| Author |
Message |
|---|
|
|
|
Yes, it's time again to rev up the engines for the PPS Sieve challenge. This year it starts on Dec 18 at 23:03 UTC (that's mid-afternoon for me, so I'll have to do another "running start").
This subproject is definitely GPU-aligned. It's basically a waste to run these units on CPU, unless you've got a rack of 64-core Xeon's and someone else is paying the electric bill. My 770 does one in something like 500 seconds, I believe. Leave a CPU core free to feed the GPU, but one can certainly crunch other projects alongside, if you like.
Total challenge run-time is 3 days. Good luck, and enjoy.
--Gary
____________
"I am he as you are he as you are me and we are all together"
87*2^3496188+1 is prime! (1052460 digits)
4 is not prime! (1 digit) |
|
|
mackerel Volunteer tester
Send message
Joined: 2 Oct 08
Posts: 2460
ID: 29980
Credit: 442,802,854
RAC: 10,291
                          
|
|
Warming up for it now. Got one GPU going but haven't figured out why another in same system isn't getting units. Still, got two evenings left to figure it out.
Edit: to answer my own question, it looks like I might need to manually add use_all_gpus setting manually. Will try that tonight... |
|
|
|
|
|
What I can't seem to figure out is why my 100+ new GPU's aren't installing.
edit: now I know, Santa's not due for another few days... dang it |
|
|
mackerelVolunteer tester
Send message
Joined: 2 Oct 08
Posts: 2460
ID: 29980
Credit: 442,802,854
RAC: 10,291
|
|
After much tinkering, I've decided against running the second GPU. It only has half the throughput of the 1st one, and as it actually drives my desktop display, I can't put up with the display lag it causes.
Right, onto setting up the 2nd computer now... not even sure what cards I have for it... |
|
|
|
|
|
As Gary has pointed out, this may as well be a GPU-only challenge, as running any CPU core(s) will be verrry slow however new they are, and obviously doesn't even yield the possibility of finding a prime. My only working, but very slow, GPU will be ticking over in the background, though. |
|
|
|
|
|
It is possible to "over commit" your GPU in BOINC with app_config.xml. That is, run more tasks at once than you actually have GPUs. This may be beneficial if you GPU isn't being fully utilized by BOINC. Create a file called "app_config.xml" in your projects/www.primegrid.com directory, with the following content:
<app_config>
<app>
<name>pps_sr2sieve</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.1</cpu_usage>
</gpu_versions>
</app>
</app_config>
(I'm not sure if all of that is actually necessary, but it worked for me)
The NON-exciting thing is that didn't improve throughput for me much... maybe 1 or 2 percent. It didn't seem to hurt, and everything is still validating. I posted briefly about this on the main board "sieving" thread.
This is on Linux with a 770. If you have a different OS and your GPU is running "cool", give it a try.
--Gary |
|
|
 Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
               
|
|
It may be necessary for us to do as much as we can. Overclock.net are on our tail(s). |
|
|
|
|
|
My experience with overcommitting the GPU has varied significantly, at least for sieving.
On a machine with dual amd 5770s, I notice little, if any benefit (in fact it may slow it down).
On machines with a gtx 680 or 980, the difference was substantial. All are running OS X, so there may be a cuda/driver difference. With a default config my 680s were running between 60 and 75% utilization. With a config as below, they run at 100%. On machines which hyper thread, I took the added measure of giving them 0.2 cpu usage, just to make sure the gpu never starves.
I would expect the 770 to be not too dissimilar.
All my (supported) GPUs but one are sieving away. Any word on an intel gpu client? |
|
|
|
|
Any word on an intel gpu client?
^What he said. IB and Haswell must be able to produce some half decent results, at least. I think they sort of crippled SB, and crippled it from a pretty poor start, too; but that must always have been the idea.
