NVIDIA Quadro RTX A6000 Ampere Card Pictured Allegedly With 10,752 CUDA Cores

NVIDIA Quadro RTX A6000
Straight from his kitchen, NVIDIA CEO Jensen Huang has served up to recipes for Ampere so far, including the A100 for machine learning chores and the GeForce RTX 30 series to satiate gamers' appetites for high-end graphics. So, what dish comes next? That will very likely be a line of Quadro cards for professionals. As we wait, a picture of a supposed Quadro RTX A6000 graphics card has emerged.

Shown above, credit goes to Twitter user @mooreslawisdead for the image. Note that it is labeled simply as "RTX 6000," which is the same model number that already exists within NVIDIA's previously generation Turing GPU stack. However, this version is said to feature a "full GA102" GPU based on Ampere, with 48GB of memory on board. Hence why we are calling it the Quadro RTX A6000.

The image comes from the same source who previous shared a screenshot of this card's specifications, as listed in the NVIDIA Control Panel. Have a look...

NVIDIA Quadro RTX A6000

If the screenshot is accurate, the Quadro RTX A6000 will wield 10,752 CUDA cores, a 1,860MHz boost clock, and a 384-bit memory bus, yielding 768.1GB/s of memory bandwidth served from the 48GB of GDDR6 memory (16Gbps).

The memory allotment is one of the more interesting developments, considering that the GeForce RTX 3080 and upcoming GeForce RTX 3090 both use GDDR6X memory instead of GDDR6. A simple explanation is that GDDR6X memory chips are not yet produced in the capacities would be needed for these kinds of cards.

This led @mooreisdead to comment that "GDDR6X may have been a mistake to utilize on the RTX 3080...," though was it really? I don't share that same sentiment. Sure, the GeForce RTX 3080 is using faster memory that what will presumably ship with the Quadro RTX A6000, but for the professionals workloads they will be tasked with handling, having more memory will prove beneficial.

For reference, here is how the Quadro RTX A6000 compares to the GeForce RTX 30 series...
  • Leaked Quadro RTX A6000: 10,752 CUDA cores, 1,860MHz boost clock, 48GB GDDR6, 384-bit bus
  • GeForce RTX 3090: 10,496 CUDA cores, 1,692MHz boost clock, 24GB GDDR6X, 384-bit bus
  • GeForce RTX 3080: 8,704 CUDA cores, 1,710MHz boost clock, 10GB GDDR6X, 320-bit bus
  • GeForce RTX 3070: 5,888 CUDA cores, 1,725MHz boost clock, 8GB GDDR6, 256-bit bus
Eventually we will see a Quadro RTX card with GDDR6X memory, most likely. That probably won't be until next year, but in the meantime, the Quadro RTX A6000 should be a decent performing card for professional chores.

Look for NVIDIA to announce new Quadro RTX cards during its 2020 GPU Technology Conference that runs from October 5-9.