I am wondering, are there things the computer is dealing with at the time influence the Seti Benchmark while it is performing ?

( Computer was very busy with other things too durings this all)


How it was:

( Due to Trux's fine 5.3.11.tx32 client. Yes I know there are upgrades by now)

Measured floating point speed 8563.22 million ops/sec
Measured integer speed 6516.89 million ops/sec

21/01/2006 18:40:55|seti@home|CC calibration: 16.36 >> 33.25 (time: 2924s >> 3322s / Gfpops: 2.20 >> 9.83)

Then:

21/01/2006 18:48:07||Suspending computation - running CPU benchmarks
21/01/2006 18:48:07|seti@home|Pausing result 25fe05aa.27968.23889.104806.1.116_2 (left in memory)
21/01/2006 18:48:10||Running CPU benchmarks
21/01/2006 18:49:09||Benchmark results:
21/01/2006 18:49:09|| Number of CPUs: 1
21/01/2006 18:49:09|| 1921 double precision MIPS (Whetstone) per CPU
21/01/2006 18:49:09|| 6517 integer MIPS (Dhrystone) per CPU
21/01/2006 18:49:09||Finished CPU benchmarks
21/01/2006 18:49:10||Resuming computation
21/01/2006 18:49:10||Rescheduling CPU: Resuming computation
21/01/2006 18:49:10|seti@home|Resuming result 25fe05aa.27968.23889.104806.1.116_2 using setiathome version 418
21/01/2006 19:28:20||Rescheduling CPU: application exited
21/01/2006 19:28:20|seti@home|Computation for result 25fe05aa.27968.23889.104806.1.116_2 finished
21/01/2006 19:28:20|seti@home|Resuming result 18se04ab.15011.30114.848592.1.26_3 using setiathome version 418
21/01/2006 19:28:20|seti@home|CC calibration: 14.33 >> 33.20 (time: 2934s >> 3521s / Gfpops: 1.92 >> 9.78)
(Mind you, this one started before the benchmark)


Then:

21/01/2006 20:19:01|seti@home|Computation for result 18se04ab.15011.30114.848592.1.26_3 finished
21/01/2006 20:19:01|seti@home beta test|Resuming result 05jl01ab.26451.1393.286082.83_1 using setiathome_enhanced version 502
21/01/2006 20:19:01|seti@home|CC calibration: 13.65 >> 27.60 (time: 2795s >> 3170s / Gfpops: 1.92 >> 8.53)

And the next one :

21/01/2006 21:30:20|seti@home|CC calibration: 14.02 >> 28.37 (time: 2871s >> 3256s / Gfpops: 1.92 >> 8.54)

Meanwile:

21/01/2006 22:20:00|seti@home|CC calibration: 13.95 >> 28.32 (time: 2857s >> 3245s / Gfpops: 1.92 >> 8.56)

Times are UTC + 1

I am wondering, are there things the computer is dealing with at the time influence the Seti Benchmark while it is performing ?

As already explained many times here and in the documentation and the readme files too, the measured benchmarks are completely irrelevant at this calibrating client. It is called calibrating just because it is not depending on the benchmarks to claim the correct credit.

---

trux
BOINC software
Freediving Team
Czech Republic

Wel, the way I put it looks I'm stupid and not reading, its my poor way of expression and the lack of good English that may be confusing.
I'm asking a question because the huge difference in mesured speed before and afterwards and I happened to see at the same time a fall back in Gfposps wondering what might have caused that.

It's just feed back. It helped you with one tiny little bug.:-)

It's just feed back. It helped you with one tiny little bug.:-)

No, sorry, there is no problem at all. As I told, the measured or reported benchmarks are irrelevant. Better told they are being adjusted exactly for the purpose of the calibration. As long as the claimed credit is all right (and it apparently is in your case), it is doing exactly what it is intended to.

---

trux
BOINC software
Freediving Team
Czech Republic

As long as the claimed credit is all right (and it apparently is in your case), it is doing exactly what it is intended to.

Define all right.;-)

Apperently its under par for a few results ever since:

20/01/2006 19:33:48|seti@home|CC calibration: 16.06 >> 32.29 (time: 2870s >> 3246s / Gfpops: 2.20 >> 9.72)

21/01/2006 23:58:49|seti@home|CC calibration: 14.01 >> 28.71 (time: 2870s >> 3272s / Gfpops: 1.92 >> 8.64)


21/01/2006 22:20:00|seti@home|CC calibration: 13.95 >> 28.32 (time: 2857s >> 3245s / Gfpops: 1.92 >> 8.56)
21/01/2006 22:54:00|seti@home|CC calibration: 13.97 >> 28.47 (time: 2860s >> 3255s / Gfpops: 1.92 >> 8.60)

Altough recovering.:-)
Not that I mind the little fall back in claim. Love the work you do.

