Posts by robertmiles
log in
1) Message boards : Windows : A Windows 10 update that blocks OpenCL use (Message 22997)
Posted 483 days ago by Profile robertmiles
Solution found if you have GEFORCE Experience installed - start it. Either tell it to install a newer driver, or click on the three dots to the right of the message telling you which driver you have installed, and tell it to reinstall the driver.

This tells it to install the Nvidia version of the driver, with the OpenCL interface, instead of the Microsoft version, which usually does not have the OpenCL interface.
2) Message boards : Windows : A Windows 10 update that blocks OpenCL use (Message 22996)
Posted 483 days ago by Profile robertmiles
WARNING - the latest 64-bit Windows 10 update blocks OpenCL access to my GT 440 graphics board. CUDA access does not seem to be affected.

I tried to report this problem to Microsoft; the normal problem reporting channels all seem to be blocked.
3) Message boards : Windows : Nvidia OpenCL problem for 364.* drivers (Message 22335)
Posted 624 days ago by Profile robertmiles
The 365.10 driver is now available, and does not fix that problem.
4) Message boards : Windows : Nvidia OpenCL problem for 364.* drivers (Message 22253)
Posted 640 days ago by Profile robertmiles
The OpenCL section of the Nvidia 364.72 driver, and earlier 364.* drivers, has a problem which can cause an entire computer to lock up, or cause a few dozen OpenCL tasks (often not all from the same BOINC project) to give a quick Compute Error. Problem not seen in the 362.00 driver.

Tasks from POEM@home seem the most likely to trigger this problem.

Threads on the problem:

https://www.primegrid.com/forum_thread.php?id=6769#94223

http://boinc.fzk.de/poem/forum_thread.php?id=1205#10896

I have not yet seem this problem for any Collatz Conjecture tasks, but you might want to watch for it.
5) Message boards : Windows : Problems with latest Windows 10 update (Message 21814)
Posted 772 days ago by Profile robertmiles
The latest Windows 10 update causes some problems affecting BOINC users.

First, it disables OpenCL use, at least for my GT 440. Installing the Nvidia version of the most recent graphics driver enables it again, though.

Second, it installs Skype. The last time I tried installing Skype on my Windows 10 desktop, just having it installed (even it not running) slowed down all other programs by about 50 %, so I uninstalled it soon afterwards. I haven't seen anything yet on whether the version this update installs fixes this slowdown problem.

Another problem with less effect on BOINC:

It changed your default browser without your permission. For me, changing it back worked, though.
6) Message boards : News : Google's Anti-Malware Detection Gives False Positives (Message 21612)
Posted 813 days ago by Profile robertmiles
How about using a different antivirus?
I have been using ESET with no issues.


As am I. I love ESET Smart Security, and have been using it for years.

But using a different security software is unrealistic. Most us don't have that kind of money sitting around.


Only if you insist on using paid antivirus programs. Some are free, such as the Microsoft Security Essentials I use under Windows Vista, and the Microsoft Windows Defender I use under Windows 10.

However, Google appears to change their anti-malware detection more than users need to change theirs, since only Google is finding the suspected virus.

Note - having more than one antivirus program installed on your computer at the same time often causes problems when one of them decides that the files the other uses to identify a particular virus are enough like that virus to treat the file as infected.
7) Message boards : News : Google's Anti-Malware Detection Gives False Positives (Message 21609)
Posted 814 days ago by Profile robertmiles
An idea on how the project can stop the false positives without very much effort:

Many BOINC projects have each workunit run from a script rather than starting the application program directly. If the project does this or can be changed to do this, just add a first step to allow sending the same information, but not the thumbprint of bytes that Google marks as a possible virus.

Write a server program to quickly change the pattern of bytes. For example, make it invert every other bit. Write a corresponding program for every type of client supported to reverse this change. Make this corresponding program the first item in the script that runs on the client.

First, apply these changes to the application programs, so that only the changed versions of the application programs can be downloaded from the server. The changed version of the application program needs a new version number, even if it has no other changes, to insure that every user will get the new version.

