Printable View
Sam commented about having poor runtime.
I run my card with lots of concurrent wu's to increase my runtime (while I can) then I came to look at the top "members" list and wondered who can tell me....
Which of these "members" really are individuals
http://www.lakecityquietpills.com/ph...3668544958.jpg
My highlighting is just those I think I know
Ditto your selection:up:
And FWIW I think Hypernova could do much more but he is not using his GPUs to anywhere near their full potential IMHO after reading he's running 10 GPU tasks and 2 cpu tasks on his HT hex cores with 7970s.
Maybe he runs Linux??? Sam says he has issues running many concurrent on Linux.
I only ran my 5870's on Linux so I cannot comment
Could be. IDK.
[EDIT: The following is incorrect, as pointed out below by Snow Crash & wplachy [/Edit]Quote:
OldChap: I run my card with lots of concurrent wu's to increase my runtime
Only CPU runtime is counted, not GPU runtime.
While the GPU is crunching an individual WU, the CPU is only spending a low % feeding it, so it's not incrementing its runtime tally much at all. You're running more tasks than when doing CPU WUs, but the total CPU run-time can only be as high as the number of CPU cores/threads times the elapsed time. There's no double-counting!
In fact, running GPU work you'll tally up less run-time than running CPU work, because with CPU work the CPU usage is at 100% 100% of the time, while with GPU work the CPU is often at less than 100% usage.
Hope I'm less confused now ... :confused:
talk about confused <wink> ... what you've posted is not true. Please take a look at your device statistics on WCG. If you are configured to concurrently process more tasks than you have actual cpu threads you WILL get more runtime than the number of threads.
Quote:
Originally Posted by BlindFreddie
WCG awards runtime as: CPU tasks get CPU time; GPU tasks get Elapsed time (wall time). To confirm I verified that my devices running 12 concurrent GPU tasks only are being awarded 11-13 days per cal day.Quote:
Originally Posted by Snow Crash
--------------------------------------------------
Bill Phttp://www.wcgsig.com/402500.gif
Yes, you guys are right, I was wrong. (Edited my post).
So you can manipulate your WCG runtime stats by simply varying the no of concurrent GPU jobs, without much effect, if any, on throughput..
Lies, damn lies & statistics ...
There is limit to how many GPU tasks you can run simultaneously on any given system before you start to lose efficiency. Depends on your hardware. Trial and error with your system(s) is the only way to know what would be the max for you.Quote:
Originally Posted by BlindFreddie
I have only done this on one rig and my observations are as follows:
q6600 @3.3 and 7950 @ 1325
the absolute best AVERAGE time when running 16 wu's concurrently is about 51 seconds per but the offset gets out of sync quickly which results in that average time dropping to 53 seconds or more.
Moving to 24 concurrent wu's best time seems to be in the order of 52 seconds and offset stays longer but still fails and then the average is maybe 54 seconds or more
With 36 concurrent wu's the best time is probably 54 or 55 seconds average but 24 hours later the offset is generally maintained and the average runtime for a wu is still around 55 seconds.
I have just set for 40 concurrent and will try to report my findings
In each instance above my offset is approximately 35-40 seconds. The idea behind that was that for at least some of the running wu's there will not be wu's in cpu mode at both the middle and end phase at the same time.
Remember this is a 4 core non ht cpu running 100% all the time.
I believe that it may be possible when flooding the cpu like this to have settings that you can set and forget for maybe a week or more at a time. So far it is not perfect but looks promising. Does the answer lie in matching the cpu and gpu runtime for each wu perhaps?
I'm now running 32 WUs with an average completion time of ~22.5-23 min per WU. This should come out to around 42-43 seconds per WU with random offsets between WU groups. This is with my 2700k @ 4.5 GHz.
What video card are you using and what are your clocks?Quote:
Originally Posted by 0ne.shot
It's amazing. Sam has a much higher point production per thread.
I do 12 million points with 195 days/day and sam does 12 million points with 135 days per day.
I am at 61'538 points/day and Sam is at 88'888 points/day. Much better efficiency per thread, as one day runtime per day is basically one thread per day.
Sam's threads are 44% more efficient than mine. That is astonishing as I am running top gear hardware. Maybe Sam runs Linux and I am with Windows.
I know that Linux is more efficient. Maybe I should change. The problem is that I am a Linux illiterate.
OdChap do you mean that on your machine you crunch an HCC GPU WU in 52 seconds with 16 or more threads.Quote:
Originally Posted by OldChap
I am between 7 and 8 minutes with a 7970 GPU and a 990X CPU.
Have I missed something !!!!!