Not that I mind the little fall back in claim. Love the work you do.

Fall back? In the four WU listed, you claim from 28.47 to 32.29, what is very close to the value of the reference WU (32.32 cobblestones, as written many times here and in the documentation). If you used to have more before, then I guess you have either a very old machine, or were manipulating the results.

---

trux
BOINC software
Freediving Team
Czech Republic

Before i say anything else, can I check that I am reading the line

21/01/2006 23:58:49|seti@home|CC calibration: 14.01 >> 28.71 (time: 2870s >> 3272s / Gfpops: 1.92 >> 8.64)

corectly.

14.01 >> 28.71 does not mean 14.01 is-much-greater-than (>>) 28.71 (Which it clearly is not).

14.01 >> 28.71 does mean that calimed credit would have been 14.01 and has been increased to 28.71 by increasing the claimed time from 2870s to 3272s, and increasing the Gfpops from 1.92 to 8.64?

I just don't want to talk out of my ass, is all.

---

Foamy is "Lord and Master".
(Oh, + some Classic WUs too.)

yup. you have it right

---

trux
BOINC software
Freediving Team
Czech Republic

then I guess you have either a very old machine, or were manipulating the results.

My machine is a 10% overclocked amd64 3000 Newcastle core.
Do'nt know what you mean with manipulating results. Just showed whats in the messenges . Just wondering what caused the gfpops go fom an average of 9.7 to 8.6 and happened to have two exact wu timings that I showed .

It al started as shown under:
Message 235223 - Posted 21 Jan 2006 21:33:03 UTC

But I'l wonder in silence further on.

Hello Trux!!!

Hello from Portugal

Recently i download core client v5.3.6.tx20 from your dowbload site.

All run ok !

But, recently to apear de fowlling message:

22-01-2006 13:23:47|seti@home|CC calibration: blocked [negative calibration limit] 43.15 >> 23.00 (time: 7876s >> 4032s / Gfpops: 2.13 >> 2.52)

What is it ?

Beste regards

Vitor

---

My machine is a 10% overclocked amd64 3000 Newcastle core.
Do'nt know what you mean with manipulating results. Just showed whats in the messenges . Just wondering what caused the gfpops go fom an average of 9.7 to 8.6 and happened to have two exact wu timings that I showed .

As far as I saw, everything in the messages you posted is perfectly OK. Credits around 30 cobblestones per WU as in your case (as long as it was a standard full-length WU) are perfectly OK, and I do not understand what you are concerned about. If you had higher credits than ~30 cobblestones per WU before, then you were definitely overclaiming. The exact purpose of the calibrating client is tat it claims fair credit regardles for each WU, regardless of te lenght of the WU, and regardless of the performance of the computer. That was the original purpose of te benchmarks, but unfortunately, due to their limited efficiency they were not sufficient.

Maybe you could specify what exactly you find wrong in the resulting claimed credit. So far I did not see anything. And please do not speak about benchmarks - as I repeatedly answered, they are completely irrelevant.

---

trux
BOINC software
Freediving Team
Czech Republic

22-01-2006 13:23:47|seti@home|CC calibration: blocked [negative calibration limit] 43.15 >> 23.00 (time: 7876s >> 4032s / Gfpops: 2.13 >> 2.52)

What is it ?

It means that your PC would claim 43.15 cobblestones if the calibration was not turned on, but with the calibration it would have to be reduced to be fair. Currently, in such case, the calibration is not used, and the full (possibly overclaimed) credit is passed to the server. Negative calibration is currently allowed at non-SETI projects only. Very possibly, I'll increase the tolerance field at the future release, and allow the negative calibration to some extend at SETI@home too.

---

trux
BOINC software
Freediving Team
Czech Republic

Trux, if you do'nt mind I let it for what it is. Obviously I can not make clear what I mean. Mixing different questions too.
Seems to me I would'nt be much of a beta tester either. :-)

Trux, if you do'nt mind I let it for what it is. Obviously I can not make clear what I mean. Mixing different questions too.
Seems to me I would'nt be much of a beta tester either. :-)

Sorry, I did not want to discourage you. I reviewed your comments again, and see what you meant. However, there is still nothing wrong in the minor change of the individual factors between the first and the second result. The calibration is a dynamical process, and there are many factors playing a role. You can neither expect that the WU's always result in the exactly same claimed credit, nor you can count with permanently growing or perfectly stable level of the calibrating coefficients.

