This is my PC:
Sabertooth Z77, i7 2600K with Noctua NHU12P (push/pull fan), SSD Samsung 840 PRO 256GB, HDD WD Caviar Black 1TB, PSU Corsair TX 750, RAM Crucial Ballistix Sport 32GB, Case Coolermaster Cosmos S 1100 (air flow is not problem.
Voltage and amps are good (passed Prime95), cooling is more than enough. All my life I am doing electronics and last 20 years IT is my hobby.
GPU is performing perfect with all games and (almost all) benchmarks. 3DMark and PC Mark were excellent but I only have small problems with Unigine Heaven 4.0.
All BOINC projects are running smooth with stock frequencies as you can see (Collatz, SETI-with lunatics, Milkyway, Primegrid).
I am very confused too!
In a normal project, perhaps yes, but not at Einstein@Home.
Yes, particularly at Einstein@Home. The fact that 'resend lost tasks' is 'on' here is irrelevant. The tasks themselves represent extremely little bandwidth. It's all the data files, support files, program files, libs, etc, that have to be replaced. If the host ever did any GW tasks, think of all the large data files that will be automatically redownloaded, even if they are no longer current but just happen to be still listed in the state file. You could end up with GBs of unnecessary downloads just because of locality scheduling which is pretty much unique to Einstein.
Quote:
The only ways to get new work here are:
1. run the work, upload, report, get new.
2. detach the project. This will delete the whole projects\einstein.phys.uwm.edu\ directory and any entry about the project in the client_state.xml file. So that when you reattached to the project, it has nothing to compare to, and will have to send you new work.
You've missed the easiest, quickest and most efficient way to get new work here:
3. In Boinc Manager, select local prefs and add (temporarily) 0.x days to the current work cache setting, where x is just enough to trigger an immediate work fetch of the required task type. Afterwards, change the value back or just return to website prefs if that's what you really want.
And as far as your implication that wasted bandwidth is unimportant is concerned, why teach people that wasting resources doesn't matter? Go waste your own if you really must! :-).
This is my PC:
Sabertooth Z77, i7 2600K with Noctua NHU12P (push/pull fan), SSD Samsung 840 PRO 256GB, HDD WD Caviar Black 1TB, PSU Corsair TX 750, RAM Crucial Ballistix Sport 32GB, Case Coolermaster Cosmos S 1100 (air flow is not problem.
Voltage and amps are good (passed Prime95), cooling is more than enough. All my life I am doing electronics and last 20 years IT is my hobby.
Can't really fault the specs, this machine should be good to go!!
Just a few initial thoughts/questions;
When you run Einstein, does it have your PC to itself or are your other projects running at the same time?
Are you able to test your card at stock clocks in another computer with this project?
What happens if you remove two sticks of RAM and run with the remaining two (in dual channel config)?
Are you using the chipset drivers that came with the motherboard or are you using the latest Intel ones?
What BIOS version for the sabertooth are you using?..
I am running LHC T4T with all my GPU BOINC project because it almost does not impact PC (CPU) performance.
Since I discovered this problem I have been thinking about testing this video card in another PC and test another video card (nVidia and AMD) in my PC. But, this video card is performing well all other tasks. I would rather test another PSU in my PC. I will try to do something in that direction in next few days.
I did not try to remove RAM. It is set @1600MHz with 9-9-9-24 and 1.5V. To be honest I do not think this could be RAM-caused problem. Motherboard has 2 phases dedicated only for memory and RAM itself has been tested before.
Latest BIOS and drivers too.
Since I switch to double WU crunching, errors are here again so I am back at single WU crunching.
OK! Couple minutes ago I snapped and I complete disassemble my PC and assemble it again. Next few hours I will crunch single WU @ 1100 MHz.
Waiting for (good?) results!
OK! Couple minutes ago I snapped and I complete disassemble my PC and assemble it again. Next few hours I will crunch single WU @ 1100 MHz.
Waiting for (good?) results!
Are you still leaving a cpu core open for the gpu to use? Your times went WAAAY up and that could be because your gpu is waiting to be fed.
4,100.88 1,181.61 10.61 pending
My own 7970 is doing units like this:
2,625.77 499.91 3.29 pending
You can see your gpu time, the 2nd number from the left, is almost 3 times mine. My gpu is pure stock right out of the box running at 995 for the gpu core clock and 1375 for the memory clock, both according to gpu-z.
Next week I'll exchange video card with my friend. I'll take his nVidia GTX 660 Ti and test it inside my PC. He is going to test mine Matrix inside his (i7)computer. My PSU is OK (tested last night).
Next week I'll exchange video card with my friend. I'll take his nVidia GTX 660 Ti and test it inside my PC. He is going to test mine Matrix inside his (i7)computer. My PSU is OK (tested last night).
OK, it seems I made it!
Sometimes, while crunching and playing games, my GPU is showing tiny amount of artifacts. I tought these ones could be connected with my problem. I was right! I found solution at asus rog forum where I discovered plenty of Matrix HD 7970 with artifacting problem caused by video memory.
The solution for my problem is to:
- set video memory clock to 6000
- set video memory voltage to 1.520V.