I believe OldChap is taking the average runtime for a WU and dividing it by the number of concurrent threads he is running. How many concurrent threads are you running (16)? I'm guessing you may get better overall numbers by upping the qty ... I've pretty much settled on 24 concurrent on my 7950 + 980X so I only have to manually intervene 1 time every 24-36 hours to make sure not too many WUs are bunched up at cpu processing.Quote:
Originally Posted by EtaCarinae
GPU OC has a lot to do with the overall runtimes ... a 7950 at 1325 is quite good. Mine maxs out at 1300. At 1325 I'll start throwing Invalids no matter how high I set the volts.
Last thought ... if you have your cache set too high you may be experiencing "High Prioity" fubar ... sometimes BOINC partially forgets I am running concurrent WUs and changes my estimated completion time to be as if I was crunching 1 at a time yet it remembers I am runing 24 "cores" so the result is I get flooded with new WUs until it settles back into the real runtime again (sometimes a couple seconds, sometimes a couple minutes). If I have .5 day cache this does not disrupt processing but if you are much higher you might have WUs preempt WUs already running and in that even the checkpoint is only at 50% so you will lose any time crunched between 0-49% and 51-99%
Yep, when I talk about AVERAGE runtime then what I mean is .....
I take the total number of threads running and suspend internet until all have completed. Today that is 40 wus so I will end up adding together all 40 times then dividing by 40 for the average time it takes a single thread to complete.(today that is 35 mins and about 13.5 seconds). However during the time it takes a single thread to complete I have completed 39 other threads too so I take the average single thread time and divide by 40 for the AVERAGE time between completed wu's. (52.83 secs for my last 40 today) with no adjustment for about 24 hours.
Is this the best way? I don't really know but it does give me something to work with when I look at daily production:
24*60*60 = 86400 seconds a day which I divide by that AVERAGE to complete a wu time of say 52.83 secs is around 1635 wu's a day so I keep an eye on this ..... It is not helped by the problem of varying PV numbers but overall I see what I need to doing this. Using this method works best for me if I run new settings for a week.
I am sure that I have explained things better than this before now but I hope you get the gist
Or just take 60 minutes divided by the average total elapsed time in BOINC for a WU and multiply by the number of concurrent WU's your running and then by 24 hours....
60/35.225*40*24=1635
I can do the math...it is just not pretty.... usually :p:
If you're using your GPUs like the post I saw at WCG (10 GPU and 2 CPU threads on a hexcore with HT) then you are not using anywhere near the potential of your 7970s. Take just one and try this so you can compare. Set BOINC to receive no new tasks. If you're still using an app_info file finish all the tasks you have in your cache first and delete any CPU tasks that are left when the GPU tasks finish. Use a separate WCG profile for that machine and set it to receive/run no CPU tasks.Quote:
Originally Posted by EtaCarinae
Upgrade your BOINC client to 7.0.56 After it starts just shut it down for now.
Copy and paste the text below into notepad. Rename it app_config.xml and place it into the WCG project folder. (That will give you 36 GPU tasks each using 1/3 of a CPU core) Delete the app-info file or at least remove it from the folder. Restart BOINC. Set you cache to 0.50 days to get started. Set BOINC to allow new tasks.
Let Boinc download maybe 100 tasks, set Boinc to receive no more tasks and suspend the ones you have(Including the ones that are running) from the tasks page not the project page. The easiest way to do that is highlight the first task, scroll down to the bottom hold down the shift key and highlight the last one. That should highlight them all then click suspend. Now highlight the first task only and click resume. Every time you see the percentage done start to move highlight and resume the next task. Do this until you have all 36 running. It will take a while but this will keep you tasks spread out for better efficiency. Once you get 36 running mass resume the rest and up your cache to whatever you'd like. Your hexcores should be able to knock out 36 tasks in 25-30 minutes, maybe less. (for a reference my 7970 does 24 tasks every 20 minutes on a 2600k running @ 4.2 GHz. with the GPU @ 115 core clock) Depends on your hardware and any overclocks you might be running. Most 7970s can run @ 1125 core clock on stock voltage. With a GPU configuration only I think you should see 1.3-1.5 mill WCG PPD on that cruncher.
