[Beowulf] Register article on Epyc (Brian Dobbins)
bill at cse.ucdavis.edu
Thu Jun 22 17:53:32 PDT 2017
On 06/22/2017 04:41 PM, mathog wrote:
> On 22-Jun-2017 15:05, Greg Lindahl wrote:
>> I don't think it hurt AMD that much in the end.
> I disagree.
It's hard to say. I agree that AMD very slowly managed to claw some small
market share from intel with the Opteron. I believe it was on the order of
10-15%. Took years for Intel to give up on Itanium and start pushing the more
competitive nehalem based servers with an on chip memory controller.
> AMD had a product that Intel could not match at the time but was unable to
> capitalize on that advantage due to non-market forces. (Ie, the game was
I don't doubt Intel was pressuring partners to stick with Intel. Financial
incentives for shipping only Intel, "marketing" budgets for vendors to use as
they see fit, etc. Then again AMD had problems with bugs, missed deadlines,
chip yields, and meeting demand. Not sure if 30% of the market had wanted
opterons that AMD could have handled it. At the time they only had their own
fabs, unlike today where they can use the split off global foundaries or any of
the other fabs.
> case having convinced Digital to axe the Alpha, SGI to drop MIPS, and HP to give
> up PA-RISC. IBM and Sun wisely avoided drinking this Kool-Aid.
IBM I believe. Seems like Sun took it hook line and sinker. My memory is
somewhat vague, but I believe they ported Solaris to Itanium and only afterwards
killed it off when they realized that Itanium was never going to be a desirable
chip for their desired market.
Not sure MIPS and PARISC were going to be competitive anyways, but losing alpha
seemed like a waste. Although I've heard discussions claiming some of the Alpha
IP ended up in one of the China Supercomputer CPUs. Dec published some very
promising papers on the performance they expected from the SMT enabled alpha...
which never shipped.
Itanium seemed "ACE" workstation all over again
More information about the Beowulf