Upgrade complete, sort of

Message boards : Number crunching : Upgrade complete, sort of

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 11 · Next

AuthorMessage
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25345 - Posted: 28 Aug 2006, 22:31:51 UTC

Well, NewEgg and UPS came through. I received my AMD Athlon X2 3800+ Dual Core and the Arctic Cooler Freezer 64 Pro this afternoon.

Little bit of a challenge to remove the stock cpu fan. No problem removing the AMD Sempron 3400+.

The X2 went right in to the Socket 939.

Installing the Arctic Cooler, well, that was an adventure unto itself.

The mobo bios had been updated.

Ran the AMD Dual Core Optimizer.

Windows XP Home booted fine.

The new cpu was recognized.

BOINC has two instances of Rosetta running at the same time.

I assume this is what is supposed to happen with dual core systems.

Strangest thing though, I seem to have lost all audio output. No, I didn't forget to plug the speaker cable back in.

Hmm on that!
ID: 25345 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
XS_The_Machine

Send message
Joined: 2 Jan 06
Posts: 47
Credit: 4,612,053
RAC: 0
Message 25350 - Posted: 28 Aug 2006, 23:08:43 UTC - in response to Message 25345.  

Well, NewEgg and UPS came through. I received my AMD Athlon X2 3800+ Dual Core and the Arctic Cooler Freezer 64 Pro this afternoon.

Little bit of a challenge to remove the stock cpu fan. No problem removing the AMD Sempron 3400+.

The X2 went right in to the Socket 939.

Installing the Arctic Cooler, well, that was an adventure unto itself.

The mobo bios had been updated.

Ran the AMD Dual Core Optimizer.

Windows XP Home booted fine.

The new cpu was recognized.

BOINC has two instances of Rosetta running at the same time.

I assume this is what is supposed to happen with dual core systems.

Strangest thing though, I seem to have lost all audio output. No, I didn't forget to plug the speaker cable back in.

Hmm on that!


Drivers?

ID: 25350 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25353 - Posted: 28 Aug 2006, 23:31:08 UTC - in response to Message 25350.  

Yeah, spent about 30 min with HP/Compaq to resolve.

Had to do a re-boot.

Funny thing, one of the two Rosetta wu's picked up at about 25 mins into project (out of about 30 mins), the second wu lost everything (~ 30 mins of work) and started over from scratch.


Drivers?

ID: 25353 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 25357 - Posted: 29 Aug 2006, 0:30:28 UTC

When you pull up taskmanager, choose the Performance tab (choose view, cpu history, one graph per cpu) do you get two cpu windows at 100%? If so, then WinXP definately deals with the change from single core to dual core better than Win2k. *grin*


ID: 25357 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25361 - Posted: 29 Aug 2006, 1:02:28 UTC - in response to Message 25357.  
Last modified: 29 Aug 2006, 1:07:27 UTC

Yes, but..... something definitely dosen't seem right.

After the reboot, one wu started at about 30 mins, and the second re-started at zero.

Now, wu1 is at cpu time 1 hour 55 mins, and wu2 is at 20 mins.

EDIT - It seems that for every 4-5 seconds that wu1 advances, wu2 advances only 1 sec - END EDIT

With no offense to anyone, something seems rotten in Denmark!

When you pull up taskmanager, choose the Performance tab (choose view, cpu history, one graph per cpu) do you get two cpu windows at 100%? If so, then WinXP definately deals with the change from single core to dual core better than Win2k. *grin*

ID: 25361 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ethan
Volunteer moderator

Send message
Joined: 22 Aug 05
Posts: 286
Credit: 9,304,700
RAC: 0
Message 25363 - Posted: 29 Aug 2006, 1:29:06 UTC

If you open task manager, then hit the processes tab, and finaly sort by cpu usage, are both rosetta.exe's getting around 50%?

It's a long shot, but something else on your system might be running a good portion of one of the cpu's.

The reason I say it's a long shot is if you have something else taking around 25%, usually windows will spread the load so that both rosetta processes would go down by the same amount.


ID: 25363 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 25364 - Posted: 29 Aug 2006, 1:34:33 UTC - in response to Message 25361.  

EDIT - It seems that for every 4-5 seconds that wu1 advances, wu2 advances only 1 sec - END EDIT


I can't explain that one, except to point out that the constraining facter on a Windows dual processor is the contention for floating point math. So the two threads don't really give you twice the crunching.

With no offense to anyone, something seems rotten in Denmark!

With regard to your time crunched and time lost on the two WUs, this sounds pretty normal. You said one crunched for 30 min, and picked up at 25 on the restart. This is because it hit a checkpoint (or end of a model) after 25min and pushed all the active information out to disk. You said the other started over from scratch... well, if you had done your restart after 24 min. they BOTH would have had to restart from scratch. The second WU did not complete a checkpoint, or model in the 30min of crunching it had done (the call for more checkpointing has already been made). Some proteins can take over an hour to reach a checkpointable state.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 25364 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
P . P . L .

Send message
Joined: 20 Aug 06
Posts: 581
Credit: 4,865,274
RAC: 0
Message 25366 - Posted: 29 Aug 2006, 1:36:48 UTC

Could it be that one W.U.is more complexed than the other

and requires more work?

Just a guess.

ID: 25366 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25367 - Posted: 29 Aug 2006, 1:43:52 UTC - in response to Message 25363.  

Didn't think of looking at it this way!

wu1 gets between 75% - 85%
wu2 gets between 15% - 25%

At any particular instant, these two numbers added together equal 100%.

