@Genoil, Thanks. I didn't realise the full nature of the --gpu-devices switch.
@antonio8. Can you do a study up on your PSU. I think your cards might be getting current limited. My rig currently has:
3x Galaxy 750Ti slim
1x emTek 750Ti
1x Gigabyte 750 OC Low profile.
pulling 43MH/s
700W Casecom PSU with 2 12v rails (CC-700W-12CM)
485W at the wall
Issues I had were mainly with the Gigabyte 750 dropping out. It doesn't seem to like the powered risers but works fine pulling power from the PCIe slot itself. Not a problem with the other cards on powered risers. Remember there's only a total of about 175W avaliable to the PCIe slots so 3 Ti's at TDP of 60W will get current limited.
I'll also note that the full size card is running 'much' cooler and faster than the low profiles.
As the 750/Ti's generally don't come with a 6/8 PCIe power connector, I'm left just powering them on Molex connectors. The question mark here is that I don't know which 12V rail the Molex connectors are on so there was a bit of trial, error and hope they weren't all on the same one which would have limited the number of cards without hacking together a PCIe -> Molex adapter.
It's been running nice now. Going from 4 to 5 cards jumped CPU usage from 200% to 400% which was unexpected.
I'm still waiting on one more Gigabyte 750Ti Low profile. I'll then rebuild the rig onto a 6 PCIe slot mobo I received yesterday. The current 5 slot mobo will be retasked with 2 (3 if I can land it) HD 7990's
Never thought about the psu being an issue as I have had no issues on any other algo. I don't have any meter to measure wattage though to see what the system is pulling.
Even running just one card and leaving the other cards idle I don't get the 750ti to go above 2 or 3 Mh/s.
Now you have me thinking and maybe it is because the cards are on powered usb risers that are causing the issue. My rig with 3 960's and 1 750ti are on the same risers but work fine.
I am running 5 EVGA SC Editions with a 700W psu on one and 5 Gigabyte OC with a 650W psu on the other. But even running one card is having the issue.
The 960's run about 50-64% Power with 99-100% gpu load and the 750ti on that rig runs at 35% power.
The 750ti rigs run about 99% load with about 35% power also. This is all from Nvidia Inspector.
I'll play around with the psu but shouldn't one card running by itself run above the 7-8 Mh/s?
@Genoil, Thanks. I didn't realise the full nature of the --gpu-devices switch.
@antonio8. Can you do a study up on your PSU. I think your cards might be getting current limited. My rig currently has:
3x Galaxy 750Ti slim
1x emTek 750Ti
1x Gigabyte 750 OC Low profile.
pulling 43MH/s
700W Casecom PSU with 2 12v rails (CC-700W-12CM)
485W at the wall
Issues I had were mainly with the Gigabyte 750 dropping out. It doesn't seem to like the powered risers but works fine pulling power from the PCIe slot itself. Not a problem with the other cards on powered risers. Remember there's only a total of about 175W avaliable to the PCIe slots so 3 Ti's at TDP of 60W will get current limited.
I'll also note that the full size card is running 'much' cooler and faster than the low profiles.
As the 750/Ti's generally don't come with a 6/8 PCIe power connector, I'm left just powering them on Molex connectors. The question mark here is that I don't know which 12V rail the Molex connectors are on so there was a bit of trial, error and hope they weren't all on the same one which would have limited the number of cards without hacking together a PCIe -> Molex adapter.
It's been running nice now. Going from 4 to 5 cards jumped CPU usage from 200% to 400% which was unexpected.
I'm still waiting on one more Gigabyte 750Ti Low profile. I'll then rebuild the rig onto a 6 PCIe slot mobo I received yesterday. The current 5 slot mobo will be retasked with 2 (3 if I can land it) HD 7990's
I have a few 6x 750ti rigs that i have been trying to get running since the release of frontier. I have spent a lot of time trying to figure out how to get these machines to run. I have successfully mined using my normal computer which has a gtx 770. However I have so far failed to successfully mine using my gpu rigs. But before i go into greater detail about my setup and the specific problems that I have been experiencing, I would love to know what settings you used to get your rig running o0ragman0o?
