P4 Hyperthreading

Message boards : Number crunching : P4 Hyperthreading

To post messages, you must log in.


Send message
Joined: 6 Apr 99
Posts: 1
Credit: 169,223
RAC: 0
Message 1068 - Posted: 24 Jun 2004, 9:25:26 UTC

I'm getting some very poor results on a P4C 2.8GHz HT machine:

455 double precision MIPS (Whetstone) per CPU
806 integer MIPS (Dhrystone) per CPU

Even if you double it for 2 logical CPUs it's still very bad. I was thinking, maybe the Boinc client doesn't understand that the two CPUs reported by the OS aren't physical and is doing some very poor load balancing choices causing cache trashing? I don't want to test with Hyperthreading off at the moment, because I can only connect remotely with SSH to the system. Has anyone else done some real HT tests?
ID: 1068 · Report as offensive
Volunteer tester

Send message
Joined: 19 Mar 02
Posts: 1856
Credit: 40,736
RAC: 0
United States
Message 1071 - Posted: 24 Jun 2004, 9:32:40 UTC

Yes, hyperthreading is known to confuse the benchmarks.

Benchmark with hyperthreading turned off for more accurate results.
ID: 1071 · Report as offensive
Volunteer tester

Send message
Joined: 29 Mar 03
Posts: 1137
Credit: 4,150,765
RAC: 0
United States
Message 1074 - Posted: 24 Jun 2004, 9:42:41 UTC
Last modified: 24 Jun 2004, 11:23:08 UTC

Are you using the latest version of Seti@Home out Majix v3.19, your benchmark score looks more like what a older version would report.

I'm getting between 1700 & 2100 double precision MIPS (Whetstone) per CPU
on my HT PC's depending on how I test it with the v3.19, with or without HT turned on.
ID: 1074 · Report as offensive
Profile DPRGI - Marzio Nieddu

Send message
Joined: 14 Feb 00
Posts: 5
Credit: 2,166,030
RAC: 0
Message 1079 - Posted: 24 Jun 2004, 10:31:13 UTC

I agree with PoorBoy , your benchmark is very low or you have something that run with BOINC, have you check the taskmanager ?

1° Italian Team in Seti@Home Classic
ID: 1079 · Report as offensive

Message boards : Number crunching : P4 Hyperthreading

©2016 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.