I don't think a broad statement would fit what I'm seeing.

Here's the same puter with new wus added. There was another VLAR added.

The first shows the result ID and both claimed credit/hour and granted credit/hour. The second shows Runtime, and the third show Granted Credit. These are the same wus shown in the same order(angle range ascending).

A few days ago I got 86 from one of 1999 tape.

---

Mandtugai!

@ Tony, WinterKnight:

If there is a shortage of information about the crunching behaviour of these VLARs, I'd be happy to join in the data-gathering team. I accept the groundrules are: reputable FLOP-reporting and non-calibrating core client and standard science app [which rules my machines out - I run Chicken - but I can pull data off the site]. Any particular chip and/or OS you'd like me to focus on? And Tony - if you could email me a copy of your spreadsheet, it would save setting-up time! initial dot surname at btinternet dot com

With the long crunch times, at least these WUs aren't getting purged very quickly. Here's a sample I can get back to quickly:

(starting from John R. Bailey's report, which seems to be the highest so far)

WU ID Name AR Credit Notes
91885463 11oc99aa.8109.20369.473568.3.213 0.084308 97.9647474788906
91885467 11oc99aa.8109.20369.473568.3.214 0.084308 97.9657265744505
91885470 11oc99aa.8109.20369.473568.3.215 0.084308 97.9681285257911
91885474 11oc99aa.8109.20369.473568.3.216 0.084308 82.4133182629421 Crunch3r's app; overflow
91885477 11oc99aa.8109.20369.473568.3.217 0.084308 97.9656189667508
91885480 11oc99aa.8109.20369.473568.3.218 0.084308 97.9720774393417
91885483 11oc99aa.8109.20369.473568.3.219 0.084308 97.9647530924282
91885487 11oc99aa.8109.20369.473568.3.220 0.084308 97.9647514553096
91885497 11oc99aa.8109.20369.473568.3.222 0.084308 21.9231178266926 overflow
91885500 11oc99aa.8109.20369.473568.3.223 0.084308 97.9656612482088
91885503 11oc99aa.8109.20369.473568.3.224 0.084308 97.9656021139623
91885504 11oc99aa.8109.20369.473568.3.225 0.084308 97.9647524939259
91885506 11oc99aa.8109.20369.473568.3.226 0.084308 97.9691631604957
91885508 11oc99aa.8109.20369.473568.3.227 0.084308 97.9647503720266
91885510 11oc99aa.8109.20369.473568.3.228 0.084308 97.9647462314034
91885511 11oc99aa.8109.20369.473568.3.229 0.084308 97.9656601188633
91885513 11oc99aa.8109.20369.473568.3.230 0.084308 97.96475109786
91885515 11oc99aa.8109.20369.473568.3.231 0.084308 97.9647480885984
91885519 11oc99aa.8109.20369.473568.3.233 0.084308 97.9656592804328
91885521 11oc99aa.8109.20369.473568.3.234 0.084308 97.964750115581
91885525 11oc99aa.8109.20369.473568.3.236 0.084308 97.9647513023108
91885527 11oc99aa.8109.20369.473568.3.237 0.084308 97.9647489681672
91885531 11oc99aa.8109.20369.473568.3.239 0.084308 97.965602686409
91885532 11oc99aa.8109.20369.473568.3.240 0.084308 97.9657009391949
91885534 11oc99aa.8109.20369.473568.3.241 0.084308 97.964750406651
91885535 11oc99aa.8109.20369.473568.3.242 0.084308 22.4446559751174 overflow
91885536 11oc99aa.8109.20369.473568.3.243 0.084308 97.9695507437456
91885537 11oc99aa.8109.20369.473568.3.244 0.084308 97.9657234390202
91885540 11oc99aa.8109.20369.473568.3.245 0.084308 97.9656639252233
91885541 11oc99aa.8109.20369.473568.3.246 0.084308 97.9647523128166
91885544 11oc99aa.8109.20369.473568.3.247 0.084308 97.9647497596823
91885547 11oc99aa.8109.20369.473568.3.249 0.084308 97.964749789305
91885549 11oc99aa.8109.20369.473568.3.250 0.084308 97.9647521251545
91885551 11oc99aa.8109.20369.473568.3.251 0.084308 97.9647510521464
91885552 11oc99aa.8109.20369.473568.3.252 0.084308 97.9656998036845
91885554 11oc99aa.8109.20369.473568.3.253 0.084308 284.039507015991 Core client 5.3.1 overclaims??
91885556 11oc99aa.8109.20369.473568.3.254 0.084308 97.9656834451382
91885557 11oc99aa.8109.20369.473568.3.255 0.084308 97.9655783037985

richard, Email sent. Did I get the addy right?


[edit] you have even more mail

richard, Email sent. Did I get the addy right?

Yes, received and replied.

This WU is claiming 86.73 credits. The output of the file is below.

Jeremy


Work Unit Info
True angle range: 0.137247
Optimized Windows SETI@Home Enhanced application
Version info: Windows P4 SSE3 32-bit V5.15 'Chicken Good!' (R-1.3|+freq|xP+)
Compiled by Simon Zadra (KWSN - Chicken of Angnor) - Member of the Knights who say Ni! (http://www.kwsn.net)
Download Updates at: http://www.zadra.org/seti_enhanced/

CPU real speed: 2793 MHz

Work Unit Info
True angle range: 0.137247

Flopcounter: 22371081691161.437000

Spike count: 0
Pulse count: 1
Triplet count: 0
Gaussian count: 0
</stderr_txt>
]]>


Validate state Initial
Claimed credit 86.7397264645727

---

I don't think a broad statement would fit what I'm seeing.

Here's the same puter with new wus added. There was another VLAR added.

