Now I have removed AMD driver with AMD cleanup tool and I installed ASUS R9 280X driver.
Running the GPU @ 1100MHz on an i7-2600k should be ok - there are others doing just that but under win7. What version of the driver(s) have you installed/tried?
You are running multiple projects but your scheduler log shows 8 cpu cores, ie none reserved for the GPU. There might be something going on here with contention over the cpu core needed to feed the GPU. The 7970 needs to be fed...
I used AMD 13.9 and 13.12 driver without any improvement. I had these problems under Windows 7 too. Now I am using ASUS driver for R9 280X GPU.
I made some changes at preferences so I need to test does it work.
I used AMD 13.9 and 13.12 driver without any improvement. I had these problems under Windows 7 too. Now I am using ASUS driver for R9 280X GPU.
I made some changes at preferences so I need to test does it work.
I have looked at ASUS website drivers and don't see anything there for a Matrix HD7970 & Win8.1 x64. There are drivers listed for other OS's, but they require a matching BIOS update which I don't think is relevant to your card. So that said I would be a bit suspicious of the driver you have chosen. Also ASUS seem to have their own driver numbering scheme and it's hard to tell just what driver they are distributing. Running GPU-Z (from TechPowerUp) would help in ID'ing the driver. TechPowerUp are a good one stop source for AMD WHQL & Beta drivers. Most of their Beta drivers are marked for Win8 and I have had good success w/ 13.11Beta6 on my OC'd 7970. Later versions of the 13.11Beta did not perform as well.
I did some checking of your results on DistrRTGen and you have a load of invalids there as well. Invalids on both projects and on the same card might be telling you something and you need to find the cause.
First step would be isolate the projects - run DistrRTGen only and monitor invalids. Then run E@H w/ BRP4 only and monitor. Then run E@H w/ BRP5 only and monitor. BRP4 & BRP5 load the GPU (and CPU) differently and may be getting in each other's way and there may also be some interference between how different projects use the PCIe3 GPU on a PCIe2 host. If everything shows valid, start recombining 2 at a time to see if you can find what makes it break.
No, there is no problems at any other projects (tested all of them). I have two small kids and you know what happened when you put together kids and computer in one room. This is the reason why I have so many invalid WUs.
There is no problem with drivers too. Now I am using (again) AMD 13.12 driver. Today I created new E@H account and Arecibo WU is crunching for 2100 but Perseus stayed at 4000. I am really confused with this ultra-speed Perseus crunching. My GPU is 85% load, CPU is not load (just overclocked, but I think this could not be problem). My PC is stable. Sometimes I am running stable 4 VM with LHC@Home - Test4Theory.
Is there way to reset preferences to default values?
OK, finally, with little help of my friends, I solved this problem. I am using MSI Afterburner to underclock GPU to 550 MHz and memory to 825 MHz and there is no errors yet.
Which hosts are you comparing against that make you say that?
I have had a quick look through the top hosts list and gave up before finding a comparable host to yours.
There are a lot of configurable options here at Einstein to help you get the best productivity from your machine/s and that can make doing direct comparisons very difficult.
As an example take a look at one of my hosts, This machine is broadly similar to yours but configured differently, if you look at the valid tasks for this machine you will see runtimes far in excess of yours (for Perseus Arm jobs), however... I am running 4 jobs at once on my GPU (as opposed to your single job) and am effectively completing each task in approx. 2550 seconds, essentially twice as fast as you ;-) and there are faster hosts out there!
I am not trying to knock you down Sale, more just trying to point out that its not that simple... You have got a lot of experimenting with task choice, numbers of concurrent tasks to run, GPU only tasks or a mix of GPU, CPU tasks to find the best balance for your computer.
I would suggest you could get faster runtimes and therefore better performance from your machine if you are willing to try :-) But, sadly, as you have had to underclock the Matrix card by so much your mileage may be limited in the grand scheme of things... but you still ought to do some fine tuning!
RE: Now I have removed AMD
)
Running the GPU @ 1100MHz on an i7-2600k should be ok - there are others doing just that but under win7. What version of the driver(s) have you installed/tried?
You are running multiple projects but your scheduler log shows 8 cpu cores, ie none reserved for the GPU. There might be something going on here with contention over the cpu core needed to feed the GPU. The 7970 needs to be fed...
