So the recommendation is to go dumpster diving behind apartments at the end of the semester in University towns for prospective (old, cheap) boinc crunchers? :)
Tom M
YES!! I have picked up many a pc on the side of the road waiting for the trash man, not all were good for crunching but they had parts I could put 'in stock' so I didn't have to buy them later on when something broke. Often I would find unwiped hard drives in them so I got a copy of Darik's Boot and Nuke program that wipes a drive of everything and anything from a cd drive
I'm putting out about 10,000,000 credits per day on E@H all on Windows machines. Granted, I've got 27 recent generation GPU's (Nvidia 3000 series and AMD 6000 series) helping. I also have 128 processors of CPU's crunching. I also crunch on Universe at home, Milkyway at home, & Asteroids at home.
So the recommendation is to go dumpster diving behind apartments at the end of the semester in University towns for prospective (old, cheap) boinc crunchers? :)
Tom M
Many major universities operate surplus stores that sell used items, including well-used computer equipment. I have purchased scientific equipment, conference tables, office chairs, desks, dressers, and vehicles from such surplus stores. Some Ebay sellers use these stores to acquire their inventory.
"I was born in a small town, and I live in a small town." - John Mellencamp
I'm putting out about 10,000,000 credits per day on E@H all on Windows machines. Granted, I've got 27 recent generation GPU's (Nvidia 3000 series and AMD 6000 series) helping. I also have 128 processors of CPU's crunching. I also crunch on Universe at home, Milkyway at home, & Asteroids at home.
just some advice.
For Einstein, Nvidia performs much better here (as long as you have recent drivers to run the newer app. you are).
For Universe. you will double (or more) your production if you switch to a modern Linux distribution.
For Milkyway, your AMD cards will probably perform better than your nvidia cards there. milkyway benefits most from FP64 performance.
For Asteroids, they only have CUDA apps, so you can't run your AMD cards there.
depending on project, you may find some bottlenecks using USB risers. a lot of projects don't have a lot of data flowing over the PCIe bus, but depending on how you end up configuring BOINC with the tasks you're running, it *could* matter and be a source of slowdown. if you want to ensure the fastest possible speed with USB risers, make sure they are at least running PCIe gen 3.0 and replace the USB cables with higher quality cables if you have to. mining never cared about USB cables and could run them at PCIe gen 1 if they wanted. but getting at least gen 3 is more important to keep good speeds on BOINC projects. supporting gen3 over USB cables usually required higher quality cables in my experience, due to the more strict signaling and shielding requirements of gen 3.0.
Thanks... I'm primarily crunching on E@H right now because discovering a Pulsar and getting a framed discovery certificate is pretty cool. On the hardware front, to keep my electricity bill from being $700 USD a month I actual underclock my GPU's by about 20%. (Plus, my house only has a 100-amp panel, so I have to monitor how much I'm running through my breakers and outlets) Even with that reduction in power I'm currently at about 10.5 million credits per day. I'm running on PCIE Gen 3 but will look into my USB cables. The two issues I'm having right now is, my 13 GPU Win 10 rig can be unstable at times (Not shocking I know) so I'm either going to split it up or switch it over to Linux. My other issue is I'm having a hell of a time getting a GPU to work on my dual cpu server board. I'm not very experienced with Linux or Servers so it will be fun trying to figure it out. Thanks again for the advice and congrats on the impressive daily credits you're pulling down. Knowing how much I have to run to get about 10.5 mil credits and you pulling in almost 28 mil a day is impressive to say the least. Hats off to you!
I’m actually doing about 45 million per day at the moment. 28 is just the recent average. There was a work shortage for FGRPB1G that restricted my production on these tasks, but that shortage stopped a few days ago and I’m back to crunching full speed again. It’ll take some time for the average to catch up.
my 7-GPU system (4x 3080Ti, 3x 2080Ti) by itself does about 23.5M per day.
That's right... I remember reading about that on a different post... Sadly its Linux only so I can't use it for now. Do you know of any Windows apps available?
Well you already have the Windows version of our optimized app. Our developer passed on the performance enhancements in the source code to the project admins who released the 1.28 (FGRPopencl2Pup-nvidia) stock app. That is a rudimentary version of our optimized app which the developer took much further later on. But for Linux only.
