The Scottish Boinc Team

http://boincstats.com/en/stats/challenge/team/chat/691

Name Coolers maximum power, infernal heat !!
Status Upcoming
Project Einstein@Home
Issued by Hammrammstein
Start time 2015-07-04 00:00 UTC
End time 2015-07-11 00:00 UTC
Late entrants allowed? Yes
Number of teams participating » 7
Number of users participating 0

Good grief! Was this challenge issued by power company shareholders?

Enough complaining. Halting further BURP crunching to bunker Einstein. Try to catch me folks

Nope, sticking with BURP at least until I get home. Heatwave forecast for UK and if you believe it then Einstein is almost , but not a suicidal as a milkyway challenge. I'll jump in when I get home and am sure I'm not cooking anything. Need to move systems out the house etc before starting IF it's actually going to be warmer than freezing ......
Mind you, with the 'summer' we've had in Scotland so far, more chance of pigs flying than heatwaves turning up

I think you'd be surprised then Alez, it's been frickin hot today and tomorrow is supposed to be even warmer

Alez wrote:Nope, sticking with BURP at least until I get home. Heatwave forecast for UK ...

Heatwave! Oh my! OK, to be fair, do you all have air conditioning?

Air conditioning ? Scotland ? Not two words that appear in the same sentence.
I've never even owned a car with air con that I've bothered to maintain. We specialise in insulation and heating. Air con means opening the windows 8)

I hate the BOINC scheduler Have a system with a 7970 and also usually a R9 270. Pulled the 270 and put in an old GTX 460 for the BURP challenge. I was crunching einstein Parkes WUs on the 7970, but after the 460 was installed, the runtimes for the Parkes WUs on the 7970 went from 1.5 hours to over 50 hours Didn't notice this until I put the R9 270 in it and tried to bunker. Even with the cache set to 10/10 days, couldn't get more than around 30 WUs. Wanted at least 200. Estimated run time for the WUs was over 48 hours (because the unknown problem) so the BOINC scheduler won't download any more. Been crunching the Parkes WUs for over a day now, with normal times, but it's painfully slow to get the estimated run times down enough to bunker. Any suggestions to fool or reset things? Would resetting the project make it start fresh?

I think resetting the project would reset it but not 100%.
There is a file in the boinc directory ( c://ProgramData/BOINC I think what you are looking for is Task duration correction factor think its
<duration_correction_factor> 1 </duration_correction_factor>
in the file sched_request_einstein.phys.uwm.edu.xml ( use notepad to edit as always with boinc xml files )
I think you set that back to 1 and then update Einstein. Not sure if you would have to alter the stated flops for the gpu in the app info further down the file. Wouldn't think so.
Can't guarantee 100 % thats what you do ( might only set the cpu task duration factor ) but 95% sure that's how it used to work. ( my laptop doesn't have a usable gpu so look at your file and see if there is a task duration factor associated with either a gpu or with the parkes app )

had to shut down the BOINC client, change the duration_correction_factor in sched_request_einstein.phys.uwm.edu.xml, client_state.xml and client_state_prev.xml, then restart the BOINC client. bunker loaded.

For any new Einstein GPU app crunchers, you can optimize your throughput by running multiple WUs on your GPU. Here's the app_config.xml I use for Einstein...

<app_config>
<app>
<name>einsteinbinary_BRP4G</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>einsteinbinary_BRP5</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>einsteinbinary_BRP6</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
</app_config>

Note, this config uses .5 CPU core per WU, which I think is still the recommended config for AMD GPUs. I think Nvidia GPUs only need .25 cores per WU. Any how, just put the config above in a file named app_config.xml stored in folder \ProgramData\BOINC\projects\einstein.phys.uwm.edu\ , then use the BOINC manager to "Read config files".

Good hunting everyone :pirate:

You can do the same without app_config.xml!
In http://einstein.phys.uwm.edu/prefs.php?subset=project
set GPU utilization factor to 0.5 for any of the three applications you want to run two instances on one GPU.
use 0.33 to run three, 0.25 to run four, etc.

Well sure, if you want to do it the easy way!

I can never remember the Einstein preferences let you control it that way.

scole250 wrote:
I can never remember the Einstein preferences let you control it that way.

You can never remember, I don't think I ever knew that!

Every day is a school day and all that.

@Scole - Really Nice drop! Congrats on that one!

Thanks. We still have our work cut out for us if we want to catch SUSA though.

I see you got the bunkering sorted out then Scole

Yeah, I managed to save up a few before the start. You should give it a try. I hope you have your new 290 warming up.

It's on it now, just need some units to validate....
Did you not notice me crashing in at 254th place

Megacruncher, Janos, Rob, Buster Gunn, Rabbie and all our other GPU crunchers...we could sure use your help in the Einstein challenge. We need to hold off B@P to stay in 3rd, but SETI.Germany is not that far ahead in 2nd. Let's make a run at it! To be honest, we have enough to take SETI.USA for 1st too, but it's going to take all of us. Let's show them who the best team is!

