What are other hosts seeing on these looprlx GP120 tasks? How does your claimed vs granted credit stack up?

I've returned only 3 looprlx GP120 tasks and:
claimed 47.67 granted 52.59
claimed 47.24 granted 50.30
claimed 47.50 granted 46.02

I've got another Tualatin Celeron (1.3GHz) that's run some and they're fine:
55795201 279343 6 Jan 2007 13:20:05 UTC 9 Jan 2007 2:47:16 UTC Over Success Done 14,297.72 24.32 17.56
55835861 279343 6 Jan 2007 18:58:34 UTC 9 Jan 2007 10:05:46 UTC Over Success Done 14,136.33 24.05 20.42

---

I've had a PILE of 'em. Here's the list I've got so far:

WU         Seconds    Claim    Granted

55705133   86,198.66  143.86   128.49 
55591588   86,199.78  143.86   137.70
55705307   86,359.00  125.13   151.95
55878685   86,264.06  143.64   144.80 
55904269   86,341.78  143.77   128.77 
55878685   86,264.06  143.64   144.80 
56484015   10,638.31   14.29    14.18 
56477369   10,755.16   14.44    12.36 
56469380   10,846.84   14.57    13.87 
56273359   10,538.11   14.15    13.70 
56260744   10,845.11   14.56    14.40 
56254647   10,470.83   14.06    13.30 
56249988   10,584.47   14.21    13.72 
56249972   10,852.09   14.57    14.14 
56249971   10,696.31   14.36    13.68 
55765648   10,691.59   14.36    15.05 

---

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/

a few more on the original comp:

56482969 50196687 10 Jan 2007 9:39:16 UTC 12 Jan 2007 3:04:38 UTC Over Success Done 8,469.35 16.64 0.43
56482867 50196589 10 Jan 2007 9:39:16 UTC 12 Jan 2007 0:02:44 UTC Over Success Done 8,389.73 16.48 0.45
56448504 50165074 10 Jan 2007 4:40:43 UTC 12 Jan 2007 0:02:44 UTC Over Success Done 7,778.94 15.28 0.38
56428941 50147278 10 Jan 2007 1:39:31 UTC 11 Jan 2007 19:10:03 UTC Over Success Done 7,491.58 14.72 0.48
56401567 50123905 9 Jan 2007 21:47:36 UTC 11 Jan 2007 17:04:18 UTC Over Success Done 8,511.39 16.72 0.62
56385298 50109917 9 Jan 2007 19:49:19 UTC 11 Jan 2007 14:41:25 UTC Over Success Done 7,460.91 14.66 0.41


There are quite a few more, but they're all the same. THey're all looprlx GP120 tasks. The other jobs run fine.

---

Mine continues to climb back slowly.
None of mine were anywhere near that drastic.

I just had a 25pt per machine drop per day. It was consistent across all 4 machines. Just all the WU's were granted lower than their normal average granted credit.

One of the has increased more that the other 3, and 2 out of the three have increased a little, and one has remained at it's low output average.

The 3 that have started to climb are crunching 8hr WU's. The one that has remained low is crunching 6hr WU's, I bumped that one up from 4hrs, to see if it would make a difference, but none so far, that is basically the same time settings I had before the drop.

---

BETA = Bahhh

Way too many errors, killing both the credit & RAC.

And I still think the (New and Improved) credit system is not ready for prime time...

dcdc, now we just need to figure out why my machine gets 32 models crunched in 10638 seconds and yours gets 1 done in 8511 seconds.

Comparing the host of mine that crunched the 32 models with the host of yours.

Mine is Intel P4 2.8 Ghz, reports 1230m float ops per second
Yours is Intel Celeron 1000Mhz, and reports 1210m float ops per second

If my machine is roughly 3x faster CPU, why are our ops per second so close? In fact your's reports double the integer ops of mine.

Seems like your benchmarks are not really reflecting the power of your machine properly. The actual work produced by your machine is not stacking up to your reported benchmarks. ...but it seems REALLY odd that your other types of WUs are apparently not affected in a similar way.

