From day one we saw early models of the RTX 30 and how the 3090 lacked the maximum GA102 chip configuration, we knew there was a superior model out there. However, it took until 18 months later to see it announced and for the RTX 3090 Ti to have a definitive date. Unfortunately, there was a casualty along the way and an NVIDIA graphics card was canceled.
One of the peculiarities that very few people have come to notice about the RTX 3090 is the fact that its configuration is not 84 SM, but 82 SM, this made it clear to us that there was an RTX 3090 Ti in development. The other clue was found in the 12-pin power connector that NVIDIA had added to its F0under edition, ie the standard models sold by NVIDIA itself and which allow much higher powers.
It was not until the power supplies were ready for the new connectors that Jensen Huang was able to launch the most powerful graphics card of the current generation on the PC. Let’s not forget that the RX 6900 XT from AMD at RRP was made to compete with the RTX 3080. Thus the RTX 3090 is without equal and the new Ti model will reign alone as master for several months.
This would be the date of the NVIDIA RTX 3090 Ti
At the moment this has not been officially confirmed, but the date of the RTX 3090 Ti would be just before the end of the current month and, therefore, the March 29, 2022. This would be a response to AMD’s RX 6000 lineup refresh where they will appear with new improved specs led by the RX 6950 XT. The NVIDIA RTX 3090 Ti is a graphics card that uses a variant of the GA102 as its primary GPU. Its configuration is:
- 84 SMs and therefore 10,752 CUDA “cores” or 32-bit floating point units.
- 336 texture units.
- 336 Color Tensor.
- 112 ROPS.
- A base speed of 1560 MHz and a Boost speed of 1860 MHz which allows it to reach more than 40 TFLOPS of power in FP32
- A 384-bit memory bus of the GDDR6X type at 21 Gbps and therefore a bandwidth of 1008 GB/s.
- 450W TGP
- 24GB of memory on the card in 12GB capacities. This means NVIDIA would drop X8 or Clamshell mode, which meant that in the RTX 3090 video RAM would be on both sides of the card, with each pair of chips sharing a data bus with the card.
The card was supposed to appear in January using the standard RTX 3090 design, but apparently there were a number of issues with the component supply. Among the notable changes is the use of a new PCB, which is an updated version of the Founders Edition of the RTX 3090 which was released a year and a half ago.
RTX 3070 Ti 16GB reportedly canceled
The other news would be that the 16GB RTX 3070 Ti It would have become vaporware and NVIDIA would not finally release it on the market although it was a project already completed and even presented. It was the only variant of GA104 in use GDDR6X memory and in an 8-chip 2GB configuration on a 256-bit bus.
The problems with the RTX 3090 Ti and the cancellation of the RTX 3070 Ti lead us to believe that the problem is in the supply of chips with 2 GB of GDDR6X capacity. We don’t know if for the next generation, NVIDIA will use these chips or adopt the new higher capacity GDDR6 memory modules. In case this is the first scenario, it is normal that a few months after the presentation and the launch of the RTX 40, NVIDIA reserves these memory chips for the new generation.