Modify the workunits to run the script that starts with running the corresponding programs.

If this doesn't stop the false positives, then some other file used by most workunits should get the same type of changes. Repeat as often as needed to stop the false positives.

With these changes, getting Google to tell you if you've stopped the false positives would help you determine how soon you can stop applying more of these changes, but really isn't required if you keep trying longer.
8) Message boards : News : Google's Anti-Malware Detection Gives False Positives (Message 21608)
Posted 814 days ago by Profile robertmiles
How about using a different antivirus?
I have been using ESET with no issues.


Should work if Google can be persuaded to use the different antivirus.

Unlikely to make much difference if only Collatz Conjecture uses the different antivirus.
9) Message boards : Number crunching : Computation Error (Sieve untis) (Message 21546)
Posted 824 days ago by Profile robertmiles
The last 7 GPU sieve workunits on one of my computers all gave the same unknown error number.

Exit status -1073741515 (0xffffffffc0000135) Unknown error number

Nothing in the log file saying why the error occurred.

Computer description in the BOINC log file:

10/15/2015 5:59:56 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 560 (driver version 358.50, CUDA version 7.5, compute capability 2.1, 1024MB, 913MB available, 1116 GFLOPS peak)
10/15/2015 5:59:56 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 560 (driver version 358.50, device version OpenCL 1.1 CUDA, 1024MB, 913MB available, 1116 GFLOPS peak)
10/15/2015 6:00:00 PM | | Host name: Robert-PC
10/15/2015 6:00:00 PM | | Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [Family 6 Model 23 Stepping 10]
10/15/2015 6:00:00 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 syscall nx lm vmx smx tm2 pbe
10/15/2015 6:00:00 PM | | OS: Microsoft Windows Vista: Home Premium x64 Edition, Service Pack 2, (06.00.6002.00)
10/15/2015 6:00:00 PM | | Memory: 8.00 GB physical, 15.66 GB virtual
10/15/2015 6:00:00 PM | | Disk: 919.67 GB total, 539.71 GB free
10) Message boards : Number crunching : Computation Error (Sieve untis) (Message 21545)
Posted 824 days ago by Profile robertmiles
The last 7 of the GPU sieve workunits on one of my computers crashed with computation errors - one after running about 60 times as long as the initial estimated runtime, and the rest within the first 10 seconds. Could you check if there's something wrong with an input files shared by at least the last 6?

All ran with the stock setup.
11) Message boards : News : False Positive: Win32/Bulta!rfn (Message 20478)
Posted 965 days ago by Profile robertmiles
I just reported the false positive to Microsoft, in case no one else has done this.
12) Message boards : Wish list : Relative sizes of the workunit types? (Message 19274)
Posted 1359 days ago by Profile robertmiles
Could you modify the preferences page for selecting which workunit types to run so that it mentions the relative sizes of those types?
13) Message boards : News : CUDA vs nVidia OpenCL User Poll (Message 18557)
Posted 1432 days ago by Profile robertmiles
I'd prefer c, unless you can add an option to let the more experienced users override what it would normally decide to do.

I've seen another BOINC project say that their CUDA application is faster on Nvidia GTX 5xx and older, but their OpenCL application is faster for Nvidia GTX 6xx and newer.
14) Message boards : Number crunching : Nvidia driver crashes (Message 15537)
Posted 1877 days ago by Profile robertmiles
The 306.23 driver gave a similar problem; progress advanced to 0.002% quickly, then didn't change for over an hour.

Nvidia doesn't seem to offer any older versions that are WHQL, so I'm now downloading the latest beta version (310.61).
15) Message boards : Number crunching : Nvidia driver crashes (Message 15536)
Posted 1877 days ago by Profile robertmiles
I'm having problems with 3.11 workunits on my laptop with a G105M.