<app_config>
<app>
<name>hcc1</name>
<gpu_versions>
<gpu_usage>.0277</gpu_usage>
<cpu_usage>0.333</cpu_usage>
</gpu_versions>
</app>
</app_config>
Let's see I will do some tests on one machine. But my past experience is that very quickly you start getting unstable systems, with some compute errors and some window restarts. I generally favor les extreme configurations but that last for weeks, months without interruption. But as GPU is coming to an end mybe it is time to try to really become an Hypernova :)Quote:
Originally Posted by nanoprobe
Running this configuration should be no sweat for your machines. I've been running my version since the app_config system came out. Not a single error, invalid or problem. One thing you may need to do is a reg fix for driver timeouts that is a known issue in windows. If you see tasks where the runtime is moving but the completion % is not moving then it has probably had a driver reset. The fix is very simple and I've done it to all my machines. Copy all the text below and paste it into notepad. Rename it fix.reg. Right click on it and open with Registry Editor. You'll get warnings about editing the registry, just click yes, and you're done.Quote:
Originally Posted by EtaCarinae
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Watchdog]
"DisableBugCheck"="1"
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Watchdog\Display]
"EaRecovery"="0"
Holy crap. 36 simultaneous tasks????:shocked: Please forgive my ignorance. I have zero time for playing/reading/learning....for at least a month now....and probably not for another month. I am not familiar with EtaCarinae's systems. Would you recommend this same app_config for a 990x with a 7970? (cpu is at 173 x 25 = 4325, gpu core seems stable at 1240)Quote:
Originally Posted by nanoprobe
ET is running 990X w/ 7970 so yes, crank your concurrent count up :up:Quote:
Originally Posted by my87csx481
You may have to back off on the GPU a tad when running 36 tasks. 1240 is pushing it pretty hard. Keep an eye on your WCG results page for errors/invalids.Quote:
Originally Posted by my87csx481
My IO rig is now crunching with 36 threads. I am still not convinced this will increase the number of results or points per day. On the other hand it should increase the runtime.
I will let now the rig run over the week-end and see the stats. If there is a sizeable increase in performance then I will propagate this change to all the rigs. Wait and see.
A single large leap to a higher number of wu's is unlikely to show an increase in points whereas small increments may do so for the duration of the move from a lesser to a larger amount.
After the rig settles the I would expect to see a similar or slightly elevated points tally and the same goes for results. The increase in runtime should be related to the percentage increase of wu's running
Early days to be certain but it may be that the number of cpu cores/threads times an odd number ie. 8*3 or 8*5 and 12*3 or 12*5 may look after the offset better for longer. This is just an observation on this rig which I have not really proven yet and may not actually happen on later hardware.
If you were running 10 GPU and 2 CPU before I can guarantee that you'll get a very noticeable increase in production.Quote:
Originally Posted by EtaCarinae
Indeed nanoprobe. I was running 10 GPU and 2 CPU. So let's see what happens. At least 36 threads make a better use of the 6 GB RAM I have on each rig.Quote:
Originally Posted by nanoprobe
And even if it is better. what is the maximum then? 100 threads, or 20 threads? Or up to the maximum that memory capacity will allow?
Running two 7970's here with 12 WU's per GPU, looks like I've got a bit of headroom on the temps.
http://img26.imageshack.us/img26/8026/vindux.jpg
Going to increase the OC on these cards during Easter when I got the time to figure out the vcore settings.
Do you think I should increase the number of WU's per GPU OC? CPU load is rarely above 60%.
That's a hard call. My maximum is 32 on my current hardware with 8GB of ram. (Your mobo ram speed will make a difference.) Any more than that saw the efficiency drop off. With an HT hex core you could possibly run up to 48 tasks. I don't know how much ram each task requires but I would think 6 GB would be enough to run 48. Running 36 will cause a backlog of PVs so I would give your current settings 4 or more days to get a baseline for results. After 36 try 42 and see what happens. If you don't see a better average per task run time than 36 is your limit. If 42 gives you better times (it may only be a few seconds per task) then bump it up to 48 and see what happens. My best guess is 48 will be your limit if 36 or 42 aren't. @ 32 my average run time per task is about 37.5 seconds. (32 tasks every 20 minutes)Quote:
Originally Posted by EtaCarinae
Are we having fun yet? :rolleyes:
Running dual cards on a system is a whole different ballgame. There is a point where you run into bottlenecks so I can't advise you on how to proceed. Trial and error is the only way to find out. Maybe someone else here who is running a dual card setup can chime in.Quote:
Originally Posted by Mydog
I did two tests. Running 36 threads and 12 threads. The results are that running 36 threads the daily point production is lower than with 10 GPU threads by -12% at least. In fact the maximum value is with 12 threads. That means that you use a full CPU core for one GPU thread. I get +8% at least over a 10+2 configuration. All this at same hardware settings.
