NVIDIA Launches A100 PCIe-Based Accelerator with 80 GB HBM2E Memory

[AleksandarK]

During this year's ISC 2021 event, as a part of the company's exhibition portfolio, NVIDIA has decided to launch an updated version of the A100 accelerator. A couple of months ago, in November, NVIDIA launched an 80 GB HBM2E version of the A100 accelerator, on the SXM2 proprietary form-factor. Today, we are getting the same upgraded GPU in the more standard dual-slot PCIe type of card. Featuring a GA100 GPU built on TSMC's 7 nm process, this SKU has 6192 CUDA cores present. To pair with the beefy amount of computing, the GPU needs appropriate memory. This time, there is as much as 80 GB of HBM2E memory. The memory achieves a bandwidth of 2039 GB/s, with memory dies running at an effective speed of 3186 Gbps. An important note is that the TDP of the GPU has been lowered to 250 Watts, compared to the 400 Watt SXM2 solution.

To pair with the new upgrade, NVIDIA made another announcement today and that is an enterprise version of Microsoft's DirectStorage, called NVIDIA GPUDirect Storage. It represents a way of allowing applications to access the massive memory pool built on the GPU, with 80 GB of super-fast HBM2E memory.


View at TechPowerUp Main Site

 

[Chomiq]

80 GB? Jesus. That's with chip shortage?

 

[Raiju]

Nvidia cut down 1/4 of the GA100 cuda cores, the yield at TSMC must be really bad with these large chips.
Also it's now made by TSMC again and not Samsung like on the RTX 3xxx. I wonder if how much more efficient is TSMC 7nm.

 

[outpt]

these are not for us mortals. part of a data center for 200,000 dollars. oxygen!

 

[Space Lynx]

I really hope TSMC gives this low priority... first the car industry chips need fixed, than enough gpu/cpu for consumers... then phones, then this...

 

[Mussels]

80 GB? Jesus. That's with chip shortage?

this was announced last nov, the update is same GPU in a new form factor, and that its capable using a fancy version of directstorage to make servers go pew pew

 

[Vya Domus]

It's insane that these are still not fully enabled by now, are the yields that bad ?

 

[StefanM]

What does "2X seismic" mean?
I see no reference in https://docs.nvidia.com/gpudirect-storage/overview-guide/index.html

 

[TheinsanegamerN]

I really hope TSMC gives this low priority... first the car industry chips need fixed, than enough gpu/cpu for consumers... then phones, then this...

In a chip shortage $200,000 server GPUs take priority over $200 bing bing wahoo toys.

 

[jabbadap]

80 GB? Jesus. That's with chip shortage?

GA100 uses HBM2 memory, which is currently not used in any high volume product. There's probably more shortages with regular components of the card itself than those memories.