The thing which will be important to us will be being able to use the OCL cores of our CPU, as well as our x86 cores, AND out GPUs. The fact that BOINC can't tell its arse from its elbow even when faced with the most basic data, doesn't encourage me. |
|
|
|
|
It is possible to "over commit" your GPU in BOINC with app_config.xml. That is, run more tasks at once than you actually have GPUs. This may be beneficial if you GPU isn't being fully utilized by BOINC. Create a file called "app_config.xml" in your projects/www.primegrid.com directory, with the following content:
<app_config>
<app>
<name>pps_sr2sieve</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.1</cpu_usage>
</gpu_versions>
</app>
</app_config>
(I'm not sure if all of that is actually necessary, but it worked for me)
The NON-exciting thing is that didn't improve throughput for me much... maybe 1 or 2 percent. It didn't seem to hurt, and everything is still validating. I posted briefly about this on the main board "sieving" thread.
This is on Linux with a 770. If you have a different OS and your GPU is running "cool", give it a try.
--Gary
Thought I'd give this a try - added the file and placed it in the folder mentioned BUT I'm only seeing 1 task running so I'm sure I did something incorrect. Wrong folder or bad copy and paste. Any ideas - I am running windows if that makes a diff. |
|
|
|
|
|
I haven't tried on a windows box, but do you see a line like this in your event log:
Thu Dec 18 13:17:02 2014 | PrimeGrid | Found app_config.xml
|
|
|
|
|
I haven't tried on a windows box, but do you see a line like this in your event log:
Thu Dec 18 13:17:02 2014 | PrimeGrid | Found app_config.xml
No I looked.. file is called app_config.xml and is in the www.primegrid.com folder and I also dropped a copy in the main boinc folder as that's where I had the cc_config.xml file stored just to see.
edit: could be the version I'm running.. but I'd figure it would at least "see" the file. I'm not going to get bent out of shape just thought I'd give it a try. |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
Just tested this app_config.
1 GPU working not 2.
You have to set use_all_gpus to 1 in cc_config.
If you have multi GPUs, e.g if 2 cards, set max_concurrent in app_config to from 2 to 4.
Benchmarking now.
Needs to be in ProgramData\BOINC\projects\www.primegrid.com. |
|
|
|
|
|
I believe it was the boinc version. Just applied the latest update and it's now running 2 instances. Not sure if it's any faster or even slower yet.. but will post results. Now onto my other pc to see if I can get that working as well.
Cheers |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
From 875 secs 1 per card to around 1740 secs 2 per card (1750, 875×2, would be no gain). So increase in speed of around 10 secs per unit for the GTX580. |
|
|
|
|
|
660ti results - 625/1 vs 1037/2 or 1250 vs 1037 for 2 wu's - pretty impressive bump
770 results - 580/1 vs 892/2 or 1160 vs 892 for wu's - even more impressive
Gpu usage was at 99% on both and temps were hovering around 70 so I think I will be using the app_config.xml file Gary so kindly posted. Won't mess with the 470.
Cheers
edit: now that I think about it, I'm just showing how sad my cpu's are. |
|
|
|
|
|
Very sorry for the confusion about BOINC version. Yeah, I think it was posted elsewhere (and I should have repeated) that you must have a relatively recent version of BOINC to use app_config.xml... 6 or 7.0 is definitely too old; I'm running 7.2.42 on my 2600K/gtx770 box where it is working... so somewhere in between is the threshold.
Interesting to see the different performance aspects of running 2-on-1 between various OS's and cards that have been reported here. I'll probably run two since I am seeing a small boost. As has been said many times before, every box is different, and you have to try out different configurations to see what works best for you. Remember to leave a CPU core unused so as to feed the GPU efficiently. I never tried running *more* than two at once, but I doubt it would be beneficial.
It sounds like everyone is crunching happily so we should be "loaded for bear" (or Cow?) for this challenge. Go AtP!
--Gary
edit: another trick I use on sieve (only) challenges is to set a big WU cache right after the start... that way I've got work to do if my network connection drops and I can't fix it for a while (sadly, a concern for me at least). |
|
|
|
|
|
I'll have all 3 card in.
