Nvidia GPUs fly out of the fabs – and right back into them

Like salmon returning to their ancestral waters, Nvidia's GPUs make their way back into chip factories. TSMC, ASML, and Synopsys are all using Nvidia's accelerators to speed up or boost computational lithography. Meanwhile, KLA Group, Applied Materials, and Hitachi are now using deep-learning code running on Nv's parallel- …

[DS999]

Fixed function

Just like GPUs are an order of magnitude better than CPUs for stuff like computational lithography or video encoding, fixed function blocks are an order of magnitude better for video encoding (that's why iPhone SoCs, Qualcomm Snapdragon etc. have fixed function blocks for this instead of relying on the CPU or GPU)

I would guess they'd get the same order of magnitude improvement for computational lithography. Even if the needs change significantly enough between process generations to where you need new ones every 2-3 years that doesn't seem like an obstacle. TSMC runs thousands of wafers that are scrapped during risk production before a node is ready for customers, if they were running wafers with fixed function computational lithography blocks at least they'd get something useful out of these test wafers instead of running a bunch of ARM A72 cores that are never used for anything beyond seeing if they meet operational parameters or not.

[Ashto5]

Machine making machines

Well that sounds perfect, I mean nothing could go wrong could it ?

Hmm let us see “I’ll be back”

[StargateSg7]

Re: Machine making machines

"This has happened before and will happen again ...."

V