The next generation of NVIDIA graphics cards is in the oven and will arrive in just a few months if all goes well. Nothing makes you think otherwise and less after having seen the attempt and try with the RTX 3090 Ti unlocked at 600 watts. But here one more added factor comes into play that must be taken into account: the new 24 Gbps GDDR6X memory. And it is that this can make a difference with AMD as the RTX 30 did against the RX 6000, for better and for worse… In any case, the RTX 4090 is already in testing.
The dates are constantly moving, the releases are beginning to be indeterminate and we are seeing it with Intel and their ARC GPUs that do not finish arriving in the case of laptops, and they do not present the desktop versions with more and more delays without date yet. In the case of NVIDIA, the fact that we still do not have a leak with performance data or even sample photos does not seem to be misguided and that is because the chip that will give life to the RTX 4090 is currently being tested by the company.
NVIDIA and 24 Gbps GDDR6X on the RTX 4090 for testing
In the first place we had a rumor going around that was neither confirmed nor denied and is nothing more than the change of nomenclature for GPUs and their series. It seems that everything will follow the normal course and these graphics cards they will be RTX 40 series and not RTX 50 as speculated, hence the naming of RTX 4090 and not RTX 5090 as such.
The leak of the tests for the AD102 is only a small preview of what is to come in a few months but at the same time it shows that the numbers are not coming out for NVIDIA and neither are they coming out for Intel.
Ignore these fanboys, let’s turn our attention to GPU.
AD102 has started testing. https://t.co/yziVJE8eFp
— kopite7kimi (@kopite7kimi) April 19, 2022
If the test period has not currently ended, it means that the production ramp is delayed to deliver the graphics cards at some point in the third quarter of this year, which moves the entire calendar being optimistic for the fourth quarter of 2022.
The problems may come from two interesting details that have slipped through the Internet: the 24 Gbps GDDR6X and the PCB design along with its weight.
The new memory and temperature problems
Almost certainly we will have to see how the RTX 4090 with its AD102 GPU will consume a total of 600 watts thanks in part to Micron’s new VRAM. This one surpasses all the previous ones and apparently we will have a bandwidth over TB/sbut this seems to imply too high a temperature factor that requires an improved PCB by NVIDIA and a heatsink that in its Founders Edition version could be massive.
It is rumored that the new NVIDIA in terms of cooling is concerned will go to nothing less than three fans with backplatesince the PCB would support memories on both sides in order to offer the appropriate ratio of power and temperature.
This would result in a GPU of more than 2 kg in your version Founders Edition and also the design would be triple slot, where in the case of manufacturers for custom models we would have 3.5 slot designs, that is, full-fledged monsters.
And it’s that cool 600 watts below 80º C It will not be easy, proof of this is that the PCBs will presumably arrive with a VRM of up to 28 power phases. If we add everything that has been said, it seems clear that prices are not going to help, because the technical and electrical level will increase again, Micron’s 24 Gbps GDDR6X will not be cheap and the dissipation systems are more advanced, not to mention of the cost of the chip itself that will skyrocket.