WU taking longer?

abinkow
abinkow
Joined: 21 Feb 05
Posts: 38
Credit: 2,724,569
RAC: 0
Topic 194093

It seems to me that WU assigned to me are taking more CPU time than they used to. I have a WU now, that is 88 hours in/11 to go (99 total hours CPU), whereas a couple of weeks ago they were all running just over 71. (And even THAT is up from a couple of months ago).

Should I be worried? Are the crunching requirements per WU increasing? Or should I be looking for problems on my end?

jedirock
jedirock
Joined: 11 Jun 06
Posts: 23
Credit: 1,517,411
RAC: 0

WU taking longer?

Quote:

It seems to me that WU assigned to me are taking more CPU time than they used to. I have a WU now, that is 88 hours in/11 to go (99 total hours CPU), whereas a couple of weeks ago they were all running just over 71. (And even THAT is up from a couple of months ago).

Should I be worried? Are the crunching requirements per WU increasing? Or should I be looking for problems on my end?


I haven't noticed any increase recently. All WUs here have been holding around 13 hours. Although I seem to remember 11 hour WUs before... If it were lower than I remembered, it's possible the ratio could be the same. Anyone from the project have answers?

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 455,521,082
RAC: 45,678

RE: It seems to me that WU

Quote:

It seems to me that WU assigned to me are taking more CPU time than they used to. I have a WU now, that is 88 hours in/11 to go (99 total hours CPU), whereas a couple of weeks ago they were all running just over 71. (And even THAT is up from a couple of months ago).

Should I be worried? Are the crunching requirements per WU increasing? Or should I be looking for problems on my end?

Well, there is a variation in runtime of the workunits (some just happen to be faster than others per se) and most of the units you have now happen to be on the slow end of the variation. The variation can be more than 30 %. Having said that, even tho the variation is nothing surprising, the absolute runtime of 80 hrs and more per WU is real slow even for a Pentium 4 in hyperthreading mode. Please let us know when the result has been reported so we can look at the output of the app.

P.S.: We are talking about this PC, right?
CU
Bikeman

tullio
tullio
Joined: 22 Jan 05
Posts: 2,038
Credit: 39,628,771
RAC: 23,957

I am averaging a steady 45k s

I am averaging a steady 45k s (~12 hours) on my Opteron 1210 at 1.8 GHz running Linux. My wingman is often an ATLAS CPU taking 25k s.
Tullio

Nothing But Idle Time
Nothing But Idl...
Joined: 24 Aug 05
Posts: 158
Credit: 289,204
RAC: 0

RE: RE: It seems to me

Message 89220 in response to message 89218

Quote:
Quote:

It seems to me that WU assigned to me are taking more CPU time than they used to. I have a WU now, that is 88 hours in/11 to go (99 total hours CPU), whereas a couple of weeks ago they were all running just over 71. (And even THAT is up from a couple of months ago).

Should I be worried? Are the crunching requirements per WU increasing? Or should I be looking for problems on my end?

Well, there is a variation in runtime of the workunits (some just happen to be faster than others per se) and most of the units you have now happen to be on the slow end of the variation. The variation can be more than 30 %. Having said that, even tho the variation is nothing surprising, the absolute runtime of 80 hrs and more per WU is real slow even for a Pentium 4 in hyperthreading mode. Please let us know when the result has been reported so we can look at the output of the app.

P.S.: We are talking about this PC, right?
CU
Bikeman


My P4 with hyperthreading active running at 3GHz is carrying completion times of about 23-24 hours, and I thought that was unacceptably slow.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 455,521,082
RAC: 45,678

RE: My P4 with

Message 89221 in response to message 89220

Quote:

My P4 with hyperthreading active running at 3GHz is carrying completion times of about 23-24 hours, and I thought that was unacceptably slow.

Well, THAT is quite normal actually. It was far a reason that Intel abandoned the Netburst architecture of the P4 in favor of the older "P6" design of the Pentium III era :-).

CU
Bikeman

Novasen169
Novasen169
Joined: 14 May 06
Posts: 43
Credit: 2,767,204
RAC: 0

RE: My P4 with

Message 89222 in response to message 89220

Quote:
My P4 with hyperthreading active running at 3GHz is carrying completion times of about 23-24 hours, and I thought that was unacceptably slow.


