P4 533 fsb 3.06 ghz HT. 4 hours for 2 Wu's

Number of CPUs: 2
1356 double precision MIPS (Whetstone) per CPU
1965 integer MIPS (Dhrystone) per CPU

---

Regards

Athlon XP 2700 aprox. 2h 55m per unit

Benchmark results:
Number of CPUs: 1
1995 double precision MIPS (Whetstone) per CPU
4797 integer MIPS (Dhrystone) per CPU

---

http://www.boincstats.com/signature/user_115848.gif

If the interest continues like this, i will turn these results into a webpage, listing CPU, wu time, overclock and so on ?
Call it a knowledge base of some kind ?

Hell i still remember my first SETI workunit in SETI-classic...
AMD k6-2 350 Mhz :-)

Fascinating what modern technology has achieved !


Keep crunching


Bukken

---

> Where can I obtain an SSE optimized verion of SETI? Also, the slowest
> Prescott is a 2.8 GHz. Is your 2.66 actually a Northwood?

Steve,

If you want to compile it yourself with no support or advice (beyond the docs already there ;)

Its on sourceforge.

I'm curious, why are the benchmarks so different for the XP2700 and the P4-3.06 that are listed a couple of posts back? Isn't the benchmark used to calculate the "claimed credit"?

Tom

---

> I'm curious, why are the benchmarks so different for the XP2700 and the
> P4-3.06 that are listed a couple of posts back? Isn't the benchmark used to
> calculate the "claimed credit"?
>
> Tom
>
Because of the HT of the P4

---

Athlon 64 3000+ approx 2h40min

---

AMD XP 1600+ 4Hrs 27 min.

I should explain better.

For example, I have a dual cpu AMD system and a single AMD system with nearly identical speed chips. The single cpu system gets a higher benchmark than the dual system. As a matter of fact the dual system's benchmark is considerably lower than the single system. I would expect the benchmarks to be much closer to each other.

Single system: XP2600 @2075Mhz, 1935 (Whetsone) 4648 (Dhrystone)
Typical workunit time 2:57 per workunit.

Dual system: MP2600, @2133Mhz, 1993 (Whestone) 1455 (Dhrystone)
Typical work unit time 3:47 per workunit per processor.

I know from seti classic that dual systems don't get the same results as 2 single systems of the same configuration. However I don't understand why the benchmark should change. For the benchmark to be lower seems like a doubt dip given that the dual system will ultimately be slower and therefore spend more time on any given workunit. Doh, I think I just answered my own question?

Tom :)

---

Amd XP barton 2600+ @2300MHz, typical workunit time 2:30
P4 2,4 GHz , Notebook, typical workunit time 4:22

---

AMD Athlon 64 XP-3500+ running all 32 bit software:
2142 Dhrystone
5928 Whetstone
2 hours 16 minutes per WU on the average.
64 bit software not too available yet. I'll try again later maybe...

Actually, has anyone managed to compile for Mac with the -mcpu=7450 -fast flags in gcc 3.3?

---

1st post on any forums...I'm running several machines; here are a few benchmarks:

Pentium 4 1.8 GHz: 4:43
Mobile Athlon 2500+ (laptop): 3:42
Athlon XP 2400+ (at 2.08 GHz, 183 FSB): 2:51
Pentium 4 2.8 GHz (no HT): 2:26
Pentium 4 2.8 GHz (w/HT): 1:43

I was amazed at the improvement with HT turned on...that really is an Athlon-crusher in this particular benchmark!

---

> Pentium 4 2.8 GHz (no HT): 2:26
> Pentium 4 2.8 GHz (w/HT): 1:43
I don't believe this time, only when it means 2 WU crunching in the same time which is 3:26 and it means 1:43 per one WU.
I have P4 2.6 GHz w/HT and it does right now 1 WU in 3:36 bat it does 2 paralel which means 1:48 per WU. If I'm right. But yesterday it did 1 WU for 4:20, depends very much from WU, they are quite different some time.
My AMD64 3200+ box (754socket, slow CL2,5 RAM) does 1 WU in 2:15.

---

ALL GLORY TO THE HYPNOTOAD!
Potrebujete pomoc?
My Stats

> 1st post on any forums...I'm running several machines; here are a few
> benchmarks:
>
> Pentium 4 1.8 GHz: 4:43
> Mobile Athlon 2500+ (laptop): 3:42
> Athlon XP 2400+ (at 2.08 GHz, 183 FSB): 2:51
> Pentium 4 2.8 GHz (no HT): 2:26
> Pentium 4 2.8 GHz (w/HT): 1:43
>
> I was amazed at the improvement with HT turned on...that really is an
> Athlon-crusher in this particular benchmark!
>
>

Which host? Best I seen of your 2.8's was 2:xxish...

---

<img src="http://boinc.mundayweb.com/seti2/stats.php?userID=2259&amp;team=off">

> > 1st post on any forums...I'm running several machines; here are a few
> > benchmarks:
> >
> > Pentium 4 1.8 GHz: 4:43
> > Mobile Athlon 2500+ (laptop): 3:42
> > Athlon XP 2400+ (at 2.08 GHz, 183 FSB): 2:51
> > Pentium 4 2.8 GHz (no HT): 2:26
> > Pentium 4 2.8 GHz (w/HT): 1:43
> >
> > I was amazed at the improvement with HT turned on...that really is an
> > Athlon-crusher in this particular benchmark!
> >
> >
>
> Which host? Best I seen of your 2.8's was 2:xxish...
>
I should clarify--the 1:43 time is for one block; the 2.8's do 2 blocks in parallel in 3:26 (which gives 1:43 per block). Here's a link to one of the hosts--if you link to the WU's, you'll see the time go up when I switched it to HT (but it is doing 2 at a time). Sorry for the confusion!

host 406081

---

> I should clarify--the 1:43 time is for one block; the 2.8's do 2 blocks in
> parallel in 3:26 (which gives 1:43 per block). Here's a link to one of the
> hosts--if you link to the WU's, you'll see the time go up when I switched it
> to HT (but it is doing 2 at a time). Sorry for the confusion!

The collision of wall clock time, processing time and throughput. So the time per WU is higher, but the throughput is greater.

One of the reasons for the wide variation can also be caused by internal contention. Though it is two "logical" processors, they can compete for the same CPU component and thereby decrease the throughput.

This is one of the other reasons why I like to have more than two projects running so that the chance of them doing the same things all of the time is lessened. And it looks like Predictor@Home is trying to come back before the end of the year (YEA!). January should be another good month with (we hope) both Predictor@Home and LHC@Home back on-line.

---