[Beowulf] XeonPhi notes.

[Vincent Diepeveen]

Seems intel has you in your pocket.

Initially you say it runs x86 code but no performance gets given of  
it - whereas already architects admitted it's a weak x86 structure
and is only fast if you rewrite your code completely vectorized -  
which gives it no advantage over gpu cards.

Later on you speak about a sample code produced by intel to run by  
the chip's architect which shows you 'at a secret location', 1 Tflop.

So when performance is as expected you post it, yet when it's about  
scripts and such then you cannot give performance numbers.
I assume becasue it runs *a *lot* slower such codes than at normal  
x64 cores...

Basically we learn nothing new here.

You don't buy this chip to get performance similar to normal x64  
cpu's which are a lot cheaper than Xeon Phi.

You buy it to get that 1 Tflop. As it seems the only thing so far  
that achieved 1 Tflop at it, is something intel cooked themselves.

Also we know some prime number guys already have larrabee, knights  
corner and all the *corner cpu's for years now, trying to get
some prime number code at work at it - and we didnt hear any reports  
from them either.

Good performance numbers always leaks... ...as good news is good  
marketing for intel here of course.

The NDA's are there just in case it doesn't perform as expected - so  
let's assume that until proven differently!



On Dec 14, 2012, at 3:55 PM, James Cuff wrote:

> Hi team,
>
> I've put a few notes up on our prerelease Phi engineering sample on  
> the blog:
>
> http://blog.jcuff.net/search/label/PHI
>
> Might be interesting to a few folks on the list.  Sorry I can't talk
> numbers, it's not fair to talk benchmarks on unreleased product but
> there are some fun things we discovered over the last week or so.
>
> Best,
>
> j.
>
> --
> dr. james cuff, director of research computing & chief technology
> architect harvard university | faculty of arts and sciences | division
> of science rm 210, thirty eight oxford street, cambridge. ma. 02138
> tel: +1 617 384 7647 | http://about.me/jcuff
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin  
> Computing
> To change your subscription (digest mode or unsubscribe) visit  
> http://www.beowulf.org/mailman/listinfo/beowulf