Gary Roberts please understand this has nothing to do with the paticular card (7950). I do use the current 13.1 drivers from AMD. Some users have said their problems went away when using the old 12.1 or by underclocking their GPU. I'm not going to reduce my card's abilities just so Einstein will work with old drivers or lower clock speeds when other projects run using current drivers just fine. It's not overclocking either and is at stock speeds and cooling is well below normal. Einstein WUs don't even challenge it as running one E@H ATI WU is usually less than 30% usage for this card. My 6950 computer has been sold since the last post which is why you see no current WUs for it.
This is a problem that several AMD GPU users here are experiencing and its getting worse week by week. This card did work fine prior to November with ATIs from Einstein. It started by flagging only one in 10 or so as invalid. That's an acceptable failure rate. But Einstein is now failing about 2 of every three ATI WUs I do. So as of today.
Look at the results of those in this thread that have brought up this problem. Their WUs fail as ATIs and pass as NVidia WUs. I have absolutely no trouble running DistrGen or Milkyway or Primegrid or World Community Grid ATI workunits. So as I stated in my last post, Einstein is too unreliable to run ATI WUs on the 7950 right now. I'll just do some CPU WUs occassionally.
My AMD 7950 works flawlessly elsewhere but not on the Einstein WUs. The only place it has trouble is where the card is not yet supported by a couple of sites. There is no problem with my card or it's settings which are at stock. I can't vouch for others cards but several people telling you about this should give you a clue that something is amiss other than lots of people having bad or overclocked hardware as you imply. I do millions of points of work just like you do and know how to setup my video card on BOINC to run. Please don't lecture me on overclocking or lack of understanding of how BOINC works. I understand this thoroughly as I've been computing on Einstein since 2009.
Please understand this IS an Einstein problem of some sort in that it is not working well with many AMD GPUs. Several people are telling Einstein there is a problem with some ATI GPUs. Trying to twist the comments to make it seem like those who have brought this up is somehow their fault for improper setup is not very productive to getting the problem fixed and makes people not want to comment at all.
Since most of the ATI WUs are now failing me after I put in several thousand seconds of work anyway, I'll just abort the ones I've got and just stick to CPUs from Einstein for now.
Gary please understand this has nothing to do with the paticular card (7950). You seem to want to blame my card for the GPU invalids as you have done with another commenter. It's not overclocking either. This is a problem that lots of AMD GPU users are experiencing and its getting worse week by week. The only reason you haven't seen much work on my 6950 lately is because I shut it off from Einstein weeks ago for this reason. Einstein is now failing about 2 of every three ATI WUs I do. So as of today. Look at the results of those in this thread that have brought up this problem. They fail as ATIs and pass as NVidia WUs. I have absolutely no trouble running DistrGen or Milkyway or Primegrid or World Community Grid ATI workunits. So as I stated in my last post, Einstein is too unreliable to run ATI WUs right now. I'll just do some CPU WUs occassionally.
My AMD 7950 works flawlessly elsewhere but not on the Einstein WUs. The only place it has trouble is where the card is not yet supported by a couple of sites. There is no problem with my card. I can't vouch for others cards but several people telling you about this should give you a clue that something is amiss other than lots of people having bad or overclocked hardware. I do millions of points of work just like you do and know how to setup my video card on BOINC to run. Please don't lecture me on overclocking or lack of understanding how BOINC works. I understand this thoroughly.
So please understand this IS an Einstein problem of some sort in that it is not working well with many AMD GPUs. Several people are telling Einstein there is a problem with some ATI GPUs. Trying to twist the facts to make it seem like those who have brought this up is somehow their fault is not very productive to getting the problem fixed and makes people not want to comment at all.
Since most of the ATI WUs are now failing after I put in several thousand seconds of work anyway, I'll just abort the ones I've got and just stick to CPUs from Einstein for now.
I certainly was not suggesting that there was anything wrong with your hardware, I was only offering my experiences with what I run in hopes that it might be of help to you or others. Sorry if you were offended in any way.
Sorry Ggesmundo, I apologize for the misunderstanding.
I was refering to Gary Roberts forum moderator and his "it's just your hardware's fault" type of comments. Not your statement. Sorry I didn't notice you were a Gary too. I should have used the full username. No insult taken at all by your comments. Didn't mean to gripe to you. I updated my post to show the full user name I meant to comment to.
