>Do you overclock the CPU and/or the GPU?
No.
>What are the temps?
Gpu-z - 43-45 C, gpu load - 0-87%.
>What driver version are you using for the Intel GPU?
10.18.10.3097. (05.08.2014) From msi site. (mb msi pc mate z97).
Couple days ago I've changed allocated memory in bios for intel gpu from 1024 mb to 512 mb. Looks like all errors are done with 512 mb. I'll try to set again at 1024 mb.
Wow, I upgraded by Intel GPU driver from 15.33.22.64.3621 to the upgraded one, 15.36.3.64.3907... every one of my workunits from BRP4 for the Intel GPU are all marked invalid or validate error. I am going to test another few workunits from other projects and see if its the driver or something worse. I have a feeling its the driver, since every workunit prior to September 3 2014 was fine, and after I upgraded the driver they all error out.
New Intel driver 10.18.10.3907 is either broken (no surprises there) or just BOINC doesn't want to play well with it.
When I "upgraded" (for HD4600 on Wintendo 7 64bit), BOINC failed to detect all OpenCL devices. Not just Intel GPU but nVidia and CPU as well!! (it's a laptop that also has a nVidia GPU). CUDA still worked. At that time I wasn't running any OpenCL BOINC applicaions (just CUDA), but I assume they would all fail.
After "downgrading" back to 10.18.10.3621, OpenCL was detected Ok again.
I won't touch Intel drivers again.
I rolled my computer back to the previous driver (10.18.10.3621) and everything is alright again with the BRP4 workunits. It was an issue with the new Intel driver (10.18.10.3907) and the BRP4 app, since the other projects using the updated Intel GPU driver with OpenCL work without any problems. I tested Seti@home, Seti@home beta and Collatz@home, (I forgot if Albert@home has active Intel GPU apps) and all of those workunits crunched just fine with the new Intel GPU driver.
Last time there was an issue was the update between Windows 8 and Windows 8.1, the Intel GPU driver was upgraded by Windows without me knowing about it, but the workunits simply errored out. This time the update caused all of the workunits to not validate or to be invalidated.
I'm leaving the Intel GPU driver alone from now on. There is something about the Intel drivers and the BRP4 app that is very funky. They don't seem to like each other very much. I guess the saying "if it ain't broke, don't fix it" applies here.
Wow, I upgraded by Intel GPU driver from 15.33.22.64.3621 to the upgraded one, 15.36.3.64.3907... every one of my workunits from BRP4 for the Intel GPU are all marked invalid or validate error. I am going to test another few workunits from other projects and see if its the driver or something worse. I have a feeling its the driver, since every workunit prior to September 3 2014 was fine, and after I upgraded the driver they all error out.
I am seeing the same thing on all of my cruncher boxes with the Intel 4600 GPU's since upgrading to the .3907 drivers - roughly a 75%-85% failure rate. In the release notes, Intel mentions some OpenCL speed enhancements, so I suspect whatever they did isn't playing nice with the Einstein app. My SETI WU's are not having any issues since upgrading the driver and have continued to have a 100% validation rate, so it is something specific to Einstein.
I am having the same problem with the broken driver. At first I thought I had a wrong setting in the bios, but since it is a locked cpu I cannot overclock it. I wish I had read this earlier, before producing like 120 or so of invalid work units. After installing the latest version 15.36.7.64.3960 and checking a few WU, they also were invalid.
So today I rolled back to driver version 10.18.10.3621. I wish I had read this earlier, before producing like 120 or so of invalid work units.
Work seems to be flowing
)
Work seems to be flowing again :)
Did you change anything, Bernd?
MrS
Scanning for our furry friends since Jan 2002
66
)
66 errors.
http://einsteinathome.org/account/tasks&offset=0&show_names=1&state=4&appid=0
rig:
intel celeron g1820
app_config:
einsteinbinary_BRP4
1
.9
1 core is used for poem@home gpu.
Another one is crunching simap and at the same time arecibo wu's on intel gpu.
Is it my fault because of such sharing?
You got 3 computation errors
)
You got 3 computation errors and 66 invalid tasks but also 125 valid tasks so not all are failing.
Do you overclock the CPU and/or the GPU?
What are the temps?
What driver version are you using for the Intel GPU?
>Do you overclock the CPU
)
>Do you overclock the CPU and/or the GPU?
No.
>What are the temps?
Gpu-z - 43-45 C, gpu load - 0-87%.
>What driver version are you using for the Intel GPU?
10.18.10.3097. (05.08.2014) From msi site. (mb msi pc mate z97).
Couple days ago I've changed allocated memory in bios for intel gpu from 1024 mb to 512 mb. Looks like all errors are done with 512 mb. I'll try to set again at 1024 mb.
Wow, I upgraded by Intel GPU
)
Wow, I upgraded by Intel GPU driver from 15.33.22.64.3621 to the upgraded one, 15.36.3.64.3907... every one of my workunits from BRP4 for the Intel GPU are all marked invalid or validate error. I am going to test another few workunits from other projects and see if its the driver or something worse. I have a feeling its the driver, since every workunit prior to September 3 2014 was fine, and after I upgraded the driver they all error out.
New Intel driver
)
New Intel driver 10.18.10.3907 is either broken (no surprises there) or just BOINC doesn't want to play well with it.
When I "upgraded" (for HD4600 on Wintendo 7 64bit), BOINC failed to detect all OpenCL devices. Not just Intel GPU but nVidia and CPU as well!! (it's a laptop that also has a nVidia GPU). CUDA still worked. At that time I wasn't running any OpenCL BOINC applicaions (just CUDA), but I assume they would all fail.
After "downgrading" back to 10.18.10.3621, OpenCL was detected Ok again.
I won't touch Intel drivers again.
I rolled my computer back to
)
I rolled my computer back to the previous driver (10.18.10.3621) and everything is alright again with the BRP4 workunits. It was an issue with the new Intel driver (10.18.10.3907) and the BRP4 app, since the other projects using the updated Intel GPU driver with OpenCL work without any problems. I tested Seti@home, Seti@home beta and Collatz@home, (I forgot if Albert@home has active Intel GPU apps) and all of those workunits crunched just fine with the new Intel GPU driver.
Last time there was an issue was the update between Windows 8 and Windows 8.1, the Intel GPU driver was upgraded by Windows without me knowing about it, but the workunits simply errored out. This time the update caused all of the workunits to not validate or to be invalidated.
I'm leaving the Intel GPU driver alone from now on. There is something about the Intel drivers and the BRP4 app that is very funky. They don't seem to like each other very much. I guess the saying "if it ain't broke, don't fix it" applies here.
RE: Wow, I upgraded by
)
I am seeing the same thing on all of my cruncher boxes with the Intel 4600 GPU's since upgrading to the .3907 drivers - roughly a 75%-85% failure rate. In the release notes, Intel mentions some OpenCL speed enhancements, so I suspect whatever they did isn't playing nice with the Einstein app. My SETI WU's are not having any issues since upgrading the driver and have continued to have a 100% validation rate, so it is something specific to Einstein.
I am having the same problem
)
I am having the same problem with the broken driver. At first I thought I had a wrong setting in the bios, but since it is a locked cpu I cannot overclock it. I wish I had read this earlier, before producing like 120 or so of invalid work units. After installing the latest version 15.36.7.64.3960 and checking a few WU, they also were invalid.
So today I rolled back to driver version 10.18.10.3621. I wish I had read this earlier, before producing like 120 or so of invalid work units.
Yes, it's a shame Intel
)
Yes, it's a shame Intel doesn't fix this.
MrS
Scanning for our furry friends since Jan 2002