As per a leaked image shared by Twitter user mooreslawisdead, NVIDIA’s next-gen Ampere based Quadro graphics card will feature the full-fledged GA102 die with as many as 84 SMs (or 10,752 shaders) paired with 48GB of memory via a 384-bit bus.
Update: A screenshot of the Quadro RTX A6000 has also been shared by the YouTuber as you can see above, in addition to the Quadro roadmap (below).
As per the above image, the new flagship will be the Quadro RTX A6000 will succeed the RTX 8000 with 48GB of memory, NVLink, PCIe Gen 4 support and a TDP of 300W+. Below that, we have the RTX A5000 with 24GB of memory and a TDP of 230W. Finally, the budget-oriented Quadro RTX A4000 will come with 16GB of memory and a TDP of just 150W. It’s unclear all three GPUs will come with GDDR6X memory, or whether the lower-end models will stick to the cheaper GDDR6 standard. The older story continues below.
As per the screenshot, the new Quadro will have a healthy boost clock of 1860MHz and a rather timid memory clock of 16Gbps for a total bandwidth of 768GB/s, just a bit higher than the RTX 3080’s 760GB/s.
In comparison, the RTX 3080 and 3090 feature 19Gbps of GDDR6X memory which is quite a bit higher than this. In fact, these clocks make you wonder if NVIDIA is actually employing the newer GDDR6X memory with the Quadro, or limiting it to the GeForce lineup. The reasons behind this could be multiple: Either to keep the price lower or limited gains in performance for creators, or perhaps both. The latter seems unlikely and if true, it put NVIDIA’s Ampere based Quadro at a disadvantage compared to AMD competitors, most notably the Radeon Pro VII which leverages HBM2 memory for a total bandwidth of 1024GB/s. Of course, this could just be an underclocked engineering/verification sample, so we can’t be too sure just yet.