@VeritasSapere, I've just come in to find my rig had crashed. A few "invalid device ordinal", reboots and some investigation later seems to tell me the 'Gigabyte 750 OC' was failing again. I've had to pull it and am hashing 35.7 on the remaining Ti's (@mrp 8.7 each, don't know what happened to my 9).
I hope this isn't a Gigabyte specific thing as I'm still waiting on shipping for a Gigabyte 750Ti. The Galaxy and emTek 750Ti's haven't skipped a beat.
I have a bunch of 750Tis running on Ubuntu; noticed speeds around 8.8Mhs to 9.8Mhs (PNY stock OC). Was able to use nvidia-settings to OC it to 1.28GHz/6.6Ghz for about 10.6Mhs. I used CUDA7.0.
The high CPU load also seem to stem from the cuda_sync blocking routines. I usually work around this by tying all my threads to a single CPU. This reduces power and cpu load and has no effect on hashing since the CPU cycles are all busy waiting.
Also, one interesting power saving trick I noticed was that if you set the power limit of the 750Ti to 30W from 38.5W, the hash rate does not drop significantly (~5%?), but the power is massively reduced. For my rig of 6 750Tis, I observed a Kill-A-Watt drop 520W to 430W, and it's still currently running at a hash rate of about 51MH/s (all stock cards) and 56MH/s (all OC cards).
I havn't been able to get 'NVIDIA X Server Settings' to work. It alerts that "You do not appear to be using the NVIDIA X driver..." Then suggests running 'nvidia-xconfig' which doesn't seem to change anything. Additional drivers show I'm using version 346.82. I downgraded from 352.21 (shell script installation) in the hope I could overclock using nvidia setting. but alas still nothing.....but I think I know where my extra 1MHs has gone now.
@ennui, have you ever tried flashing the TI's bios to break the 38W limit so as to allow for the full 60W? Very interested to see how far these Ti's can be pushed...
So I dual booted linux on my rig. about 5h later I managed to compile the miner Now my 750Ti mines @ 9.5Mh/s and as you can see from the pic gets much warmer. 50degC vs 65 (win10/linux mint)
good to see it works in Linux. just made a half harted attempt at installing ubuntu, but dropped out after NVidia driver install led me to a blinking cursor of death. god i hate Linux..
good to see it works in Linux. just made a half harted attempt at installing ubuntu, but dropped out after NVidia driver install led me to a blinking cursor of death. god i hate Linux..
I don't exactly hate linux, but it is a big PITA indeed. For me drivers were easy, but CUDA installing gave me the biggest headaches .
BTW I had to remove -Werror from EthCompilerSettings.cmake or it wouldn't compile.
Hey Genoil - thanks for building this cuda version.
the blinking cursor may be caused by saying 'yes' to the dkms stuff in the NVidia driver.... it may be NVidia's implementation crew you hates on that one ....
Anyway, maybe someone can enlighten me -
I'm running Ubuntu 14.04.04 and have been successfully (haz blocks) running the official, repo version of ethminer on my GTX780 Ti.
I had cuda-5.0 installed from a while back and after upgrading to cuda-7.0 I see maybe a slight rise in hashrate, maybe not. I usually see around 16MH/s - 17MH/s, and with the updated cuda version it seems the same.
So I installed cuda-7.0 according to NVidia's instructions and pulled your repo, branch == 'cudaminer', compiled with '-DGUI=0 -DETHASHCL=1'. Seems to compile fine.
Running yours with 'ethminer -G' the hashrate seems pretty much identical to the current repo version.
But when I use 'ethminer -U' it ...really slows down, I get like 150KH/s .
Not clear where I fell off the tracks.
System installed opencl-headers show as 1.2-2013.10.23-1.
Your CMakeCache.txt references cuda-7.0, the official repo's does not. Both yours and the repo are linked to libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1
Does this mean that I am already using cuda with the stock, repo build AND when I don't specify '-U' to your build? Probly not, right?
Your build has instructions for using cuda, the official build shows nothing of the sort..
