OK, there it is and it works better: the R9 270X card GPU % utilization on the slower PC is much more uniform from minute to minute: in addition I let one of the 4 cores be idle. The utilization is 33%: running 1CPU + 1 ATI GPU with "Gamma Ray Pulsar Search #3". 33% GPU is less than I would like but for now it is OK since it is still running off the old Benicia motherboard.
As long as your running "Gamma Ray Pulsar Search #3" tasks you probably won't see much better utilization than that, the app is under development and only part of it is done on the GPU, quite a lot of the calculations are still done on the CPU, but it's still faster than running them purely on CPU. BRP4 and BRP5 task should show much better utilization of the GPU.
For BOINC socket 2011 is not worth it. For about the same cost you could get 2 regular Haswell i7 systems (if you already have everything but CPU, RAM and mainboard). Which would give you 2 more cores with the newer and more energy efficient architecture. The main benefits of LGA2011 are:
- maximum CPU performance in a single relatively affordable socket
- twice as much memory capacity than mainstream platform
- massive PCIe bandwidth (40 lanes)
By the way, for those wondering why the AMD R9 270X is being recognized as a HD 7850/7870 (Pitcairn), this is because these new cards no longer support CAL. So BOINC is a bit confused. BOINC its GPU name detection is based on the CAL support being available in both the drivers and the GPU itself. It will detect some GPU's OpenCL names, but even show those wrong.
I'm sure that with CAL support now definitely being dropped from new GPUs and it never coming back, that the developers will look into alternate ways of getting the correct name on the GPU again.
The R9 270X is not so good as I was hopping. Here is a times for BRP4G tasks:
1. One WUs ~ 40 minutes
2. Two WUs - 01:08
3. 4 WUs - 02:04
4. 5 WUs - 02:35
All four CPUs are free.
So effective time is about 31 minutes for one WU in a best case. My passed away GTX660Ti was providing the same result.
However, to be on bright side - I can't hear noise of this card under the heavy load at all even in the open box.
The R9 270X is not so good as I was hopping. Here is a times for BRP4G tasks:
1. One WUs ~ 40 minutes
2. Two WUs - 01:08
3. 4 WUs - 02:04
4. 5 WUs - 02:35
All four CPUs are free.
So effective time is about 31 minutes for one WU in a best case. My passed away GTX660Ti was providing the same result.
However, to be on bright side - I can't hear noise of this card under the heavy load at all even in the open box.
Yep, and half of two dozen WUs have not passed validation. I've used last stable Catalyst and run this video card in pretty nice conditions - open box, open window (it is minus 3C outside). The temperature was about 45C according to MSI afterburner. Power supply is good too.
I'v been crunching Einstein for more then 4 years and never have experienced more than one validation error in couple of mouths.
Congratulations to AMD since I'm coming back to NVidia.
The R9 270X is not so good as I was hopping. Here is a times for BRP4G tasks:
1. One WUs ~ 40 minutes
2. Two WUs - 01:08
3. 4 WUs - 02:04
4. 5 WUs - 02:35
All four CPUs are free.
So effective time is about 31 minutes for one WU in a best case. My passed away GTX660Ti was providing the same result.
However, to be on bright side - I can't hear noise of this card under the heavy load at all even in the open box.
Yep, and half of two dozen WUs have not passed validation. I've used last stable Catalyst and run this video card in pretty nice conditions - open box, open window (it is minus 3C outside). The temperature was about 45C according to MSI afterburner. Power supply is good too.
I'v been crunching Einstein for more then 4 years and never have experienced more than one validation error in couple of mouths.
Congratulations to AMD since I'm coming back to NVidia.
Try a project like Moo, Collatz, PrimeGrid or DistRTgen for your high end AMD cards, they love them. They work well on MilkyWay too, just not as well as the others I mentioned. As for your invalids are your still running more then two units per gpu? If so drop back and see if that is the problem, if you are only running 2 at a time drop back to one and see if that fixes it, faster but less reliable isn't helpful in the end.
As for your old Nvidia 660 doing better then your new R9-270 that is expected at Einstein, Einstein favors Nvidia gpu's in their programming. I have an Nvidia 780 and it is MUCH faster then my AMD 7970 that was crunching here!!
As for your old Nvidia 660 doing better then your new R9-270 that is expected at Einstein, Einstein favors Nvidia gpu's in their programming. I have an Nvidia 780 and it is MUCH faster then my AMD 7970 that was crunching here!!
I think you'll find that not being due to the project favoring one brand of GPUs over another, but because CUDA (Nvidia) is at this moment so much faster than OpenCL (AMD/Intel/the rest). Nvidia is also only developing for CUDA, no longer for OpenCL. Maybe that that decision will bite them one day in the future, when OpenCL overtakes.