ATI 7970, GTX 770 and 470.
1 hour 45 minutes to start! |
|
|
|
|
|
Have idled my rigs except for the testing. Rebooted but not had time to clean out the dust bunnies.
As Gary said, we should be ready for some steak on stick.
Cheers |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
Yes I've 'preset' a 3-day cache setting for when we go boom + un-NNT.
Be interesting to see how the users/teams positions dance around during the challenge before things settle back down afterwards.
Good luck all.
Er we've just topped 7blln + are about to overtake [Lard]COP. |
|
|
|
|
All my (supported) GPUs but one are sieving away. Any word on an intel gpu client?
The current OCL apps should run just fine on Intel GPUs. I tested the stock Mac app with my Iris Pro GPU, and it worked OK. I need to finish some CPU WUs on the mac, and I'll try to set up an app_info.xml in the morning. I will start from what is posted here.
If you want to try it, let me know how you get on!
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
|
In addition to my bad planning re: the CPU and Intel GPU, I have also ended up with a GFN-WR running on one of my two AMD FirePro cards, which will run for the duration of the challenge. Still the other has started up OK and takes about 880s per task. I upgraded my  to  already during the warmup, and think I should be able to hit the next level midway through the challenge.
Keep running rats!
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
|
I've never had the slightest luck running Intel integrated GPUs on BOINC on Linux. Last I checked (which admittedly was a couple of months ago) on the Intel website and chat rooms, there were OpenCL drivers available for CPU but not GPU. There was a reference to some sort of open-source project for GPU, but that seemed a bit dicey to me.
My 3770K's integrated GPU is disabled in BIOS, just to save power, since I can't get it to crunch.
G |
|
|
|
|
|
My favorite user names on the team standings board, now just 5 hours or so into it:
(AtP teammates are automatically excluded from consideration)
Johnny Rotten
Hans Sveen (just seems to roll off the tongue)
ExtraTerrestrial Apes
Some bizarre (I almost used a different adjective) user names:
Brovarets Yuriy(Astronom2003)[Ukraine-Sarny team]
S2RiJtUVgYTRJDbMvtPjwHEPWT83zBfeiQ
gachagachacute@jisaku
And I see that there is now a team named "Tequila Sunrise". Seems like a nice place to party. As opposed to the weird "HUNGARY - HAJRA MAGYARORSZAG! HAJRA MAGYAROK!" team. I tried to use Google Translate on that one and my computer exploded. We also have teams "sleeping dogs" and "the assimilated", neither of whom seem to have many computers. But to each, their own.
Gary |
|
|
|
|
We also have team "sleeping dogs"
I saw them, too. We should just let them lie, I think.
Also, right now the Christians are one place ahead of Blasfemias in spots 27 & 28. Will be interesting to see how *that* plays out...
____________
Proud member of Team Aggie the Pew
"Wir müssen wissen. Wir werden wissen."
"We must know, we shall know."
- David Hilbert, 1930 |
|
|
|
|
All my (supported) GPUs but one are sieving away. Any word on an intel gpu client?
The current OCL apps should run just fine on Intel GPUs. I tested the stock Mac app with my Iris Pro GPU, and it worked OK. I need to finish some CPU WUs on the mac, and I'll try to set up an app_info.xml in the morning. I will start from what is posted here.
If you want to try it, let me know how you get on!
I was able to get PPS Sieve running on my Intel integrated GPU, looks like about 80 mins per WU, which is OK. The trick in the thread above worked where you defined two app_versions, one which is for CPU and the other with an Intel GPU co-processor. Of course this doesn't help any if you don't have an OpenCL stack that will run on your card at all, sorry!
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
|
i was able to get tasks downloaded and assigned to the gpu, all of which immediately resulted in computation errors, so I suspect I'm doin' it wrong.
Side note, don't experiment on machines which have long-running work units nearing completion. #$&$@!