If you open task manager, then hit the processes tab, and finaly sort by cpu usage, are both rosetta.exe's getting around 50%?

It's a long shot, but something else on your system might be running a good portion of one of the cpu's.

The reason I say it's a long shot is if you have something else taking around 25%, usually windows will spread the load so that both rosetta processes would go down by the same amount.

ID: 25367 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 25368 - Posted: 29 Aug 2006, 1:44:02 UTC - in response to Message 25366.  

Could it be that one W.U.is more complexed than the other
and requires more work?


Well, the WUs do vary in complexity. But he's saying that 10 seconds go by on his watch, and one WU gets 10 seconds, and the other only gets 3. On a dual core machine, unless it's busy doing some higher priority tasks (which should also show on the displays he's looking at) the two should be very nearly equal.

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 25368 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25369 - Posted: 29 Aug 2006, 1:48:42 UTC - in response to Message 25366.  

I would think that this would result in one wu taking longer to complete in absolute time. But for every advancement of 1 sec on wu1, I would "expect" an advancement of 1 sec on wu2.

Could it be that one W.U.is more complexed than the other and requires more work?
Just a guess.

ID: 25369 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ethan
Volunteer moderator

Send message
Joined: 22 Aug 05
Posts: 286
Credit: 9,304,700
RAC: 0
Message 25372 - Posted: 29 Aug 2006, 2:03:56 UTC - in response to Message 25367.  
Last modified: 29 Aug 2006, 2:05:00 UTC

Didn't think of looking at it this way!

wu1 gets between 75% - 85%
wu2 gets between 15% - 25%

At any particular instant, these two numbers added together equal 100%.



Unfortunately, Rosetta isn't multi-threaded. There is no way for it to use more than 100% of a single cpu (in my experience with the dual cores at home and in my office). It would be a great feature to have since it would require less memory from the system (running 1 wu twice as fast vs 2 at once). Hopefully memory technology keeps up, in 3-4 years we'll have cpus with 16 or 32 cores!
ID: 25372 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 25373 - Posted: 29 Aug 2006, 2:04:15 UTC
Last modified: 29 Aug 2006, 2:10:36 UTC

You should have two Rosetta processes in the process list of Task Manager (you do) Is there any other process in the task list that is getting more than 0 percent on a steady basis? (If they exist, list the percentage of cpu time given to them.)

Wait a sec.. if you're seeing one instance of Rosetta getting 75-85% and another getting 25-15%.. then they're both running on the same core. With a dual core Athlon system running Win2k or WinXP, the max you get is 50% for a single thread app.

Now it's time to switch WinXP to the multiprocessor HAL.
ID: 25373 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25375 - Posted: 29 Aug 2006, 2:11:08 UTC - in response to Message 25373.  

Most recent observation:

wu1 = ~ 75% to 90%
wu2 = ~ 10% to 25%

sporadically @ 1% = boinc.exe, iexplore.exe, taskmgr.exe


You should have two Rosetta processes in the process list of Task Manager - what are their percentages? And is there any other process in the task list that is getting more than 0 percent on a steady basis? (If they exist, list the percentage of cpu time given to them.)

ID: 25375 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25376 - Posted: 29 Aug 2006, 2:14:18 UTC - in response to Message 25373.  

I guess that why I figured the upgrade was "sort of" complete.

Switching to multiprocessor HAL?

Um, is this gonna hurt (lol)?

Now it's time to switch WinXP to the multiprocessor HAL.

ID: 25376 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25377 - Posted: 29 Aug 2006, 2:16:20 UTC - in response to Message 25373.  

But TaskManager tab Performance shows two cpu usage at 100%.

Wait a sec.. if you're seeing one instance of Rosetta getting 75-85% and another getting 25-15%.. then they're both running on the same core. With a dual core Athlon system running Win2k or WinXP, the max you get is 50% for a single thread app.

ID: 25377 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25378 - Posted: 29 Aug 2006, 2:23:32 UTC - in response to Message 25377.  

The most interesting thing to note, not that it makes any sense, is that prior to my re-installing the audio driver, wu1 and wu2 were in synch, 1 sec advance per 1 sec advance. i.e. when I did the re-boot, wu1 "checkpointed" and saved about 25 mins of work, while wu2 apparently didn't "checkpoint" and I lost about 25 mins of work.

What would an audio driver have to do with this?!
ID: 25378 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 25380 - Posted: 29 Aug 2006, 2:38:24 UTC

For some reason, you're running on 1 core now.

Perhaps this site will help, but similar threads exist on many of the hardware tech sites (or DC team sites.)
Site1
a short list at AMD forums



ID: 25380 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25384 - Posted: 29 Aug 2006, 2:52:45 UTC - in response to Message 25380.  

I'm willing to entertain that it might be this HAL. So long as it's not HAL-9000...

For some reason, you're running on 1 core now.

Perhaps this site will help, but similar threads exist on many of the hardware tech sites (or DC team sites.)
Site1
a short list at AMD forums



ID: 25384 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 25389 - Posted: 29 Aug 2006, 3:14:39 UTC - in response to Message 25384.  

Even more interesting, if I suspend one of the two wu's, the remaining wu is at 15% - 25%, with System Idle Process at 75% - 85%!
ID: 25389 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · 3 · 4 . . . 11 · Next

Message boards : Number crunching : Upgrade complete, sort of



©2024 University of Washington
https://www.bakerlab.org