@afogu you need to add the -DETHASHCU flag on compilation. In your specific case, I would cmake -DBUNDLE=miner -DETHASHCU=1 -DCOMPUTE=35 ..
"35" is for your Compute version. You should be able to pull over 18MH/s with a 780Ti. Use --gpu-workgroup-size 128 --gpu-batch-size 20
The whole opencl / cuda thing is a bit confusing, I agree. The thing is that for OpenCL on NVidia, you need CUDA, whereas or CUDA on NVidia, you don't need OpenCL. So OpenCL on NVidia is built on top of CUDA. My cudaminer is not using OpenCL, but native CUDA. That's also the reason why it is a bit faster.
Genoli -- many thanks for this thread. Took me a couple of days of tweaking to get this up and running, but I have one rig with 6 x 750 ti, linux mint 17, 8gb ddr3, 3.1gz haswell cpu running at 54M+ and best of all, it's coming in at 390 watts at the wall... love that 750 ti!
ℹ 23:58:25|ethminer Mining on PoWhash #9e3b5328… : 54143339 H/s = 471859200 hashes / 8.715 s ℹ 23:58:26|ethminer Mining on PoWhash #9e3b5328… : 54143291 H/s = 508297216 hashes / 9.388 s ℹ 23:58:27|ethminer Mining on PoWhash #9e3b5328… : 54155964 H/s = 545783808 hashes / 10.078 s ℹ 23:58:27|ethminer Mining on PoWhash #9e3b5328… : 54110616 H/s = 581959680 hashes / 10.755 s ℹ 23:58:28|ethminer Mining on PoWhash #9e3b5328… : 54052437 H/s = 617873408 hashes / 11.431 s ℹ 23:58:29|ethminer Mining on PoWhash #9e3b5328… : 54060572 H/s = 654835712 hashes / 12.113 s ℹ 23:58:29|ethminer Mining on PoWhash #9e3b5328… : 54061303 H/s = 691011584 hashes / 12.782 s ℹ 23:58:30|ethminer Mining on PoWhash #9e3b5328… : 54058932 H/s = 728498176 hashes / 13.476 s
@SuchFakeAccount hm weird...i don't even have that file on my system. I did notice however that a step was missing from my windows build instructions on page 5 of this thread. After downloading the source, you need to run 'getstuff.bat' in cpp-ethereum/extdep. Never mind I see thatbyou already did that... What cmake command did you run?
The high CPU load also seem to stem from the cuda_sync blocking routines. I usually work around this by tying all my threads to a single CPU. This reduces power and cpu load and has no effect on hashing since the CPU cycles are all busy waiting.
@ennui, I'm all ears here... How does one go about tying my threads to a single CPU (single core?)
I havn't been able to get 'NVIDIA X Server Settings' to work. It alerts that "You do not appear to be using the NVIDIA X driver..." Then suggests running 'nvidia-xconfig' which doesn't seem to change anything. Additional drivers show I'm using version 346.82. I downgraded from 352.21 (shell script installation) in the hope I could overclock using nvidia setting. but alas still nothing.....but I think I know where my extra 1MHs has gone now.
OK figured it out.... Naturally I'm running my rig headless and remoting in over a persistent xRDP sesman xVNC session. The nvidia-settings worked fine once I plugged in a monitor. I'm guessing XRDP doesn't hook into the NVIDIA X driver. I've installed Xming on my windows box to see if it could work by going through Putty, but no. Neither have I been able to connect xRDP through the sesman-x11RDP.... but this is all getting well off topic.
I'm getting clued up on nvidia-smi now instead.
Downgrading to 346.82 proved to be a really bad move and 4 of my 5 cards became very unstable, something I've only recovered from now after purging and re-running the 352.21 scripted install.
The high CPU load also seem to stem from the cuda_sync blocking routines. I usually work around this by tying all my threads to a single CPU. This reduces power and cpu load and has no effect on hashing since the CPU cycles are all busy waiting.
@ennui, I'm all ears here... How does one go about tying my threads to a single CPU (single core?)