Tom M wrote: So the
)
YES!! I have picked up many a pc on the side of the road waiting for the trash man, not all were good for crunching but they had parts I could put 'in stock' so I didn't have to buy them later on when something broke. Often I would find unwiped hard drives in them so I got a copy of Darik's Boot and Nuke program that wipes a drive of everything and anything from a cd drive
I'm putting out about
)
I'm putting out about 10,000,000 credits per day on E@H all on Windows machines. Granted, I've got 27 recent generation GPU's (Nvidia 3000 series and AMD 6000 series) helping. I also have 128 processors of CPU's crunching. I also crunch on Universe at home, Milkyway at home, & Asteroids at home.
Tom M wrote: So the
)
Many major universities operate surplus stores that sell used items, including well-used computer equipment. I have purchased scientific equipment, conference tables, office chairs, desks, dressers, and vehicles from such surplus stores. Some Ebay sellers use these stores to acquire their inventory.
"I was born in a small town, and I live in a small town." - John Mellencamp
Locations Unknown wrote: I'm
)
just some advice.
For Einstein, Nvidia performs much better here (as long as you have recent drivers to run the newer app. you are).
For Universe. you will double (or more) your production if you switch to a modern Linux distribution.
For Milkyway, your AMD cards will probably perform better than your nvidia cards there. milkyway benefits most from FP64 performance.
For Asteroids, they only have CUDA apps, so you can't run your AMD cards there.
depending on project, you may find some bottlenecks using USB risers. a lot of projects don't have a lot of data flowing over the PCIe bus, but depending on how you end up configuring BOINC with the tasks you're running, it *could* matter and be a source of slowdown. if you want to ensure the fastest possible speed with USB risers, make sure they are at least running PCIe gen 3.0 and replace the USB cables with higher quality cables if you have to. mining never cared about USB cables and could run them at PCIe gen 1 if they wanted. but getting at least gen 3 is more important to keep good speeds on BOINC projects. supporting gen3 over USB cables usually required higher quality cables in my experience, due to the more strict signaling and shielding requirements of gen 3.0.
_________________________________________________________________________
Thanks... I'm primarily
)
Thanks... I'm primarily crunching on E@H right now because discovering a Pulsar and getting a framed discovery certificate is pretty cool. On the hardware front, to keep my electricity bill from being $700 USD a month I actual underclock my GPU's by about 20%. (Plus, my house only has a 100-amp panel, so I have to monitor how much I'm running through my breakers and outlets) Even with that reduction in power I'm currently at about 10.5 million credits per day. I'm running on PCIE Gen 3 but will look into my USB cables. The two issues I'm having right now is, my 13 GPU Win 10 rig can be unstable at times (Not shocking I know) so I'm either going to split it up or switch it over to Linux. My other issue is I'm having a hell of a time getting a GPU to work on my dual cpu server board. I'm not very experienced with Linux or Servers so it will be fun trying to figure it out. Thanks again for the advice and congrats on the impressive daily credits you're pulling down. Knowing how much I have to run to get about 10.5 mil credits and you pulling in almost 28 mil a day is impressive to say the least. Hats off to you!
I’m actually doing about 45
)
I’m actually doing about 45 million per day at the moment. 28 is just the recent average. There was a work shortage for FGRPB1G that restricted my production on these tasks, but that shortage stopped a few days ago and I’m back to crunching full speed again. It’ll take some time for the average to catch up.
my 7-GPU system (4x 3080Ti, 3x 2080Ti) by itself does about 23.5M per day.
_________________________________________________________________________
Why do you prefer the FGRPB1G
)
Why do you prefer the FGRPB1G over the other tasks?
We have a optimized app that
)
We have a optimized app that runs them faster than anything else. So more daily production and credits with that type of work.
That's right... I remember
)
That's right... I remember reading about that on a different post... Sadly its Linux only so I can't use it for now. Do you know of any Windows apps available?
Well you already have the
)
Well you already have the Windows version of our optimized app. Our developer passed on the performance enhancements in the source code to the project admins who released the 1.28 (FGRPopencl2Pup-nvidia) stock app. That is a rudimentary version of our optimized app which the developer took much further later on. But for Linux only.