Are the bionc preferences recognised across the board. i.e., if I change them on one system, will they affect the others?

Preferences set at the website, yes, except on those systems running a recent BOINC client that have local preferences turned on, or those that are otherwise overriding the preferences. And, of course, those that are in a different “venue” from that for which the prefs were set.

Your BOINC general preferences can be updated at any project website where you have hosts attached; they’ll eventually propagate to all your other projects.

Thanks Odysseus.

Regards

PJ

---

Hi,

Oh, and the main difference - the chicken apps are faster. Much faster.

Unfortunately, not on my system ... ;-)

Two new results:

5.27 stock app:
result #595481563, AR 0.386386, CPU time: 26459 secs

KWSN_2.4V_MB_SSE (the new one downloaded from Crunch3r's site):
result #595481573, AR 0.386386, CPU time: 26863 secs

Regards,
Carsten

Well, there's a challenge.

Unusually for me, I can't substantiate the claim I made that the Chicken/Crunch3r apps are faster at the moment. I based my claim on (a) experince with Chicken 2.2B compared with the equivalent stock 5.17: and (b) Crunch3r's claim that he had been able to find an extra 5% from somewhere.

So I'm going to try and check it out. Two new Q6600s have just taken up residence in my shrubbery. Identical hardware, both running for the purposes of testing at stock speed, software environment as near identical as I can manage (both running 32b XP Home).

Host 3751792 started running the stock Berkeley 5.27 last night.
Host 3755243 started about an hour ago, and had Crunch3r's 2.4v for SSSE3 (the 26 August build) running from the very start.

I'm going to chart the speed comparison by angle range, and we'll see point by point which app works best where (provided I can get a range of ARs, that is - they all seem to be VLAR shorties today).

Once I've got a reasonable range of ARs, I'll switch one over to Chicken 2.4, and the other to Chicken 2.2B, so we'll see how they compare too. [Before anyone complains, Joe Segur has prepared a binary-hacked copy of 2.2B for testing purposes: it'll do the crunching at exactly the same speed as the old 2.2B, but it has the current multiplier so the credit claims will be directly comparable to the current apps].

We'll wait and see. Watch this space.

Hi,

Oh, and the main difference - the chicken apps are faster. Much faster.

Unfortunately, not on my system ... ;-)

Two new results:

5.27 stock app:
result #595481563, AR 0.386386, CPU time: 26459 secs

KWSN_2.4V_MB_SSE (the new one downloaded from Crunch3r's site):
result #595481573, AR 0.386386, CPU time: 26863 secs

Regards,
Carsten

Well, there's a challenge.

Unusually for me, I can't substantiate the claim I made that the Chicken/Crunch3r apps are faster at the moment. I based my claim on (a) experince with Chicken 2.2B compared with the equivalent stock 5.17: and (b) Crunch3r's claim that he had been able to find an extra 5% from somewhere.

So I'm going to try and check it out. Two new Q6600s have just taken up residence in my shrubbery. Identical hardware, both running for the purposes of testing at stock speed, software environment as near identical as I can manage (both running 32b XP Home).

Host 3751792 started running the stock Berkeley 5.27 last night.
Host 3755243 started about an hour ago, and had Crunch3r's 2.4v for SSSE3 (the 26 August build) running from the very start.

I'm going to chart the speed comparison by angle range, and we'll see point by point which app works best where (provided I can get a range of ARs, that is - they all seem to be VLAR shorties today).

Once I've got a reasonable range of ARs, I'll switch one over to Chicken 2.4, and the other to Chicken 2.2B, so we'll see how they compare too. [Before anyone complains, Joe Segur has prepared a binary-hacked copy of 2.2B for testing purposes: it'll do the crunching at exactly the same speed as the old 2.2B, but it has the current multiplier so the credit claims will be directly comparable to the current apps].

We'll wait and see. Watch this space.

Verrry interesting. I currently have a test run going on my x64 quad rig comparing all of the most recent non GFX builds by crunch3r against the 2.2b as reference.
Will post it in the pre release test section of Simon's site when it is complete.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

Verrry interesting. I currently have a test run going on my x64 quad rig comparing all of the most recent non GFX builds by crunch3r against the 2.2b as reference.
Will post it in the pre release test section of Simon's site when it is complete.

