We've been waiting for Nvidia's Ampere GPUs to make their way into desktop PC GeForceware, and now the Green Team has released a version of the top graphics silicon GA100 that actually works in a gaming PC. Sort of. I mean, it won't run "Red Dead Redemption 2" at 1000 frames per second, but it sure can compute like a gun.

The new Nvidia Ampere A100 PCIe (via ComputerBase) is a sleekly designed silicon slice that plugs into a PCIe motherboard slot. Attach it to an AMD X570 and you have a powerful compute machine. It has no ports, so a separate video card is required for actual output.

It also runs at a much lower TDP than the previously released Nvidia A100 for HGX, which uses the SXM board form factor. The TDP of the first Ampere card was 400W, but this PCIe version is much more reasonable at 250W.

It has the same GA100 Ampere GPU, with the same HBM2 memory, tremendous bandwidth, and all sorts of floating point and integer arithmetic smarts. In fact, despite the 150W drop in maximum TDP, Nvidia reports only a 10% drop in "top application" performance.

For Ampere, it is probably almost a desktop Max-Q design.

A very tempting story, but this is PC Gamer, not Comput Nerd Monthly. According to the latest rumors, the GeForce Ampere top card will record 350W of total board power (TBP), which initially seemed ludicrously high for a 7nm part.

This is slightly different from the thermal design point (TDP), which should be used to refer to the GPU itself and how much power the attached cooler needs to dissipate. TBP, on the other hand, is categorized as how much power the graphics card consumes by itself. This is also known as total graphics power (TGP), but these terms are virtually interchangeable.

Unfortunately, TDP and TBP/TGP are often incorrectly confused. For example, the TDP of the Nvidia RTX 2080 is regularly listed as 215W, which is actually what Nvidia rates as TDP; the actual TDP of the RTX 2080 is much lower at 148W, and the difference between these two numbers is made up by the power demands of memory and power circuits

Turing.

This difference for the Turing card is about 45%, with the TBP being 45% higher than the TDP. If this follows Ampere, then the rumored 350W and 320W TBP numbers for the first three cards appear to be about right in practice.

If the TDP of a GA100 GPU in PCIe spec is 250W, it is not unreasonable to assume that the GA102 version with GDDR6/X instead of HBM2, whether in Titan or RTX 3090 color, will have a TDP of around 350W. the highest GA102 TDP will be around 240W, and the RTX 3080's TDP could be around 220W.

And if so, Nvidia may go back to just using TDPs to talk about new graphics cards in marketing materials.