HCC GPU Launched.

[0ne.shot]

And it stopped again.

But now I figured it out. I run this cruncher without monitor and keyboard and use Remote Desktop to contol it. And as soon as I connect to it, it loses the GPU. Even GPU-z doesn't show all info on the sensor tab.

Now without RDC connected I can see it working fine in BoincTasks. So, after all.

I monitor my other cruncher that way when I'm too lazy to walk to the other room. I noticed it says GPU is missing when I connect. When I actually walk over to it and check, the GPU is back. I assumed it was Remote Desktop causing the issue.

[nanoprobe]

Ok, I adjusted my 2D clocks to never go below 600/800 and I'm hoping that solves my OC issues I was having OCing this card. I think the low 2D idle state of 300/150 are just to low and cause problems on some systems.

So, I just upped my 7870 from stock clocks 1050/1250 to 1200/1400. Going to see what diff it makes. Max temps only go up 1 of 2C so that's no biggy and wattage is up 7-8W more than stock. Looks like I'll be able to shave 10-15s off depending on WU.

EDIT: MF'er. I guess that didn't work either, card crashed again. I can't get this card to OC for nothing ):

EDIT 2: I may just RMA this card and get 2x 7770 instead.

Just an FYI. Overclocking your GPU ram will have negligible if any effect with performance on this app. The mobo memory handles the load. Overclocking that memory will help, just don't try to overdue it. And FWIW I downclock my GPU memory by at least 50% and there is no measurable change in crunch times. What that will do is save you a bunch of power and lower your temps.

[Snow Crash]

RDP bad ... VNC good :-)

[pirogue]

10/15/2012 0:000:17:36:03 104,456 458
10/14/2012 0:001:12:48:41 208,592 921
10/13/2012 0:001:10:54:03 181,188 806

A little OC tweaking an adding the app_info made a bit of a difference.
10/16/2012 0:001:18:23:44 322,309 1,415
<added app_info + a little OC>
10/15/2012 0:001:06:56:13 222,159 972
10/14/2012 0:001:12:48:41 208,592 921
10/13/2012 0:001:10:54:03 181,188 806

[fallwind]

^^same here!

Before: stock BOINC settings(1 GPU task at time), stock GTX 460 at 675MHz= 10 WU/hour, 56000 points/day.

After: added app_info to run 3 tasks simultaneously and OC'd the GPU to 850MHz= 16 WU/hour, 88,000 points/day.

2 problems though: 1) I had to exclude my display GPU from use, the lag was unbearable. I'll let it run while I'm sleeping. Secondary GPU runs 24/7. 2) I missed the courier truck with my shiny new 7870 in it....lol!!

[Mydog]

Finally got the app_info to work on my 3770K @4,7GHz + HD7970 @stock rig running 4 tasks simultaneously and by the looks of it each WU takes 130 sec at most.
If I calculate 2.5 minute on each I then get 160 WU/hour *24*30 points(average looks more like 35) that's 115,200 Boinc PPD or 806,400 WCG PPD from one rig pulling 230W


Edit:
Just checked the Pending Validation on that rig and it has gone up from 22 pages on hour ago to 29 pages now, 7 pages in on hour and by the looks of it so has the Valid ones.

[bluestang]

MyDog: Is that 4 GPU tasks + 4 CPU tasks your running or strictly 4 GPU only? Care to share that app_info...I'd like to try 2 simultaneous on my 7870.

[Conquistador SW]

Finally got the app_info to work on my 3770K @4,7GHz + HD7970 @stock rig running 4 tasks simultaneously and by the looks of it each WU takes 130 sec at most.
If I calculate 2.5 minute on each I then get 160 WU/hour *24*30 points(average looks more like 35) that's 115,200 Boinc PPD or 806,400 WCG PPD from one rig pulling 230W


Edit:
Just checked the Pending Validation on that rig and it has gone up from 22 pages on hour ago to 29 pages now, 7 pages in on hour and by the looks of it so has the Valid ones.

Wow, that is crazy!!!

Are you sure you calculated that correctly though?
Because if 1 WU takes 2.5 minutes, that is 24 per hour. 24 x 4 = 96 WU's per hour with 4 threads. 96 x 24 hours is 2304 WU's per day.
This would be ~69000 boinc points per day or about 480,000 WCG point per day, which is still a crazy amount, but not as much as you predicted.

[Conquistador SW]

I fired up my old baby: E6600 at 3.2GHz + 5870 at 900/1200.

I got these results only running GPU WU's:


I calculated the average elapsed time for these WU's and came to 223 sec/WU = 16 WU's per hour.
16 x 24 = 384 WU's per day
384 x 33 points = 12672 boinc points/day = 88700 WCG points per day.

