Windows S5R4 App 6.06 available for Beta Test

Rudy
Rudy
Joined: 12 Dec 05
Posts: 33
Credit: 2956403
RAC: 7886

yep 1709839 Seeing this

yep

1709839

Seeing this thread, for fun I started up this machine that I keep around for old dos games, print server. etc.

Based on the current % complete it will complete a standard app task in about 12 days.

Waiting 2 weeks for a task result might however be a bit longer than Bernd wants to wait.

I "retired" this machine from crunching a few years ago since it was always missing deadline (on Seti) as my daughters gaming machine.

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

Next one on the AMD64 came in

Next one on the AMD64 came in with a similar time as the previous opti and should have the result from the XP Athlon soon

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109396060035
RAC: 35790180

In case people haven't

In case people haven't noticed, I'd like to draw attention to an edit that Bernd has placed in the opening message.

Quote:
[Edit] It looks like this App has a serious bug in the checkpointing (see below) Only use it if you are sure of what you are doing, e.g. keep the App in memory all the time through a task. [/Edit]

I've converted a number of "SSE" machines running Windows to the 6.06 app so I now have direct experience with the new checkpointing problem. Since I use the preference to "keep apps in memory when preempted" and the machine runs 24/7, there is mostly no problem with the fact that only the very first checkpoint for a task ever gets written. This is also true when the machine switches between projects, since once again the preempted task is kept in memory until the next time slice commences. It is fortunate that the task always residing in memory means that a checkpoint isn't usually going to be needed. If you don't have that preference set, any cycling between projects would not allow useful progress to be made on the EAH task as every time a new time slice commenced, the original (and very much out of date) checkpoint would always be reloaded.

There are at least two other situations in which this new bug will give problems. Firstly, if you don't run 24/7, every time you power off you will lose all progress, since the next startup will be right back at square one. Secondly if you have any type of crash or if you get one of those "zero exit status but no finished file" situations (as one of my machines does from time to time) you will lose all progress.

People basically have three choices:-

1. Revert to the previous app and wait for Bernd to release 6.07
2. Keep the app in memory and make sure you don't lose power or crash :-).
3. Make your checkpoint file "auto-update" by user intervention.

For the third option, if you look in the slots directory where the crunching is occurring, you will immediately see that the problem is characterised by the presence of a way-out-of-date checkpoint file (.cpt extension) and a very-up-to-date temporary file (.cpt.tmp extension) that is supposed to be replacing the .cpt file every time there is a new checkpoint available to be saved. If you delete the .cpt file, the next time a new checkpoint is ready, it will be able to be written as a file since there will be nothing to block the .tmp file from being renamed to the .cpt file. I'm not suggesting that anyone do this manually. It should be possible to have a fairly simple Windows batch file do it for you automatically at regular intervals.

BTW, I have actually confirmed that deleting the out-of-date checkpoint does allow a fresh up-to-date one to be written. So now my "zero exit status but no finished file" machine is actually making some progress again. It has so far clocked up 38 hours on a task that used to take less than 19 under the old app. It has just reached 67% finished and I've just "refreshed" its checkpoint file which was 2 hours out-of-date again. Of course, I could revert to the previous app to finish it off but that would take away all the fun of getting this task "over the line". If the OS were Linux, I'd write a shell script but I've got no desire whatsoever to do the same job in Windows since I'm completely ignorant about batch files. I don't even know if Windows has a sleep command - surely it must have, I guess.

Cheers,
Gary.

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

Ok AMD Athlon XP 2200+ just

Ok AMD Athlon XP 2200+ just finished the first task in just over 20 hours on 6.06. Will post the results of the second one when it comes in tomorrow.

Edit: I can't believe this pc returned a result before my wingman with the Core 2 Duo CPU T7300

BTW I also have a Duron that I could run the test on too if it's wanted.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109396060035
RAC: 35790180

RE: Ok AMD Athlon XP 2200+

Message 88719 in response to message 88718

Quote:
Ok AMD Athlon XP 2200+ just finished the first task in just over 20 hours on 6.06. Will post the results of the second one when it comes in tomorrow.


Athlon XPs were great CPUs in their day. The Thoroughbred or Barton cores overclock quite nicely too. I've got three Athlon XP hosts that have been converted to 6.06. I also have a little history on 6.04 so it should be possible to get a reasonable guesstimate for any speedup.