Gord
I used AMD 13.9 and 13.12
)
I used AMD 13.9 and 13.12 driver without any improvement. I had these problems under Windows 7 too. Now I am using ASUS driver for R9 280X GPU.
I made some changes at preferences so I need to test does it work.
Why my GPU is crunching WU
)
Why my GPU is crunching WU double-fast than fastest E@H GPU-s? My crunching time for Perseus 3600 - 4500 and Arecibo 1100-1400. This is very strange.
RE: I used AMD 13.9 and
)
I have looked at ASUS website drivers and don't see anything there for a Matrix HD7970 & Win8.1 x64. There are drivers listed for other OS's, but they require a matching BIOS update which I don't think is relevant to your card. So that said I would be a bit suspicious of the driver you have chosen. Also ASUS seem to have their own driver numbering scheme and it's hard to tell just what driver they are distributing. Running GPU-Z (from TechPowerUp) would help in ID'ing the driver. TechPowerUp are a good one stop source for AMD WHQL & Beta drivers. Most of their Beta drivers are marked for Win8 and I have had good success w/ 13.11Beta6 on my OC'd 7970. Later versions of the 13.11Beta did not perform as well.
I did some checking of your results on DistrRTGen and you have a load of invalids there as well. Invalids on both projects and on the same card might be telling you something and you need to find the cause.
First step would be isolate the projects - run DistrRTGen only and monitor invalids. Then run E@H w/ BRP4 only and monitor. Then run E@H w/ BRP5 only and monitor. BRP4 & BRP5 load the GPU (and CPU) differently and may be getting in each other's way and there may also be some interference between how different projects use the PCIe3 GPU on a PCIe2 host. If everything shows valid, start recombining 2 at a time to see if you can find what makes it break.
Gord
No, there is no problems at
)
No, there is no problems at any other projects (tested all of them). I have two small kids and you know what happened when you put together kids and computer in one room. This is the reason why I have so many invalid WUs.
There is no problem with drivers too. Now I am using (again) AMD 13.12 driver. Today I created new E@H account and Arecibo WU is crunching for 2100 but Perseus stayed at 4000. I am really confused with this ultra-speed Perseus crunching. My GPU is 85% load, CPU is not load (just overclocked, but I think this could not be problem). My PC is stable. Sometimes I am running stable 4 VM with LHC@Home - Test4Theory.
Is there way to reset preferences to default values?
RE: No, there is no
)
Have a look here...
As you can see there is no
)
As you can see there is no problem with MilkyWay, PrimeGrid and DistRTgen.
http://www.primegrid.com/results.php?userid=72330
http://milkyway.cs.rpi.edu/milkyway/results.php?userid=28548
http://boinc.freerainbowtables.com/distrrtgen/results.php?userid=23052&offset=0&show_names=0&state=4&appid=
Last E@H session was disaster!
http://einsteinathome.org/account/tasks&offset=0&show_names=1&state=4&appid=0
OK, finally, with little help
)
OK, finally, with little help of my friends, I solved this problem. I am using MSI Afterburner to underclock GPU to 550 MHz and memory to 825 MHz and there is no errors yet.
http://einsteinathome.org/account/tasks
Thank you friends!
As I can see, my GPU is doing
)
As I can see, my GPU is doing fastest WU run time even with extreme underclocking....
http://einsteinathome.org/account/tasks
RE: As I can see, my GPU is
)
Which hosts are you comparing against that make you say that?
I have had a quick look through the top hosts list and gave up before finding a comparable host to yours.
There are a lot of configurable options here at Einstein to help you get the best productivity from your machine/s and that can make doing direct comparisons very difficult.
As an example take a look at one of my hosts, This machine is broadly similar to yours but configured differently, if you look at the valid tasks for this machine you will see runtimes far in excess of yours (for Perseus Arm jobs), however... I am running 4 jobs at once on my GPU (as opposed to your single job) and am effectively completing each task in approx. 2550 seconds, essentially twice as fast as you ;-) and there are faster hosts out there!
I am not trying to knock you down Sale, more just trying to point out that its not that simple... You have got a lot of experimenting with task choice, numbers of concurrent tasks to run, GPU only tasks or a mix of GPU, CPU tasks to find the best balance for your computer.
I would suggest you could get faster runtimes and therefore better performance from your machine if you are willing to try :-) But, sadly, as you have had to underclock the Matrix card by so much your mileage may be limited in the grand scheme of things... but you still ought to do some fine tuning!
Good luck.
Gavin.