As most reading this will be familiar with the more in-depth parts of the RTX family, we will try and keep this a relatively short read. In the spirit of transparency, as this was made pre-release, some tests were not fully functional, and have not been included.
While pricing of this particular Gigabyte RTX3080Ti card is unknown at the time of writing, Nvidia has put it at 1200 Euro, with the normal RTX3080 at 720 Euro, and the RTX3090 at 1549 Euro. This is way above what we had hoped for, but in all honesty, so is the performance.
To our surprise, this card, being made to fit both GeForce experience gaming drivers, and content creator Studio drivers, is only slightly overclocked by Gigabyte with the base clock being at 1710Mhz instead of 1665Mhz, and it still managed to match, and even on a few occasions beat the RTX3090. This is amazingly done with half the VRAM of the RTX3090, but the pricing is also matching this, with actual retail prices most likely being around the 4000 Euro mark, as RTX3090 cards have been seen going as high as 4500 Euro with select retailers.
Your question might be "do we need another card?" - and for Nvidia the answer is yes, the AMD 6900XT will beat the RTX3080 and even the 3090 in some cases, despite being offered - at least in theory, at a bit over 1000 Euro. Nvidia needs a card that can compete with it, and be in the same price range. The last isn't met to my disappointment, but the performance is what Nvidia needs.
The RTX3080Ti is by all means much closer to the RTX3090 than the RTX3080. It uses the GA102-225 GPU, with the RTX3090 and RTX3080 using the 100 and 300 variants.
Sporting 10240 Cuda cores puts it almost on par with the 3090, and far ahead of the 8704 of the RTX3080. it even rocks 12GB of GDDR6X VRAM, and has a 1.67Ghz boost clock. While this is slower then the regular RTX3080, it has a 384 bit memory bandwidth, with the normal 3080 having just 320 bit. It has 320 Tensor cores and 80 RT cores, with a memory bandwidth of 912.4 GB/s.
It has however, kept the 2x8pin PSU design.
This Vision G comes in a fantastic white colour that will make it very attractive for many PC builders. The shroud of the card is in a very high quality aluminium finish that doesn't look cheap at all. And its still pretty compact - for an RTX 30 series card.
It uses the well-tested Windforce 3X cooling system with two 90mm fans, and the middle 80mm fan spinning the other way to reduce noise. It has a massive backplate and Nvidia's 30 series signature screen cooling, meaning that some of the backplate has been removed to increase airflow on the extended part of the card that doesn't have PCB under it. As usual, graphene nano lubricant is used, so is large copper plating on the GPU, and heat sinks that connect the VRAM and composite heat pipes are used, as we expect from Gigabyte. There is of course, also RGB, but I turned it off during testing to avoid it contributing to heat build-up.
During testing 35dB max noise level was measured, which isn't much, and the card did 65 degrees when under load, 76 under heavy load, and hit 80 at maximum synthetic load. Perhaps a few degrees less had been more desirable, but that would have affected noise levels.
I would might have liked some larger fans, but that can be relieved with some aggressive fan curves. However, this is somewhat Gigabyte's "stock" card, and not their total-over-the top overclocked model.
But lets get to the main event, the Benchmarks.
All games and software is at max possible settings.
Time Spy: 17256
Time Spy Extreme:9429
DirectX Raytracing: 55.78
DLSS 2.0 1440p: 60.08/ 129.34
Fire Strike: 34155
Fire Strike Extreme: 22520
Fire Strike Ultra: 12483
Port Royal: 12930
These are interesting, as they are on par with a stock RTX3090, slightly over, or slightly under, but with so little difference that it doesn't matter for either card.
4.32.20 - This is actually the only case where the RTX3080Ti beats not only the RTX3090 in both stock and overclocked versions, but also the RX6900XT, by 0.13 seconds, but its still a win.
Heaven might be old, but its also measuring very precisely. The RTX3080Ti didn't beat any of the competing cards, but it was a very close call in 1080p with the RX6900XT with only 130 points difference.
Games - FPS
Total War: Warhammer II, 1440p: 127.2
Total War: Warhammer II, 4K: 82.6
Hitman 2, 1440p: 106.97
Hitman 2, 4K: 94.48
AC Valhalla, 1440p: 80
AC Valhalla, 4K: 59
But there are some games where an actual difference is seen consistently, and where it isn't on par with the RTX3090
Red Dead Redemption 2, 1440p: 108.65
Red Dead Redemption 2. 4K: 67.25
RTX3090 got 74.93FPS in 4K for the Gigabyte Gaming OC, while an overclocked card from a competitor got 78.76FPS.
Doom Eternal 4K: 268
Horizon Zero Dawn, 1440p: 118
Horizon Zero Dawn, 4K: 76
Dirt 5: 104.7 - with more than 30% better FPS, this is the one game where the RTX3090 outperforms the RTX3080Ti at a level where the price difference makes sense.
Total War: Troy, 1440p: 102.1
Total War: Troy, 4K: 58.1
Control 1440p: 91
Control 1440p + DLSS: 146
Control 4k: 46
Control 4K+ DLSS:82
In Control, the double amount of VRAM gives the RTX3090 a 10% higher FPS rate.
Cyberpunk 2077, 1440p: 92
Cyberpunk 2077, 4K: 84
While the difference in 4K was about 10%, the non 4K gave more than 18% difference.
While I must congratulate Gigabyte on having made a beautiful card with high performance, I dread the price. The result isn't as clear-cut as I had hoped, and while the RTX3080Ti GPU can in some cases match and even beat the RTX3090, its still extremely expensive, and only very rarely able to beat the AMD RX6900XT in anything but Nvidia optimised scenarios. If they could just lower the price to the same level as the competition, the better Ray Tracing of the RTX3080Ti would make it a much more appealing offer for many gamers, but even at a theoretical MSRP of 1300 Euro or so, it is still too expensive for most, and currently, you may not even be able to get your hands on one. On the bright side, for pure gaming, buying an RTX3090 makes no sense anymore, even if you almost exclusively game with DLSS and Ray Tracing.
Would I buy one myself if I could afford it? Yes. Yes I would. If I could get one.