Seti benchmark on P4 with HT...updated

[Duvie]

I ran this for Dapunisher and Mechbgon....If you want the file please PM mechbgon....He will give you a setibench.zip file and you will need to download the client file...From what I read this is not the most optimised one to use currently but it gives a good comparison by holding same work unit and defaults....


The seti exe I ran is clearly no HT enabled as running 1 instance only loads the cpu to 50%....

My time w/ 1 instance using rig below...

2 hours 6 minutes


I then erased the sah files and then unzipped the original zip file and exe to another sub-directory and ran 2 instances of the program

My time w/ 2 instances using rig below....

2 hours 59 minutes / 2 = 1 hour 29.5 minutes per unit....

That represents a 40% increase using HT


Other run it and list what you get...AMD64 users welcomed...I would like to see if others get same HT increases....

I run a hardware monitor and occasionally open the state.sah file to see how far along it is...I also did not close any apps below so I have a bit of items in the taskbar but under task manager the system is idle...

Updated:

I ran 1 instance while gaming on TMPGenc for 3 hours (oH my brain!@!!!)....

2 hours 52 minutes or 35% increase over SETI alone....The sacrifice was 11% in avergae fps as per fraps program benchmark...

 

[THUGSROOK]

was that no HT thru the program, or no HT thru bios and the program?

 

[Duvie]

HT bios on but no HT was being taking advantage of since it only maxed at 50%...I will run a try with it with HT off in the bios to make sure no funny business....

 

[Jeff7181]

What client are you using Duvie? I wanna compare =)

 

[mechBgon]

If I'm fortunate enough to get an AMD64 system I'll be sure to report my results too. Duvie, so far your results seem to jive with what's normal for HT Pentium4 DCDDR systems. What would be interesting is to see:

1) Athlon 64 running two separate clients at the same time
2) Opteron (dual-channel goodness) running two separate clients at the same time
3) Dual Opterons running two separate clients at the same time

I have an extra-large paycheck coming on the 15th thanks to some extra hours I've worked, but I think I may need to spend some of that on a nice pair of Nokian studded tires for my mountain bike so I make it through the winter with all my bones intact. Been snowin' all day here...

Hmmm... Athlon 64 or intact bones... Maybe I need to post a poll!

 

[Duvie]

setiathome-3.08.i386-winnt-cmdline.exe

This is the one mechbgon told me to get at seti@home's webpage...

 

[Tom93R1]

I dont remember exactly what my non-HT seti runs take, but it was pretty close to 2.5 hours or 2 every 5 hours. With HT I get 2 done in about 3.5 hours.

I have P4 2.4c @ 3.0 GHz and a single 512MB stick of memory so dual channel or whatever it is called.

 

[DAPUNISHER]

Thanks Duvie :beer: I just PM'd Mech for the benchie and I'll post results in a couple days using a Barton@2.43ghz and A64 3000+@ whatever I can juice out of it.

 

[Jeff7181]

My Barton @ 2.2 Ghz did it in 2 hours 31 minutes. Sounded unusually high to me, so I used the 3.03 client that I've been running for about 6 months now and it did it in 2 hours 7 minutes... wonder why the client makes THAT big a difference?

Didn't test at 2.3 Ghz... I don't feel like it today, lol, I'd rather be doing other things =)

 

[Confused]

The 3.08 client should just be some kind of "security update" for a hole they found that could possibly be exploited if someone was able to get into your system and adjust the client and then it might just then not upload the results...but they haven't had any instances of it happening.

AFAIK, they didn't change ANYTHING else, and definately not anything that should cause it to take so much longer on "normal" Work Units. For High Angle Rate Work Units, it does display a slight decrease in CPU time over the 3.03 client.

Anyway...you will get better times on normal (~0.417AR) work units, like the test unit, and the majority of units coming from Berkeley are


Confused

 

[maumoon]

I'm trying to decide between a P4 3.06 and a AMD64 3000+ so have been reading a few threads.

I also do SETI so this was of interest.

I currently have an AMD Athlon 1200 machine with 768MB, I though I would test out running two processes on my machine out of interest, also because I read elsewhere that running two processes on non-HT machine would still result in lower overall processing times due to the way SETI was coded.

I didn't run the specific workunit mentioned, but on average my machine runs a workunit in 5h10 to 5h30 and SETI reports 99% efficient on average. Running two at the same time resulted in the same sort of CPU time however SETI reports only 48% efficient on average. I haven't been running anything else on the machine either.

Confirms (at least in my opinion) that any gain is strictly down to the HT technology. I think I might buy the P4 just to play with the HT technology.

 

[Duvie]

BUmp...More ppl please....