What about Itaniums?

[Shelnutt2]

I was just browsing fleebay and I found a few 900MHz Itaniums for $35 each. I'm still looking for a motherboard but my gosh thats pretty darn cheap!

Now only if I can find a motherboard for the cheap. But could I get folding to run on it? Would I be able to get the SMP client up and chugging on a IA64 linux distro? Acording to wikipedia, which I know isn't the best source "Itanium running at 800 MHz can produce four floating-point results a cycle for a peak 64-bit performance rating of 3.2 billion floating-point operations per second (GFLOPS)". The best info I can find on a C2D is that the X6800 offers ~2.2 GFLOPS.

Should I go for this? It seems like a great deal, folding wise It would crank out the SMP folding. If I ran in a quad setup..might have to go dual depending on what motherboard I can find on the cheap.

[dropadrop]

Don't think it would work. Though it's 64bit, it's a complete different command set.

[David]

Weren't these referred to as the "Itanics"? It would have to emulate the x86 instruction set surely?

[Shelnutt2]

Weren't these referred to as the "Itanics"? It would have to emulate the x86 instruction set surely?

If I am remembering correctly the Itaniums can/do emulate the x86 instruction set. I need to go read some white docs.

[Hookj5]

Have you seen these things? They are Huge! Or the heatsink is huge, or something.
55538

[Shelnutt2]

Have you seen these things? They are Huge! Or the heatsink is huge, or something.
55538

They are "large" they look like old slot 1 chips!


http://news.com.com/Intel+scraps+once-crucial+Itanium+feature/2100-1006_3-6028817.html
It looks like Intel has dropped hardware support for X86 and are just using software emulation now...but the chips I'm looking at are old enough to have the hardware emulation.

[David]

If I am remembering correctly the Itaniums can/do emulate the x86 instruction set. I need to go read some white docs.

The question is I guess how fast are they having emulated x86. As fast as a comparibly priced XP or P4 rig?

[Shelnutt2]

The question is I guess how fast are they having emulated x86. As fast as a comparibly priced XP or P4 rig?

From what I'm reading the hardware emulation was not that great. a 1.5GHz Itanium 2 would be about the speed of a 1.5GHz Northwood (/Prestonia Xeon).

I'm wondering about the software emulation though, it seems to be better. I'm reading some more on it now.

Also I posted over at the FAH forums about it. I'm wondering if stanford would rework the SMP client to work on the IA64 architecture. As a novice programmer I offered my services to rework the client for this architecture.

[David]

From what I'm reading the hardware emulation was not that great. a 1.5GHz Itanium 2 would be about the speed of a 1.5GHz Northwood (/Prestonia Xeon).

I'm wondering about the software emulation though, it seems to be better. I'm reading some more on it now.

Also I posted over at the FAH forums about it. I'm wondering if stanford would rework the SMP client to work on the IA64 architecture. As a novice programmer I offered my services to rework the client for this architecture.

I hate to be a killjoy but I seriously doubt it. I suspect there are very few Itanium rigs around that people would run Folding on, and they won't likely let anyone see the source code for obvious reasons (risk of introducing/finding exploits) etc.

It would be too much work for too little benefit.

[Shelnutt2]

I hate to be a killjoy but I seriously doubt it. I suspect there are very few Itanium rigs around that people would run Folding on, and they won't likely let anyone see the source code for obvious reasons (risk of introducing/finding exploits) etc.

It would be too much work for too little benefit.

Well Bruce over at the FAH forums has come to the conclusion that there will never be an IA64 client and that I would be better spending my money on a Conroe and a 1950XTX...Although I was checking and gromacs already offers a IA64 based version of the gromacs core...so to me it seems like the hardest part of the work has already been done for stanford.

It would be a huge security risk for them to let me see the source code..I mean besides my lack of experience in C, along with not being in any way associated with Stanford...but hey I offered my help and no one can say that I wasn't willing to do the work myself.

[WarriorII]

....no one can say that I wasn't willing to do the work myself.



NEVER GIVE UP ! NEVER SURENDER !

Fight the fight worth fighting !!!

:beer:

[David]

Well Bruce over at the FAH forums has come to the conclusion that there will never be an IA64 client and that I would be better spending my money on a Conroe and a 1950XTX...Although I was checking and gromacs already offers a IA64 based version of the gromacs core...so to me it seems like the hardest part of the work has already been done for stanford.

FaH uses a modified gromacs core, and the client would need to be redone in IA64. That's the minimum work needing done. You also have the overhead in terms of maintaining, bugfixing etc. It wont work out as cost effective for Stanford to maintain.

[Sleepy_Steve]

Cant they just throw TA's at the problem till it goes away??