The reason is that HCC WU are still CPU intensive. The CPU is havily used at the end of each diffraction image, this means two times per WU, to which you have to add all the logistics, like loading and unloading the WUs on the GPU and all PC functions on the networks etc... If one would like to add over 12 threads then there would be a way to have better results, but this would mean to be able to define the starting time of each WU, so to avoid collisions and have multiple WUs using at the same time CPU resources.
The reason I am using the 10+2 configuration is that in case there is a shortage of GPU WUs and the cache is empty (it happened already) I avoid having an powered but idle machines and the CPU crunches at least two WUs, better than nothing.
If we run some CPU WUs in addition to HCC GPU ones, I think we have to do some micromanagement of the work cache to keep things running optimally.Quote:
Originally Posted by EtaCarinae
You have to have "HCC" ticked in your machine's WCG Device Profile on order to get GPU work. This means that you get some HCC CPU work as well as the GPU work. IMHO, crunching HCC WUs with a CPU is now a huge waste of energy & computing resources, because it's so much more efficient to crunch them with a GPU. To avoid this wastage, you need to abort all HCC CPU WUs received! The CPU that you would have used for HCC should be used for a CPU-only project instead - for those who want to crunch cancer work, HFCC work is still available.
Interesting that you found 12 GPU threads to be best. I'm running a Q9650 with 1 x HD7870, no CPU WUs, and 12 threads seems to be best with it too. I just tried 15, 14 and 13, but throughput was worse. Earlier I tried all numbers from 4 to 12, plus 16, and chose 12. Running 1 thread per CPU core (4) was definitely worse, ie there was no "magic" in having 1 CPU core per GPU thread.
I'm beginning to think the methodology for max. production numbers being posted is flawed. There is no doubt that running CPU tasks with GPU will slow down production, that is a given and makes sense.
However on the subject of GPU only crunching something isn't adding up!
A while back I was running 16 wu's on 79xx and 8 on 7770's. Someone said that 10 was optimal for 79xx cards so I dropped mine to 10 and added another 7950 and 7770 to the farm. After adding those cards my point production barely went up and stayed around the 530k mark for. So I'm reading this thread a couple of weeks ago and decided to put the 79xx's back to 16 wu's, then 24 were they have been running for just over a week. I've seen my PPD production jump up significantly since then averaging well over 620k ppd, nothing else changed(have a look at the pie chart). I'm not doing any fancy math trying to figure out immediate production or anything, just looking at BOINC stats and free-dc. Also my pending where increasing throughout that last period, I see today that pending's are finally coming down from the 600 page mark.
So what the heck is going on, my PPD is not jiving with what other people are seeing/saying ? Like I've always said I really care less about points other than their indication of work done, but I find this to be odd.
FWIW my current farm consists of 1-7970, 2-7950's and 5 7770's, all running on Quad core's with hyperthreading.
JohnMark,
Your latest posted numbers for the 7970 and 7950 are good. The dual 7770 rig you run appears to a low scaling compare to other single 7770 rigs that you operate.
I believe that each rig has its own sweet spot insofar as the number of WUs to run for best results #s or optimum runtime. The rigs components and whether you are in a single, Xfire or Tri-fire video card set up will dictate the ideal # of concurrent WUs to run for best efficiency. You must find tune each rig individually. There is no magic WU # that can be applied to a single video card series. What folks provide in this forum are good WU benchmarks #s to use to fine tune your own rigs individually. Even if you have the same hardware as someone else you may not exactly match their PPD as variables such as applications versions, running background processes, carrying out daily WUs spacing and internet speed will influence your result output.
How am I ever going to catch up to you with them new Xtreme results you're now putting out...??? :D
What type of quads?Quote:
Originally Posted by Johnmark
Nano..., List of JohnMark's rigs hereQuote:
Originally Posted by nanoprobe
I hear what your saying jeanguy2, it's just interesting there is such a significant difference on output!
Some of it makes sense, however I can't help feeling I'm missing something in this whole GPU/W.U. vs production thing.
You've got a slice or two of my pie already, isn't that good enough :)
nanoprobe, the CPU's are Intel Quads a 950,920,860,875k and 4 2600k's.
Quote:
Originally Posted by jeanguy2
Here's my take FWIW. I did a lot of testing from the time the WCG GPU app started. 4 of my 7970s run on 2600k's @ 4.2 Ghz. All have 8 GB of 1600 MHz. memory. The core clock on all the cards is 1125 Mhz on stock voltage with the memory downclocked to 1000 MHz. (I found no advantage to running the memory higher and trying to run any lower would cause the card to revert back to factory default) I stated with 8 tasks each and worked my way up 4 tasks at a time until I reached the 32 tasks per card I'm running now. The amount of tasks completed and PPD stayed about the same from 16 on up. What did increase as the amount of tasks run above 16 was the amount of run time I accumulated. Running more than 32 caused less tasks completed and PPD. The "trick" with running that many is getting the proper spacing between each task so the GPU stays as close to its' maximum load all the time.Quote:
Originally Posted by Johnmark
Overclocking your CPU/GPU and mobo memory will produce more results and more points but I'm looking for longevity so I chose not to run mine any harder.
