I've installed the new version but now run only one project with cpu idle at 50%.
The Pentium D has one CPU but manage the HyperThread and with 5.7 version I calculated 2 WU at time.
Why ?

Thanks.

With BOINC 6.2 a new option on the web sites got into full swing.
Instead of the option On multiprocessors, use at most 16 processors, the new option On multiprocessors, use at most 100% of the processors is used. If your number is 50% here, it will use only 1 CPU.

Also check your Advanced Local Preferences for the same option, in case you're using those (and not the preferences of the web site).

---

Honestly, on 50%, 1 thread on the physical CPU you probably get better performance. The second core is virtual with risk of lots of disk and memory i/o and less optimal responsiveness when actually using the system.

Well... If it is really a Pentium D it has two physical cores.
The Pentium 4 had HT, the Pentium D not... ;)

---

Well... If it is really a Pentium D it has two physical cores.
The Pentium 4 had HT, the Pentium D not... ;)

Ok, I didn't know. Thanks.

the new option On multiprocessors, use at most 100% of the processors is used

Yessss. Now work at 100%.
Thanks a lot.

I've installed the new version but now run only one project with cpu idle at 50%.
The Pentium D has one CPU but manage the HyperThread and with 5.7 version I calculated 2 WU at time.
Why ?

Thanks.

Honestly, on 50%, 1 thread on the physical CPU you probably get better performance. The second core is virtual with risk of lots of disk and memory i/o and less optimal responsiveness when actually using the system.

It's odd that if you were on 16 cores default before or 2 manually adjusted, that it switched to 50% (1 core use).... little gnomes? 5.7 far as I remember had no local preferences menu, so would expect it to show on the website profile.... and which projects have adopted the new notation of % use? Rosetta this moment shows: On multiprocessors, use at most 2 processors in one of my profiles.

The experiment was done a while ago. The throughput was about 30% better in total for HT versus single CPU on the same system. Each of the two tasks would run about 65% as fast as a single task would. This leads to an increase in the total throughput to 130% of what a single CPU setup got. Yes, each of the two tasks got done more slowly, but there were two tasks working and the two got finished in less than double the amount of time that one took.

---

BOINC WIKI

I've installed the new version but now run only one project with cpu idle at 50%.
The Pentium D has one CPU but manage the HyperThread and with 5.7 version I calculated 2 WU at time.
Why ?

Thanks.

Honestly, on 50%, 1 thread on the physical CPU you probably get better performance. The second core is virtual with risk of lots of disk and memory i/o and less optimal responsiveness when actually using the system.

It's odd that if you were on 16 cores default before or 2 manually adjusted, that it switched to 50% (1 core use).... little gnomes? 5.7 far as I remember had no local preferences menu, so would expect it to show on the website profile.... and which projects have adopted the new notation of % use? Rosetta this moment shows: On multiprocessors, use at most 2 processors in one of my profiles.

The experiment was done a while ago. The throughput was about 30% better in total for HT versus single CPU on the same system. Each of the two tasks would run about 65% as fast as a single task would. This leads to an increase in the total throughput to 130% of what a single CPU setup got. Yes, each of the two tasks got done more slowly, but there were two tasks working and the two got finished in less than double the amount of time that one took.

That depends on the projects in question. I think you're remembering old readings with SETI on one HT 'core', and Einstein on the other.

There has been a recent report (2 days ago) that with CPDN and Astropulse sharing a HT processor, the Astrolpulse task runs seven times slower than when it runs on its own (no figures for what happened to the CPDN task).

I've installed the new version but now run only one project with cpu idle at 50%.
The Pentium D has one CPU but manage the HyperThread and with 5.7 version I calculated 2 WU at time.
Why ?

Thanks.

Honestly, on 50%, 1 thread on the physical CPU you probably get better performance. The second core is virtual with risk of lots of disk and memory i/o and less optimal responsiveness when actually using the system.

It's odd that if you were on 16 cores default before or 2 manually adjusted, that it switched to 50% (1 core use).... little gnomes? 5.7 far as I remember had no local preferences menu, so would expect it to show on the website profile.... and which projects have adopted the new notation of % use? Rosetta this moment shows: On multiprocessors, use at most 2 processors in one of my profiles.

The experiment was done a while ago. The throughput was about 30% better in total for HT versus single CPU on the same system. Each of the two tasks would run about 65% as fast as a single task would. This leads to an increase in the total throughput to 130% of what a single CPU setup got. Yes, each of the two tasks got done more slowly, but there were two tasks working and the two got finished in less than double the amount of time that one took.

That depends on the projects in question. I think you're remembering old readings with SETI on one HT 'core', and Einstein on the other.

There has been a recent report (2 days ago) that with CPDN and Astropulse sharing a HT processor, the Astrolpulse task runs seven times slower than when it runs on its own (no figures for what happened to the CPDN task).

Did it have sufficient RAM? If not, this would explain the massive slow down.

---

BOINC WIKI

I've installed the new version but now run only one project with cpu idle at 50%.
The Pentium D has one CPU but manage the HyperThread and with 5.7 version I calculated 2 WU at time.
Why ?

Thanks.

Honestly, on 50%, 1 thread on the physical CPU you probably get better performance. The second core is virtual with risk of lots of disk and memory i/o and less optimal responsiveness when actually using the system.

It's odd that if you were on 16 cores default before or 2 manually adjusted, that it switched to 50% (1 core use).... little gnomes? 5.7 far as I remember had no local preferences menu, so would expect it to show on the website profile.... and which projects have adopted the new notation of % use? Rosetta this moment shows: On multiprocessors, use at most 2 processors in one of my profiles.

The experiment was done a while ago. The throughput was about 30% better in total for HT versus single CPU on the same system. Each of the two tasks would run about 65% as fast as a single task would. This leads to an increase in the total throughput to 130% of what a single CPU setup got. Yes, each of the two tasks got done more slowly, but there were two tasks working and the two got finished in less than double the amount of time that one took.

That depends on the projects in question. I think you're remembering old readings with SETI on one HT 'core', and Einstein on the other.

There has been a recent report (2 days ago) that with CPDN and Astropulse sharing a HT processor, the Astrolpulse task runs seven times slower than when it runs on its own (no figures for what happened to the CPDN task).

Did it have sufficient RAM? If not, this would explain the massive slow down.

If you follow the link, you'll find that the user's only computer is a P4 @ 2.80 GHz, with 1GB of RAM and running Windows XP.

It's running the least demanding of the CPDN models, an SM3 for which the recommended memory statement is:

57MB used by the model, 256MB RAM recommended. 

Astropulse has a similar memory footprint:

Minimum CPU: 1.6 GHz 
Minimum RAM: 256 MB 
Miniumum disk space: 128 MB

and it occupies about 48 MB while running.

Doesn't sound like memory starvation to me.