I found two very similar named non-looprlx WUs between our two machines:
Mine 11 models, 9,941 seconds, claimed 13.24, granted 16.15
Yours 16 models, 14,489 seconds, claimed 28.12, granted 25.33

There must be some rather dramatic differences in how the two types of WUs run that hit a bottleneck of some kind on your machine. Are you doing a lot of page faults on the looprlx tasks?

---

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/

It's a 1GHz Celeron but running at 133 x 10 giving 1.333GHz, so it's pretty much half the speed of yours, but I'd expect it to be around 2/3's the speed of your 2.8GHz P4 on the tiny BOINC benchmarks because of the higher efficiency of the P3 architecture. I'm surprised it gets a higher score though, but my mum's celeron 1.3GHz (stock speed) gets pretty much the same score as this one (although it's got a lower FSB and therefore RAM speed @100MHz):

Measured floating point speed 1137.54 million ops/sec
Measured integer speed 1957.56 million ops/sec

Maybe a corrupt file? I guess I should start a new thread for this - it looks like a separate error. There doesn't seem to be anything wrong here:

2007-01-12 06:18:50 [---] Rescheduling CPU: files downloaded
2007-01-12 06:53:25 [---] Rescheduling CPU: application exited
2007-01-12 06:53:25 [rosetta@home] Computation for task 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_4IGNORE_THE_REST_0001_1475_2964_0 finished
2007-01-12 06:53:25 [rosetta@home] Starting task 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_10IGNORE_THE_REST_0001_1475_3512_0 using rosetta version 543
2007-01-12 06:53:27 [rosetta@home] Started upload of file 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_4IGNORE_THE_REST_0001_1475_2964_0_0
2007-01-12 06:53:31 [rosetta@home] Finished upload of file 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_4IGNORE_THE_REST_0001_1475_2964_0_0
2007-01-12 06:53:31 [rosetta@home] Throughput 17976 bytes/sec
2007-01-12 06:53:35 [rosetta@home] Sending scheduler request to https://boinc.bakerlab.org/rosetta_cgi/cgi
2007-01-12 06:53:35 [rosetta@home] Reason: To report completed tasks
2007-01-12 06:53:35 [rosetta@home] Reporting 1 tasks
2007-01-12 06:53:40 [rosetta@home] Scheduler request succeeded
2007-01-12 10:43:46 [---] Rescheduling CPU: application exited
2007-01-12 10:43:46 [rosetta@home] Computation for task 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_10IGNORE_THE_REST_0001_1475_3512_0 finished
2007-01-12 10:43:46 [rosetta@home] Starting task s013__BOINC_ABRELAX_SAVE_ALL_OUT_hom001__1405_109582_0 using rosetta version 543
2007-01-12 10:43:49 [rosetta@home] Started upload of file 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_10IGNORE_THE_REST_0001_1475_3512_0_0
2007-01-12 10:43:55 [rosetta@home] Finished upload of file 1fvv_1_NMRREF_1_1fvv_1_yyidrenum_10IGNORE_THE_REST_0001_1475_3512_0_0
2007-01-12 10:43:55 [rosetta@home] Throughput 17898 bytes/sec
2007-01-12 10:43:56 [rosetta@home] Sending scheduler request to https://boinc.bakerlab.org/rosetta_cgi/cgi
2007-01-12 10:43:56 [rosetta@home] Reason: To report completed tasks
2007-01-12 10:43:56 [rosetta@home] Reporting 1 tasks
2007-01-12 10:44:01 [rosetta@home] Scheduler request succeeded
2007-01-12 14:39:24 [---] Rescheduling CPU: application exited
2007-01-12 14:39:24 [rosetta@home] Computation for task s013__BOINC_ABRELAX_SAVE_ALL_OUT_hom001__1405_109582_0 finished
2007-01-12 14:39:24 [rosetta@home] Starting task s014__BOINC_LOOP_RELAX_IGNORE_THE_REST_hom001__IGNORE_THE_REST_mapback_hom006_S_00003_0004165_0_1447_303_0 using rosetta version 543
2007-01-12 14:39:26 [rosetta@home] Started upload of file s013__BOINC_ABRELAX_SAVE_ALL_OUT_hom001__1405_109582_0_0
2007-01-12 14:39:33 [rosetta@home] Finished upload of file s013__BOINC_ABRELAX_SAVE_ALL_OUT_hom001__1405_109582_0_0
2007-01-12 14:39:33 [rosetta@home] Throughput 8932 bytes/sec
2007-01-12 14:39:38 [rosetta@home] Sending scheduler request to https://boinc.bakerlab.org/rosetta_cgi/cgi
2007-01-12 14:39:38 [rosetta@home] Reason: To report completed tasks
2007-01-12 14:39:38 [rosetta@home] Reporting 1 tasks
2007-01-12 14:39:43 [rosetta@home] Scheduler request succeeded
2007-01-12 18:27:26 [---] Rescheduling CPU: application exited
2007-01-12 18:27:26 [rosetta@home] Computation for task 