Shortly after the workunit starts, the driver crashes and apparantly recovers. The workunit then runs very slowly, such as 0.002% progress in 56 hours.

http://boinc.thesonntags.com/collatz/result.php?resultid=128955949



The running slowly afterwards is because the cpu is crunching the unit not the gpu. This was discussed ages ago and I don't remember if a solution was ever found or not. I would back off the driver and/or the Boinc version as a trial though. I use version 7.0.25 for my 560Ti, not here, and it works best for me. It DOES run ALL Nvidia tasks at 'high priority' though so if you run multiple projects that is not a good thing.


The same BOINC version worked with 3.11 previously, but I may have upgraded the driver to 306.97 before the problem appeared.

I run many BOINC projects.

That laptop appears to frequently download too many workunits for the time to the deadlines for those workunits; that appears related to the amount of slowdown I have to set Tthrottle to apply in order to prevent shutdowns from overheating, and the three days of waiting workunits I must have in order to prevent BOINC from ceasing to ask for any more workunits at all, even after it finishes the last one it has. No similar problems on my two desktops, though.
16) Message boards : Number crunching : Nvidia driver crashes (Message 15526)
Posted 1880 days ago by Profile robertmiles
I'm having problems with 3.11 workunits on my laptop with a G105M.

Shortly after the workunit starts, the driver crashes and apparantly recovers. The workunit then runs very slowly, such as 0.002% progress in 56 hours.

http://boinc.thesonntags.com/collatz/result.php?resultid=128955949

17) Message boards : Windows : Nvidia 301.32+ drivers (Message 14056)
Posted 2062 days ago by Profile robertmiles
GPUGRID has announced that they are about to release a new version of their application that requires the 301.32 or higher Nvidia driver. I've already downloaded 301.42. Will you be ready for users who want to share their GPUs between your project and GPUGRID?
18) Message boards : Number crunching : Extra load on Collatz (Message 12180)
Posted 2465 days ago by Profile robertmiles
Is they any way that I can tell ONLY Collatz to reduce the size of my queue? If not, could you ask the BOINC developers for a change that will make the queue length for CPU workunits independent from the queue length for GPU workunits?


The best way to get that to happen is to subscribe to the BOINC_DEV mailing list via http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev and request it. If enough people request it, the boinc development team adds it to their list of "To Do's" for a future release. That's the best way. If a single project admin requests it, that's only one person asking for it. Getting the whole community to request a change along with one or more project admins agreeing will make it a higher priority.

FYI, all of the BOINC email lists and their descriptions can be found at http://boinc.berkeley.edu/email_lists.php


Done.
19) Message boards : Number crunching : Extra load on Collatz (Message 12179)
Posted 2465 days ago by Profile robertmiles
Is they any way that I can tell ONLY Collatz to reduce the size of my queue? If not, could you ask the BOINC developers for a change that will make the queue length for CPU workunits independent from the queue length for GPU workunits?


If you crunch for different projects for the cpu and gpu this won't happen. If you have multiple pc's you can flip flop the projects running Collatz as a gpu project on one pc and then as a cpu project on the other pc, with a different project using the other resources.


I've been crunching for different projects for the CPU and GPU for about as long as I've had a suitable GPU. It wasn't adequate to avoid that problem.
20) Message boards : Number crunching : Extra load on Collatz (Message 12160)
Posted 2467 days ago by Profile robertmiles
Something I've noticed that puts an extra load on the Collatz server:

Whenever my laptop asks for more Collatz workunits, it usually gets about twice as many as it can complete by the deadline, so I have to abort around half of them before they start. Collatz is the only BOINC project it is connected to that is currently offering GPU workunits.

I'm using this add-on to prevent the laptop from frequently overheating and going into sleep mode:

http://efmer.eu/boinc/

In the past, I've found that I have to tell BOINC to request a 3-day queue of workunits in order to get it to download CPU workunits for that laptop at all.

TThrottle appears to slow down at least the CPU workunits enough that that queue actually takes something like a week to execute.

Is they any way that I can tell ONLY Collatz to reduce the size of my queue? If not, could you ask the BOINC developers for a change that will make the queue length for CPU workunits independent from the queue length for GPU workunits?


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.