It was a matter of time and we finally know almost everything about NVIDIA’s new flagship for 2022, at least until the RTX 40 make an appearance. The leaked reveals not only the presentation and launch data, but it gives a very concrete figure of what we are going to see in performance over the original RTX 3090, as well as its price of course. How much will the RTX 3090 Ti cost?
Why is there so much fuss with this graphics card? In the first place because it is the most powerful that we are going to see in Ampere and secondly because it should integrate the LHR restrictions like the rest of the models that have been released after the originals, which gives users hope in being able to get hold of their services.
The NVIDIA RTX 3090 Ti arrives earlier than expected
As filtered will be the day 5 in the middle of CES when NVIDIA finally unveils this long-awaited GPU for gamers and miners. The best of the best of architecture with certain modifications to give it a little boost in performance, which we can discover for ourselves on another unexpected date: the 27th of the same month of January.
The rush doesn’t seem like such, mainly because there isn’t much to show here beyond small improvements, but this GPU is needed on the market to get the RTX 3090 and its zero mining limiting capabilities out of the way.
As it happened with other models, what NVIDIA tries is not only to distance itself from AMD and possibly Intel, but to stop selling GPUs without mining capacity and in view of the time that has passed and the validity of the limitations, it is partially succeeding.
The GA102 finally in full
With the RTX 3090 Ti we will see all the potential that this chip could develop in its full version, since there was physical space to house more SMs. Thus, this new GPU will have a whopping 10,752 Shaders (+ 2.45% compared to his sister) thanks to 84 SM (2 SM more than its predecessor) which will include 336 TMU, 114 ROP, 336 Tensor Cores and 84 RT Cores where they will all work with 6MB of L2.
Where there will be more changes will be in its VRAM, which now arrives with the highest speed recorded for a GDDR6X: 20 Gbps effective or 1250 MHz true, which will be kept in its 24 GB of memory except that in this case it will not be 1 GB per module, but two. This is a key advantage that will reduce the temperature and consumption of the GPU.
And it is that the increase in benefits will have a direct consequence on the TGP, since NVIDIA will notify a whopping 450 watts for a single GPU, which is 100 more than the RTX 3090. What do we get for it? About a 5% more performance and in some cases possibly around 7%, but not more than there.
The only good thing is that NVIDIA has set an MSRP of 1499 euros, the same as in the RTX 3090, so those lucky enough to be able to get an original NVIDIA model will be in luck, the rest will have to wait to see the speculation, which already estimates it above 2,500 euros.