I'll run the xp memtest util on it this weekend to make sure there's nothing wrong with it.

---

Mine continues to climb back slowly.
None of mine were anywhere near that drastic.

I just had a 25pt per machine drop per day. It was consistent across all 4 machines. Just all the WU's were granted lower than their normal average granted credit.

One of the has increased more that the other 3, and 2 out of the three have increased a little, and one has remained at it's low output average.

The 3 that have started to climb are crunching 8hr WU's. The one that has remained low is crunching 6hr WU's, I bumped that one up from 4hrs, to see if it would make a difference, but none so far, that is basically the same time settings I had before the drop.

I putted back the V8 online, let s see if you start dropping again ... ;)

who?

Mine continues to climb back slowly.
None of mine were anywhere near that drastic.

I just had a 25pt per machine drop per day. It was consistent across all 4 machines. Just all the WU's were granted lower than their normal average granted credit.

One of the has increased more that the other 3, and 2 out of the three have increased a little, and one has remained at it's low output average.

The 3 that have started to climb are crunching 8hr WU's. The one that has remained low is crunching 6hr WU's, I bumped that one up from 4hrs, to see if it would make a difference, but none so far, that is basically the same time settings I had before the drop.

I putted back the V8 online, let s see if you start dropping again ... ;)

who?

What do I know...but seems like the two of you would have to be processing the same proteins/tasks to postulate that Who? has any effect on Stevea?

dcdc, now we just need to figure out why my machine gets 32 models crunched in 10638 seconds and yours gets 1 done in 8511 seconds.

Comparing the host of mine that crunched the 32 models with the host of yours.......

DCDC's machine has less than 1/2 the memory of Feet1st, though the cache is the same. This have some bearing on the disparity?

What do I know...but seems like the two of you would have to be processing the same proteins/tasks to postulate that Who? has any effect on Stevea?

I never thought that who? had anything to do with it..

---

BETA = Bahhh

Way too many errors, killing both the credit & RAC.

And I still think the (New and Improved) credit system is not ready for prime time...

What do I know...but seems like the two of you would have to be processing the same proteins/tasks to postulate that Who? has any effect on Stevea?

I never thought that who? had anything to do with it..

now, i don t think so.
one week without processing ... my RAC went down seriously ;)
I am playing catch up

who?

Mine continues to climb back slowly.
None of mine were anywhere near that drastic.

I just had a 25pt per machine drop per day. It was consistent across all 4 machines. Just all the WU's were granted lower than their normal average granted credit.

One of the has increased more that the other 3, and 2 out of the three have increased a little, and one has remained at it's low output average.

The 3 that have started to climb are crunching 8hr WU's. The one that has remained low is crunching 6hr WU's, I bumped that one up from 4hrs, to see if it would make a difference, but none so far, that is basically the same time settings I had before the drop.

I putted back the V8 online, let s see if you start dropping again ... ;)

who?

What do I know...but seems like the two of you would have to be processing the same proteins/tasks to postulate that Who? has any effect on Stevea?

