PRESCOTT - Whats the effect of P4E on DC results?

[MadRat]

I figure that Intel's P4E (Prescott) design has several crucial changes that should effect DC projects. The "Improved Shift" is 50% faster than comapred to Northwood. The trace and L2 caches are double what they were over Prescott. The branch predictor is .1% better than Northwood, which makes a difference when we are talking literally BILLIONS of computations. The imul Latency is 200% faster. I'm sure there are more but I'm going by the numbers from here.

Does all this make a difference for any DC projects?

 

[Robor]

I'm curious to see some benchmarks of the Prescott running SETI. Especially an overclocked Prescott - something in the upper 3Ghz range would be nice.

 

[Accord99]

http://www.computerbase.de/artikel/hardware/prozessoren/intel_pentium_4_prescott/18

Noticeably faster than the Northwood, slightly slower than equivalent clocked P4EE.

 

[ViRGE]

Originally posted by: Accord99
http://www.computerbase.de/artikel/hardware/prozessoren/intel_pentium_4_prescott/18

Noticeably faster than the Northwood, slightly slower than equivalent clocked P4EE.

It's faster? That doesn't make much sense; unless SETI really values L2 cache so much that it disreguards latency and pipelines, it should be as fast, if not slower.

 

[Baldy18]

Any increase in L2 cache up to 1MB dramatically affects the speed of SETI.

 

[Engineer]

kokomo has reported that his 2.8GHz Prescott clocked to 3.1GHz chruns out 22 to 24 WU's per day. My 3.2GHz P4 does 17. That's a huge difference....hell, that's huge from one processor period! :Q

 

[Robor]

Are those times running 2 instances at once? If not I'm unimpressed. My P4-2.4C @ 2192 does WU's in the 2:46 hour range running 2 at a time and that includes using the box for web browsing, Email, usenet, casual gaming, etc. I know that's not using the benchmark WU but it is after 500 WU's as of now.

 

[Accord99]

Yep, its two instances.

 

[Robor]

Originally posted by: Accord99
Yep, its two instances.

A 3.1 Prescott can do 22-24 WU's per day? Okay, now I'm impressed!

 

[Overkiller]

It does decently WORSE than P4C in DPAD @ similar clocks

 

[Kokomo]

Originally posted by: Overkiller
It does decently WORSE than P4C in DPAD @ similar clocks

Uhhh... How you know that? Do you have one, were you taking my word - or someone else's for it - or were you merely "making posts?" Mine is 2nd from the bottom in my "NewRigs" sig, and I've had it for quite some time. Got it from the Philippines, only a few weeks after Anand and his people got one. I paid $300+ shipping.

When I made my original post regarding the then "P4P," some "sooper jeenyus" here remarked, 'He doesn't have one." I doubt that was you, but what's your point? Buy one, run it, THEN tell us about it...

For anyone truly interested, the 1MB L2 cache makes it a real screamer at SETI, and that's about it for DC as far as I know. I understand it also screams at SoB, but I haven't tried it there. SSE3 has not yet arrived, and it's unlikely that our "Brit Friends" will soon code their stuff to accomodate it. Hell, they can't even accomodate SSE2, so what's the point?

Flame me all you want, but if you're nothing but a "parrot," SD,STFU.

 

[Overkiller]

unless i'm completely mistaken you yourself said that it screams at seti but is slower than a comparably clocked northwood

 

[Kokomo]

Originally posted by: Overkiller
unless i'm completely mistaken you yourself said that it screams at seti but is slower than a comparably clocked northwood

Twist it anyway you like, do you own one, or are you merely a parrot? If you own a P4E, please share your results. If not --- Post on...

 

[Engineer]

Originally posted by: Kokomo

Originally posted by: Overkiller
unless i'm completely mistaken you yourself said that it screams at seti but is slower than a comparably clocked northwood

Twist it anyway you like, do you own one, or are you merely a parrot? If you own a P4E, please share your results. If not --- Post on...

What exactly are the WU averages in SETI?

And I don't think Overkiller posted anything wrong....other than to state what had been posted before. Nothing wrong with that.

 

[Kokomo]

Originally posted by: Engineer

Originally posted by: Kokomo

Originally posted by: Overkiller
unless i'm completely mistaken you yourself said that it screams at seti but is slower than a comparably clocked northwood

Twist it anyway you like, do you own one, or are you merely a parrot? If you own a P4E, please share your results. If not --- Post on...

What exactly are the WU averages in SETI?

And I don't think Overkiller posted anything wrong....other than to state what had been posted before. Nothing wrong with that.

And I have no problem with that, save for the "twist," but who cares anymore, as long as one can improve one's stats, and climb the "posting ladder."

I have no idea what the "averages" are in SETI. I have a fair idea what various machines can do, but no idea whatsoever what anyone with "control" over many machines --- with or without permission --- can do. Do you?

I think that's quite enough for tonight, gents. For my part in this, I apologize. Tomorrow, however, I will read some reviews, post remarks, and gain "fame." That seems to be the way of it...

