Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I mean, trying and failing isn't the same as doing nothing. But regardless, necessity is the mother of invention, so I'm sure they'll figure something out. They have 20b of cash (1/4 of AMD's total market cap) and their profits are damn good.

Make no mistake, Intel is very much aware this is a do or die situation, and I expect them to deploy all the resources at their disposal to fight both AMD and ARM. They don't have much debt and can borrow heavily and at great rates if necessary.



Why is it do or die just because they can't make faster CPUs? AMD can't make them either. It's the end of an entire era technologically. We all knew this day would come.It's a hard problem that requires a whole new paradigm probably. Hopefully both AMD and Intel will have enough money to keep doing research by trying and failing with various radical approaches or we may never see faster CPUs again. Why would they need to worry about ARM? That is low power mobile devices, a sector they got out of a while ago.


> Why would they need to worry about ARM?

Because ARM is coming (has come) to the server (eg Ampere, Graviton). x86 was a moat around a server product with high margins. Now they have lost the process lead and there is the possibility of using ARM on the server that business is very vulnerable.


That's not true. Top of the Top500: https://www.top500.org/system/179807/


Apple is moving to ARM, and ARM is very much beginning to make inroads in the data center space. ARM is a big threat to x86 (so to both Intel and AMD).


I believe the current super computer leader uses ARM.


And 2. and 3. use the IBM Power architecture. Only 5. place has Intel.


With crazy wide vector SIMD, which gives you nothing but Linpack Flops.


The Fujitsu ARM chip has an even MORE impressive HPCG score than Linpack actually.

I expect the Fujitsu ARM to be the greatest HPCG computer for years to come. No other design comes close.


Is HPC moving away from purely ALU throughput benchmarks? If so, I think that would be good news.


HPCG has been reported for years.

Linpack is the "legacy" benchmark on pure FLOPs throughput. But HPCG is the more realistic benchmark that better tests the interconnects.

HPCG is a sparse-matrix, which more accurately represents many HPC problems. Linpack is a bit of a legacy, but still represents dense-compute pretty well.

Its still a GFLOPs / TFLops style measurement, but HPCG has a LOT more data movement.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: