Is it possible that this is a Rosetta wu problem?

One wu seems to have errored out, and I suspended the other "NMR_"

I now am running two "Bench_Ab_Relax_Save_All_Out", and they are in synch!

Using AMD's utility "DAshboard", shows two cores @ 100% each.

Possibly stupid question, but you have made sure that your BOINC preferences are specified to allow more than one processor to be used on multiprocessor systems?

By the sounds of it the operating system sees and is using both cores, but BOINC and Rosetta are only using one core.

---

Good question to ask. BOINC preference is set up to use up to 2 processors, up to 100%.

Now that I have two of the same "type" of wu's (as per naming convention), Windows Task Manager reports each Rosetta around 50/50. Fluctuates 45/55, but near 50/50 with a 1 sec to 1 sec synch.

So, I'm "assuming" that my setup was correct the entire time, and the question then becomes why would two different types of wu's (as per naming conventions) cause BOINC manager to report the disparity I observed?

Possibly stupid question, but you have made sure that your BOINC preferences are specified to allow more than one processor to be used on multiprocessor systems?

By the sounds of it the operating system sees and is using both cores, but BOINC and Rosetta are only using one core.

You may be on to something. Using the Clock in "Date & Time Properties", it took about 1 min 40+ secs for the Boinc cpu time to advance 1 min on the wu's.

By the sounds of it the operating system sees and is using both cores, but BOINC and Rosetta are only using one core.

With my dual core Athlon 64 x2 3800+ at work, the highest a task is reported in Task Manager is 50%. If you're seeing tasks eating up 55% or more of the cpu in task manager, it's running as a single core system. If the system was setup correctly for multiple cores - and Boinc was setup to use just 1 core for 2 Rosetta apps then Task Manager would max out at 50% between the two apps. (25/25, 20/30, etc)

Did you run through the page I linked and verify that you've got the right HAL selected? Right now, it sure seems like your system is saying, "I'm sorry Dave, but I'm afraid I can't do that." when you ask it to use one core per Rosetta task. ;)

---

If by HAL you mean "ACPI Multiprocessor PC", then "yes", this is what appears inside Device Manager / (expand) Computer.

With my dual core Athlon 64 x2 3800+ at work, the highest a task is reported in Task Manager is 50%. If you're seeing tasks eating up 55% or more of the cpu in task manager, it's running as a single core system. If the system was setup correctly for multiple cores - and Boinc was setup to use just 1 core for 2 Rosetta apps then Task Manager would max out at 50% between the two apps. (25/25, 20/30, etc)

Did you run through the page I linked and verify that you've got the right HAL selected? Right now, it sure seems like your system is saying, "I'm sorry Dave, but I'm afraid I can't do that." when you ask it to use one core per Rosetta task. ;)

My Rosetta preference is currently set for 6 hour wu's.

The two same-name-type wu's that I started 7 hours ago are each about 75% done (about 4.5 hours cpu time each).

Now, I'm not that greatest at math, but at that rate it seems that it will take about 9 hours to run two 6 hour wu's with the dual core.

So, that seems like a 33% gain.

Was I expecting too much with a 100% gain (also keeping in mind the switch from Sempron 3400+ to Athlon X2 3800+) ?

But then again, the wu is not over, and it may take the full 12 hours.

I'll keep an eye on this. Results should be telling.

My Rosetta preference is currently set for 6 hour wu's.

The two same-name-type wu's that I started 7 hours ago are each about 75% done (about 4.5 hours cpu time each).

Now, I'm not that greatest at math, but at that rate it seems that it will take about 9 hours to run two 6 hour wu's with the dual core.

So, that seems like a 33% gain.

Was I expecting too much with a 100% gain (also keeping in mind the switch from Sempron 3400+ to Athlon X2 3800+) ?

But then again, the wu is not over, and it may take the full 12 hours.

I'll keep an eye on this. Results should be telling.

Hi,

have a look at this host. It has the same specs than your computer and uses the standard client and has similar benches (not overclocked). It gets about 40 credits/3 hours ~ 13/hour/core ~ 624 credits/day, if its crunching 24 hours. You should achieve similar numbers if your host is working fine. From what you've reported it seems both cores don't get utilized fully or there is some other problem preventing your comp from achieving its normal performance. There is a BIOS setting, enable ACPI or something, can you check that? You also can change your general preference so that it uses only one core. Then you should still get 13credits/hour but only with one working core it should be something like 320 credits/day. This would allow you to check whether running on one core gives normal performance.

Hey, thanx for info and benchmarks!

I went through all BIOS settings, and there are NONE that deal with single/dual processors/cores. The only things BIOS confirms for me is that the cpu is an AMD Athlon X2 3800+.

Device Manager confirms "ACPI Multiprocessor PC".

Here's what I don't understand: Hardware / System-software "seems" to recognize the dual core. Boinc settings are for 2 cpus, 100%.

If the issue was that Rosetta was really running two instances on one core, then why would the second core also show 100% utilization?

Wouldn't it show 1 core at 100% (i.e., the two instances of Rosetta), and the other core at some nominal low value (remembering that TaskManager/Processes shows only Rosetta with a value greater than 1%).

I agree that the results I am seeing appear to indicate that only one core is being (fully) utilized. But all the other evidence (system/utilities) seem to indicate that two cores are being used 100% for two instances of Rosetta.

There's a contradiction here somewhere, and I'm just not seeing it.

Hi,

have a look at this host. It has the same specs than your computer and uses the standard client and has similar benches (not overclocked). It gets about 40 credits/3 hours ~ 13/hour/core ~ 624 credits/day, if its crunching 24 hours. You should achieve similar numbers if your host is working fine. From what you've reported it seems both cores don't get utilized fully or there is some other problem preventing your comp from achieving its normal performance. There is a BIOS setting, enable ACPI or something, can you check that? You also can change your general preference so that it uses only one core. Then you should still get 13credits/hour but only with one working core it should be something like 320 credits/day. This would allow you to check whether running on one core gives normal performance.

Hey, thanx for info and benchmarks!

I went through all BIOS settings, and there are NONE that deal with single/dual processors/cores. The only things BIOS confirms for me is that the cpu is an AMD Athlon X2 3800+.

Device Manager confirms "ACPI Multiprocessor PC".

Here's what I don't understand: Hardware / System-software "seems" to recognize the dual core. Boinc settings are for 2 cpus, 100%.

If the issue was that Rosetta was really running two instances on one core, then why would the second core also show 100% utilization?

Wouldn't it show 1 core at 100% (i.e., the two instances of Rosetta), and the other core at some nominal low value (remembering that TaskManager/Processes shows only Rosetta with a value greater than 1%).

I agree that the results I am seeing appear to indicate that only one core is being (fully) utilized. But all the other evidence (system/utilities) seem to indicate that two cores are being used 100% for two instances of Rosetta.

There's a contradiction here somewhere, and I'm just not seeing it.

It seems odd I agree. However I wouldn't spend too much time thinking about what indicates what but just trying to figure out how to fix it. First to check is whether one instance on one core brings the expected performance (about 13 credits/hour). If that is the case try activating the second core and checking what happens than to the credits/hour. It might be that due to the different RAM sticks your RAM performance is somehow degraded and can't serve both cores effectively. You can try to remove the old 256 MB stick and see if the performance increases.

Running some benchmarks and checking whether they match other Athlon 64 X2 3800+ could also point to possible problems. For example:

http://www.sisoftware.net/redirect/dload.php?id=113

wu1 just finished, over 10 hours for a 6 hour Rosetta prefernce. About 68 credits, in line with what a 6 hour wu on the Sempron got.

wu2 should be finishing shortly, and expect similar.

So, for 10 hours of work, about 135 credits. Extrapolated out, about 325 credits for 24 hours worth of work.

This would be in line with one cpu/core, not two. (i.e., one-half of your benchmark on a similar platform).

So, the issue does seem to been that only one core in actually being utilized, in spite of what system/utility software indicates.

Don't think it's a RAM problem, but I can remove the 256mb stick to rule out this as a possibility.

EDIT -> I should also mention, now that wu1 is completed, wu2 is running at 20%-25%, with system idle at about 75%-80%. Don't know what to make of this.

It seems odd I agree. However I wouldn't spend too much time thinking about what indicates what but just trying to figure out how to fix it. First to check is whether one instance on one core brings the expected performance (about 13 credits/hour). If that is the case try activating the second core and checking what happens than to the credits/hour. It might be that due to the different RAM sticks your RAM performance is somehow degraded and can't serve both cores effectively. You can try to remove the old 256 MB stick and see if the performance increases.

Running some benchmarks and checking whether they match other Athlon 64 X2 3800+ could also point to possible problems. For example:

http://www.sisoftware.net/redirect/dload.php?id=113

EDIT -> I should also mention, now that wu1 is completed, wu2 is running at 20%-25%, with system idle at about 75%-80%. Don't know what to make of this.

That's not good. You should some benchmarks with Sisoft Sandra and check whether your system can achieve normal dual-core performance on benchmarks. Furthermore I definitely would try setting BOINC to one core only and finish a short WU (3 hours) and check how long it takes and how many credits you receive.

Good ideas. Will implement, and advise.

EDIT -> I should also mention, now that wu1 is completed, wu2 is running at 20%-25%, with system idle at about 75%-80%. Don't know what to make of this.

That's not good. You should some benchmarks with Sisoft Sandra and check whether your system can achieve normal dual-core performance on benchmarks. Furthermore I definitely would try setting BOINC to one core only and finish a short WU (3 hours) and check how long it takes and how many credits you receive.

According to SiSoft Sandra Processor/Arithmetic benchmark, I am UNDERPERFORMING against an Athlon XP 2200+, not to mention other more powerful processors.

That begs the question, why would an X2 3800+ underperform benchmarks against an XP2200+?

I'll "assume" that if I mucked-up the physical install of the cpu, either the computer wouldn't boot, or Windows wouldn't start.

So, given that the physical install was properly done, where do I turn next?

Good ideas. Will implement, and advise.

EDIT -> I should also mention, now that wu1 is completed, wu2 is running at 20%-25%, with system idle at about 75%-80%. Don't know what to make of this.

That's not good. You should some benchmarks with Sisoft Sandra and check whether your system can achieve normal dual-core performance on benchmarks. Furthermore I definitely would try setting BOINC to one core only and finish a short WU (3 hours) and check how long it takes and how many credits you receive.

According to SiSoft Sandra Processor/Arithmetic benchmark, I am UNDERPERFORMING against an Athlon XP 2200+, not to mention other more powerful processors.

That begs the question, why would an X2 3800+ underperform benchmarks against an XP2200+?

I'll "assume" that if I mucked-up the physical install of the cpu, either the computer wouldn't boot, or Windows wouldn't start.

So, given that the physical install was properly done, where do I turn next?

Not necessarily. If the CPU is overheating due to a badly seated heatsink etc then it will throttle down. Are you sure it's running at full speed? I don't know if they can throttle independantly, but if one is running cooler than the other and therefore faster, it might explain some of what you're seeing.

I'd recommend re-doing the heatsink installation.

---

I'll look into this, badly situated heatsink.

I reset BOINC to 1 cpu, 100%. Updated the Rosetta project with these new Boinc preferences.

Still showing 75% idle, 25% Rosetta.

Taking about 6-7 mins of "real time" to do about 2 mins of "cpu time" on Rosetta.

AMD's software utility "DashBoard" shows two cores being utilized less than 100%, and cpu temperature is 40C. (So, not certain it is bad placement of heatsink).

CPU speed is at 90% most of the time (system idle?), and briefly spikes to 100% (when Rosetta "kicks-in" ?).

Not necessarily. If the CPU is overheating due to a badly seated heatsink etc then it will throttle down. Are you sure it's running at full speed? I don't know if they can throttle independantly, but if one is running cooler than the other and therefore faster, it might explain some of what you're seeing.

I'd recommend re-doing the heatsink installation.

Perhaps it's time to reformat the system and setup WinXP from scratch.

---

Another dumb question, though:

have you set the AMD Cool&Quiet Drivers to allow for full speed or are they set to throttle down the system? As every BOINC task has the lowest priority to make sure it really only uses the unused cycles, Cool&Quiet does not regard Rosetta as a task that needs full CPU speed and throttles down to something around 1 GHz, which is 50% of maximum speed.

This should be somewhere in the Control Panel and under "Power Options Properties" under XP. In Win2000 you find it in the "Power Optins Properties" as a separate tab labeled "AMDs Cool&Quiet Technology" or something (I have the German version, so I do not know the exact names for you).

You need to set the properties to that one meaning "always on" or "maximum performance". Another prossibility is to simply switch C&Q support off in the BIOS.

Regards,

Christoph

Have not installed C&Q.

The heatsink on cpu is properly situated.

I'd really rather not to an XP re-install, (groan!).

Trying over at AMD's board, see if they know whats going on.

Another dumb question, though: have you set the AMD Cool&Quiet Drivers to allow for full speed or are they set to throttle down the system?

AMD says Windows re-install. Easy for them to say!

I don't believe them, but I'll give it a shot.

Glad I have this laptop as a (Rosetta) back-up.

I'll keep ya posted.