Comments
My 750Ti running the shuffle version only ~2.2MHs on Win10
My 750ti is mining 2.38 to 2.4 Mh/s and Mem conroller load is @ 28%
@antonio8. Can you do a study up on your PSU. I think your cards might be getting current limited.
My rig currently has:
- 3x Galaxy 750Ti slim
- 1x emTek 750Ti
- 1x Gigabyte 750 OC Low profile.
- pulling 43MH/s
- 700W Casecom PSU with 2 12v rails (CC-700W-12CM)
- 485W at the wall
Issues I had were mainly with the Gigabyte 750 dropping out. It doesn't seem to like the powered risers but works fine pulling power from the PCIe slot itself. Not a problem with the other cards on powered risers. Remember there's only a total of about 175W avaliable to the PCIe slots so 3 Ti's at TDP of 60W will get current limited.I'll also note that the full size card is running 'much' cooler and faster than the low profiles.
As the 750/Ti's generally don't come with a 6/8 PCIe power connector, I'm left just powering them on Molex connectors. The question mark here is that I don't know which 12V rail the Molex connectors are on so there was a bit of trial, error and hope they weren't all on the same one which would have limited the number of cards without hacking together a PCIe -> Molex adapter.
It's been running nice now. Going from 4 to 5 cards jumped CPU usage from 200% to 400% which was unexpected.
I'm still waiting on one more Gigabyte 750Ti Low profile. I'll then rebuild the rig onto a 6 PCIe slot mobo I received yesterday. The current 5 slot mobo will be retasked with 2 (3 if I can land it) HD 7990's
Never thought about the psu being an issue as I have had no issues on any other algo. I don't have any meter to measure wattage though to see what the system is pulling.
Even running just one card and leaving the other cards idle I don't get the 750ti to go above 2 or 3 Mh/s.
Now you have me thinking and maybe it is because the cards are on powered usb risers that are causing the issue. My rig with 3 960's and 1 750ti are on the same risers but work fine.
I am running 5 EVGA SC Editions with a 700W psu on one and 5 Gigabyte OC with a 650W psu on the other. But even running one card is having the issue.
The 960's run about 50-64% Power with 99-100% gpu load and the 750ti on that rig runs at 35% power.
The 750ti rigs run about 99% load with about 35% power also. This is all from Nvidia Inspector.
I'll play around with the psu but shouldn't one card running by itself run above the 7-8 Mh/s?
I hope this isn't a Gigabyte specific thing as I'm still waiting on shipping for a Gigabyte 750Ti. The Galaxy and emTek 750Ti's haven't skipped a beat.
I'm running Ubuntu 14.04
@genoil's ethminer ver. 0.9.23
Gigabyte GA-MA770T-UD3P mobo
Phenom II x4 925
4GB DD3
Think we need to start a new thread specifically for rig build issues.....
The high CPU load also seem to stem from the cuda_sync blocking routines. I usually work around this by tying all my threads to a single CPU. This reduces power and cpu load and has no effect on hashing since the CPU cycles are all busy waiting.
Also, one interesting power saving trick I noticed was that if you set the power limit of the 750Ti to 30W from 38.5W, the hash rate does not drop significantly (~5%?), but the power is massively reduced. For my rig of 6 750Tis, I observed a Kill-A-Watt drop 520W to 430W, and it's still currently running at a hash rate of about 51MH/s (all stock cards) and 56MH/s (all OC cards).
Additional drivers show I'm using version 346.82. I downgraded from 352.21 (shell script installation) in the hope I could overclock using nvidia setting. but alas still nothing.....but I think I know where my extra 1MHs has gone now.
@o0ragman0o Heard of it but never tried that; might allow for greater memory OC headroom.
So I dual booted linux on my rig. about 5h later I managed to compile the miner
BTW I had to remove -Werror from EthCompilerSettings.cmake or it wouldn't compile.
the blinking cursor may be caused by saying 'yes' to the dkms stuff
in the NVidia driver.... it may be NVidia's implementation crew
you hates on that one ....
