JoelC707 Premium Member join:2002-07-09 Lanett, AL |
to Nanoprobe3
Re: HCC GPU app close to going live IT HAS LAUNCHEDSweet! I've been noticing that for a while, even before the GPU crunching launched. It annoyed me but it hadn't gotten bad enough for me to actually go look for the solution. Thanks! |
|
glennaxl Premium Member join:2007-02-16 Zion
1 recommendation |
to Nanoprobe3
From: GTX 580 (x2) & GTX 570 to: HD 7970 (x2)
10/22/2012 0:012:00:55:21 936,639 4,057 = huge bump plus less electricity 10/21/2012 0:007:06:54:39 407,522 1,753 10/20/2012 0:008:19:35:32 198,845 815 10/19/2012 0:012:11:54:27 429,446 1,844 10/18/2012 0:008:01:09:27 424,621 1,851 10/17/2012 0:009:13:21:59 409,057 1,783 10/16/2012 0:010:20:15:02 445,146 1,915 10/15/2012 0:007:00:15:13 413,278 1,797 10/14/2012 0:010:11:31:11 455,032 1,945 10/13/2012 0:011:22:34:07 433,707 1,814 |
|
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Oct-22 8:38 pm
Nice jump! You were using nearly top of the line nvidia cards too so that just goes to show how much better AMD cards are at this. |
|
MacBridgerLate to the party Premium Member join:2001-01-11 Morgantown, WV |
Can you see when GPU units started? | How about here? |
Here is a quick look at my crunching stats. Due to the fact that I only have one, severely outdated PC, I am only crunching while my PC is not in use. 15 hours of crunching was reported for Oct 23. This is why I am so excited for GPU crunching. I'm actually able to benefit the cause again! |
|
MerinXCrunching for Cures Premium Member join:2011-02-03
1 recommendation |
MerinX
Premium Member
2012-Oct-25 1:18 pm
We may not all have AMD 7970s but we can crunch with what we got |
|
MacBridgerLate to the party Premium Member join:2001-01-11 Morgantown, WV |
20 minutes on my GT430 sure beats 28 hours on my P4. |
|
glennaxl Premium Member join:2007-02-16 Zion |
to Nanoprobe3
Looks like 12.11 beta is faster. When installing to Win 2008 R2 x64, App SDK can never be found in the install wizard. Solution: Just install the OpenCL.MSI manually. Its located here for x64: \Packages\Apps\OpenCL64\OpenCL.msi With that said, XP machines can probably do this. |
|
glennaxl |
to Nanoprobe3
After upgrading PC2 to 12.11 beta4 |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
What program are you using to show these stats? |
|
glennaxl Premium Member join:2007-02-16 Zion |
glennaxl
Premium Member
2012-Oct-28 5:15 pm
said by Nanoprobe3:What program are you using to show these stats? Im using BOINC Task then export them to Excel. |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
Gotcha. Was wondering what was giving you the average times. Been using BoincTasks for a while. Great program. I'll have to give the Excel export a go. Thanks |
|
Nanoprobe3
1 recommendation |
Just an FYI to everyone running the HCC GPU tasks. The newer WUs now contain 2 tasks each. If you notice your crunch times increasing the double sized WUs are the reason. Crunch on. |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Nov-13 6:27 pm
So this means people with modified app_info files to run 2 tasks at a time are now running 4 tasks at a time? If so, should they re-examine their GPU load and such to make sure they aren't overloading them? Either way I can see this being a better choice as I was frequently at the task limit per computer and it was constantly transferring data back and forth with how quickly it was churning through them. |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
said by JoelC707:So this means people with modified app_info files to run 2 tasks at a time are now running 4 tasks at a time? If so, should they re-examine their GPU load and such to make sure they aren't overloading them? Either way I can see this being a better choice as I was frequently at the task limit per computer and it was constantly transferring data back and forth with how quickly it was churning through them. All this change means is each WU contains twice the amount of work. It will not change the app_info settings. The task limit per GPU is now 4500 per day. Are you sure that's the problem? Try increasing your cache settings. |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Nov-14 10:50 am
Ahhh OK so it's not that it's running two tasks concurrently, but it's running double the amount of work. To be honest, I haven't looked at my cache settings in BOINC. I have noticed over the last while that when I look through the logs it tries to fetch work and gets told "no new tasks for project" and "this computer has reached the limit of in-progress tasks". Admittedly, this could have been fixed with this update as I rebooted from a power outage the other day and don't see any such messages in the logs now. |
|
MacBridgerLate to the party Premium Member join:2001-01-11 Morgantown, WV |
to Nanoprobe3
Have they recently changed the amount of work per GPU WU? I just noticed that it's taking me twice as long to crunch them. |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Nov-20 11:39 am
psst. Look two and four messages up from yours, this was just mentioned. |
|
MacBridgerLate to the party Premium Member join:2001-01-11 Morgantown, WV |
lmao. Ok. I must have still had some old WUs waiting to crunch because I just noticed it. |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Nov-20 1:31 pm
It was a fairly recent change so depending on how quickly you go through your cached WUs, it could take a few days to see the new ones. I haven't calculated mine against the new ones yet but with the old WUs, I calculated that if I shut off getting new tasks it would be out of GPU WUs in about 8 hours with a "full" task log (well, full as in what was shown in the task list).
