By: Passing Through (ireland.delete@this.web.ie), August 3, 2018 11:02 am
Room: Moderated Discussions
Michael S (already5chosen.delete@this.yahoo.com) on August 25, 2015 2:58 pm wrote:
> And here you understood why AVX512 is a huge mistake. Core-based
> GPGPU killer absolutely needs AVX-1024. Or wider.
>
>
Is that the full story Michael, do you think now in retrospect (from 2018)?
Or had it anything to do with failures of a company like Intel, to support the younger AI scholars and post-grad researchers, in building some kind of a community around that technology that you mentioned (giving the designers at Intel benefit of doubt for sake of argument, in that they wanted to cheap more KNL parts cheaper to more third level institutes for research, at cheaper prices).
I mean, it's a hard thing to debate, or to review in cold, hard engineering terms. But how many hardware engineering products have failed over that inability 'to build' a sufficient community interest around it? And is that where Intel ultimately failed badly, rather than in the sense that you explain?
What Linus referred here in early July, in relation to other hardware as an ecosystem?
https://www.realworldtech.com/forum/?threadid=177724&curpostid=178263
Like, it wasn't a 100% GIVEN, that Linux community would have even gotten built around x86 as it did. Intel could just as easily have missed that opportunity as well. We then to see it as inevitable looking at it in retrospect. But really, it was a collection of random events that meant it worked out like that. Maybe it was the failure of Intel, to fully understand how 'random' these eco-system, or community building success stories are in 2014, that was really at issue. Not the one of AVX512, versus AVX-1024.
Look at all of the generations of upgrade and evolution that x86 has gone through, starting from its humble beginnings, and how faithful that industry has remained in building it's innovation around x86 hardware.
> And here you understood why AVX512 is a huge mistake. Core-based
> GPGPU killer absolutely needs AVX-1024. Or wider.
>
>
Is that the full story Michael, do you think now in retrospect (from 2018)?
Or had it anything to do with failures of a company like Intel, to support the younger AI scholars and post-grad researchers, in building some kind of a community around that technology that you mentioned (giving the designers at Intel benefit of doubt for sake of argument, in that they wanted to cheap more KNL parts cheaper to more third level institutes for research, at cheaper prices).
I mean, it's a hard thing to debate, or to review in cold, hard engineering terms. But how many hardware engineering products have failed over that inability 'to build' a sufficient community interest around it? And is that where Intel ultimately failed badly, rather than in the sense that you explain?
What Linus referred here in early July, in relation to other hardware as an ecosystem?
https://www.realworldtech.com/forum/?threadid=177724&curpostid=178263
Like, it wasn't a 100% GIVEN, that Linux community would have even gotten built around x86 as it did. Intel could just as easily have missed that opportunity as well. We then to see it as inevitable looking at it in retrospect. But really, it was a collection of random events that meant it worked out like that. Maybe it was the failure of Intel, to fully understand how 'random' these eco-system, or community building success stories are in 2014, that was really at issue. Not the one of AVX512, versus AVX-1024.
Look at all of the generations of upgrade and evolution that x86 has gone through, starting from its humble beginnings, and how faithful that industry has remained in building it's innovation around x86 hardware.