How's your P4E doing, BTW? Any cooling problems?

 

[Engineer]

http://www.ocforums.com/showthread.php?s=&threadid=270884

http://episteme.arstechnica.com/eve/ubb.x?a=tpc&s=50009562&f=122097561&m=809001074031

 

[Kokomo]

Mad, sorry 'bout all this. To answer your question, my P4E 2.8, no overclock, runs with my Dual Xeon 2.4 on SETI. Call it 20 WUs+ a day. It sux, however, on DPAD, that's all I know from experience, and I don't SETI anymore. I'm at 5,500+ feet above sea level, air cooling is tough up here, and while the processor will easily overclock to 3.4 or so, I can't keep it cool/stable with air. I've made arrangements to swap it with a friend who lives closer to sea level, and if you're interested I'll have him keep you posted.

Again, my apologies.

 

[Ronin]

Ok, this is gonna be a bit messy, but closely accurate. This is a P4HT 2.8E Prescott running 2 processes of Seti at stock speed. There are some days that the box wasn't up all day, but you can get a good idea of where it sits WU wise. On second thought, on looking on how it would look if I posted the log, I've uploaded it instead for you to look at.

SetiSpy log

The average time is 2.966 (just a hair under 3 hours) for each unit. That says the box should do around 16. I'm somewhat surprised, even overclocked, someone's stating they're getting 24WU/day. That seems a bit high (my dual xeon box doesn't do that many, which is a 2.8GHz HT running 4 processes). I'm gonna pull the log off that box and get some averages for it as well.

 

[Ronin]

Ok, I've taken a look at the logs for the dual xeon box, and this is what I've got:

Avg WU time is 4.71 (4 hours, 42 minutes)
Most WU processed in a day is 22.

A Dual Xeon HT 2.8GHz running 4 processes does 22 WU, and an OCed P4 2.8E running 2 processes does 24? I'm not gettin something here.

Dual Xeon SetiSpy log

 

[Kokomo]

This is the only log I have left for the P4E from NEP (Never Ending Project), and I'm not about to run it again merely for the sake of argument. It's done a little better, depending on WU AR, mainboard, and memory...

The Xeon is on a Supermicro MBD-X5DAL-6-0 mainboard with RAMBUS memory, and has ALWAYS run slightly better than 20 WU a day, even with a string of low ARs WUs mixed in.


******* SETI Cache Spy v2.44 WU Log Calculations *******

Cache Name:

Log File: C:\Documents and Settings\Farrell\Desktop\SETI@Home\SETILog.csv
Log Section:
Log Contents: 15 WU's counted; 0 WU's dropped

AR # Min Max Avg
High (>0.8) 8 02:01 02:52 02:17
Med (0.4-0.8) 7 02:42 02:57 02:49
Low (<0.4) 0 00:00 00:00 00:00

Real Average WU Time: 02:32 (9.47 WU's per day)
SCS Weighted Average WU Time: N/A
Overall Average AR: 2.132

(MULTIPLY X2 TO ALLOW FOR HYPERTHREADING)

 

[Kokomo]

Found log for the Xeon from the last time I ran NEP. Only 7 WUs, but there it is. Multiple x4 to allow for hyperthreading.

******* SETI Cache Spy v2.44 WU Log Calculations *******
Cache Name:
Log File: C:\Program Files\SETI@Home\SETILog.csv
Log Section:
Log Contents: 7 WU's counted; 1 WU dropped

AR # Min Max Avg
High (>0.8) 2 04:01 04:26 04:13
Med (0.4-0.8) 5 04:40 05:09 04:51
Low (<0.4) 0 00:00 00:00 00:00

Real Average WU Time: 04:40 (5.14 WU's per day)
SCS Weighted Average WU Time: N/A
Overall Average AR: 1.003

 

[Overkiller]

in this thread:

http://forums.anandtech.com/messageview.cfm?catid=39&threadid=1240285&FTVAR_MSGDBTABLE=

you clearly say that the 2.8C "runs Rings" around a 2.8E in Dpad while it "screams" in Seti.

When something "runs rings" around something else i assume the difference to be quite tangible.

Am i wrong?

 

[Kokomo]

Originally posted by: Overkiller
in this thread: http://forums.anandtech.com/messageview.cfm?catid=39&threadid=1240285&FTVAR_MSGDBTABLE=

you clearly say that the 2.8C "runs Rings" around a 2.8E in Dpad while it "screams" in Seti. When something "runs rings" around something else i assume the difference to be quite tangible. Am i wrong?

Well, OverK, considering what I paid for the P4E (about twice what they sell for now), and my high expectations for it... Belay that. No, you are not wrong. Poor choice of words on my part. Just really bummed out over the whole deal, and I apologize for being such an *ss.

 

[Overkiller]

well that's life, at least it screams at seti right

have a :beer: on me

 

[MadRat]

So just SETI and super pi seem to be better on P4E?