Some more food for thought, I had my 7950 running at 1050 Mhz and Milkway ran fine. I had to set it at 1010 Mhz to get Seti to work without errors and as I recently discovered, 990 Mhz to get Einstein to run without errors. So if you are overclocking the fact that other projects run fine does not mean you won't have problems with others. I thought that Milkways dependence on double precision meant that it would be the most finicky about clock speeds. Go figure.
I have absolutely no trouble running DistrGen or Milkyway or Primegrid or World Community Grid ATI workunits. So as I stated in my last post, Einstein is too unreliable to run ATI WUs on the 7950 right now. I'll just do some CPU WUs occassionally.
I experience the opposite, when running Milkyway on ATI 1 in 3 or 1 in 4 are marked invalid. With Einstein only 1 in 25 to 1 in 30.
On nVidia still nor invalids for both projects. And in my case it is not hardware either. I keep my systems clean, in good shape and with plenty of (liquid) cooling. It's the OpenCL thing.
Gary Roberts please
)
Gary Roberts please understand this has nothing to do with the paticular card (7950). I do use the current 13.1 drivers from AMD. Some users have said their problems went away when using the old 12.1 or by underclocking their GPU. I'm not going to reduce my card's abilities just so Einstein will work with old drivers or lower clock speeds when other projects run using current drivers just fine. It's not overclocking either and is at stock speeds and cooling is well below normal. Einstein WUs don't even challenge it as running one E@H ATI WU is usually less than 30% usage for this card. My 6950 computer has been sold since the last post which is why you see no current WUs for it.
This is a problem that several AMD GPU users here are experiencing and its getting worse week by week. This card did work fine prior to November with ATIs from Einstein. It started by flagging only one in 10 or so as invalid. That's an acceptable failure rate. But Einstein is now failing about 2 of every three ATI WUs I do. So as of today.
Look at the results of those in this thread that have brought up this problem. Their WUs fail as ATIs and pass as NVidia WUs. I have absolutely no trouble running DistrGen or Milkyway or Primegrid or World Community Grid ATI workunits. So as I stated in my last post, Einstein is too unreliable to run ATI WUs on the 7950 right now. I'll just do some CPU WUs occassionally.
My AMD 7950 works flawlessly elsewhere but not on the Einstein WUs. The only place it has trouble is where the card is not yet supported by a couple of sites. There is no problem with my card or it's settings which are at stock. I can't vouch for others cards but several people telling you about this should give you a clue that something is amiss other than lots of people having bad or overclocked hardware as you imply. I do millions of points of work just like you do and know how to setup my video card on BOINC to run. Please don't lecture me on overclocking or lack of understanding of how BOINC works. I understand this thoroughly as I've been computing on Einstein since 2009.
Please understand this IS an Einstein problem of some sort in that it is not working well with many AMD GPUs. Several people are telling Einstein there is a problem with some ATI GPUs. Trying to twist the comments to make it seem like those who have brought this up is somehow their fault for improper setup is not very productive to getting the problem fixed and makes people not want to comment at all.
Since most of the ATI WUs are now failing me after I put in several thousand seconds of work anyway, I'll just abort the ones I've got and just stick to CPUs from Einstein for now.
RE: Gary please understand
)
I certainly was not suggesting that there was anything wrong with your hardware, I was only offering my experiences with what I run in hopes that it might be of help to you or others. Sorry if you were offended in any way.
Sorry Ggesmundo, I apologize
)
Sorry Ggesmundo, I apologize for the misunderstanding.
I was refering to Gary Roberts forum moderator and his "it's just your hardware's fault" type of comments. Not your statement. Sorry I didn't notice you were a Gary too. I should have used the full username. No insult taken at all by your comments. Didn't mean to gripe to you. I updated my post to show the full user name I meant to comment to.
Some more food for thought, I
)
Some more food for thought, I had my 7950 running at 1050 Mhz and Milkway ran fine. I had to set it at 1010 Mhz to get Seti to work without errors and as I recently discovered, 990 Mhz to get Einstein to run without errors. So if you are overclocking the fact that other projects run fine does not mean you won't have problems with others. I thought that Milkways dependence on double precision meant that it would be the most finicky about clock speeds. Go figure.
RE: I have absolutely no
)
I experience the opposite, when running Milkyway on ATI 1 in 3 or 1 in 4 are marked invalid. With Einstein only 1 in 25 to 1 in 30.
On nVidia still nor invalids for both projects. And in my case it is not hardware either. I keep my systems clean, in good shape and with plenty of (liquid) cooling. It's the OpenCL thing.
Greetings from
TJ