Countdown to HCC Completion Thread!

[jeanguy2]

Just got more GPU WU's; was enjoying silky smooth computer too XD

Something is skewed with your "WUs to be processed" cache. You should not run out of WUs in your cue for a short 2 to 6 hours server hiccup. There should be enough WUs in your cue(s) to go for 1.5 to 2 + days before exhausting your WUs to be processed cache(s). What are your present cache settings for that rig? (Min Work Buffer/Max Work Buffer)

Go to your WCG/My Grid/Results Status webpage and select the "In Progress" option for rig in question. (Post the number of Pages of in Progress?) Fifteen WUs are displayed per pages. A single AMD high end video card should have 200 to 250 pages while a Xfire setup should have 475 to 550 pages of in Progress WUs. If your numbers are low try changing your cache settings. I use 2.10 / 0.15 for all of my rigs which seems to give me the max cache possible without running into the "Max number of WUs to be processed has been reached" snag.

FYC,
Jean-Guy

[Red Maw]

Something is skewed with your "WUs to be processed" cache. You should not run out of WUs in your cue for a short 2 to 6 hours server hiccup. There should be enough WUs in your cue(s) to go for 1.5 to 2 + days before exhausting your WUs to be processed cache(s). What are your present cache settings for that rig? (Min Work Buffer/Max Work Buffer)

Go to your WCG/My Grid/Results Status webpage and select the "In Progress" option for rig in question. (Post the number of Pages of in Progress?) Fifteen WUs are displayed per pages. A single AMD high end video card should have 200 to 250 pages while a Xfire setup should have 475 to 550 pages of in Progress WUs. If your numbers are low try changing your cache settings. I use 2.10 / 0.15 for all of my rigs which seems to give me the max cache possible without running into the "Max number of WUs to be processed has been reached" snag.

FYC,
Jean-Guy

I am running a 0 day cache to avoid an issue where WU's were getting returned late when running CPU with intermittent GPU.

I have never actually run out of WU's on with a 0 day cache (yay old HW XD) so with HCC GPU project coming to a close soon it isn't worth fixing for me.

[0ne.shot]

I set my cache to 2.10/.15 and ended up with 500 pages of In Progress WUs. If anyone wants to start hording WUs as the project winds down, here is your chance thanks to the above post by JG2.

[Johnmark]

Something is skewed with your "WUs to be processed" cache. You should not run out of WUs in your cue for a short 2 to 6 hours server hiccup. There should be enough WUs in your cue(s) to go for 1.5 to 2 + days before exhausting your WUs to be processed cache(s). What are your present cache settings for that rig? (Min Work Buffer/Max Work Buffer)

Go to your WCG/My Grid/Results Status webpage and select the "In Progress" option for rig in question. (Post the number of Pages of in Progress?) Fifteen WUs are displayed per pages. A single AMD high end video card should have 200 to 250 pages while a Xfire setup should have 475 to 550 pages of in Progress WUs. If your numbers are low try changing your cache settings. I use 2.10 / 0.15 for all of my rigs which seems to give me the max cache possible without running into the "Max number of WUs to be processed has been reached" snag.


FYC,
Jean-Guy

Strange Jean, as on all my 79xx machines if I go above 1 day min. (regardless of max.) my running w.u.'s go into "running high priority" mode. Which needless to say causes a lot of havoc on GPU W.U.'s.
On the 7770 setup's any combination on min/max does not cause this issue?

[Mumak]

I recently run into a similar problem. Have set my cache to 0.10 / 0.10 and now I have 501 pages of In Progress. The client is responding very slow now and some WUs are getting stuck after a while (don't know if this is because of so many WUs in cache). Has anyone experienced stuck WUs? I've been running without issues so far with 7950+7770, after replacing the 7770 with another 7950 such problems start to appear...