Please keep me posted. Joe has nominated me for access to the pre-release area, but unfortunately he can't upgrade me himself - it has to wait until Simon gets back.

It might be useful if all the testers - you, me, Tony, Alinator, and probably others - agreed amongst ourselves about the best way of graphing the results, and came up with some sort of common standard.

Verrry interesting. I currently have a test run going on my x64 quad rig comparing all of the most recent non GFX builds by crunch3r against the 2.2b as reference.
Will post it in the pre release test section of Simon's site when it is complete.

Please keep me posted. Joe has nominated me for access to the pre-release area, but unfortunately he can't upgrade me himself - it has to wait until Simon gets back.

It might be useful if all the testers - you, me, Tony, Alinator, and probably others - agreed amongst ourselves about the best way of graphing the results, and came up with some sort of common standard.

I'll post the results in the general access Windows area as well, so you can have a look see. Simon had posted something about a result database for test results, but I am not sure how far that has progressed yet.

EDIT....I don't think anybody objects to me posting the test results in the general access area, do they?

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

OK... So a glib comment is that I would expect the optimised 'no graphics' apps to be faster than the Berkeley stock app if only simply for there being no graphics overheads.

Aside: Will Berkeley be doing a native linux 64bit app rather than sending out the 32bit app to 64bit systems?

Keep up the good hackings!

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

A snippet from Mr. Segur in another thread.

OTOH, the Beta Applications page shows both 5.21 and 5.27 for Linux/x86_64 and both those builds have the optimizations.

Now I don't know if these are truly 64b linux or 32b wannabees.

full text from whence the snippet comes

OTOH, the Beta Applications page shows both 5.21 and 5.27 for Linux/x86_64 and both those builds have the optimizations.

Now I don't know if these are truly 64b linux or 32b wannabees.

Well, for a 64b Linux system, I'm seeing:

<core_client_version>5.8.15</core_client_version>
<![CDATA[
<message>
process exited with code 22 (0x16)
</message>
<stderr_txt>
execv: No such file or directory

</stderr_txt>
]]>

Which is suggestive that it is a 32bit app trying to access non-existant libraries on the all 64bit Linux.

(I could add the 32bit libraries but would prefer to see fully 64bit apps on there.)


Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

you did see where his snippet says "beta" didn't you? I wasn't aware they were back in bidness yet. Hmmm, got some checking to do.

Richard, Astro, Alinator.....
Have finished my test run for the recent crunch3r app builds, and the results have been posted over on Lunatics. Posted a copy in the general access Windows thread as well.
Let me know what you think, it's late and I will not take the time to sort through what the results were until tomorrow.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

Hi,

Oh, and the main difference - the chicken apps are faster. Much faster.

Unfortunately, not on my system ... ;-)

Two new results:

5.27 stock app:
result #595481563, AR 0.386386, CPU time: 26459 secs

KWSN_2.4V_MB_SSE (the new one downloaded from Crunch3r's site):
result #595481573, AR 0.386386, CPU time: 26863 secs

Regards,
Carsten

Well, there's a challenge.

Unusually for me, I can't substantiate the claim I made that the Chicken/Crunch3r apps are faster at the moment. I based my claim on (a) experince with Chicken 2.2B compared with the equivalent stock 5.17: and (b) Crunch3r's claim that he had been able to find an extra 5% from somewhere.

So I'm going to try and check it out. Two new Q6600s have just taken up residence in my shrubbery. Identical hardware, both running for the purposes of testing at stock speed, software environment as near identical as I can manage (both running 32b XP Home).

Host 3751792 started running the stock Berkeley 5.27 last night.
Host 3755243 started about an hour ago, and had Crunch3r's 2.4v for SSSE3 (the 26 August build) running from the very start.

I'm going to chart the speed comparison by angle range, and we'll see point by point which app works best where (provided I can get a range of ARs, that is - they all seem to be VLAR shorties today).

