No multicore grahics from Intel, Intel withdraws Larrabee!

  • Thread starter Thread starter Jan Panteltje
  • Start date Start date
Jan said:
No multicore grahics from Intel, Intel withdraws Larrabee!
http://news.cnet.com/8301-13924_3-10409715-64.html?tag=newsEditorsPicksArea.0

Its speed was a lot slower then current available AMD and Nvidia products,
its software support not up to it...
AMD does 4.7 Tera-flops, while Intel could hardly get 1TF at peaks in Larrabee.

See also, in German:
http://www.heise.de/newsticker/meldung/Intel-macht-Rueckzieher-bei-Larrabee-878138.html
Click to expand...

Intel showed off an overclocked Larrabee with the SGEMM benchmark, where
it barely reached 1 TF. The SGEMM benchmark is actually just one of the
components of the Linpack benchmark, the single-precision
matrix-multiply component actually. They made a big deal out of it, as
if the SGEMM was a great accomplishment, as if it was as important as
the overall Linpack, but then the next day they announced they're
cancelling it.

Yousuf Khan
 
Intel showed off an overclocked Larrabee with the SGEMM benchmark, where
it barely reached 1 TF. The SGEMM benchmark is actually just one of the
components of the Linpack benchmark, the single-precision
matrix-multiply component actually. They made a big deal out of it, as
if the SGEMM was a great accomplishment, as if it was as important as
the overall Linpack, but then the next day they announced they're
cancelling it.

        Yousuf Khan
Click to expand...

I think that at that point it's not matter of performance, it's a
matter of software and programmability. For me it looks like Intel is
going to promote and optimize open standard heterogeneous multicore
programming (OpenCL). They are probably giving Larrabee to software
companies that can do it better than Intel (Codeplay for example). I
don't think it's a falure. It could be a smart move in a long run and
good for everybody.

Jenia.
 
"Evgenia Zborowsky" <[email protected]> skrev i meddelelsen
Intel showed off an overclocked Larrabee with the SGEMM benchmark,
where
it barely reached 1 TF. The SGEMM benchmark is actually just one of
the
components of the Linpack benchmark, the single-precision
matrix-multiply component actually. They made a big deal out of it, as
if the SGEMM was a great accomplishment, as if it was as important as
the overall Linpack, but then the next day they announced they're
cancelling it.

Yousuf Khan
Click to expand...

I think that at that point it's not matter of performance, it's a
matter of software and programmability. For me it looks like Intel is
going to promote and optimize open standard heterogeneous multicore
programming (OpenCL). They are probably giving Larrabee to software
companies that can do it better than Intel (Codeplay for example). I
don't think it's a falure. It could be a smart move in a long run and
good for everybody.

Jenia.
---------------------

Does anyone know if AMD/ATI are going to endorse OpenCL?

btw: Jenia it looks like your news reader isent set up correctly, or
perhaps its
mine, but it dosent insert the > in front of the quoted text you have
written! As it does with
other peoples! On second thougth it looks like you are posting from
Google so its
their mistake!

Path:
news.eternal-september.org!feeder.eternal-september.org!eternal-september.org!news.glorb.com!news2.glorb.com!postnews.google.com!j9g2000vbp.googlegroups.com!not-for-mail
From: Evgenia Zborowsky <[email protected]>
Newsgroups: comp.sys.ibm.pc.hardware.chips
Subject: Re: No multicore grahics from Intel, Intel withdraws Larrabee!
Date: Mon, 7 Dec 2009 00:28:40 -0800 (PST)
Organization: http://groups.google.com
Lines: 35
Message-ID:
<[email protected]>
References: <[email protected]> <[email protected]>
NNTP-Posting-Host: 79.179.204.185
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
X-Trace: posting.google.com 1260174520 4346 127.0.0.1 (7 Dec 2009
08:28:40 GMT)
X-Complaints-To: (e-mail address removed)
NNTP-Posting-Date: Mon, 7 Dec 2009 08:28:40 +0000 (UTC)
Complaints-To: (e-mail address removed)
Injection-Info: j9g2000vbp.googlegroups.com;
posting-host=79.179.204.185;
posting-account=tX73zQoAAADjN4HIGVYxvV3ZIWKdmoAv
User-Agent: G2/1.0
X-HTTP-UserAgent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US;
rv:1.9.1.3)
Gecko/20090824 Firefox/3.5.3 (.NET CLR 3.5.30729),gzip(gfe),gzip(gfe)
Xref: eternal-september.org comp.sys.ibm.pc.hardware.chips:511
 
Peter said:
Does anyone know if AMD/ATI are going to endorse OpenCL?
Click to expand...


Yup, AMD and Nvidia have both released final-release versions of OpenCL
drivers.

Yousuf Khan
 
Back
Top