Dang, if I wasn't running native Linux I'd kick 3 7970s your way and help kick SUSA butt. Unfortunately, I don't have a clue as to how to get them to work on Linux and I'm bunkering Denis so I can't switch to Windows yet.

We're sitting 3rd behind SETI.germany and SETI.usa. Not a bad position at all for so few of us compared to those behemoth teams .

Scole is in 1st place, well done sir.
I'm in 8th, Lee Carver is in 90th
Peter Teago 190th
Reeltime 197th
Paul of TSBT 202nd 52 places gained so far
pd 236th
Skywalker TSBT 278th
Giles 283rd
is our top 300 users. Plenty of scope for moving up that board before the end of this one. Keep it up all

Great going team, keep up the fantastic effort!
Scole is putting in some ridiculous numbers at the top

Does anyone else have a huge amount of pending units?
If I could get some units to validate I might manage to bust my way inside the top 200

See what kind of mayhem you can cause if you plan a little bit :)

I have 880 valid, 300 pending.

scole250 wrote:See what kind of mayhem you can cause if you plan a little bit :)

I have 880 valid, 300 pending.

If I was organised Scole, I'd be dangerous

Less than 20% of my uits have validated so far, here's hoping they come through during the comp!

We're holding 3rd, but BOINC@Poland will not let up! We did take a nice leap ahead of them which should hold them off for a little while, but we need more Einstein crunchers. Come on TSBT!

Carpet bombing at it's finest Mr Scole .......

Half way point and we are holding 3rd but it will be a struggle to hold of Boinc@Poland. Any units crunched just now by any member will help regardless of how small the contribution. Lets try and stay on this and hold our line.

If someone would be kind enough to kick the server so some of my units validate that'd be great

Your WUs are going invalid Over heating, driver issue? How many WUs at a time?

Only a percentage are invalid and of those a few have been invalid for at least 2 people. There is far more valids to invalids and of course the tons of pendings as is the usual progress in Einstein.
I actually can't see any difference between the valid units and the invalid. Nothing jumps out that would be an obvious problem.
Plenty of valids so not driver, maybe temp but would expect a lot more errors. Might be memory / pagefile use or just that there is always an error rate on certain projects. Use open hardware monitor or similar and just check that you are not over comitting on memory / cpu / temp etc. Could also just be that windows needs a reboot.
You actually had far more invalids in the first two days and since then it has dropped considerably. Having checked my account your error rate is maybe a bit higher that mine, but I have invalids too. Think there is just simply an error rate on this project.
Certain systems seem to just suit different projects for no reason whatsoever.

Not your WUs. Paul's are going invalid. 9 valid, 28 invalid. Something is wrong.

Wasn't looking at my units, was looking at Pauls, but was looking at the number of pendings to invalids.

This is Pauls system http://einstein.phys.uwm.edu/results.ph ... d=11755274

For comparison here is my system Orac with 2 7970's on this. http://einstein.phys.uwm.edu/results.php?hostid=6068445

My completion times are less than half so I suspect he is running 2 units at a time. The new Parkes is far better optimised than previous, I think the issue is trying to run 2 at a time on the same card.
He also does have a huge number of pendings.

Actually scratch all that, this system of mine Beagle with 1 7970. http://einstein.phys.uwm.edu/results.ph ... d=11723934 has nothing but validation errors, not a single valid.

Going to have to look into this

Right all my nvidias have basically a zero error rate.
With AMD Orac, Heimdal excreta have near enough a zero error rate to. They are all running driver 1.4.1848

Beagle is 100% invalid and is running driver 1.4.1741

I will check the driver version and post it before upgrading. The site doesn't report which driver Paul is using.

All my AMD GPUs are running the 13.251.0.0 driver under WIndows. Running 2 per GPU, very few invalids.

All mine are on 14.1x except Beagle with all the failures which was on 12.8. Updated now , so will see if I start getting valids now.
All my nVidia cards are running Linux using the 340 driver from the Ubuntu repo.

Obviously Einstein no long works with 12.8, but I think if I remember that the new Parkes apps are optimised for the 7 series / R9 AMD's and Maxwell nVidias so the new drivers required would make sense.

Under Paul's account it simply states CAL Hawaii as the driver. Think it points to he has a bad driver install or Einstein is not reading his driver version correctly ???
I've also seen some posts on Einstein that suggests it's only possible to run 1 unit at a time on a R290 or they will go invalid.
Now we must await a responce from the man himself.

I spend a few hours jumping in puddles and the TSBT detectives have fairly been on the case!

I think it might have been a driver issue, windows had flagged up an issue saying my graphics driver had crashed 4+ times in the last day or two. Driver now re installed.
I was also running 3 units at a time, my old card was quite happy with that but that was on the older wu's, dropped it down to 2 now so again I'll keep an eye on the next couple of units and see what happens.

Thanks for the assistance guys

I think it's a good reminder for anyone crunching GPU projects to take a look at the status of WUs at the project website occasionally to make sure everything is validating sucessfully. It's a bummer to run a ton of WUs only to find out they error out. Best to find out sooner than later and figure out what's going on. Has happened to all of us. Of course, all here are glad to help if we can.