Once I've got a reasonable range of ARs, I'll switch one over to Chicken 2.4, and the other to Chicken 2.2B, so we'll see how they compare too. [Before anyone complains, Joe Segur has prepared a binary-hacked copy of 2.2B for testing purposes: it'll do the crunching at exactly the same speed as the old 2.2B, but it has the current multiplier so the credit claims will be directly comparable to the current apps].

We'll wait and see. Watch this space.

Replying to myself again, but it's the easiest way to bring new readers up to speed.

I ran the new quaddies over the long weekend. Unfortunately, the work being split was very repetitive (lots of WUs with very nearly the same Angle Range), but I suppose that gives added confidence to the results. Anyway, with 192 results in for the stock 32b Windows app, and 263 results for Crunch3r's 2.4v SSSE3 32b Windows app (the 26 August version), here are the results:


(direct link)

I think my original assertion is vindicated: the optimised apps (Crunch3r's, in this case) are much faster.

The sharp-eyed amongst you will have noticed that I've switched the machines over, and I'm now data-gathering for the Chicken apps - both 2.4 and 2.2B. Or I will be as soon as I can get some ****** WUs. More details later. NB I'm using Joe's hacked 2.2B so I won't screw up the credit claims.

I'm also building up the credit / AR graph nicely. This contains more data points (944, if my maths is accurate) - without timing information, I can combine logs from different machines. Here it is so far:


(direct link)

Any resemblance to Joe's theoretical graph is purely coincidental! Extra data points from the Chicken timing runs will also be added to this graph.

Richard, how are you getting the AR, CPU, and credit information to your graphing application.

As it happens I was building very similar graphs, but I'm hand harvesting the three numbers from each result page on the web site, which is infernally labor intensive, and risks missing some data at the shorter recent retention times.

Come to think of it, I should probably use the BOINCView log as my data source.

---

archae86, I don't know if he still uses it, but sent him my excel file, and he amended it to his needs (I guess). You can get the idea about how/with what by reading this post and the next couple from me to get the idea. PM me and I'll send you what I have. I liked his comparison chart and requested he send it to me....I see I have a PM. have to go check it out.

Richard, how are you getting the AR, CPU, and credit information to your graphing application.

As it happens I was building very similar graphs, but I'm hand harvesting the three numbers from each result page on the web site, which is infernally labor intensive, and risks missing some data at the shorter recent retention times.

Come to think of it, I should probably use the BOINCView log as my data source.

The 'graphing application' is M$ Excel.

I copy and paste in a whole page of 20 results, then click on each result in turn to open the result screen to get the AR and check for -9s. That's the only manual bit - if anyone knows how to get AR out of a standard logging tool, I'm sure we'd all be grateful.

PM me with an email adress if you'd like a copy of the spreadsheet (with or without current data)

if anyone knows how to get AR out of a standard logging tool, I'm sure we'd all be grateful.

Not BoincView, as you doubtless knew. Probably I confounded it with a memory of logs left by a management tool I used in the classic days. I distinctly recall building a graph or two covering really large numbers of result, sometime quite a while ago.

---

I've just started using KWSN_2.4V_SSE_MB_GFX.

On 1.45xxxx to 1.48xxxx AR it is slower! I saw this first yesterday with other tasks running. Did 2 more WUs overnight with slightly better results, but not as fast as stock.

11mr07ad.13682.12751.3.5.177_1
http://setiathome.berkeley.edu/result.php?resultid=603887054
CPU time 8885.9375
<core_client_version>5.10.7</core_client_version>
<![CDATA[
<stderr_txt>
setiathome_enhanced 5.27 DevC++/MinGW

Work Unit Info:
...............
WU true angle range is : 1.486741
Optimal function choices:
-----------------------------------------------------
name
-----------------------------------------------------
v_BaseLineSmooth (no other)
v_vGetPowerSpectrum 0.00265 0.00000
sse1_ChirpData_ak 0.02506 0.00000
v_vTranspose4x16ntw 0.01513 0.00000
BH SSE folding 0.00252 0.00000
Restarted at 38.80 percent.
Restarted at 78.06 percent.