Quote:
Edit: I can't believe this pc returned a result before my wingman with the Core 2 Duo CPU T7300


If your wingman's core 2 host has a largish cache, you should always easily beat him - no surprise there :-).

Quote:
BTW I also have a Duron that I could run the test on too if it's wanted.


Without some crunching history on 6.04, how are you going to measure the approximate speedup?

Cheers,
Gary.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4267
Credit: 244926081
RAC: 16571

I issued a SSE2 part of a new

I issued a SSE2 part of a new App 6.07 that is intended to fix the checkpointing problem. However my possibilities for testing are very limited during the weeked, so I rely more or less on you for that. Please take a look and report.

BM

BM

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: RE: Edit: I can't

Message 88721 in response to message 88719

Quote:
Quote:
Edit: I can't believe this pc returned a result before my wingman with the Core 2 Duo CPU T7300

If your wingman's core 2 host has a largish cache, you should always easily beat him - no surprise there :-).

Quote:
BTW I also have a Duron that I could run the test on too if it's wanted.

Without some crunching history on 6.04, how are you going to measure the approximate speedup?


He only had one task, that's why I was surprised.

I'd run a few 6.04 and then the 6.06 on the Duron. I doubt they'll need the info though. Between just what myself and especially you (since you would be running more than I can, which btw I'm interested in seeing too) as well as any others out there could return on the XP's, I'm guessing will give them what they're looking for in reults. I just thought I'd offer.

Edit:

Quote:

I issued a SSE2 part of a new App 6.07 that is intended to fix the checkpointing problem. However my possibilities for testing are very limited during the weeked, so I rely more or less on you for that. Please take a look and report.

BM

OK, got it installed and ready to run. Will be switching tasks between the opti AP and the 6.07

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109396060035
RAC: 35790180

RE: I issued a SSE2 part of

Message 88722 in response to message 88720

Quote:
I issued a SSE2 part of a new App 6.07 that is intended to fix the checkpointing problem. However my possibilities for testing are very limited during the weeked, so I rely more or less on you for that. Please take a look and report.

Wouldn't you just know it!

Not long after my previous message I went to bed as it was close to midnight here. I didn't check the boards or else I'd have seen your message.

It's now close to 7.00AM and I've been up for about an hour. The new app is installed and crunching on the only suitable machine I have here at home - a Q6600 with 3 cores running E@H. The new app is creating checkpoints correctly and one of the tasks in progress will finish in about 20 mins.

I'll report the result in the 6.07 thread once it's been uploaded.

If you put up an SSE version, I have a few more machines at home (Coppermine PIII, Tualatin PIII, Athlon XP) that are currently showing the problem that I could use to test the fix more thoroughly.

Thanks for the quick fix.

Cheers,
Gary.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109396060035
RAC: 35790180

RE: He only had one task,

Message 88723 in response to message 88721

Quote:
He only had one task, that's why I was surprised.


I guess because he's probably running something like Seti as the main project with E@H as a very low resource share backup :-).

Quote:
I'd run a few 6.04 and then the 6.06 on the Duron.


Ahhh... That would work nicely! As you say, there is probably not much point as there will be plenty of others who could give a faster answer.

Although I'd long ago converted the bulk of my machines to Linux, there are probably around 10 SSE machines in my fleet still running Windows that will get upgraded to 6.07. Most of these had been recently upgraded to 6.06. I'll certainly report about these once there is something significant to report.

Cheers,
Gary.

RandyC
RandyC
Joined: 18 Jan 05
Posts: 6005
Credit: 111139797
RAC: 0

RE: RE: He only had one

Message 88724 in response to message 88723

Quote:
Quote:
He only had one task, that's why I was surprised.

I guess because he's probably running something like Seti as the main project with E@H as a very low resource share backup :-).

Quote:
I'd run a few 6.04 and then the 6.06 on the Duron.

Ahhh... That would work nicely! As you say, there is probably not much point as there will be plenty of others who could give a faster answer.

Although I'd long ago converted the bulk of my machines to Linux, there are probably around 10 SSE machines in my fleet still running Windows that will get upgraded to 6.07. Most of these had been recently upgraded to 6.06. I'll certainly report about these once there is something significant to report.

v6.07 is for SSE2 only.

Seti Classic Final Total: 11446 WU.

Comment viewing options

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