If we thought that NVIDIA’s switch to the Quadro series would be the only one, we don’t know the company yet. The Tesla series is also saying goodbye definitely and as in the case of the Quadro, it will be replaced by a new batch of graphics cards with the designation Axx. Therefore, both series fall short of the GA100 A100 accelerator already shown above and complement the company’s high-performance product line.
NVIDIA RTX A6000, the data-devouring monster mercilessly
It is the only one and will therefore be NVIDIA’s spearhead in the enterprise sector. The RTX A6000 is the model based on the full GA102, which means giving this card a whopping 10,752 Shaders in FP32 with an overall compute capacity of 38.7 TFLOPS, or whatever is equal is 3.1 TFLOP. faster than the desktop RTX 3090.
As expected, NVIDIA has put all the meat on the grill to provide this RTX A6000 with as much VRAM as possible, so in this case we’re talking about 48GB of GDDR6.
And yes, this is not a typo, it will be GDDR6 and not GDDR6X the type of memory that will be used in this card, as Samsung and Micron do not have the high capacity modules available under the most extreme version.
Better yet, despite being GDDR6, the design of the NVIDIA heatsink goes back to the typical high performance fan, as the chips selected were the best of the wafers and this gives them greater efficiency in combining, to the point being the full GA102. , the card only consumes a maximum of 300 watts.
They will not integrate HDMI, and there are a large number of models available.
Design amazing products, design state-of-the-art buildings, make scientific breakthroughs and create
immersive entertainment with the world’s most powerful graphics solution.
the #NVIDIARTX A6000 is now available.
– NVIDIA design (@NVIDIADesign) December 15, 2020
There will be no different BOMs this time around, but only one RTX A6000 model with different versions of VRAM. Thus, we can choose between the aforementioned 1 GB and 48 GB with the resulting price differences.
One of the most striking sections is the lack of HDMI 2.1, as these cards will be based on DisplayPort 1.4, which is also odd as it doesn’t use the new specification for that interface. In any case and being a novelty in itself, only two cards can be installed in tandem using the new low profile NVLink bridge, so support is also limited compared to previous Quadro options.
And the price? As expected, they’re not cheap at all, but at the same time they’re a step up from what NVIDIA has showcased with Turing. It’s priced at $ 4,650 MSRP, less than half the price of the Quadro RTX 8000, with the latter being considerably slower.
We’re guessing the new AMD Instinct has a lot to do with these prices, so the fight will be fun without a doubt.