Just a little over 3 days to go, so how about little update...

SETI.Germany and SETI.USA are slugging it out for the top 1 an 2 positions.

Meanwhile, TSBT is keeping ahead of BOINC@Poland but they are slowly getting a little closer each update. Those sneaky Polish crunchers

Everyone else is, well, behind us!

Here's the TSBT standings so far (name/position/total) . A good turn out for this challenge ...

davidSandilands/1062/63
Robert Hendry/950/466
Thinboab/911/583
Wolfsschanze/826/1123
tiggsy/679/2788
DonTex/661/3442
b166er/653/3560
Kostoglotov/649/3716
Jeremy Roebuck/637/4188
Mark Mair/636/4286
David Jackson/476/7400
Andrew McCluskey/447/8800
B McT/396/10454
Simon/379/11787
Giles/375/12082
ejfj/344/13200
pd/306/17600
Skywalker TSBT/288/17928
Reeltime/244/26400
Paul of TSBT/197/39600
Peter Teago/193/39725
Lee Carver/144/61600
alez [ TSBT's Pirate ]/6/1504800
scole of TSBT/1/5592400

Paul, you're still getting errors.

Should have thought about this and asked earlier, but did you use the AMD Clean Un-install utility to remove all traces of CCC and the driver first? I've had trouble getting the expected driver installed if I didn't use the utility first.

Other than that, might try 1 WU at a time, maybe downclock the gpu cores.

As Alez reported, multiple reports of errors on R9 290Xs. Recommend 1 WU per GPU and downclock gpu core and memory speed by 20% and see if that works. If so, you can step up until errors occur.

(could always sell it on Ebay and replace with a 7970. Might even be able to get 2 if it sells for enough)

B@P has closed within 700,000 credits and they're getting closer each hour. There's still over 2 days left, but without more einstein crunchers, it doesn't look like we'll be able to hold 3rd. Come on TSBT we can do it!

scole250 wrote:Paul, you're still getting errors.

Should have thought about this and asked earlier, but did you use the AMD Clean Un-install utility to remove all traces of CCC and the driver first? I've had trouble getting the expected driver installed if I didn't use the utility first.

Other than that, might try 1 WU at a time, maybe downclock the gpu cores.

As Alez reported, multiple reports of errors on R9 290Xs. Recommend 1 WU per GPU and downclock gpu core and memory speed by 20% and see if that works. If so, you can step up until errors occur.

(could always sell it on Ebay and replace with a 7970. Might even be able to get 2 if it sells for enough)

I knew there was something I was missing when I re installed the driver. I'll try it again with the un install utility later on.

I'm just on my pc now and owing to a massive $@%& up with my scheduling/project planning and config files not playing the game my computer has been trying to use 5 cores when I only have 4 so I have made little progress all afternoon on Einstein

Now configured to run 1 wu at a time and we'll see what happens.........

Silver wrote:Now configured to run 1 wu at a time and we'll see what happens.........

The last 4 WU completed look good. Only one has gone from Pending to Valid, but the others are "Completed, waiting for validation", not "validation inconclusive".

Only 5 days late.....

Hey, new hardware takes a thorough shakedown :)

OK! Less than 24 hours to go, TSBT is holding 3rd, but do not let up! We are kicking butt!

For those with the challenge bug who just can't stop crunching a challenge without careening into a ditch, at 00:00 UTC, suspend Einstein and let's hit PrimeGrid with everything we got for 39 hours. :pirate:

BTW, I'm going to put a curse on the next team that issues a GPU challenge before September

@alez - I personally think you should go 100% on the PG Challenge...no ulterior motive here at all. :mrgreen: That's just my opinion of course...that would make me... Kidding! You need to be where you need to be, but TSBT isn't showing well in the PG challenge...Hint...hint...!

Don't take it the wrong way Sphynx. I think you were collateral damage. Alez appears to be going for 3rd.

I was going for 3rd too
I may have left my challenge slightly late though and I may also be lacking several dozen gpu's too

scole250 wrote:Don't take it the wrong way Sphynx. I think you were collateral damage. Alez appears to be going for 3rd.

Not the first nor last time I'll have footprints on my back, especially in a gpu challenge. :)

TSBT finished 3 from 14

Individual kudos
scole of TSBT 1st
alez [ TSBT's Pirate ] 5th
Paul of TSBT 98th
Lee Carver 129th
Peter Teago 138th
Reeltime 220th
Skywalker TSBT 305th
Megacruncher TSBT 344th
ejfj 379th
pd 390th
B McT 399th
Andrew McCluskey 432nd
Giles 441st
Simon 444th
David Jackson 592nd
Mark Mair 600th
Jeremy Roebuck 615th
Kostoglotov 623rd
DonTex 656th
tiggsy 780th
b166er 812th
Wolfsschanze 895th
Allan McBain 911th
Thinboab 1,082nd
Robert Hendry 1,115th
davidSandilands 1,221st

Well done all :salute:

Well done everyone, a cracking effort from the team there!
Special mention for Scole setting the pace from the front