15/10-20/10 Full Moon* Rising - MilkyWay@home

[Alez]

http://boincstats.com/en/stats/challenge/team/chat/716

Name Full Moon* Rising
Status Upcoming
Project MilkyWay@home
Issued by Sicituradastra.
Start time 2015-10-15 00:00 UTC
End time 2015-10-20 00:00 UTC
Late entrants allowed? Yes
Number of teams participating » 9
Number of users participating 0

[scole of TSBT]

Hopefully it will cool off some more by then. Hit 88 here today. Have a tahiti gpu down too, Fans dead. Have a new set. Just need to replace them and do the whole PM on the heat sinks in the process.

[Silver]

Ooh, we haven't had a milkyway comp for a while.

[scole of TSBT]

Well I've fired up the heaters for a test run before the comp. Anyone recall the best app_config.xml settings for these apps? And I'm not sure this project will give you enough WUs for a significant bunker either. I can't seem to get more than 40 per GPU and that doesn't last long.

[Silver]

I honestly can't remember now Scole.
I was thinking that I'll need to see if my card will play the game straight away or need some assistance to get crunching.
Once I've clocked up the last few hours I need on Primegrid I'm going to have a look.

[Alez]

Optimisation is simply ... plenty of cooling, wear earplugs and use AMD cards. Tahiti cards rule. Increase power limit, down clock mem and overclock till the power limit.
Don't even think of doing the above without cleaning the card of all dust and ensuring you have the cooling capacity.
My 7970's can do a unit in 12 secs I believe from memory the best I managed was 8 but that was hard work. Will try and find the old threads about it .

[Alez]

I might even dig out my 6990 just for a blast from the past
From a previous MW comp

If you want to get the HD6xxx series to work on all milkyway units.
run amdcleanuputility
install AMD-APP-SDK-v2.8-Windows-64
install 12-10_vista_win7_win8_64_dd_ccc_whql_net4

Should do the trick Worked for me anyway on two different boxes running win 7, one with a HD6990 and one with a 6970 plus a pair of gigabyte 7770's

For the 7970 on win 8 I have no idea as I have not got and I'm trying never to have to have anything to do with win 8 or the vampire that is the win DDM. My 7970 runs fine with the latest drivers under win 7. No reason not to try the above mentioned drivers. Alternatively you could try the fix from this post http://milkyway.cs.rpi.edu/milkyway/for ... stid=61344

more info here http://milkyway.cs.rpi.edu/milkyway/for ... =6&start=0

and then there's always the optimisation thread I wrote way back in the day here

[scole of TSBT]

On mine, the MilkyWay@Home v1.02 take about 53 sec, but the MilkyWay@Home Separation (Modified Fit) might only take 15 sec or less. The pay is better on the plain MilkyWay@Home v1.02 app though.

[scole of TSBT]

Milkyway@home optimization here...http://www.dunadd.co.uk/seti/forum/view ... 172&t=2759

[Dirk Broer]

Full Moon rising...

[Alez]

Full Moon rising...

I am really surprised it has taken this long for a comment to be made...... Back of the class Dirk, turn round and face the wall, naughty step for you :bootyshake:

[scole of TSBT]

Full Moon rising...

I can't believe you just did that! Gang, we're not going to take that from the Lad from Leiden are we? Tune up those GPUs TSBT!

[Silver]

I suspect I have some driver issues, not picking up any work

[Alez]

I suspect I have some driver issues, not picking up any work

That's the R 290 effect again. I'll have a dig and see what the cure is this time .......
Seems the issue is being blamed on older server software that thinks you have a 3800 series card.

app_info.xml time again I'm afraid. Dig out notepad and make the file.

<app_info>
<app>
<name>milkyway_separation__modified_fit</name>
</app>
<file_info>
<name>milkyway_separation__modified_fit_1.36_windows_x86_64__opencl_ati_101.exe</name>
<executable />
</file_info>
<app_version>
<app_name>milkyway_separation__modified_fit</app_name>
<version_num>136</version_num>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.567833</max_ncpus>
<plan_class>opencl_ati_101</plan_class>
<file_ref>
<file_name>milkyway_separation__modified_fit_1.36_windows_x86_64__opencl_ati_101.exe</file_name>
<main_program/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1.0000</count>
</coproc>
</app_version>
<file_info>
<name>milkyway_separation__modified_fit_1.36_windows_x86_64.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway_separation__modified_fit</app_name>
<version_num>136</version_num>
<file_ref>
<file_name>milkyway_separation__modified_fit_1.36_windows_x86_64.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>milkyway</name>
</app>
<file_info>
<name>milkyway_separation_1.20_windows_x86_64__opencl_amd_ati.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway</app_name>
<version_num>120</version_num>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.567833</max_ncpus>
<plan_class>opencl_ati</plan_class>
<file_ref>
<file_name>milkyway_separation_1.20_windows_x86_64__opencl_amd_ati.exe</file_name>
<main_program/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1.0000</count>
</coproc>
</app_version>
<file_info>
<name>milkyway_separation_1.20_windows_x86_64.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway</app_name>
<version_num>120</version_num>
<file_ref>
<file_name>milkyway_separation_1.20_windows_x86_64.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>


You know the drill with where it needs to go etc. C:\ProgramData\BOINC\projects\milkyway.cs.rpi.edu_milkyway
You might also have to download the program executables or it will moan that app_info refers to non existant file etc.
File downloads are here

http://milkyway.cs.rpi.edu/milkyway/dow ... ti_101.exe
http://milkyway.cs.rpi.edu/milkyway/dow ... x86_64.exe
http://milkyway.cs.rpi.edu/milkyway/dow ... md_ati.exe
http://milkyway.cs.rpi.edu/milkyway/dow ... x86_64.exe