Here are my latest stats for 1 machine and why I think my set up is optimum for me. They were lower for 3/13 because the machine was offline part of the day for dust removal and on 3/22 for a re-image on a new drive.
http://i303.photobucket.com/albums/n...psd069fbe9.jpg
JohnMark, I went down to the Fraser river today and asked the local Sto:Lo medicine man about what to do to challenge you. He told me to move my wapiti rack from the north side of the cabin to the south side of the cabin so that it faces the spirit of the river. I am also to hang my snowshoes beneath it with their pointed heels facing upwards. I will do so tomorrow.Quote:
You've got a slice or two of my pie already, isn't that good enough:)
Hypotetically as MM would say, you could be in trouble as this is very strong medicine.... :D
:rofl: Well I guess I asked for it :)
Quote:
Originally Posted by jeanguy2
Rob, that machine is running at 4600 mhz w/8gigs RAM and the HIS GPU at 1200 at stock voltage. It's was running 8 wu's from day one and I bumped it up to 12 the other day. That's all Rob !
Quote:
Originally Posted by Rob_B
Thanks. I will play a little with the number of WU's and try to overclock the GPU.
One more question. How many WU's does that 7770 crunch each day? Mine does 650 WU's on average.
I though that with 4 WU's it completed the WU's the quickest. I tried 6 and that lowered the total WU's each day because it took much longer to crunch them.
800-900 a day, last few days have been 900-1000.
Quote:
Originally Posted by Rob_B
With a 2600k I would do 8 WU's and space them apart 2 at a time.Quote:
Originally Posted by Rob_B
Also, why is that GPU only running at 1000? ;)
Not sure how that happened ;) But will fix that today. Just don't think it will do 1200, but it had done 1100 just fine before.Quote:
Originally Posted by bluestang
my MSI 7770 is running 1200 at stock w/o invalids.
As soon as I run it at 1200 the driver crashes. But I have it at 1150 now and running 16 WU's without errors. Doing a WU in 20-21 minutes it should do a lot better than before.
I just don't see it yet. Run time has gone from 4 to 7 days and a few more WU's over the last few days. But still not seeing the points I'm hoping for. The PV's are going higher so will have to see what happens in another 2-3 days.
I have the same experiences with my card. I have lots of invalids when running close to 1200. I have several pages over a day, but with around 1500 WUs I'm not sure it's that big of an issue. My PV list seems to be growing and not getting smaller.
When you say your 7770 card is running at stock volts, what is GPUZ giving you for a voltage?
1.S,Quote:
Originally Posted by 0ne.shot
You maybe faced with a catch 22 situation... getting invalids will definitely raise the Pending Verification #s for that rig... It may also slow down your validation process until invalid situation is resolved and WCG server trust that rig again. Most of your invalids will have been or be re-processed by you and/or another rig to confirm status. Operating a crunching rig with invalids being produced reduces your efficiency and that of others rigs as they have to re-do or re-check your work.
Is that a question for me? I didn't say anything about voltage. But I'm running stock voltage and GPUZ shows 1.200 volts.Quote:
Originally Posted by Johnmark
Thanks JG2. I raised up my voltage to 1.25 and lowered core clock to 1160, and I only have 2 invalids today so far.Quote:
Originally Posted by jeanguy2
Well done 1.S Keep an eye on that rig and get invalids to zero. Note the date of the invalids you are getting. Some of them could be from before you raised voltage and lowered core speed. Do not take any further action on invalids pre-dating these afore mentioned changes. These would come from your accumulated pending validation and pending verification files.Quote:
Thanks JG2. I raised up my voltage to 1.25 and lowered core clock to 1160, and I only have 2 invalids today so far.
Sure :) Thanks.Quote:
Originally Posted by Rob_B
Just curious as to what stock voltage is for these cards. For some reason my 7770 cards, from different vendors and of course models are all over the place. I only really noticed the HIS stock volts which was 1v using their OC tool. Then I just used Afterburner to put them all at 1200 mhz after installing them.
However I'm finding the various tools are giving me various voltages, and it seems that some OC tools are causing the volts to raise on on their own when an OC is applied? I just noticed this when looking at results for the various cards when replying to OC's thread as that was the first card I installed.