In case you're wondering why I calculated that, I wanted to run out the saved WUs so I could reformat a computer and didn't want to just abort the tasks. I wanted to see how long I would have to wait before I could proceed. So I simply decided to do it the next day, set it to not get tasks before I went to bed and by morning it was sitting idle. |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
FYI! WCG HCC GPU app running low on work. » secure.worldcommunitygri ··· lastpost |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL |
JoelC707
Premium Member
2012-Nov-20 3:51 pm
Well, that would explain why I only have 19 pending GPU tasks on this computer. Honestly I expected this would happen. At the rate we are churning through them (and I mean all of us on WCG as a whole), it's no wonder they are running low. This isn't like CPU crunching where you have an hour or several turnaround time per WU and can keep a steady stream flowing in both directions. I'll bet all the uploading and downloading of the WUs in such short order is wreaking havoc on their bandwidth and storage system. They even mentioned 100 days of WUs is 5TB on their servers. I'm sure they have plenty of space to process all the completed WUs on their back-end systems, but they probably didn't plan for that much space being needed on front-end systems. I certainly wouldn't have expected that much but I'm also not working on the project either |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
I'm out of tasks on 4 of 6. On to something else for now. |
|
MacBridgerLate to the party Premium Member join:2001-01-11 Morgantown, WV |
to JoelC707
I keep a 3 day cache and only crunch when the PC is not in use. That gives it about 12-16 hours a day of crunching. So probably about 5 days to burn through my WUs. |
|
Nanoprobe3Looking for cures in memory of Mom Premium Member join:2003-05-11 Crab Nebula |
said by MacBridger:I keep a 3 day cache and only crunch when the PC is not in use. That gives it about 12-16 hours a day of crunching. So probably about 5 days to burn through my WUs. Hopefully they'll get things going again by then. |
|
Nanoprobe3 |
Latest update from the techs at WCG on the WU shortage. quote: Great news everyone. We are in the process of downloading new work from the researchers. Instead of sending out a trickle as they arrive because everyone is starving for work today. We will more than likely be able to release them and set the project back to full weight tomorrow. This is barring any delays of course. We will try to keep you up to date and informed of any changes that may happen.
Thank you for your patience, -Uplinger
|
|
Nanoprobe3
1 recommendation |
HCC GPU work is flowing again. |
|
spm9999Fluffy Farkletush Premium Member join:2003-11-17 Imperial, MO |
spm9999
Premium Member
2012-Nov-24 8:08 pm
Super!! My NVIDIA machine never ran out. Takes longer too for that one.. |
|
michieru Premium Member join:2009-07-25 Denver, CO |
to Nanoprobe3
Machine is on overdrive and using 400W of power. Keeps me nice and warm during the winter. Heater uses slightly more. |
|
MerinXCrunching for Cures Premium Member join:2011-02-03 |
MerinX
Premium Member
2012-Nov-25 9:54 pm
Damn i love the Watt Meter on the LCD on your PC case |
|
JoelC707 Premium Member join:2002-07-09 Lanett, AL
1 recommendation |
JoelC707
Premium Member
2012-Nov-25 10:19 pm
It says 396W in case it's hard to read |
Actually that's on an APC battery backup. Here's mine That's actually with two i5-3450 machines (one has a 5830, the other a 6870) and three 22" LCD monitors. |
|