[Silver]

Thanks Alez! I'll give that a go.
I guessed it was the same as the primegrid problem.
I'll need to try and work out how the app info file works so I can do it myself next time, although I did have a bash at the primegrid one and failed

[Alez]

It's not difficult to do, I did post somewhere explaining but that was ages ago. Feel free to try and I can debug it for you if you want. The difficult part is that you have to be very presice. Any syntax in the wrong place and it will simply fail. There is no error tolerance.
The app_info you have is for 1 GPU running 1 task per GPU.
Do the optimisation on your account options as per pirate first and see what you get. See if you can beat my scores. Obviously you can go down the running 2, 3 or more units at the same time but be very careful before you do as the R 280 and 290 have a reputation for running very hot on MW. Use open hardware monitor, afterburner etc to confirm the card temps, don't simply trust catalyst.

[scole of TSBT]

1.5 days left to tune up those GPUs. Like Alez said, make sure they don't overheat. Especially the AMD Pitcairn, Tahiti and Hawaii GPUS. Not much of a bunker can be saved on the GPU apps. Project will only give you 40 WUs per GPU. I guess 42k at the start is better than nothing.

[Silver]

Well after the above helpful advice I've got milkyway running again but....
Anyone else getting quite a few validation inconclusive units?

[Alez]

No zero invalids is usual for me, Do you have a core free ? If not do so, if so try downclocking the GPU slightly.
What percentage are we talking, how fast is it running a unit at how much % GPU usage and what temp is the GPU running at ?

[Silver]

3 cores on yoyo
4th is "split" 0.5 cpu + 0.5 gpu for milkyway and also 0.1cpu for miners (although according to resource monitor the miners use less than 0.01%)
191 valid 48 inconclusive at the moment
Run times seem to be around 24 seconds for a seperation modified fit unit, 80 seconds for an @home.
temp is sitting at 63C, gpu usage 100%

[scole of TSBT]

You only have 3 invalid though. I think most end up validating. I have a tahiti doing that too. Also, unless you're looking for WUProp hours, skip Milkyway@Home Separation (Modified Fit) and stick with plain MilkyWay@Home. It pays better per sec.

[scole of TSBT]

Ugh...I forgot about how finicky the Boinc Scheduler can be with Milkyway. I've having trouble keeping work on my systems.

[Alez]

I forgot how much noise and heat is produced.....don't think I can stand this 24/7 for a week.

[scole of TSBT]

I forgot how much noise and heat is produced.....don't think I can stand this 24/7 for a week.

Yeah! Try to host a conf call in with two other people in your office with the door shut! I had to pass out paper towels!

[Silver]

Looking at my stats I guess my pc has crashed, a few hours after I left yesterday, and I think I'm away till Sunday....

[Janos (retired)]

Yeah! Try to host a conf call in with two other people in your office with the door shut! I had to pass out paper towels!

Dude, that sounds like a really interesting conf call

[Alez]

Looking at my stats I guess my pc has crashed, a few hours after I left yesterday, and I think I'm away till Sunday....

I know 25 mins into the comp, I started getting project has no tasks available and so it picked up collatz as the BU project. Not a great start at all. The 2 x 7970's in the living room will be getting turned off soon. Too much noise. Kid off school so his 7970 occupied all day . Looks like it will be 1 main 7970 during the day and the rest at night.

[Alez]

Answers on a post card please.

Orac, i7 no HT 2 x 7970
Milkyway, ave work done 38066, resources share 7500
Collatz, ave work done 646460, resource share 100

result from scheduler - Milkyway Not requesting tasks AMD/ATI GPU not highest project priority.

Just what is the point / logic behind the scheduler , I see none. Absolute joke.
I either have to set all GPU projects bar Milkyway to NNT and abort any remaining work to get Milkyway to work or it does not do what I want. The scheduler has always been bad, but this is now rediculous.

All my systems currently munching on BU project Collatz. and I don't have time to go babysit them today.

[Silver]

Not that I want to try and bring some logic and reason to the scheduler but does it take runtime into account as well as average work done?
I'm sure I've had the same issue as you Alez, I try and run down other work and just leave the project I want to run, appreciate that I've got the advantage of usually being at home and you'll need to have a better plan owing to your work.

[Alez]

TSBT down in 6th Russia Team here we come

[scole of TSBT]

Keep it up gang. I hope my rigs keep running smooth. I'm away on another scout backing trip. I do have cell service so I am keeping up on the challenge.

[Alez]

Didn't realise Orac had crashed ... again. On the plus side we are now up to 5th in the teams.

[Alez]

Chomping away ... my next victim will be another of moon's disciples No more chocolate cake and baileys for you .... Should get me into top 20, aiming for 18th at least.

[Alez]

Made 18th but an Itallian missile when past me like a ... well, missile So now I'm not in 18th

[scole of TSBT]

TSBT finished 5th of 24 teams. Not too shabby gang.

There were 1,166 participants!

TSBT individual positions...
scole of TSBT 8th
alez [ TSBT's Pirate ] 19th
Janos of TSBT 21st
Paul of TSBT 110th
Iain 140th
Megacruncher TSBT 231st
Laird o' th' wee white hoose 284th
allan 297th
Reeltime 344th
John Black 531st
Philippe Demeur 533rd
Garry Ross 791st
Martin Green 850th
Kostoglotov 878th
Grum 944th
Shazzadouglas 1,027th
Radio 1,121st
MarkRBright 1,163rd

Well done all! :salute:

[Janos (retired)]

Nice work guys.

[Silver]

Good job team
Nice to see Janos on the list too :)