the other hand in 4K we double the frame rate and hit 74 FPS at 1 percentiles making this card very capable at 4K gaming.
Next we have Horizon Zero Dawn and this game is a challenge to run on any PC. With RTX2080 Super at best you could run it at 1080p and even then the 1% are below 60fps. RTX3080 on the other hand improves 1080p score by 60% and 1440p score by 54% pulling off a really good 1440p gaming experience.
Another challenging game is Total War Three Kingdoms and here we have a moderate 16% improvement at 1080p and a whopping double frame rate at 1440p. Just like in the last game 4K is still somewhat a challenge but even so it is very impressive to jump a whole resolution class.
Next we have a game that supports Ray Tracing and also DLSS. It’s Metro Exodus and while this game may not have the best ray tracing effects, at the same time it does provide us with a great representation of how it taxes your system and how much performance can be recovered using DLSS. First we need a baseline where both are turned off. We have 40% improvement by using RTX3080 while at 1080p, 48% while at 1440p and 55% when moving up to 4k. Unfortunately the raw frame rate just barely passes an acceptable level at 1080p.
Setting ray tracing to ultra immediately destroys both graphics cards but even then we still see that performance from RTX3080 is 56% better at 1080p, 61% better at 1440p and 76% better at 4k.
With DLSS enabled things change, and there is 38 to 46% improvement between the resolutions, actually making this game playable at 1080p and borderline playable at 1440p while ray tracing is on. This truly highlights the power of the new generation of RT and Tensor cores.
To quickly recap, I believe in all the games we covered in order to have a truly good experience the frame rate at 1 percentile should be at least 60 and above. With this in mind in most games that we benchmarked, if you take 2080 Super and swap it with 3080 – you can safely turn up the settings by at least 1 resolution, or like in Ghost Recon Breakpoint – we jumped from 65fps at 1080p to 60fps at 4k which is really impressive.
If we focus on ray tracing, there has definitely been a massive improvement and together with DLSS this generational leap allows you to either choose 1 tier up in terms of resolution or enable ray tracing and at least not lose out on reasonable frame rate at current resolution. This provides users with a better choice. By the way if you would like us to do a more dedicated video about ray tracing, let us know, we can test more titles and do a deep dive into that.
Gaming aside, with 10GB of very fast VRAM and clearly great computational power it should do well in productivity tasks such as video editing, transcoding and rendering. We ran a few tests to check it out, first one being Handbrake. As you can see there is no difference in performance, but this is due to a CPU bottleneck.
On the other hand when running Blender, the results are simply astonishing. In BMW test we have more than halved the rendering times.
When doing a longer Classroom test, the difference is even more prominent, and this is where I can see that direct 2 times improvement over 2080. To be fair, here we have much more than just 2 times if we consider that 2080 super is about 10% faster than the standard 2080.
We also have a very special test for the video editors out there, particularly people using BlackMagic Davinci Resolve, but the relative performance difference is useful for everyone. In this test created by Puget systems the performance increase between the two cards is considerable. We are looking at 24% improvement on overall score while editing at 4K. Elements that require a lot of GPU computational power like OpenFX or noise reduction have improved as high as 50% making this card a considerable bump in performance for all creators.
Next we wanted to check how loud this card gets and once again we used this in comparison with 2080 Super. This test is only to check the relative loudness in an open air bench so depending on individual cases the final noise results will be different for everybody.
We tested the fans at three speeds: 60, 80 and 100%.
While at 60% the RTX3080 is actually below our room floor level at 34.5db, so the sound meter registers only ambient sound here. RTX2080 Super on the other hand is at 46 db and is certainly audible. When turning both to 80%, we have 45 db on the 3080 and 56 db on the 2080 Super. Considering sound loudness doubles every 10db, this is now a massive difference. The noise coming from 3080 is more like a gentle hum, but 2080 Super is very loud. When we set fans to max of course both of them are screaming but 3080 is still only 55db while 2080 super is at 60db.
Since we are on a subject of noise – in Assassin’s Creed Odyssey while benchmarking at 1080p we could hear the coil noise from the RTX3080. It was audible while on the bench right next to us, and only for a small part of the test. We didn’t hear it in any other scenarios, but will keep testing.
Right, that’s a lot of information on this new card. To be honest I was very skeptical when hearing the phrase “Greatest generational leap ever”, but after carrying out the tests, we found compelling arguments in multiple scenarios, and in some it does indeed deliver 2x the performance of 2080. It is definitely a very exciting product and it finally opens up the extremely premium performance for a more “reasonably premium” price. I can’t wait to test all the other upcoming cards!
Affiliate disclosure: as an Amazon Associate, we may earn commissions from qualifying purchases from Amazon.
Comments