The first shows the result ID and both claimed credit/hour and granted credit/hour. The second shows Runtime, and the third show Granted Credit. These are the same wus shown in the same order(angle range ascending).

Yep...It does seem a little more random than I first thought. Too tired at the moment to take it all in though. (Toddler in the household)

---

I've gotten a few more VLARs of various lengths. It looks like they all are in the "moderately slow" category compared to the average unit. The fastest ones are the shortest-to-crunch, in general, with the longest angle ranges. The worst are the 0.7 to 1.0 degree and the average ones are the 0.39 ones. Basis is credits per hour. But I've noticed that some "half-hour" units give less credits per hour, even for identical angle ranges.

---

The highest credit I've gotten so far is a WU valued at 88.15 and so far of the 4 users 3 are using v5 (5.2.13, 5.4.11 & 5.5.0) Boinc client software, I have no idea what the 4th user is using yet of course.

http://setiathome.berkeley.edu/workunit.php?wuid=91880182

---

Savoir-Faire is everywhere!
The T1 Trust, T1 Class 4-4-4-4 #5550, America's First HST

I've gotten a few more VLARs of various lengths. It looks like they all are in the "moderately slow" category compared to the average unit. The fastest ones are the shortest-to-crunch, in general, with the longest angle ranges. The worst are the 0.7 to 1.0 degree and the average ones are the 0.39 ones. Basis is credits per hour. But I've noticed that some "half-hour" units give less credits per hour, even for identical angle ranges.

I agree that the 0.7 to 1.0 are worst possibly, my results from Beta just before it was released here on main site are shown here

Andy

http://setiathome.berkeley.edu/workunit.php?wuid=91883957

Check this one out:
My machine claimed 94.73
Another machine claimed a lot less, but had an error
A third machine has claimed 175.69

OMG!

I wonder how Berkeley will resolve this one in terms of how much credit will get awarded?

I suppose it will depend on the third valid result, and a canonical result will be taken as the middle value?

---

@ Richard, Tony:

If there is a shortage of information about the crunching behaviour of these VLARs, I'd be happy to join in the data-gathering team. I accept the groundrules are: reputable FLOP-reporting and non-calibrating core client and standard science app....

I'd like to join too, but my cruncher (notebook) is running at varying speed, thus disqualifying itself as a non-reputable FLOP-reporting source. I just recalled I stored once a rare VLAR with AR=00700245, lower than these in Tony's data set. I can send it per email for recrunching if interested.

Peter

@ Richard, Tony:

If there is a shortage of information about the crunching behaviour of these VLARs, I'd be happy to join in the data-gathering team. I accept the groundrules are: reputable FLOP-reporting and non-calibrating core client and standard science app....

I'd like to join too, but my cruncher (notebook) is running at varying speed, thus disqualifying itself as a non-reputable FLOP-reporting source. I just recalled I stored once a rare VLAR with AR=00700245, lower than these in Tony's data set. I can send it per email for recrunching if interested.

Thanks, Peter. I think there are two parts to the VLAR question.

1) Does the splitter/scheduler predict the crunch times accurately, or does it get it wrong, messing up cache sizes/deadlines/DCF for future WUs etc.? I think it gets it wrong.

2) Does the credit claim, based on reported FLOPs, make sense when put alongside the time spent crunching? It seems to me that the claims seem low, but whether this is because of difficult FLOPs (eg more divisions than usual), mis-counting FLOPs, a bug in code wasting time on something other than countable FLOPs, or anything else, I don't yet have a view.

Your variable speed would make it difficult to gather data re (1), but you might still be able to count a few FLOPs for us!

I've had a few email contacts off-board with Tony and others, and I've said I'm willing to help gather (and coordinate) data if it would help. However, I'm still waiting to hear if the project views this as an issue important enough to devote scarce programming time to, or whether it's just a minor oddity they can live with while they get on with more important matters. I would be reluctant to spend a lot of my time gathering data without some positive indication that it would be a help [rather than a nuisance!] to Berkeley.

Your variable speed would make it difficult to gather data re (1), but you might still be able to count a few FLOPs for us!

Yes, it would still be possible to gather at least the FLOPS/AR ratios with my mentioned host. Originally I was not thinking about this possibility, as we were then, few months ago on Beta, trying to find the AR/time/FLOPS/CC ratio for various HW and app versions.

Peter

I just completed one of those ULAR's (0.007), and they definitely seem to score lower on the credit/hr. than the VLAR's. Looks to be about the same as the VHAR's for the machine that ran it.

Alinator

got 97.97 for this unit.

angle range is : 0.0843


[edit] am i the winner? ;)

---

got 97.97 for this unit.

angle range is : 0.0843

[edit] am i the winner? ;)

I'll let you know when this WU finally gets validated! lol

http://setiathome.berkeley.edu/workunit.php?wuid=91883957

My machine claimed 94.73 (Quite a bit less than you)
Another machine claimed a lot less, but had an error
A third machine has claimed 175.69 (HEAPS more than yours, but once a canonical result has been determined, it could be closer to the amount claimed by my machine)

---

I just wonder what was so special on the 11th of October 1999... Something was, that's for sure... :)

---

"The Truth Is Out There!" (X-Files)

Alinator thanks for the compliment I checked your computer and I think it's a damn good 1 just needing a ram upgrade you'll notice better perfomance

Alinator thanks for the compliment I checked your computer and I think it's a damn good 1 just needing a ram upgrade you'll notice better perfomance

LOL, most of mine are "Old Timers", but they do all right for themselves. ;-)

You must be crunching part time or else keeping that Opty busy doing other stuff. You might want to give it some of Simon's "Chicken Soup" though. Mostly likely the optimized app would knock a healthy chunk off those crunch times. :-)

Alinator