Welcome to MilkyWay@home

Posts by SLRE

1) Message boards : News : Deprecation of Linux 32bit (Message 63930)
Posted 13 Sep 2015 by SLRE
Post:
Some info that may or may not be relevant to this post:
My Linux Mint 17 32-bit has shown recent clutches of errors on 1.52 N-body - in fact all the N-body units have failed since 11 Sept, though - but the opencl versions of Separation modfit (1.36) and MW@home (1.02) have been consistently OK.

All the N-body fails have been reporting as computing errors but at 00:00:00 elapsed time, which I'd have tagged as a startup error.

The failures don't look as if they go back far, and the error start date did not coincide with a system update at my end.

I had a pretty difficult time getting nvidia driver versions working on any distro (it seems to be a lottery getting a working combo of cuda, opencl, kernel and nvidia drivers), and the 32-bit Mint build was the first that worked. It probably took a week to get opencl working, so I don't plan to rebuild soon.

When are you proposing to cease distributing 32-bit?
2) Message boards : News : server issues (Message 63777)
Posted 28 Jun 2015 by SLRE
Post:
Correction to above; MW doesnt want to run on a gt430. The tasks that ran OK were cpu tasks. So CPU only on that machine more or less indefinitely unless the gt750 issue resolves itself, maybe in a future update.
3) Message boards : News : server issues (Message 63776)
Posted 28 Jun 2015 by SLRE
Post:
For some reason, your computer and only your computer seems to be returning a different results than everyone else. ....

....
I have a couple of exploration options. ...

A bit more exploration.
MW runs OK in single task mode on the GT430 on this machine, albeit with a sprinkling of compute errors. But putting single tasks on the GTX750 has generated another string of 100% validate errors.

I'm switching MW to run solely on the GT430, which will cut throughput a lot (750 is maybe six times faster and the 640 it replaced was also at least 3x the 430 throughput) but better than nothing. SETI and Einstein can have the 750.
If anyone finds an explanation, or has smilar issues with a GTX750 and finds a fix, please post here and I'll open up the 750 again.
4) Message boards : News : server issues (Message 63748)
Posted 21 Jun 2015 by SLRE
Post:
For some reason, your computer and only your computer seems to be returning a different results than everyone else.

...
I have a couple of exploration options. One is to see if MW runs OK on another PC with the same model 750.


Well, I tried that, on host 486179. Looks like it happens _sometimes_ there too - but I see more compute errors rather than validation errors.

Apart from the gpu and one other thing (both are set to run double gpu tasks for ME) the pcs are fairly different - amd vs intel, win 64 vs win32 and different NV driver versions.

Anybody else out there running an ASUS GTX 750?
5) Message boards : News : server issues (Message 63745)
Posted 20 Jun 2015 by SLRE
Post:
For some reason, your computer and only your computer seems to be returning a different results than everyone else. Do you have any custom settings, or did you change graphics drivers recently or anything that might cause this?

Maybe ...
It's a brand new card and replaces a GT640 so driver obviously updated for that (that is, it'll have switched in 750 support - no driver upgrade needed as the NV drivers bundle across ranges). Its also sharing with a GT430, but MW is restricted to the GTX750.
SETI and Einstein seem to be running OK on the 750 though...
I have a couple of exploration options. One is to see if MW runs OK on another PC with the same model 750. Another is to reset, aborting existing MW tasks, and see if newly downloaded tasks run OK...?
6) Message boards : News : server issues (Message 63741)
Posted 18 Jun 2015 by SLRE
Post:
Seem to have a lot of 'validation inconclusive' nvidia opencl units ...


This is normal. The results are simply waiting to be confirmed by other crunchers.


Ermm...pretty much the whole batch shows as 'validate errors'. ...


You asked about "validation inconclusive".

To clarify: When I first posted, they were all showing as 'validation inconclusive' (roughly, 'pending'). By next post, all had moved to 'validation error'. I should clearly have put a 'now' in the Ermmm...

But sitting tight as suggested. Suspended MW on that pc in the mean time as it seems silly to burn gpu until I know what's causing the validation errors. Will watch board for suggestions.
7) Message boards : News : server issues (Message 63734)
Posted 17 Jun 2015 by SLRE
Post:
Ermm...pretty much the whole batch shows as 'validate errors'. In fact everything back to the server dropout shows as a validation error. All but two out of 188 tasks show as validation errors. That seems much less normal ...

To be fair, I put a new card in a week or so back - a new Asus GTX 750. But that model ran fine on another machine and is handling SETI with no problems.

Comments?
8) Message boards : News : server issues (Message 63730)
Posted 16 Jun 2015 by SLRE
Post:
Seem to have a lot of 'validation inconclusive' nvidia opencl units - about 50 over the last couple of days since the wu's started coming in again. Not sure whether that's an anomaly, but it seems unusual enough to let you guys know ...
9) Message boards : News : New Version of Separation Modfit 1.34 (Message 62412)
Posted 28 Sep 2014 by SLRE
Post:
I'm also seeing essentially 100% computation error on 1.34 tasks. Machine is Win 7 running a geforce GT 640.
I'm letting them error naturally so they show up in your logs; they aren't using time.
10) Message boards : News : New Version of Separation Modified Fit (1.32) (Message 62353)
Posted 16 Sep 2014 by SLRE
Post:
For info: The following (representative) jobs all errored out over the last couple of days.

ps_modfit_15_3s_130_wrap_const_1_1405680903_7765724_3
de_modfit_15_3s_132_wrap_8_1410552780_140958_1
de_modfit_15_3s_132_wrap_8_1410552780_141748_0
de_modfit_15_3s_132_wrap_8_1410552780_131333_0
de_modfit_15_3s_132_wrap_8_1410552780_133902_0
de_modfit_15_3s_132_wrap_7_1410552780_130778_1
ps_modfit_15_3s_132_wrap_3_1410552780_136533_0
de_modfit_15_3s_132_wrap_8_1410552780_131960_2

This isn't occasional;it's endemic across all modfit jobs on my Windows machine.
Machine is Win7, Nvidia GT640. Currently running MW jobs doubled up, so that may be contributing ...

On the linux box (Mint, 32-bit, Geforce GTX 660) most 1.30 modfit and some 1.32 teststars jobs are erroring out as other folk have reported.
11) Message boards : Number crunching : PC with AMD GPUs and the LinuxMint 14 64bit Cinnamon system (Message 61914)
Posted 18 Jun 2014 by SLRE
Post:
Any clues as to how to get the nvidia cards working would be welcome: my client is also repoerting 'no usable gpus' on suse 12.3 with nvidia drivers installed and working.
I'm obviously missing something - but what?

12) Message boards : Number crunching : Where's the work? (Message 56325)
Posted 29 Nov 2012 by SLRE
Post:
I have the same issue: No work units for days. Is something amiss?




©2022 Astroinformatics Group