- set GPU voltage to 1.238V
- set default GPU clock (1100 MHz)
OK, it seems I made it!
Sometimes, while crunching and playing games, my GPU is showing tiny amount of artifacts. I tought these ones could be connected with my problem. I was right! I found solution at asus rog forum where I discovered plenty of Matrix HD 7970 with artifacting problem caused by video memory.
The solution for my problem is to:
- set video memory clock to 6000
- set video memory voltage to 1.520V.
- set GPU voltage to 1.238V
- set default GPU clock (1100 MHz)
This is my PC: Sabertooth
)
This is my PC:
Sabertooth Z77, i7 2600K with Noctua NHU12P (push/pull fan), SSD Samsung 840 PRO 256GB, HDD WD Caviar Black 1TB, PSU Corsair TX 750, RAM Crucial Ballistix Sport 32GB, Case Coolermaster Cosmos S 1100 (air flow is not problem.
Voltage and amps are good (passed Prime95), cooling is more than enough. All my life I am doing electronics and last 20 years IT is my hobby.
GPU is performing perfect with all games and (almost all) benchmarks. 3DMark and PC Mark were excellent but I only have small problems with Unigine Heaven 4.0.
All BOINC projects are running smooth with stock frequencies as you can see (Collatz, SETI-with lunatics, Milkyway, Primegrid).
I am very confused too!
RE: In a normal project,
)
Yes, particularly at Einstein@Home. The fact that 'resend lost tasks' is 'on' here is irrelevant. The tasks themselves represent extremely little bandwidth. It's all the data files, support files, program files, libs, etc, that have to be replaced. If the host ever did any GW tasks, think of all the large data files that will be automatically redownloaded, even if they are no longer current but just happen to be still listed in the state file. You could end up with GBs of unnecessary downloads just because of locality scheduling which is pretty much unique to Einstein.
You've missed the easiest, quickest and most efficient way to get new work here:
3. In Boinc Manager, select local prefs and add (temporarily) 0.x days to the current work cache setting, where x is just enough to trigger an immediate work fetch of the required task type. Afterwards, change the value back or just return to website prefs if that's what you really want.
And as far as your implication that wasted bandwidth is unimportant is concerned, why teach people that wasting resources doesn't matter? Go waste your own if you really must! :-).
Cheers,
Gary.
RE: This is my
)
Can't really fault the specs, this machine should be good to go!!
Just a few initial thoughts/questions;
When you run Einstein, does it have your PC to itself or are your other projects running at the same time?
Are you able to test your card at stock clocks in another computer with this project?
What happens if you remove two sticks of RAM and run with the remaining two (in dual channel config)?
Are you using the chipset drivers that came with the motherboard or are you using the latest Intel ones?
What BIOS version for the sabertooth are you using?..
Gavin.
I am running LHC T4T with all
)
I am running LHC T4T with all my GPU BOINC project because it almost does not impact PC (CPU) performance.
Since I discovered this problem I have been thinking about testing this video card in another PC and test another video card (nVidia and AMD) in my PC. But, this video card is performing well all other tasks. I would rather test another PSU in my PC. I will try to do something in that direction in next few days.
I did not try to remove RAM. It is set @1600MHz with 9-9-9-24 and 1.5V. To be honest I do not think this could be RAM-caused problem. Motherboard has 2 phases dedicated only for memory and RAM itself has been tested before.
Latest BIOS and drivers too.
Since I switch to double WU crunching, errors are here again so I am back at single WU crunching.
OK! Couple minutes ago I
)
OK! Couple minutes ago I snapped and I complete disassemble my PC and assemble it again. Next few hours I will crunch single WU @ 1100 MHz.
Waiting for (good?) results!
RE: OK! Couple minutes ago
)
Are you still leaving a cpu core open for the gpu to use? Your times went WAAAY up and that could be because your gpu is waiting to be fed.
4,100.88 1,181.61 10.61 pending
My own 7970 is doing units like this:
2,625.77 499.91 3.29 pending
You can see your gpu time, the 2nd number from the left, is almost 3 times mine. My gpu is pure stock right out of the box running at 995 for the gpu core clock and 1375 for the memory clock, both according to gpu-z.
Next week I'll exchange video
)
Next week I'll exchange video card with my friend. I'll take his nVidia GTX 660 Ti and test it inside my PC. He is going to test mine Matrix inside his (i7)computer. My PSU is OK (tested last night).
RE: Next week I'll exchange
)
Sounds like a good option!
OK, it seems I made
)
OK, it seems I made it!
Sometimes, while crunching and playing games, my GPU is showing tiny amount of artifacts. I tought these ones could be connected with my problem. I was right! I found solution at asus rog forum where I discovered plenty of Matrix HD 7970 with artifacting problem caused by video memory.
The solution for my problem is to:
- set video memory clock to 6000
- set video memory voltage to 1.520V.
- set GPU voltage to 1.238V
- set default GPU clock (1100 MHz)
E@H running smooth. No even pending WU!!!
Thank you friends!
RE: OK, it seems I made
)
Are those settings up or down from the defaults?