Flopcounter: 5851456053899.197300

Spike count: 2
Pulse count: 0
Triplet count: 0
Gaussian count: 0
Claimed credit 19.3181131407555
Granted credit 19.3129513050006

11mr07ac.16034.154445.14.5.34_1
http://setiathome.berkeley.edu/result.php?resultid=603887056
CPU time 13811.3125
stderr out <core_client_version>5.10.7</core_client_version>
<![CDATA[
<stderr_txt>
Optimized SETI@Home Enhanced application
Optimizers: Ben Herndon, Josef Segur, Alex Kan, Simon Zadra
Version: Windows SSE 32-bit based on S@H V5.15 'Noo? No - Ni!'
Revision: R-2.4V|xK|FFT:IPP_SSE|Ben-Joe
CPUID: AMD Athlon(tm) XP 2200+
Speed: 1 x 1798 MHz
Cache: L1=64K L2=256K
Features: MMX SSE

Work Unit Info
WU Credit multi. is: 2.85
WU True angle range: 1.453009
Restarted at 26.83 percent.
Restarted at 47.00 percent.
Restarted at 69.53 percent.
Restarted at 78.31 percent.
Restarted at 97.34 percent.

Spikes Pulses Triplets Gaussians Flops
4 0 1 0 5854098477014
Claimed credit 19.3268294670031
Granted credit 19.3268294670031

11mr07ac.12699.159762.11.5.25_1
CPU time 12354.265625
stderr out <core_client_version>5.10.7</core_client_version>
<![CDATA[
<stderr_txt>
Optimized SETI@Home Enhanced application
Optimizers: Ben Herndon, Josef Segur, Alex Kan, Simon Zadra
Version: Windows SSE 32-bit based on S@H V5.15 'Noo? No - Ni!'
Revision: R-2.4V|xK|FFT:IPP_SSE|Ben-Joe
CPUID: AMD Athlon(tm) XP 2200+
Speed: 1 x 1799 MHz
Cache: L1=64K L2=256K
Features: MMX SSE

Work Unit Info
WU Credit multi. is: 2.85
WU True angle range: 1.486794
Restarted at 68.06 percent.

Spikes Pulses Triplets Gaussians Flops
2 0 0 0 5849842869343
Claimed credit 19.3127918722553
Granted credit 19.3127918722553

[edit]
http://setiathome.berkeley.edu/results.php?hostid=3001110

---

Sir Arthur C Clarke 1917-2008

I'm seeing the same thing with Simon's 64b optimized vs 32b stock. The stock is faster on the high angle ranges (above 1.2). You should compare at other angle ranges and see what you get. See the high angle range on the charts in this post

I'm seeing the same thing with Simon's 64b optimized vs 32b stock. The stock is faster on the high angle ranges (above 1.2). You should compare at other angle ranges and see what you get. See the high angle range on the charts in this post

Thanks for the confirmation Tony.

I'm not knocking the Chicken apps, I've seen great results with 2.2b but there does seem to be a problem with the latest apps.

I currently have 2 more high AR workunits to crunch before I get to one with <true_angle_range>0.0085430016715714</true_angle_range>. I will stay with this app for now, until I see some comments from Joe or Simon or Crunch3r

I noticed that we are both using AMD processors, but I did not think the patching issue applied to SSE, only SSE2 and SSE3.

---

Sir Arthur C Clarke 1917-2008

It's not a cut and dry thing. To be sure of anything, many samples must be taken. For example on my AMD64 3700 sandiego, I've been running stock 5.27 (32b) and KWSN_2.4_SSE2-AMD_MB(32b). I am so far seeing Simon's as faster, but more samples are needed (see highlighted areas for comparison)



Actually the right hand half of the larger ellipse can't be compared really either, but the left 1/2 is dang close)

I also see two comparable points on my AMD64 X2 4800 which is running the same apps as my 3700, and simon's "seems" to be faster on that too, but I'll save everyone the bandwidth for the chart until more samples are recorded.