The gpu works fine for collatz and otherwise appears to be OpenCL-driven
Fri Dec 19 06:58:09 2014 | | OpenCL: Intel GPU 0: HD Graphics 4000 (driver version 1.2(Sep 25 2014 22:26:04), device version OpenCL 1.2, 1024MB, 1024MB available, 154 GFLOPS peak)
Fri Dec 19 06:58:09 2014 | | OpenCL CPU: Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz (OpenCL driver vendor: Apple, driver version 1.1, device version OpenCL 1.2)
my (unsuccessful) app_info.xml:
vans-Mac-mini-i7:www.primegrid.com vzimmerman$ vi app_info.txt
<app_info>
<app>
<name>pps_sr2sieve</name>
</app>
<file_info>
<name>primegrid_tpsieve_1.39_x86_64-apple-darwin__cpuPPSsieve</name>
<executable/>
</file_info>
<app_version>
<app_name>pps_sr2sieve</app_name>
<version_num>139</version_num>
<platform>x86_64-apple-darwin</platform>
<avg_ncpus>1.00</avg_ncpus>
<max_ncpus>1.00</max_ncpus>
<plan_class>opencl_intel_100</plan_class>
<cmdline></cmdline>
<coproc>
<type>intel_gpu</type>
<count>1</count>
</coproc>
<file_ref>
<file_name>primegrid_tpsieve_1.39_x86_64-apple-darwin__cpuPPSsieve</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>pps_sr2sieve</app_name>
<version_num>139</version_num>
<platform>x86_64-apple-darwin</platform>
<file_ref>
<file_name>primegrid_tpsieve_1.39_x86_64-apple-darwin__cpuPPSsieve</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>
|
|
|
|
|
|
Iain, Tim and Nick are locked in a running battle. |
|
|
|
|
Iain, Tim and Nick are locked in a running battle.
I don't know what Nick is using but Iain and I have very different hardware which makes it even more interesting than usual |
|
|
|
|
I don't know what Nick is using but Iain and I have very different hardware which makes it even more interesting than usual
Not to mention hair cuts. |
|
|
|
|
I don't know what Nick is using but Iain and I have very different hardware which makes it even more interesting than usual
Not to mention hair cuts.
Hah Genesis 27:11, |
|
|
|
|
|
Oh, brother. |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
Can we beat Canada by the end of tomorrow (before the French get us as well)? I feel we're in a sandwich. |
|
|
|
|
Can we beat Canada by the end of tomorrow (before the French get us as well)? I feel we're in a sandwich.
My dog is paddling as fast as it can. Just glad Gary supplied the app_config.xml file and comment. It's working out rather well.
Cheers |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
I understand you can't change or take out app_info once it's in place without first emptying the cache, otherwise it'll blow units up (+ I have SoBs going as 'luck' should have it).
I've blown up 4 PPS svs due to overclocking (783 standard factory o/c, graphics spat dummy out @ 846 (7% o/c on top of the 783) then 837MHz (6%), now 822 (5%), wasting 13,484 units for myself & the team. |
|
|
|
|
I understand you can't change or take out app_info once it's in place without first emptying the cache, otherwise it'll blow units up (+ I have SoBs going as 'luck' should have it).
I've blown up 4 PPS svs due to overclocking (783 standard factory o/c, graphics spat dummy out @ 846 (7% o/c on top of the 783) then 837MHz (6%), now 822 (5%), wasting 13,484 units for myself & the team.
Sorry, but I'm not sure if I understood - this is O/C on a gpu? You can also reduce memory on those to help with errors and more importantly heating. i have my cards fairly o/c'd at the moment.
As for the app_config.xml file, since this is my first time using I would've stopped boinc, installed the file and restarted but you are saying doing that will remove or disrupt what you had loaded? I understand about the sob you have running but as for the sieves, who cares? Sorry not trying to sound harsh at all.. just making sure I don't kill some important jobs in the future. |
|
|
|
|
Hah Genesis 27:11,
glad you got that out of your system. Zimm got it right "Oh, brother"
|
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
I understand you can't change or take out app_info once it's in place without first emptying the cache, otherwise it'll blow units up (+ I have SoBs going as 'luck' should have it).
