Even as Silicon Graphics trumpeted on Tuesday a new speed record with the Columbia supercomputer it built for NASA, CNET News.com has learned, it quietly submitted another, faster result: 51.9 trillion calculations per second. SGI also plans to announce a new Linux computer Nov. 1, a machine that uses Intel’s newest Itanium 2 processor and packs the chips twice as compactly as current machines do.
SGI supercomputer: Two records in one day
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
21 Comments
2004-10-27 1:54 pm
2004-10-27 3:16 pm
2004-10-27 3:47 pm
Anonymous
I miss the time when I was working at SGI. I had fun helping a team build a 64P Onyx 2 system. I can only imagine how much fun it was setting up a 512P Altix 3000+, or 20 of ’em. Never got my hands on a Cray Metarouter or anything that cool (T3E), dunno if they’re needed with those Altix systems. But from what I’ve seen, all of SGI’s architecture it truely fascinating and fun to work with. They actually make computing fun.
But its not yet perfect, you can’t run a single thread across all 512 processors simultaneously. I want to see them doing that before I’ll buy one. 
er, well, I’ll never be able to afford one, but its a nice thought.
2004-10-27 5:09 pm
2004-10-27 6:13 pm
2004-10-27 7:00 pm
2004-10-27 11:27 pm
2004-10-28 6:07 am
But still good news for SGI
<sarcasm>I thought Itanuim was dead?</sarcasm>
Yea, it’s a super fast machine, but it also cost something around $50 million. For the money you 12,000 dual xserves which would blow the SGI numbers away.
Itanium is dead as a general software platform. It will survice as a niche platform for those doing mathematically intensive work and modelling.
> <sarcasm>I thought Itanuim was dead?</sarcasm>
The really sad part is that the number of Itanium CPU’s in this supercomputer alone will acount for half of all Itanium CPU’s shipped this quarter — not a very positive statistic. It doesn’t look like Itanic is gaining *any* traction nowadays.
12,000 dual XServes will not provide this level of performance. Even using the fastest available clustering interconnect (ie Infiniband, Myrnet, etc).
It does the job better than anything out there. Price is really more of a concern for dweebs who think white boxes or clustered dual proc systems can do everything. People who do “REAL WORK”™ are willing to pay the price.
Well if you can create an xserve based platform that offers a single image solution and the same amount of bandwidth, then maybe you had a case. This is a cluster of supers, not a cluster of PC’s… two very different things. There are problems that are not easily mapped to a multi-image system, and vice versa.
At least one company is really making a difference using Itanium. HP could learn a thing or two from these boys.
Supercomputing, just a matter of money.
Congrats to SGI for adapting their business to the 21st century. As was previously mentioned, HP as well as Sun, SCO, MSFT, and RIAA could learn a thing or two from companies like IBM, SGI, and some digital camera companies.
“Congrats to SGI for adapting their business to the 21st century. As was previously mentioned, HP as well as Sun, SCO, MSFT, and RIAA could learn a thing or two from companies like IBM, SGI, and some digital camera companies.”
Hell yes. Without rock-bottom share prices and dwindling sales, how will Microsoft ever survive?
In related news, SGI presents hardware raytracing in their workstations:
http://uk.biz.yahoo.com/041007/183/f440s.html
Itanium is dead as a general software platform. It will survice as a niche platform for those doing mathematically intensive work and modelling.
Niche like the kind of customers SGI traditionally gets? They weren’t exactly “mainstream” with Irix/MIPS.
@Stew
In related news, SGI presents hardware raytracing in their workstations: