There are indications that lead us to think that NVIDIA could launch an even more powerful graphics card than the current RTX 3090 . Although it is normal for new architectures to come out and increasingly faster, in the case at hand there are enough clues to think that NVIDIA could have in the bedroom a more powerful graphics in the form of RTX 3090 Ti, and who knows if a complete renovation of the entire range for the right moment.
The brand new RTX 3090 is at the time of writing this article, the most powerful graphics card that NVIDIA has on the market within the GeForce range; however, there are indications that when the time comes a somewhat more powerful graphics card may appear.
The RTX 3090 Ti would use the full configuration of the GA102 chip
The core chip of the RTX 3090 is a beast called the GA102-300 with a 628mm 2 area and 84 SMs in total.
What if we told you that the RTX 3090 does not use the full 84 SM configuration but one with 82 SM? We’re actually talking about just over 2.5% extra performance at the same speed in all respects, but it’s still 2 disabled SM’s that could add some performance boost.
We have to keep in mind, however, that as manufacturing polishes it will be easier to get graphics cards that have a GPU with full configuration and that are stable at speeds much higher than those of the current RTX 3090. We know that at the moment only 10% of the graphics produced are used to manufacture the Founder Edition models, but as time goes by, this percentage will rise as the manufacturing node is polished.
The RTX 3090 Ti would make use of the faster GDDR6X
In terms of memory, graphics cards are not affected by latency but are vulnerable to the bandwidth available to them. The reason for this is that in stream processors, their performance is related to bandwidth, and if this is not large enough then the productivity obtained from the GPU is not 100%. The effect of using faster memory was seen with the GeForce RTX 20 × 0 Super range that had a faster GDDR6.
NVIDIA may end up launching a new lineup next year taking advantage of faster and even higher density GDDR6X memory. But rather this should be expected by 2021 in the form of a relaunch of the recent RTX 3000 with a faster GDDR6X memory because the standard supports up to about 21 Gbps of speed at the moment. Currently GDDR6X chips are integrated at 19 Gbps and 19.5 Gbps.
21 Gbps GDDR6X chips have a very low success rate; This means that of the chips that come out of each wafer, those that can support that specification are too few at the moment to make a graphics card with that memory speed viable.
Another problem is the fact that GDDR6X is currently sold in densities of 1GB per chip when 2GB GDDR6 is already available. This makes GDDR6X memory enormously expensive and makes it necessary for assemblers to decide for the moment between the higher capacity of the GDDR6 and the higher speed of the GDDR6X, something that should be solved with the launch of GDDR6X chips with 2 GB of density.
The optimization of the consumption of the GDDR6X is important for the eventual adoption in the GPUs of the RTX 3000 range that still work with GDDR6, and not only can we expect an RTX 3090 Ti and 3080 Ti with GDDR6X memory at a higher speed but even eventual RTX 3060 Ti and RTX 3070 Ti that replace the GDDR6 controller with a GDDR6X one.
Ti or Super?
We do not know if the improved versions of the RTX 3000 will come out with the Ti or Super surname, since one thing or the other can happen. We do not believe that NVIDIA will launch the renewed version of the RTX 3000 before launching all the representatives of that range.