Not bad for this old machine

[Mydog]

This is what I got running 4 GPU WU's simultaneously

[bluestang]

10/17/2012 0:003:16:07:01 112,883 429 -- so far today
10/16/2012 0:006:12:47:23 215,444 831
10/15/2012 0:007:17:35:24 214,910 781
10/14/2012 0:003:13:31:42 172,025 695

So, I was getting 40K WCG PPD previously running 8 GFAM tasks.

Now, I should do about 215K WCG PPD on a machine running 7 GFAM + 1 HCC GPU.

+175K WCG PPD for the 7870 (and I've had an hour or 2 each day down time).

[Rob_B]

Looks like they raised the number of WU you can have on a cruncher. Up till today I had a max of 60 WU and that is now 100 for both GPU crunchers that I have. That is still only a little over 2 hour cached on my 7770.

[OldChap]

yeah, I wish I had noticed that limit before I set my normal internet use hours

[Mydog]

Hmm just checked on my 7970 Cruncher and it had been stuck in 40 mins on the same four WU's(stuck on the CPU part of the work on all the WU's), restarted Boinc and now it's running normal again.
I've seen other people have that problem so just wondering if it's a app_data or WU problem, anyone here have a clue?

[Rob_B]

I had the same problem about 2 hours ago. It had been running fine all day with 1 WU. But after 1 or 2 hours running 2 WU's it didn't finish the WU's and was stuck for over one hour.

I'm back to 1 WU now. Just to be save and keep crunching.

[Mydog]

Then I need to keep an eye on all rigs running the app_data script I guess. Anyone of you guys using BoincTasks know if it can produce a web site displaying progress?
I remember doing that running F@H in a similar program to BoincTasks.

[pirogue]

Hmm just checked on my 7970 Cruncher and it had been stuck in 40 mins on the same four WU's(stuck on the CPU part of the work on all the WU's), restarted Boinc and now it's running normal again.
I've seen other people have that problem so just wondering if it's a app_data or WU problem, anyone here have a clue?

I think it has something to do with the driver. I've seen it a couple of times. The worst part is that the WUs normally result in an error, which screws up the quota.

[Mydog]

I think it has something to do with the driver. I've seen it a couple of times. The worst part is that the WUs normally result in an error, which screws up the quota.

Got no errors but I keep getting a lot of invalids, maybe between 5% and 10% of the WU's from the looks of it.
Might go back to CPU and GPU crunching again but gonna give it another day to see how it goes points wise.

[pirogue]

Got no errors but I keep getting a lot of invalids, maybe between 5% and 10% of the WU's from the looks of it.
Might go back to CPU and GPU crunching again but gonna give it another day to see how it goes points wise.

If you're only getting invalids, you're lucky. I got a couple of errors and all of a sudden the machine that can do 180 per hour is limited to 150 per day. YEOUCH!

[Mydog]

If you're only getting invalids, you're lucky. I got a couple of errors and all of a sudden the machine that can do 180 per hour is limited to 150 per day. YEOUCH!

Oh no

I think I might caught my problem, the rig just displayed that "Display driver faild and recoverd" thingy and it looks like I have an old driver on the 7970(4-12) so I dowloaded the newest one I could find(8-12)
Gonna watch that rig closely the next 24 hours.

[bluestang]

I had a WU stuck when the driver crashed on me trying to OC my GPU core. After the driver recovered, I suspended and restarted BOINC and I was good to go again. Either way this 7870 is going back after I get my (2) 7770 cards.

[=[PULSAR]=]

How are you guys running multiple GPU tasks at once?

[Snow Crash]

How are you guys running multiple GPU tasks at once?

custom app_info.xml file controls which projects you want to run and also how many GPUGRID tasks at a time. Depending on your card type ATI is on page 3, Nvidia page 4, Nvidia and HCC on CPU as well as Nvidia plus GFAM are on page 9 of this thread.

[Otis11]

So I think I'm giving in... only have a 460 ATM but have to fight off the dogs as best I can! (That stats page is looking pretty scary right now...)

What's the best way to maximize output? It's on a 2700k with a GTX460.

Edit: Also, anyone know what kinda points I can expect out of that?

[pirogue]

So I think I'm giving in... only have a 460 ATM but have to fight off the dogs as best I can! (That stats page is looking pretty scary right now...)

What's the best way to maximize output? It's on a 2700k with a GTX460.

Edit: Also, anyone know what kinda points I can expect out of that?

There was a new list of GPUs and times posted today.
http://www.worldcommunitygrid.org/fo...d_thread,34078

A good ballpark is to multiply your expected WUs completed by 30pts.

It looks like that card is averaging a workunit per a little over 6 minutes. That's around 10/hour * 30, so 300 PPH.