I've blown up 4 PPS svs due to overclocking (783 standard factory o/c, graphics spat dummy out @ 846 (7% o/c on top of the 783) then 837MHz (6%), now 822 (5%), wasting 13,484 units for myself & the team.
Sorry, but I'm not sure if I understood - this is O/C on a gpu? You can also reduce memory on those to help with errors and more importantly heating. i have my cards fairly o/c'd at the moment.
As for the app_config.xml file, since this is my first time using I would've stopped boinc, installed the file and restarted but you are saying doing that will remove or disrupt what you had loaded? I understand about the sob you have running but as for the sieves, who cares? Sorry not trying to sound harsh at all.. just making sure I don't kill some important jobs in the future.
Yes GPU. Didn't think to lower the mem clocks. Will leave it for now.
I haven't blown up the SoBs - just going by some advice I saw elsewhere on the fora this week. Obviously I stop + start BOINC for a change to a system file anyway, but I'm under the impression changing app_info requires an empty cache. No I know not to be too bothered about a sieving cache.
I might try adjusting the mem down + GPU cores up actually. Any problems, if I were not able to hear the fans dip down momentarily at such an event, show up in eventvwr as "Windows error reporting". |
|
|
TylerProject administrator
Volunteer tester
Send message
Joined: 4 Dec 12
Posts: 1077
ID: 183129
Credit: 1,280,170,555
RAC: 7,054
            
|
|
I've got my gtx 760 running...
stock 980MHz core clock, running at 1280MHz without issue. Temps a reasonable 73C with high fan speed. I run 2 WUs at a time, each allocated 50% of the GPU. CPU usage is low, 2-3% per instance. GPU usage is always maxxed out. I'm getting around 1,150sec/2WUs, so equivalent of 575sec/Wu, Just under 10 minutes. Not bad for a GTX 760 IMO.
____________
275*2^3585539+1 is prime!!! (1079358 digits)
 Proud member of Aggie the Pew
 |
|
|
|
|
|
It is completely safe to change app_config.xml with running or cached work units. They won't be trashed. This is not true for app_info.xml... you must have no cache and nothing running for that to be changed or put in place or removed. Don't you just love computers? :-p After changing these .xml files, I do a complete reboot of the box.
I'm running two-at-once on my 770, although the drivers/OS seem very efficient in the first place, so it provides very little benefit, as I posted earlier (but it *is* there!). My 570 is just running 1 at a time. Even my iMac is "in", plodding along at like 7 hours per unit (it's almost as old as I am). I broke my own rule and am running a couple of CPU units on each of the Linux machines. They take roughly 12 hours each (i7, mod o/c, HT off).
As I type this, we're #5, but 4, 5, and 6 are *really* close together.
--Gary |
|
|
|
|
Hah Genesis 27:11,
glad you got that out of your system. Zimm got it right "Oh, brother"
I suggest the last sentence of each of Matthew 27:5 and Luke 10:37. |
|
|
|
|
|
|
|
|
|
|
Trust you to bring a token stereotypical Scotsman into the thread. Ahem...
Back later gotta molest some sheep.
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
Trust you to bring a token stereotypical Scotsman into the thread. Ahem...
Back later gotta molest some sheep.
Crumbs. I knew about the deep fried Mars bars, the excessive drinking, the miserliness, the hatred of the English, the religious bigotry and the inability to play soccer but I thought "animal husbandry" was a vice reserved for the Welsh.
|
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
May the 4th be with you.
As in: we're in 4th place. |
|
|
|
|
Crumbs. I knew about the deep fried Mars bars, the excessive drinking, the miserliness, the hatred of the English, the religious bigotry and the inability to play soccer but I thought "animal husbandry" was a vice reserved for the Welsh.
To be honest, that particular proclivity is not unique to the Welsh or the Scots, but to any out-in-the-sticks farming' country. In addition, allegations of cousin marriage, webbed feet are also common - think of the West Country!