[jeanguy2]

Strange Jean, as on all my 79xx machines if I go above 1 day min. (regardless of max.) my running w.u.'s go into "running high priority" mode. Which needless to say causes a lot of havoc on GPU W.U.'s.
On the 7770 setup's any combination on min/max does not cause this issue?

Strange one JM, my 7970 & 7950 Xfire rigs are both using BOINC 7.0.56. The 7970 runs CAT 13.1 and the 7950 CAT 12.11 beta 4. Both of these rigs profile preferences are set to disallow CPU WUs. Confirm that your rigs have the same CPU setting. How many pending WUs do you have in your cache at present?

[jeanguy2]

I recently run into a similar problem. Have set my cache to 0.10 / 0.10 and now I have 501 pages of In Progress. The client is responding very slow now and some WUs are getting stuck after a while (don't know if this is because of so many WUs in cache). Has anyone experienced stuck WUs? I've been running without issues so far with 7950+7770, after replacing the 7770 with another 7950 such problems start to appear...

Mumak, Suggest that you do a full uninstall of AMD CAT drivers and then re-install them. Select Xfire upon restart and then re-start BOINC. I also once experienced sluggish BOINC and windows 7 performance when I added Xfire card without doing full driver re-install. Are you getting an error readout with your stuck WUs? If so what is it?

[Mumak]

Mumak, Suggest that you do a full uninstall of AMD CAT drivers and then re-install them. Select Xfire upon restart and then re-start BOINC. I also once experienced sluggish BOINC and windows 7 performance when I added Xfire card without doing full driver re-install. Are you getting an error readout with your stuck WUs? If so what is it?

I did a complete CAT reinstall when I put the new card in, also enabled Xfire. Now I set it to No New Tasks, performed another reinstall + reboot, will see... Decreased from 16 WUs/GPU to 12/GPU. There are few errors per day of all sorts like "Invalid kernel name", "Kernel execution time estimate too high, exiting.", "Extracting GLCM features... ".
What do you think, is it better to crunch in Xfire or not ? Not sure how Xfire is used for computing, but WUs are assigned to both GPUs in this config anyway, so theoretically it should not be needed to run in Xfire.

[Johnmark]

Strange one JM, my 7970 & 7950 Xfire rigs are both using BOINC 7.0.56. The 7970 runs CAT 13.1 and the 7950 CAT 12.11 beta 4. Both of these rigs profile preferences are set to disallow CPU WUs. Confirm that your rigs have the same CPU setting. How many pending WUs do you have in your cache at present?

Noticed this behavior many weeks ago and figured it was a normal glitch(for GPU wu's). Didn't think anything of it until your post, so I changed my config. to match yours and presto back to running high priority units.
The three machines I'm referring to all have Cat. 13.1 using boinc vr. .042/.054
WU's in progress is between 400-500 pages depending on machine. Pending validation is between 100-165 pages per machine. In the end there is lots of work in the Que., so all is well.

BTW what is the max. number of wu's that are given out these days?

[OldChap]

I don't think they changed it after the shutdown to upgrade their software because the previous limit was 1200 and I have >2000 just now on this slow rig. (Does around 1500/day)

I never did determine whether that change for the outage was max 5000 wu's or 7500 wu's because their settings were:

<daily_result_quota>500</daily_result_quota>cpu work
<gpu_multiplier>15</gpu_multiplier>GPU work This would equate to 7500 per gpu
<initial_daily_result_quota>5</initial_daily_result_quota>
<max_wus_to_send>50</max_wus_to_send>per download tranch
<max_wus_in_progress>500</max_wus_in_progress>cpu work
<max_wus_in_progress_gpu>5000</max_wus_in_progress_gpu>GPU workThis would be 5000

Your rig will calculate the time per wu then give you the number wu's in a day multiplied by the number you have entered as a minimum cache .... or at least that is the theory.

I started incrementing by 0.1 today in order that I see where the max is before getting "priority" issues and having the cache as full as possible before the end of HCC.