My P4 with hyperthreading (3.2 GHz) used to have roughly the same results so yes that's normal, but 100 hours is way too long...

Nothing But Idle Time
Nothing But Idl...
Joined: 24 Aug 05
Posts: 158
Credit: 289,204
RAC: 0

RE: RE: My P4 with

Message 89223 in response to message 89221

Quote:
Quote:

My P4 with hyperthreading active running at 3GHz is carrying completion times of about 23-24 hours, and I thought that was unacceptably slow.

Well, THAT is quite normal actually. It was far a reason that Intel abandoned the Netburst architecture of the P4 in favor of the older "P6" design of the Pentium III era :-).

CU
Bikeman

Then I scratch my head and wonder how such brilliant people in a very competitive arena could actually develop, test and market such ineffective architecture. And worse, I'm stuck with this (then) cutting edge architecture until I can afford something else. At which time I still won't know what to buy and still risk buying another red herring.

mikey
mikey
Joined: 22 Jan 05
Posts: 6,196
Credit: 548,414,294
RAC: 117,204

RE: RE: RE: My P4 with

Message 89224 in response to message 89223

Quote:
Quote:
Quote:

My P4 with hyperthreading active running at 3GHz is carrying completion times of about 23-24 hours, and I thought that was unacceptably slow.

Well, THAT is quite normal actually. It was far a reason that Intel abandoned the Netburst architecture of the P4 in favor of the older "P6" design of the Pentium III era :-).

CU
Bikeman

Then I scratch my head and wonder how such brilliant people in a very competitive arena could actually develop, test and market such ineffective architecture. And worse, I'm stuck with this (then) cutting edge architecture until I can afford something else. At which time I still won't know what to buy and still risk buying another red herring.

I have both an Intel P4 dual core with HT on each core and a true AMD Phenom quad core and the Intel is producing more RAC, and both are on the same Project.
Here is a link to the Intel
http://abcathome.com/show_host_detail.php?hostid=65041
The Intel is a 32.ghz

and the AMD http://abcathome.com/show_host_detail.php?hostid=65177
The AMD is a 9850

Both are running Linux and both are running the latest version of Ubuntu. Both have not been updated as often as they could be but are fairly current. They are both Boinc only machines, so they do nothing else. So don't cut the older HT Intel's too short just yet, they seem to have a bunch of life left in them.

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,313
Credit: 1,718,024,411
RAC: 940,730

RE: Then I scratch my head

Message 89225 in response to message 89223

Quote:
Then I scratch my head and wonder how such brilliant people in a very competitive arena could actually develop, test and market such ineffective architecture. And worse, I'm stuck with this (then) cutting edge architecture until I can afford something else. At which time I still won't know what to buy and still risk buying another red herring.

If you remember the late 90's intel slides where they projected CPU power growth and had our current time running somewhere between nuclear reactor and the sun in terms of heat. The networst design team just assumed *something* would be done to abate the problem, and as it turned out they were never able to reach design performance levels at acceptable power/thermal envelopes.

abinkow
abinkow
Joined: 21 Feb 05
Posts: 38
Credit: 2,724,569
RAC: 0

RE: RE: It seems to me

Message 89226 in response to message 89218

Quote:
Quote:

It seems to me that WU assigned to me are taking more CPU time than they used to. I have a WU now, that is 88 hours in/11 to go (99 total hours CPU), whereas a couple of weeks ago they were all running just over 71. (And even THAT is up from a couple of months ago).

Should I be worried? Are the crunching requirements per WU increasing? Or should I be looking for problems on my end?

Well, there is a variation in runtime of the workunits (some just happen to be faster than others per se) and most of the units you have now happen to be on the slow end of the variation. The variation can be more than 30 %. Having said that, even tho the variation is nothing surprising, the absolute runtime of 80 hrs and more per WU is real slow even for a Pentium 4 in hyperthreading mode. Please let us know when the result has been reported so we can look at the output of the app.

P.S.: We are talking about this PC, right?
CU
Bikeman

That's correct -- that is the right PC. That result, and another 90+ hour result, were posted a few days ago. (I have two processors, so I take two at a time). The two I'm working on right now are predicted to come in at 86 and 88 hours, respectively.

Comment viewing options

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