U41.xx Observation Thread

Clay Ruth
Clay Ruth
Joined: 28 Dec 05
Posts: 17
Credit: 662723
RAC: 0

RE: Running HT on my

Message 30176 in response to message 30174

Quote:
Running HT on my Prescott 3.0 ghz, I found that S39L and S40.04 was the fastest ones. ALL other bins were slower.


I also found S39L and S40.04 to be essentially equal and faster than any others I've tried on the Prescott with HT enabled. The differences between those two client versions are less than the noise level (differences between WUs) on that processor. I'm currently running S40.04 on my HT-enabled systems except for one on which I've installed U41.04 for testing.

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

RE: Running S40.04 in HT

Message 30177 in response to message 30174

Quote:

Running S40.04 in HT mode, a wu took approx. 2.30 hours. After "disabling" HT mode S41.06 took 55 mins.

This means that in HT mode a wu took approx. 1.15 mins, now 55 mins.

Hope this clarifies my situation.

It is absolutely correct.
Both HT threads use the same FPU circuits, and both threads can load it near to 100% of their time, so both threads wait for the other. That means the processor is not able to run two threads paralell because it doesn't have more resources. It also means that the code use near to 100% of the resources.

The difference of the two time (1.15 vs 55) is very talkative.
It gives you the efficiency of HT system (in this case), so your system waste about 27% of the resources (time) to handle HT.

Erik
Erik
Joined: 14 Feb 06
Posts: 2815
Credit: 2645600
RAC: 0

How does one go about

How does one go about disabling HT on a P4 (XP)? I would like to give it a try for comparison purposes.

-edit- nevermind, I've clued myself in

-edit- If one is only running Einstein, would it be best to leave on the HT as opposed to setting the affinity to one processor per project if running, for example, Einstein and Seti?

LucaB76 - BOINC.Italy
LucaB76 - BOINC...
Joined: 16 Jan 06
Posts: 14
Credit: 754232
RAC: 0

RE: It would be really good

Message 30179 in response to message 30175

Quote:
It would be really good to clarify the speedups (or speeddowns) on Prescotts. Thanks.


Updates in Stats for my Intel P4 3.0Ghz Prescott (with SSE3) Hyper-Threaded:

S41.06 => 5273s (mean of 12 results) => 1.00
S41.07 => 4858s (mean of 5 results) => 0.921 (8% gain)
U41.04 => 4833s (mean of 5 results) => 0.916 (8% gain)

One thread on Einstein and the other one on Seti to reduce overhead and bottlenecks...

I can't see any significative difference between U41.04 and S41.07. Clearly the SSE3 instructions set are not so powerful on my Intel! My decision is to crunch with the U41.04 and I'll wait for a specific Intel version...

I'm sorry but at the moment I can't test S39L and S40.04 cause I will be away from home 'til monday. So next test are delayed!

Thanks again, Akos, ever for your clear explanations!! ;-)

sygopet
sygopet
Joined: 23 Jul 05
Posts: 2
Credit: 2080872
RAC: 0

RE: It would be really good

Message 30180 in response to message 30179

Quote:
It would be really good to clarify the speedups (or speeddowns) on Prescotts. Thanks.


I don't have the faciliites (or time) to do a full statistical comparison but, for my Pentium 4 (Prescott) 3.0 GHz machine I was taking about 150 minutes using S38, about 120 minutes with S40.12, while early tests with U41.04 are coming in at around 80 minutes.
The "official" version was taking 9 hours!
Needless to say I'm impressed.
Now if only Climate Prediction could be speeded up by a similar factor...

archae86
archae86
Joined: 6 Dec 05
Posts: 3157
Credit: 7223034931
RAC: 976834

RE: [One thread on Einstein

Message 30181 in response to message 30179

Quote:
[One thread on Einstein and the other one on Seti to reduce overhead and bottlenecks...

How are you doing that? I badly want to do that on my Gallatin host, so I can raise the SETI resource fraction without losing the huge mixed-ap benefit. I run tx36 client.

Erik
Erik
Joined: 14 Feb 06
Posts: 2815
Credit: 2645600
RAC: 0

RE: How are you doing that?

Message 30182 in response to message 30181

Quote:
How are you doing that? I badly want to do that on my Gallatin host, so I can raise the SETI resource fraction without losing the huge mixed-ap benefit. I run tx36 client.

I just figured this out today myself. Open Windows Task Manager (XP) and right click on the app exe. It will open a menu with "Set Affinity" at the bottom. Set one app to CPU 0 and the other to CPU 1.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 578226874
RAC: 199674

This works but you have to do

This works but you have to do this for every new WU - not very practical in the long run.

MrS

Scanning for our furry friends since Jan 2002

CSL.48
CSL.48
Joined: 16 Mar 05
Posts: 7
Credit: 9738
RAC: 0

Add the line to

Add the line to truxoft_prefs.xml file.

Erik
Erik
Joined: 14 Feb 06
Posts: 2815
Credit: 2645600
RAC: 0

RE: This works but you have

Message 30185 in response to message 30183

Quote:
This works but you have to do this for every new WU - not very practical in the long run.

Didn't know that. Thank you for the heads up.

Quote:
Add the line to truxoft_prefs.xml file.

Mike Hewson posted another way of changing HT through BIOS.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.