UIUC's 640-node Xserve cluster wins on price, speed

 Technology - MacCentral via Yahoo! ^ | Fri Feb 4,12:29 PM ET | By Peter Cohen

[Swordmaker]

Another Mac Supercomputer... PING!!!

If you want on or off the Mac Ping list, Freepmail me.

[HangnJudge]

I want one
It followed me home
Can I keep it pleeeeeeze??

[Barney59]

Too bad the little Mac Mini is only 100 base T, how cool to stack up a bunch of them and string together an XGrid of Mini's...

[SunkenCiv]

Somebody from the left has hacked CYNDISLIST.COM
CYNDISLIST.COM (Genealogy) ^ | 2/4/05 | Unknown
Posted on 02/03/2005 9:44:30 PM PST by FUMETTI
http://www.freerepublic.com/focus/f-news/1335633/posts

[ProudVet77]

That's when you get into the serious supercomputer hardware. The backplane has to be wide (128bits prefer 256) and interleaved. That's what makes a Cray a Cray.

[Khurkris]

How does this set-up compare to what Cray offers - apples to apples?

[ProudVet77]

It doesn't. It's an apples to oranges kind of thing. But when I hear Apple folks say they have the fastest super computer I like to remind them. FWIW - I go back with apple to Lisa days.

[Khurkris]

OK..thanks for the follow up.

[solitas]

"Sweeet!"
                 E. Cartman

[solitas]

Yeah - it could have been a nasssty little machine for clustering if they's only given it a gigabit ethernet chipset. How much more could it have added to the price?

[Barney59]

"How much more could it have added to the price?"----

Not much and I'm sure it was in the initial designs, just they want those customers to buy XServer's and Rack'em (like VTech)...

[Barney59]

..."That's what makes a Cray a Cray." ----

Ah, yes BUT have you seen what those tricky programmers have made (actually a year ago it was released but I digress)?

http://www.apple.com/acg/xgrid/

Virginia Tech bought 1100 Xservers, clustered them using XGrid and immediately became the 3rd fastest computer in the world...


Check out their server support pages for the various options on clustering.

[Hoplite]

The Cray's CPU's crunch faster than a G5 does:

504 CPU Cray X1 = 5,895 GFlops, or 11.70 GFlops per CPU.

2200 2.3MHz G5s = 12,250 GFlops, or 5.568 GFlops per CPU

My 1.8 MHz G5 = (RAverage) .27 GFlops, if my ~ 4 Hrs per 3.87 TFlop Seti@Home work unit processing speed is any indication.

I feel so slow.

[coconutt2000]

Have you thought of using IP over Firewire?

[Barney59]

"Have you thought of using IP over Firewire?"

----

You're right, I think the new Mini has 2 FireWire 400 ports...

http://developer.apple.com/devicedrivers/firewire/ip_over_firewire.html

I already have a Gigabit network at my studio, but the firewire option would be a great cost saver...

[coconutt2000]

You might still be better off waiting for G5 minis.

[Barney59]

Mini's aren't on the drawing board for us, G5 PowerBooks, yes!

[antiRepublicrat]

Too bad the little Mac Mini is only 100 base T, how cool to stack up a bunch of them and string together an XGrid of Mini's...

There are a whole bunch of other reasons not to.

• The more processors talking to each other over high-bandwidth, low-latency connections the better, XServes have two processors per, the mini only one, so XServe is more efficient
• You can fit the same amount in a rack as XServes, but each is far less powerful than an XServe -- wasted racks and floorspace
• They don't have the same hardware monitoring with notification of conditions and failures
• Nothing in a mini is swappable, have to swap the whole thing
• The mini isn't 64-bit, which helps a lot on scientific applications
• Many more...

But somebody did come up with the idea of doing Mac mini colocation:

[terabyte]

University of Illinois AND terabyte POING!

SEVEN terabytes actually...could you handle it?