I am still as of yet to find a single product that uses NPU's for any measurable ANY **** THING. GPU's even integrated... yep. But stand alone NPU cores capable of 40+ teraflops... in my experience. 100% marketing GIMMIC.
We the DIY community now get well and truly hosed by the rest of the industry. Until AI datacenter production slows... or stops or hopefully reverses... we won't see product targeted to us other than stuff that was made but doesn't pass muster for enterprise consumption. Then it will take...
Yea, and it's not like AMD is the half starved pit bull of the compute creation arena any longer. They are one of the big boys and don't have to negotiate like they used to.
I am 100% not holding my breath for any new tech that allows better I/O (PCIE lanes) or anything high bandwidth compute to trickle down to our market. ESPECIALLY since AMD has now broken through the 1 trillion dollar company barrier.
Credit where credit is due... Nvidia was 10-20x AMD's...
Wait isn't this the second or third time that halo has passed between studios? Do we even REALLY care at this point. All of the same regional programmers will end up working on whatever project is in their region again.
*blinks* Is this really news? I always thought the news was IT...