However, fans of Aberdeen FC have taken this to heart and are widely known as the 'Sheep Shaggers'. I refer you to (very NSFW, and not for those of a gentle disposition) : https://www.youtube.com/watch?v=oYNrHLfIx_4. Once, I bought one of these for a friend who was an Aberdeen supporter. He looked surprisingly happy to receive it!
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
mackerelVolunteer tester
Send message
Joined: 2 Oct 08
Posts: 2460
ID: 29980
Credit: 442,802,854
RAC: 10,291
|
|
Woke up to find my main box crashed this morning. Fortunately seemed to only crash not long before I checked so little crunching time lost.
The box normally runs 24/7 so I was a surprised to see it fall over. That's when I noticed a possible reason why. CPU temps were pushing 85C! I know my case doesn't have the best airflow and a fully loaded GPU is heating everything up. Backed off CPU crunching now and temps are relatively cool below 70C. Just have to wait for challenge end before putting things back to normal, as even with 4 LLR tasks the CPU temps don't normally get much above 60C. |
|
|
mackerelVolunteer tester
Send message
Joined: 2 Oct 08
Posts: 2460
ID: 29980
Credit: 442,802,854
RAC: 10,291
|
|
Just had two more crashes while I was using the computer. I've turned off GPU use now so my production will drop but hopefully system will remain working normally... |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
Aren't PCs effing b sometimes.
Just turned piano on to do some music - + the phone goes. Dad can't get his PC on. Suspected RAM or onboard graphics failure. Showing online now however so don't know until I visit them.
Then I was plugging something else into my UPS + my main 248648 machine goes off abruptly in full flow. Nothing lost bar some component wear + tear.
Good luck with your machine Matt. Did you check eventvwr? Mayb try underclocking or something. We're safely in 4th place anyway. |
|
|
mackerelVolunteer tester
Send message
Joined: 2 Oct 08
Posts: 2460
ID: 29980
Credit: 442,802,854
RAC: 10,291
|
|
Gave the box a good dusting now. The heatsink/fan for the CPU is bolted to the motherboard and I'm not taking the whole thing apart to clean it. But the dust clouds when I manually blew backwards through it suggested there may be a fair bit of blockage. Gave the other case filters a clean too and CPU temps while doing LLR are now sub 60C! Compares well with the 70+ it was at just before.
I took out the nvidia GPU though. Haven't cleaned that yet. Will put it in another box as dedicated cruncher another time. |
|
|
|
|
|
I've got to buy a couple of cans of compressed air too, whilst yanking out a fan (clearly cheap rubbish that came with the case) the other day, I noticed that my heatsink now resembles an old-fashioned hoover bag. |
|
|
|
|
|
Well done ATP! I will continue on this sieve for a bit. We hammered it really hard.
Cheers Rick |
|
|
|
|
|
My GPUs are still sieving away but I've put the CPUs on TRP-LLR for a while.
Regarding blasting out the cases; yes, thanks for the reminder.
In typical fashion, I started fairly high in the challenge rankings and slowly slid downwards. Did manage to hold on to a top-100 spot though. :-)
I almost said something like, "I have to get one of those Titan cards to keep up", but it strikes me that the Titan has been out there for months now, and something new is probably right around the corner. Chasing technology is futile, at least until what you've got is clearly obsolete, or have a bazillion dollars to toss around.
--Gary |
|
|
|
|
|
AtP finished 6th and here's the individual stats. Job WELL DONE!
Rank Name Score
3 zunewantan 31,185,121
48 Van Zimmerman 5,639,683
67 Rick Reynolds 4,102,507
77 mattozan 3,573,260
83 brinktastee 3,246,273
97 Gary Craig 2,669,832
130 Dave 2,063,052
214 Iain Bethune 1,125,914
226 1998golfer 1,061,865
227 NickBurrlock 1,058,494
238 ~~~(,__,)^'> 1,024,784
294 mackerel 768,588
452 Neo 296,648
611 gazzyk1ns 128,098
Will now make this unsticky as AtP almost never has invalid units show up during the cleanup process :)
Cheers |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
Well done all, good push.