---

trux
BOINC software
Freediving Team
Czech Republic

Hello Trux,

Very possibly, I'll increase the tolerance field at the future release, and allow the negative calibration to some extend at SETI@home too.

Did you already implement negative calibration at SETI@home?

Every SETI-WU I have crunched with your client up to now is claiming less credit than it would claim without calibration. This is my most recent result:

2006-01-23 01:36:10 [SETI@home] Computation for result 20se04aa.18132.13264.603396.1.136_0 finished
2006-01-23 01:36:10 [SETI@home] CC calibration: 30.19 >> 22.70 (time: 8139s >> 5464s / Gfpops: 1.64 >> 2.41)

Here are the corresponding result details on the server: #207541726

And there is another thing, I do not understand: If I'm getting the line from the message tab right, it means, that I would claim 30.19 without calibration and 22.70 with calibration. The original CPU time is 8139s and I'm reporting 5464s. But as you can see in the result details I am claiming 23.88 instead of 22.70 and I am reporting 5650s instead of 5464s. These values are slightly higher than the initially calibrated values but they are still lower than the original values.

Do you know, why this all happens?

Regards,
Carsten

Hello again,

last I night I noticed the same problem, which Zap de Ridder has already reported previously:

After the restart of the client finished but not uploaded results are calibrated again. As (for whatever reason) the negative calibration blocking for SETI-WUs is not active in my client, this leads to a further reduction of the reported cpu time and the claimed credit:

2006-01-23 03:37:43 [SETI@home] Computation for result 17se04aa.6145.13312.253418.1.164_0 finished
2006-01-23 03:37:43 [SETI@home] CC calibration: 27.05 >> 21.40 (time: 7293s >> 5063s / Gfpops: 1.64 >> 2.54)
[...]
2006-01-23 10:49:46 [SETI@home] CC calibration: 18.78 >> 18.23 (time: 5063s >> 3998s / Gfpops: 1.64 >> 3.11)

Result details: #207541524


And another thing, I have noticed after the restart of the client, but I don't know, if it is intended: It looks like, that after the restart of the client the results are not calibrated immediately after they were finished but shortly before the upload:

2006-01-23 04:33:52 [---] Starting BOINC client version 5.3.12.tx36 for windows_intelx86
2006-01-23 04:33:52 [---] libcurl/7.14.0 OpenSSL/0.9.8 zlib/1.2.3
2006-01-23 04:33:52 [---] Data directory: D:\\PROGRAMME\\BOINC
2006-01-23 04:33:52 [SETI@home] Found app_info.xml; using anonymous platform
2006-01-23 04:33:52 [---] Processor: 1 AuthenticAMD AMD Athlon(tm) XP 2000+
2006-01-23 04:33:52 [---] Memory: 223.43 MB physical, 1.78 GB virtual
2006-01-23 04:33:52 [---] Disk: 24.31 GB total, 20.25 GB free
2006-01-23 04:33:52 [SETI@home] Computer ID: 2058665; location: home; project prefs: home
2006-01-23 04:33:52 [Einstein@Home] Computer ID: 518119; location: home; project prefs: home
2006-01-23 04:33:52 [SETI@home Beta Test] Computer ID: 3996; location: home; project prefs: home
2006-01-23 04:33:52 [---] General prefs: from SETI@home (last modified 2006-01-22 18:38:01)
2006-01-23 04:33:52 [---] General prefs: using separate prefs for home
2006-01-23 04:33:52 [---] Remote control not allowed
2006-01-23 04:33:52 [---] Listening on port 31416
2006-01-23 04:33:52 [---] truXoft add-on: calibrate_credit = all
2006-01-23 04:33:52 [SETI@home Beta Test] Deferring computation for result 05jl01ab.26451.1634.704834.2_2
2006-01-23 04:33:52 [SETI@home] Resuming computation for result 24se04ab.6540.16880.503418.1.243_3 using setiathome version 418
2006-01-23 04:33:52 [---] Suspending network activity - user request
2006-01-23 04:53:53 [---] Rescheduling CPU: application exited
2006-01-23 04:53:53 [SETI@home] Computation for result 24se04ab.6540.16880.503418.1.243_3 finished
2006-01-23 04:53:53 [SETI@home] Starting result 17se04aa.6145.22193.354830.1.38_3 using setiathome version 418
2006-01-23 06:45:48 [---] Rescheduling CPU: application exited
2006-01-23 06:45:48 [SETI@home] Computation for result 17se04aa.6145.22193.354830.1.38_3 finished
2006-01-23 06:45:48 [SETI@home] Starting result 25se04aa.16928.6705.104806.1.174_2 using setiathome version 418
2006-01-23 08:49:20 [---] Rescheduling CPU: application exited
2006-01-23 08:49:20 [SETI@home] Computation for result 25se04aa.16928.6705.104806.1.174_2 finished
2006-01-23 08:49:20 [SETI@home] Starting result 18se04aa.18724.16352.109660.1.87_1 using setiathome version 418
2006-01-23 08:51:26 [---] Rescheduling CPU: application exited
2006-01-23 08:51:26 [SETI@home] Computation for result 18se04aa.18724.16352.109660.1.87_1 finished
2006-01-23 08:51:26 [SETI@home] Starting result 20se04aa.18132.19233.1022148.1.213_0 using setiathome version 418
2006-01-23 10:49:46 [---] Resuming network activity
2006-01-23 10:49:46 [SETI@home] Sending scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2006-01-23 10:49:46 [SETI@home] Reason: To fetch work
2006-01-23 10:49:46 [SETI@home] Requesting 153778 seconds of new work
2006-01-23 10:49:46 [SETI@home] CC calibration: 18.78 >> 18.23 (time: 5063s >> 3998s / Gfpops: 1.64 >> 3.11)
2006-01-23 10:49:46 [SETI@home] CC calibration: 14.46 >> 14.04 (time: 3897s >> 3078s / Gfpops: 1.64 >> 3.11)
2006-01-23 10:49:46 [SETI@home] CC calibration: 24.90 >> 24.18 (time: 6713s >> 5302s / Gfpops: 1.64 >> 3.11)
2006-01-23 10:49:46 [SETI@home] CC calibration: 27.50 >> 26.70 (time: 7412s >> 5854s / Gfpops: 1.64 >> 3.11)
2006-01-23 10:49:46 [SETI@home] CC calibration: 0.45 >> 0.44 (time: 122s >> 96s / Gfpops: 1.64 >> 3.11)
2006-01-23 10:49:47 [SETI@home] Started upload of 17se04aa.6145.13312.253418.1.164_0_0
[...]


