Last year under S@H classic I had computers with average result times:
1857XP IP42.66 1762XP 1800XP 1819XP 1807XP
2.682 2.8569 3.0437 3.184 3.1936 3.3073

where XP is AMD and the number nnnn is actual MHz. The IP42.66 is an Intel Pentium 4 @ 2.66 MHz. Now the same computers under Boinc S@H have average times of
3.120 3.647 3.4225 3.4806 3.517 3.6351

In other words, the 2.66 P4 was second of all these computers and now it is last. The only thing that has changed is the S/W. I checked with several other 2.66 MHz P4s and my benchmarks are nearly the same (1377 FOPS and 4080 IOPS). Has anyone else noticed that their P4s are no longer as fast as their AMD XP+s? Has anyone commented on this on the message boards? Does anyone know why this is so?

---

Do you mean 2660 MHz / 2.66 GHz?

---

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

Didn't think that they made P4's with 2.66Mhz, however with regard to your P4 times they will slow down because with an HT enabled CPU it will crunch 2 WU at the same time. Therefore it will take just a little longer than normal.

---

<a href="http://www.timtaylor.net/kwsn"><img border="0" src="http://www.boinc.dk/auto.php?user=916957&amp;project=sah&amp;input=&amp;layout="></a>

> In other words, the 2.66 P4 was second of all these computers and now it is
> last. The only thing that has changed is the S/W. I checked with several
> other 2.66 MHz P4s and my benchmarks are nearly the same (1377 FOPS and 4080
> IOPS). Has anyone else noticed that their P4s are no longer as fast as their
> AMD XP+s? Has anyone commented on this on the message boards? Does anyone
> know why this is so?


My 2.8 GHz P4's are doing blocks in slightly less time than yours (about 3.4 hours; my AMD XP at 2.08 GHz does one in about 2.9), but as KWSN_Dagger pointed out, they crunch 2 at a time with hyperthreading enabled. So your effective throughput is 2 blocks per 3.647 hours, or one every 1.82 hours. So you're really doing more blocks in the same amount of time than you were with classic. You should be able to see it crunching 2 at a time when you pull up the BOINC GUI on the Work tab. I can't really explain the time increases for the AMD chips; mine actually seems to be quicker under BOINC--it averaged just over 3 hours in classic. I would be interested in other benchmarks of AMD chips (classic vs. BOINC) to see if others are noticing an increase.

---

My 2.53 GHz P4 non ht is completing wu in 3 hrs.

---

Fred
BOINC Alpha, BOINC Beta, LHC Alpha, Einstein Alpha

My 2.0 GHz P4 M is completing WU in just under 2 hours on average. It has 2 Mb L2 Cache and 1 Gb DDR RAM on a 400 MHz bus.

---

Was wondering how you could be 15th overall, and 23rd in your team?

Edit: sorry I guess that your team must be 23rd.

I complete a single WU in 1hr and 50 mins with my Athlon 64 at 2.3GHz, but that's only if i only run boinc.

---

<a href="http://www.timtaylor.net/kwsn"><img border="0" src="http://www.boinc.dk/auto.php?user=916957&amp;project=sah&amp;input=&amp;layout="></a>

> Do you mean 2660 MHz / 2.66 GHz?
>

Obviously, my stupid error. Sorry.

---

> My 2.53 GHz P4 non ht is completing wu in 3 hrs.
>

Our times are very similar. That's my point. The AMD XP CPU has one more floating point unit than the Intel P4. Boinc S@H uses a highly optimizing compiler. What I think is happening is that the new S@H client is able to use the AMD XP CPU to maximum advantage, whereas the classic S/W did not. That is the only factor I can think of. I ran several diagnostics on my 2.66 GHz CPU and apparently nothing is wrong. Yet, whereas it used to produce the same WU times as an AMD XP 2400+, now it produces about the same times as a AMD XP 2100+, slightly overclocked.

---

> Was wondering how you could be 15th overall, and 23rd in your team?
>
> Edit: sorry I guess that your team must be 23rd.
>

I believe as of 12/31/2004 2:59:07 PM I was 72nd overall. On the Web site http://setiweb.ssl.berkeley.edu/team_display.php?teamid=30188, the entries are not sorted, as far as I can tell. I have no idea what my team rank is.

How I do it from a technical standpoint is that I have 21 computers, most of which are PS/MB/HD/NIC/Video card only combinations, placed on old newspapers in a metal bookcase / utility stand in my living room with a 20 inch box fan directed at it. The furnace is turned off (Philadelphia, PA). I use the computers to heat my very small house. Otherwise, as a CS grad student, I could never afford it.

---

>
> > In other words, the 2.66 P4 was second of all these computers and now it
> is
> > last. The only thing that has changed is the S/W. I checked with
> several
> > other 2.66 MHz P4s and my benchmarks are nearly the same (1377 FOPS and
> 4080
> > IOPS). Has anyone else noticed that their P4s are no longer as fast as
> their
> > AMD XP+s? Has anyone commented on this on the message boards? Does
> anyone
> > know why this is so?
>
>
> My 2.8 GHz P4's are doing blocks in slightly less time than yours (about 3.4
> hours; my AMD XP at 2.08 GHz does one in about 2.9), but as KWSN_Dagger
> pointed out, they crunch 2 at a time with hyperthreading enabled. So your
> effective throughput is 2 blocks per 3.647 hours, or one every 1.82
> hours. So you're really doing more blocks in the same amount of time than you
> were with classic. You should be able to see it crunching 2 at a time when you
> pull up the BOINC GUI on the Work tab. I can't really explain the time
> increases for the AMD chips; mine actually seems to be quicker under BOINC--it
> averaged just over 3 hours in classic. I would be interested in other
> benchmarks of AMD chips (classic vs. BOINC) to see if others are noticing an
> increase.
>

I don't have hyperthreading.

---

my pentium M 1.7 uses between 3 and 10 hours on s@h classic workunits from where the difference is I don't know and for boinc the benchmarks are:
1490 MIPS whetstone
4233 MIPS Dhrystone

---