[Johnmark]

Thanks Old Chap. I was playing around a bit today and noticed on my 7970 rig that as soon as I went to .60/10 from .50/10 a couple of wu's went to high priority. Put it back to .5 min and priority is normal. Counted 6480 wu's in the Que on the 7970 rig.

Also noted that on two identical 2600k platforms both running at 4600 Mhz as well, one using the Matrix 7970(1230Mhz) and the other an Asus 7950(1100Mhz) there was a vast difference in CPU usage that I can not understand(both running 24 wu's).

Made sure both are running the the newest vr. of Boinc and using the same video driver, and not running other background tasks, then started boinc creating an offset between wu's. Machine #1 uses ~70% cpu while machine #2 is ~28% when no wu's are starting/finishing.

So I look at task manager to see who is using up the CPU and notice that machine #1 is using 1-3% per thread up to 12% when wu's are transitioning. Machine #2 on the other hand seems to be barely working a sweat 0-1% per thread and up to 8% when transitioning.
Strange

[OldChap]

My settings just now are minimum work buffer 1.5 days and additional work buffer 0.1 days for a total of 2835

Are these truly identical rigs?....same everything?

If you have no wu's in starting/finishing phase at any one time then you maybe need to run more concurrent wu's

Show your app_config please

[Johnmark]

I just find it odd there is such a difference between the two machines. Yes they are identical (except psu), although something is different and the only thing I haven't really looked into is the BIOS.

The wu's will be bunching up real soon , I did one run on each machine for test purpose's before they bunch up.

<app_config>
<app>
<name>hcc1</name>
<max_concurrent>24</max_concurrent>
<gpu_versions>
<gpu_usage>0.0416</gpu_usage>
<cpu_usage>0.3333</cpu_usage>
</gpu_versions>
</app>
</app_config>

[Gandalf]

Unless the server is down, I think WCG is out of HCC1 GPU tasks. One of my computers can't Update and get any more.

[0ne.shot]

Unless the server is down, I think WCG is out of HCC1 GPU tasks. One of my computers can't Update and get any more.

There have been several issues in the past few days with the servers giving out WUs. I hope this is the case this time!

[Gandalf]

I've been preparing for this by adding tasks from other Projects on a minimum resource basis. I just didn't think I'd be using these preparations quite so soon.

[nanoprobe]

I've seen couple of those http server error messages today but they all seemed to have cleared up and I have full caches on all machines ATM.

[Sparky]

My queue is full again. Reminds of when you get to the last bit of ketchup in the bottle and it sputters a bit.

[nanoprobe]

My queue is full again. Reminds of when you get to the last bit of ketchup in the bottle and it sputters a bit.

When that happens I usually store the bottle upside down to get it all to the top. Wonder if that works with 'puters?

[0ne.shot]

When that happens I usually store the bottle upside down to get it all to the top. Wonder if that works with 'puters?

I flipped my case upside down. Now we wait to see if it works the same as Ketchup! Only time will tell.

[Gandalf]

When that happens I usually store the bottle upside down to get it all to the top. Wonder if that works with 'puters?

I did that with my last DIY computer and found the screw driver I thought I'd lost.

[0ne.shot]

Time is winding down. There are an estimated 8 days of new WUs remaining. Time to crank everything up to 12.

[Rob_B]

Just 8 days? Lets see if that is enough.

[bluestang]

Well, with summer coming I was going to have to tone it down some anyways. I have a large amount of photos I need to get cracking on editing and organizing too. Pretty good timing. Maybe the "new project" will start just in time for the cooler months.

[0ne.shot]

Well, with summer coming I was going to have to tone it down some anyways. I have a large amount of photos I need to get cracking on editing and organizing too. Pretty good timing. Maybe the "new project" will start just in time for the cooler months.

Oh, I can do the photo thing for you. In fact, I'll take the 300mm f/4 off your hands for no charge.