All results, which are finished afterwards, are calibrated immediately after they were finished:

2006-01-23 10:57:49 [SETI@home] Computation for result 20se04aa.18132.19233.1022148.1.213_0 finished
2006-01-23 10:57:49 [SETI@home] CC calibration: 28.13 >> 27.31 (time: 7582s >> 5988s / Gfpops: 1.64 >> 3.11)

Regards,
Carsten

As written already many times here and in the documentation, you have to complete a dozen or two of WU's before the calibration is efficient. Looking at your results, this is exactly what is happening in your case. The calibration is being done permanently and continuosly, but dispayed only once, hence the final result may be slightly different, and it is normal that it is being displayed again when you restart the client. There is nothing wrong about it.

---

trux
BOINC software
Freediving Team
Czech Republic

Hi,
I started to use the version tx36 on one of my computers and I've had some new problem: I'm running on this computer SETI, LHC and CPDN and four times I've received error

2006-01-23 13:24:36 [SETI@home] Result 17se04aa.6145.25137.848566.1.59_2 exited with zero status but no 'finished' file
2006-01-23 13:24:06 [SETI@home] If this happens repeatedly you may need to reset the project.

Two times on SETI and two times on LHC. On one of the slots folders I found sterr.txt which said "no heatbeat from core client for 31 seconds".
I am running on this computer also a 3D simulation program which can take 99% of CPU time for hours at a time. I believe this might be the reason for this error. Anyway I have run several previous versions of Boinc clients (standard and optimized but not calibrating) without seeing this problem before although I have been running the 3D simulation with them also.

So I think that this version is somehow more sensitive to other software using a lot of CPU power.

Harri Liljeroos

---

Hello trux,

The calibration is being done permanently and continuosly, but dispayed only once

Ok, that's the sentence I missed in the countless amount of information ...
Sorry for the inconvenience and thank you for giving some order to my confusion ... ;-)

But I'm still confused about the main reason for my first posting: As you mentioned many times, you cannot claim less credit with calibration than without calibration at SETI@home. But why *do* I claim less credit?

Regards,
Carsten

Result 17se04aa.6145.25137.848566.1.59_2 exited with zero status but no 'finished' file

Not sure if it has any relation to the modification in the client. I saw it at the LHC project on one of my machines that does not run the new core client, just 2-3 days ago.

---

trux
BOINC software
Freediving Team
Czech Republic