anyone want to know how seti runs on one of these?

[Adul]

http://www.rs6000.ibm.com/hardware/workstations/44p_270.html

We have the single proc model. I know they run rc5 very well, about the speed of a p3 550, but I would like to benchmark seti on it. So you seti guys, tell me how to do it.


*take note of the 4 MB of L2 cache. I believe that is on die.

 

[Wiz]

I can't say I 'know' but I'll bet it will do 5 WU/day!
(my P3@1Ghz does about 3.5 - 4 / day)

 

[Adul]

I know seti like bandwidth, and with that much L2 cache, it has to be able to run pretty fast.

 

[Sukhoi]

Uh, what OS is on that? Hopefully you can find a client on this page. BTW, if that machine doesn't have net access you can run a SETIQ server on a windows machine and have that one connect to it; the different OSes don't matter.

http://setiathome.ssl.berkeley.edu/unix.html

 

[Adul]

it runs on IBm version of unix called AIX.

 

[Poof]

That might be able to run the AS/400 client then.

 

[Sukhoi]

Just what I was thinkin'.

 

[Viztech]

Will you be able to assimilate it, even for a weekend Adul?

Everybody would love to see what it would do.

viz

 

[Assimilator1]

Adul
take note of the 4 MB of L2 cache. I believe that is on die.

F**k!:Q

I hope you do manage to run SETI on that

 

[RaySun2Be]

I'd probably try a Unix version before I tried the AS/400 version. But it would be great to see how the 4M cache helped.

I know that Netfinity server kicks butt spitting out WUs with the dual Xeon 2MB cache!

 

[Assimilator1]

How many are you getting per day? & what are the WU times?

 

[toft-dk]

<< I know that Netfinity server kicks butt spitting out WUs with the dual Xeon 2MB cache! >>




Hey Ray You don't have to rub it in. :Q

 

[Slahr Dzhe]

From what I recall (this is off the top of my head, as access to the old UBB threads is not available?) an increase in L2 cache beyond 1MB does not help speed up the client. The reason (remember, this is off the top of my head) for this is the client(instructions and whatnot) and data(whatever that might be) are approximately 524K in size, thus not fitting in the 512K L2 cache(Older Intel and AMD processors), forcing the client to fall back to the system bus. I believe due to the RISC(sorta) architecture of the CPU and the quality of the components therein along side the full speed L2 cache is what makes these type of systems fly through WU's. I believe the fastest WU on record was done by an Alpha 21264 machine(another RISC type quality machine). The Xeons(>512MB L2) rock at crunching WU's due to the client and data fitting nicely inside the full speed L2 cache.

At least that's what I remember.

Could be wrong though.

SD

 

[networkman]

You're not wrong - 1meg cache was the sweet spot, anything more was just fluff and not worth the extra invesstment(in regard to SETI) cuz you wouldn't see any better performance(speaking from experience with Xeons with 1meg).

 

[Adul]

Didn't get a chance to try it out guys, I will try next week to just see how it does.

 

[Adul]

ok, I plan to hit up the unix admin today and see how it runs.

 

[Assimilator1]

Good luck

 

[Orange Kid]

Hit up -------- hit on -------- do what you gotta do, but get that thing running some seti

 

[Sukhoi]

LMAO Orange Kid!

 

[BurntKooshie]

If you guys like those, you will drop your jaws at the POWER4.... Let's just put it this way: it is a bandwidth behemoth, with CMP chips, 4 chips to an MCM, and 4 MCM's to a typical system - that's 32 processors tightly bound in one box folks.... Too bad IBM has no plans to sell them outside it's own server range. Oh yeah, plus the fact that it'll be shipping on the higher side of 1ghz....

 

[Assimilator1]

Wow!

So you assimilated 1 BK?

 

[BurntKooshie]

Assim - they aren't even out yet I doubt I'll ever have access to them while they are still in use....(I'd first have to work at IBM or something).

 

[Adul]

I read about the power4. Pretty damn impressive if you ask me. A real server chip.