NBIT you would need to read the whole thread to understand this part.
But generelly, Who? thought he might be causing the problem so we jested it's all his fault, aftera ll he does have the 2 fastest crunching computer at rosetta :-D

---

Team mauisun.org

Actually "idle" has it right. Who?'s machine would have to have crunched the same work to have any effect on any given other machine. That's because the averages are calculated on a per WU basis. But, given the scale of that beast, I'm sure it crunched a LOT of EVERYTHING. Again, a fast machine would ONLY effect the credits if it's claim were vastly different from the claims of others for the same number of models in the same time. And even then, it would ONLY, POSSIBLY be a factor very early in the life of a WU. After that, the masses outweight the Who?. ...and it would effect everyone, not just dcdc.

And yes, dcdc's machine has half the memory, but both machines have about twice as much as Rosetta actually needs while it is crunching. So, I doubt that is a measurable factor. Also, there still remains the mystery of why other WUs are not effected. If memory were the issue, then we should see the WUs dcdc scores poorly on using more memory. I believe I had checked for that while some of them were running and did not see it to be the case.

---

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/

that machine is almost always just running Rosetta - it's my living-room media centre, but i shut the MCE interface down when it's not in use to save cycles for rosie ;) So I don't think it's the memory that's the problem. I need to test the memory on it though - have only just got back from London and have a hangover so won't get round to it tonight!

---

I noticed the degradation on 10 Dec and mentioned it on my team forum.

I have 2 FX-60s, overclocked, running 24/7. Daily production has dropped from the low 1600s to the mid 1400s since then.

Here are two recent uploads from one of the boxes.

57191300 50836422 14 Jan 2007 15:49:27 UTC 15 Jan 2007 4:04:39 UTC Over Success Done 10,451.06 46.75 45.50

57180944 50827223 14 Jan 2007 14:18:50 UTC 15 Jan 2007 1:53:11 UTC Over Success Done 10,864.11 48.60 1.70

---

Here are two WUs that stalled. The second one stalled at 88.541%.
I suspended then resumed and it started at 88.000%, crept back up to 88.541%, and stalled again. I suspect a huge value in one loop or an infinite loop. I didn't want to waste my cycles to find out.

57239311 50879456 14 Jan 2007 22:36:33 UTC 15 Jan 2007 7:56:01 UTC Over Client error Compute error 4,722.81 21.39 ---


57223113 50865015 14 Jan 2007 20:08:59 UTC 15 Jan 2007 7:55:51 UTC Over Client error Compute error 9,663.22 43.22 ---

Looks like I'm stalled at 1250-1260.

Had my first wu crash since Dec 19th on Jan 17th. So my systems are pretty reliable. 1 crash per month is livable. And it was a graphics error, even though I don't look at the graphics or run the screen saver.

Still 75ppd lower than before?

Any more info why this is still happening?

---

BETA = Bahhh

Way too many errors, killing both the credit & RAC.

And I still think the (New and Improved) credit system is not ready for prime time...

Stevea, I started collecting data on the 16th, should have a comparative chart available soon. comparing old "new" rosetta and today.

tony

dcdc, now we just need to figure out why my machine gets 32 models crunched in 10638 seconds and yours gets 1 done in 8511 seconds.

Comparing the host of mine that crunched the 32 models with the host of yours.......

DCDC's machine has less than 1/2 the memory of Feet1st, though the cache is the same. This have some bearing on the disparity?

Actually, the cache size you're seeing (976.56 KB) is the default reported by some versions of the BOINC software. My P-4 based Celeron 2.3 GHz has 128KB L2, but reports the same value, 976.56KB, for cache size on some versions of the BOINC software.

P-III Celerons never got higher than 256KB cache and most had 128KB cache. If dcdc's 1 GHz Celeron is a Coppermine, it should have a 128KB cache. If it's a Tualatin, it should have 256KB. The FSB is 100 MHz for a 1 GHz on both Coppermine and Tualatin.

A P-4 2.8 should have 512KB cache (Northwood) or 1MB cache (Prescott / 5xx series). The Pentium 6xx series has 2 MB cache, but I think it starts at 3.0 Ghz. The FSB should be either 533 MHz (early Northwood) or 800 MHz.

HTH,

-- David

---

Have you read a good Science Fiction book lately?