As for your old Nvidia 660 doing better then your new R9-270 that is expected at Einstein, Einstein favors Nvidia gpu's in their programming. I have an Nvidia 780 and it is MUCH faster then my AMD 7970 that was crunching here!!
I need to correct this, I do NOT have a 780, I have a 760 gpu!!
Jord I think you are correct, but as you said only time will tell.
As for your old Nvidia 660 doing better then your new R9-270 that is expected at Einstein, Einstein favors Nvidia gpu's in their programming. I have an Nvidia 780 and it is MUCH faster then my AMD 7970 that was crunching here!!
I just took a look at the statistic page - first 10 lines are computers with AMD 7970/280X and decided to try something not so powerful(literally or not) and expensive as 7970/R9 280X.
I'm considering my next step so could you please provide WUs time for your 760 Gpu?
RE: Thanks Holmis! I was
)
As long as your running "Gamma Ray Pulsar Search #3" tasks you probably won't see much better utilization than that, the app is under development and only part of it is done on the GPU, quite a lot of the calculations are still done on the CPU, but it's still faster than running them purely on CPU. BRP4 and BRP5 task should show much better utilization of the GPU.
For BOINC socket 2011 is not
)
For BOINC socket 2011 is not worth it. For about the same cost you could get 2 regular Haswell i7 systems (if you already have everything but CPU, RAM and mainboard). Which would give you 2 more cores with the newer and more energy efficient architecture. The main benefits of LGA2011 are:
- maximum CPU performance in a single relatively affordable socket
- twice as much memory capacity than mainstream platform
- massive PCIe bandwidth (40 lanes)
MrS
Scanning for our furry friends since Jan 2002
By the way, for those
)
By the way, for those wondering why the AMD R9 270X is being recognized as a HD 7850/7870 (Pitcairn), this is because these new cards no longer support CAL. So BOINC is a bit confused. BOINC its GPU name detection is based on the CAL support being available in both the drivers and the GPU itself. It will detect some GPU's OpenCL names, but even show those wrong.
I'm sure that with CAL support now definitely being dropped from new GPUs and it never coming back, that the developers will look into alternate ways of getting the correct name on the GPU again.
Well, the detection isn't so
)
Well, the detection isn't so bad: R9 270X is basically a HD7870 :D
MrS
Scanning for our furry friends since Jan 2002
The R9 270X is not so good as
)
The R9 270X is not so good as I was hopping. Here is a times for BRP4G tasks:
1. One WUs ~ 40 minutes
2. Two WUs - 01:08
3. 4 WUs - 02:04
4. 5 WUs - 02:35
All four CPUs are free.
So effective time is about 31 minutes for one WU in a best case. My passed away GTX660Ti was providing the same result.
However, to be on bright side - I can't hear noise of this card under the heavy load at all even in the open box.
RE: The R9 270X is not so
)
Yep, and half of two dozen WUs have not passed validation. I've used last stable Catalyst and run this video card in pretty nice conditions - open box, open window (it is minus 3C outside). The temperature was about 45C according to MSI afterburner. Power supply is good too.
I'v been crunching Einstein for more then 4 years and never have experienced more than one validation error in couple of mouths.
Congratulations to AMD since I'm coming back to NVidia.
RE: RE: The R9 270X is
)
Try a project like Moo, Collatz, PrimeGrid or DistRTgen for your high end AMD cards, they love them. They work well on MilkyWay too, just not as well as the others I mentioned. As for your invalids are your still running more then two units per gpu? If so drop back and see if that is the problem, if you are only running 2 at a time drop back to one and see if that fixes it, faster but less reliable isn't helpful in the end.
As for your old Nvidia 660 doing better then your new R9-270 that is expected at Einstein, Einstein favors Nvidia gpu's in their programming. I have an Nvidia 780 and it is MUCH faster then my AMD 7970 that was crunching here!!
RE: As for your old Nvidia
)
I think you'll find that not being due to the project favoring one brand of GPUs over another, but because CUDA (Nvidia) is at this moment so much faster than OpenCL (AMD/Intel/the rest). Nvidia is also only developing for CUDA, no longer for OpenCL. Maybe that that decision will bite them one day in the future, when OpenCL overtakes.
RE: As for your old Nvidia
)
I need to correct this, I do NOT have a 780, I have a 760 gpu!!
Jord I think you are correct, but as you said only time will tell.
RE: As for your old Nvidia
)
I just took a look at the statistic page - first 10 lines are computers with AMD 7970/280X and decided to try something not so powerful(literally or not) and expensive as 7970/R9 280X.
I'm considering my next step so could you please provide WUs time for your 760 Gpu?