I didn't see where we were in terms of the challenge team positions, but the efforts did bring us up to 4th in the overall team positions (non-challenge).
Now then, onto SoB... Who can do them under 4 days to fit 6 or 8 in? I know I can't. |
|
|
|
|
<app_config>
<app>
<name>pps_sr2sieve</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.1</cpu_usage>
</gpu_versions>
</app>
</app_config>
--Gary
Gary or anyone in the know - can you explain to me what the .5 and .1 figures mean and how they affect the gpu and cpu? I like using this file but have noticed it affecting my actual cpu jobs.
Cheers Rick |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 5,071
|
|
0.5 means it uses 0.5 of a GPU i.e it allows 2 threads per GPU.
0.1 uses 0.1 of a core, meaning if 4 cores total of 0.4 of a core, hence best to keep 1 core free (or 0.6 of a core lol) when using this file.
I haven't tried amending either to be honest. Be interesting to know what effect on GPU tasks any CPU usage amendment would have, |
|
|
|
|
0.5 means it uses 0.5 of a GPU i.e it allows 2 threads per GPU.
0.1 uses 0.1 of a core, meaning if 4 cores total of 0.4 of a core, hence best to keep 1 core free (or 0.6 of a core lol) when using this file.
I haven't tried amending either to be honest. Be interesting to know what effect on GPU tasks any CPU usage amendment would have,
.5 makes sense as explained. the .1 makes no sense. Seems like it should translate to a percentage which I was thinking the .1 was. ie 4 cores and I want to use at least 1 core the number would be .25 as the .1 would seem to "me" it will only use part of 1 core.
Will make some adjustments and see if anything changes. |
|
|
|
|
|
From the useless links collection under the porch ...
Nearly Everything You Never Wanted To Know About BOINC Client Configuration Including cc_config and app_config
____________
There's someone in our head but it's not us. |
|
|
|
|
From the useless links collection under the porch ...
Nearly Everything You Never Wanted To Know About BOINC Client Configuration Including cc_config and app_config
It's good to have storage space :)
I've made a change to .4 on a 4 core system - from the looks of things it didn't make an difference but then again I might have to let it run for a few. And yes, I did exit out of Boinc and restart so the app_config file would be read in again.
I do notice that running 2 instances on my gpu does slow my cpu job but the task manager shows only a 40% usage overall. |
|
|
|
|
|
I observe that the "Christians" finished the challenge ahead of "Blasfemias". Noticeably far down in the standings this time are PrimeSearchTeam and the two Polish national teams, usually strong performers.
Other entertaining team names that I've just noticed:
Hug & Dance Ltd.
H.I.K. Humans in Kansas
Don't Panic Labs
Brethern of Swill
--------------------------
Regarding app_config.xml, on my 770 box, I used the file just as I posted it, except I changed the <max_concurrent> value to 4. I set client CPU usage to 75%, and set preferences to run PPS Sieve on both GPU and CPU. This consistently caused two units to run on the GPU, and two more to run on CPU. I was a little afraid that it might do something silly like run 4 on CPU and leave the GPU idle, so I checked in on it occasionally. But it all worked fine.
Running these on CPU I think moved me up *one* spot in the standings, but still it's probably a waste of energy.
g |
|
|
NeoVolunteer tester
Send message
Joined: 28 Oct 10
Posts: 710
ID: 71509
Credit: 91,178,992
RAC: 0
                   
|
|
Wow...
I'm a tad bit disappointed in my finishing at #452.. I did have several computer reboots/downtime, but even if I didn't, I don't think I would have caught Mackerel at #292....
Would have thought the 512 cuda cores in the 750GTX would have done more damage than the 92 cuda cores in my retired Nvidia 430GT... I guess that's the difference in the Maxwell vs. Kepler chipsets.
Oh well.
Neo
|
|
|
