Nvidia's bet on transformers acceleration was a good one. They had to make that commitment in silicon well before transformers became the monster it is today.
But I bet they are kicking themselves over the 6xHBM VRAM limits, just like everyone else in the AI space.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
8 Comments
Back to Article
Threska - Monday, May 29, 2023 - link
Does each one come with a nanosecond bundle?https://americanhistory.si.edu/collections/search/...
Dante Verizon - Monday, May 29, 2023 - link
More ARM toys. Nice.Bulat Ziganshin - Monday, May 29, 2023 - link
correction: 1000 tflops is 1 pflop, not 1 exaflop (which is 1000 pflops)lemurbutton - Tuesday, May 30, 2023 - link
Nvidia buying ARM would have led to some very interesting products and a 3rd competitor to AMD and Intel in SoCs in the PC/HPC space.lazarpandar - Tuesday, May 30, 2023 - link
They can still compete there without owning ARM outright....hix - Monday, June 5, 2023 - link
well to compete in cpu battle u need to know how to make one ,nvidia deff cantbrucethemoose - Tuesday, May 30, 2023 - link
Nvidia's bet on transformers acceleration was a good one. They had to make that commitment in silicon well before transformers became the monster it is today.But I bet they are kicking themselves over the 6xHBM VRAM limits, just like everyone else in the AI space.
CoreLogicCom - Tuesday, May 30, 2023 - link
How long until this chip shows up in a Cyberdyne Systems Model 101?