Anyway, maybe someone can enlighten me -
I'm running Ubuntu 14.04.04 and have been successfully (haz blocks) running the official,
repo version of ethminer on my GTX780 Ti.
I had cuda-5.0 installed from a while back and after upgrading to cuda-7.0 I see
maybe a slight rise in hashrate, maybe not.
I usually see around 16MH/s - 17MH/s, and with the updated cuda version it seems the same.
So I installed cuda-7.0 according to NVidia's instructions and pulled your repo,
branch == 'cudaminer', compiled with '-DGUI=0 -DETHASHCL=1'.
Seems to compile fine.
Running yours with 'ethminer -G' the hashrate seems pretty much identical to the
current repo version.
But when I use 'ethminer -U' it ...really slows down, I get like 150KH/s .
Not clear where I fell off the tracks.
System installed opencl-headers show as 1.2-2013.10.23-1.
Your CMakeCache.txt references cuda-7.0, the official repo's does not.
Both yours and the repo are linked to
libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1
Does this mean that I am already using cuda with the stock,
repo build AND when I don't specify '-U' to your build? Probly not,
right?
Your build has instructions for using cuda,
the official build shows nothing of the sort..
"35" is for your Compute version. You should be able to pull over 18MH/s with a 780Ti. Use --gpu-workgroup-size 128 --gpu-batch-size 20
The whole opencl / cuda thing is a bit confusing, I agree. The thing is that for OpenCL on NVidia, you need CUDA, whereas or CUDA on NVidia, you don't need OpenCL. So OpenCL on NVidia is built on top of CUDA. My cudaminer is not using OpenCL, but native CUDA. That's also the reason why it is a bit faster.
ℹ 23:58:25|ethminer Mining on PoWhash #9e3b5328… : 54143339 H/s = 471859200 hashes / 8.715 s
ℹ 23:58:26|ethminer Mining on PoWhash #9e3b5328… : 54143291 H/s = 508297216 hashes / 9.388 s
ℹ 23:58:27|ethminer Mining on PoWhash #9e3b5328… : 54155964 H/s = 545783808 hashes / 10.078 s
ℹ 23:58:27|ethminer Mining on PoWhash #9e3b5328… : 54110616 H/s = 581959680 hashes / 10.755 s
ℹ 23:58:28|ethminer Mining on PoWhash #9e3b5328… : 54052437 H/s = 617873408 hashes / 11.431 s
ℹ 23:58:29|ethminer Mining on PoWhash #9e3b5328… : 54060572 H/s = 654835712 hashes / 12.113 s
ℹ 23:58:29|ethminer Mining on PoWhash #9e3b5328… : 54061303 H/s = 691011584 hashes / 12.782 s
ℹ 23:58:30|ethminer Mining on PoWhash #9e3b5328… : 54058932 H/s = 728498176 hashes / 13.476 s
Error 117 error LNK1181: cannot open input file 'pthread.lib'
I did notice however that a step was missing from my windows build instructions on page 5 of this thread. After downloading the source, you need to run 'getstuff.bat' in cpp-ethereum/extdep.Never mind I see thatbyou already did that...What cmake command did you run?
cmake -DBUNDLE=miner -DETHASHCU=1 -DCOMPUTE=50 -G "Visual Studio 12 2013 Win64" ..
me the newbie will appreaciate it a lot !
I'm getting clued up on
nvidia-sminow instead.Downgrading to 346.82 proved to be a really bad move and 4 of my 5 cards became very unstable, something I've only recovered from now after purging and re-running the 352.21 scripted install.
1. Extract
2. Open terminal
3. cd /to the folder were you extracted it.
4. launch it: ./ethminer -F http://us1.ethpool.org/miner/_your_wallet_address_/10 -U
BTW It was compiled for 750Ti: cmake -DBUNDLE=miner -DETHASHCU=1 -DCOMPUTE=50 -DGUI=0
If other compute version is needed then ask!
UPDATE: Added missing libraries
You mentioned a version of the cuda you are running but I have not been able to find it. Do you have a link for it?
ps -eL | grep cudaminer | grep -v grep | awk '{ print "taskset -cp 1 " $2 }' | bash