Thank you MAGIC for your detailed and well documented post. I now see how to set the PREFERENCES to maximize hardware utilization. I must confess however that I'm not sure I followed your reasoning regarding the "GPU utilization factor" setting. If I understand you, a ".5" setting, for example, will permit two GPU tasks to run simultaneously. Presumably the default "1" setting runs 1 task at a time. If this in fact is the correct interpretation of the "GPU utilization factor" setting, I would guess that 2 tasks running simultaneously would take much longer to run then 1 running task. I'm using the default setting of "1", and my GPUs run between 50-80% capacity most of the time. But of course, I won't know for sure until I try it, and that's what I intend to do. (I guess this might cause some thermal issues and maybe this is what the GPU utilization WARNING is about, "only change this if you know what you're doing". Do any of us really know what we're doing? Any way I have after market coolers on both my CPU and GPU, so I guess I qualify as a man who knows what he is doing.) Thanks once again, and please don't feel obligated to respond. I am quite comfortable in a state of doubt. I'll work myself though this. As Socrates himself said, "an unexamined doubt is not worth having" or something like that.
That's the right interpretation, the utilization factor tells Boinc that the tasks will use 0.5 of the GPU, ie half of it, so Boinc starts 2 tasks to fully occupy the GPU. Same goes for 0.33 but then we're talking thirds.
Download and run some kind of monitoring software to keep tabs on the temp of both CPU and GPU and you should be fine. I use GPU-Z to check on my GPUs and "Core Temp" for the CPU.
And if you set it at .5 you can run 2 GPU tasks at the same time (I pretty much only run the Perseus Arm Survey tasks) which are 3,333 credits each so you can test and see how many tasks X2 it can run in a 24 hour period to see if you get more tasks per 24 hours done running 2 at a time instead of 1 and I can almost guarantee you will get more done running 2 of the Perseus Arm Survey tasks at a time and you could even run a CPU task at the same time (or even 2 or 3 at the same time) And yes most of us run the GPU-Z free program to tell us just what the video card is clocked at and it's running temp.
I try to keep mine running at 60C or better but then the one in my laptop runs 2 GPU's and 8 CPU tasks at the same time with the card temp usually running about 83C and it has been running 24/7 for almost 2 years now.
And as I mentioned my OC'd (actually super-clocked) GeForce 660Ti runs 4 GPU tasks at the same time and one CPU task and I don't use any fancy coolers and my CPU's are not new top of the line as you can see if you check my hosts and it averages 40,000 credits each day by itself (set at .25)
All I have for cooling is what comes with the card (and use PrecisionX to set that fan to run at 50%) and the side panels off with a 20in fan blowing across all 6 pc's and the CPU's are only using stock fans and heat sinks.
Don't worry about that "warning" you see on the preference page.......just use GPU-Z and make sure the temp of your card is in the lower 60's C
The only extra thing I did was all my power supplies are outside the box and run at room temp. and don't take up room or add heat inside the box.
Hi again MAGIC. I think I'll try to run 2 Perseus Arms on the GTX 480 and let Gravitational Wave S6 run on the cpu threads. Turns out that this is a perfect time to start experimenting. I'm having a driver problem on my einstein machine. It will run, at most, 36 hours, until it either blue sceens or stops responding in a dead freeze. (Gruesome analogy, I know.) Since I did not have this problem until I activated the intel graphics, my guess is that NVidia and intel drivers cannot get along. But thanks to my fellow E@H participants, I can effectively deal with this problem. BRP arecibo v1.34 runs on the intel graphics thread yeilding a wopping 62.50 credits per task. Meanhile BRP arecibo v1.39 and BRP perseus arm run on the GTX 480 yeilding 1000 and 3333 credits per task respectively. Obviously, integrated intel graphics are a long way from replacing high end video cards, assuming my experience has been typical. Anyway, I'm done with the intel graphics. About my GTX 480, no way it will run in the low 60c's with stock cooling. I put the 480 online about a year or so ago and can still remember my shock at seeing it running at 92c, and it wasn't even under full load. I knew that computer components were very inefficient users of power, but I wasn't prepared for that. Only a fraction of the power used goes into processing graphics files, the bulk of the power goes into producing unusable heat adding more entropy into an already disordered universe. (Sorry about that, but I do get a little carried away at times.) Anyway I took it apart, cleaned it, added some decent thermal paste, and replaced the casing with a hybrid cooler, and have been running it ever since without temperatures exceeding ~62c. Your GTX 660 is a very good choice. I'm guessing it has more than 1000 cuda cores and probably a central processing core in the 900-1100 MHz range. I'm also guessing that it doesn't require too much support from your power supplies' 12 volt rail (one six plug, or maybe two six plugs). Since you are running around 60c with stock cooling, that's about as effecient as these things get. My GTX 480 is not nearly as efficient as your GTX 660, but you might be pleased to know that I also run two very efficient cards. I run a biology project on a GTX 650. Asus makes a 650 that is completely powered by the PCIE bus. I believe they rate it at 75 watts. It has 384 cuda cores and nets me ~90,000 credits daily and has never exceeded 55c. But the one I'm most proud of is Sapphire's version of the radeon HD 7970. Its an absolute beast. It has 32 compute units containing 2048 stream processors. (I don't know how to compare this to cuda cores, but suffice it to say, it would be equivalent to lots and lots of cuda cores. I run this with stock cooling and have never seen it hotter than 62c. It usually runs in the low to mid 50c's. I run this in an AMD fx-4300 machine which also runs with stock cooling at ~45c. I run milkway@home on this card and the credits come raining down like manna from heaven - about 245,000 per day. One more thing, the new intel icore processors run hot. If you ever think about upgrading to one of these to take advantage of the hperthreading technology, plan on buying an aftermarket cooler. They are very inefficient users of power. (By the way, I know that CPU-Z is very popular, but I've always used 'Open Hardware Monitor'.) Well thanks once again to you and to the others who have posted to this thread, for taking the time to instruct me and for giving me an opportunity to lecture on computer hardware. You have all enlightened me far beyond my expectations and made me very happy that I asked for help.
.... the bulk of the power goes into producing unusable heat adding more entropy into an already disordered universe. (Sorry about that, but I do get a little carried away at times.) ....
Hey! Entropy rules! Our computers are merely heaters that calculate. :-)
[ However your name is now recorded on the list of those to blame in the far future when the Universe dies of it's Heat Death. ]
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Thank you MAGIC for your
)
Thank you MAGIC for your detailed and well documented post. I now see how to set the PREFERENCES to maximize hardware utilization. I must confess however that I'm not sure I followed your reasoning regarding the "GPU utilization factor" setting. If I understand you, a ".5" setting, for example, will permit two GPU tasks to run simultaneously. Presumably the default "1" setting runs 1 task at a time. If this in fact is the correct interpretation of the "GPU utilization factor" setting, I would guess that 2 tasks running simultaneously would take much longer to run then 1 running task. I'm using the default setting of "1", and my GPUs run between 50-80% capacity most of the time. But of course, I won't know for sure until I try it, and that's what I intend to do. (I guess this might cause some thermal issues and maybe this is what the GPU utilization WARNING is about, "only change this if you know what you're doing". Do any of us really know what we're doing? Any way I have after market coolers on both my CPU and GPU, so I guess I qualify as a man who knows what he is doing.) Thanks once again, and please don't feel obligated to respond. I am quite comfortable in a state of doubt. I'll work myself though this. As Socrates himself said, "an unexamined doubt is not worth having" or something like that.
That's the right
)
That's the right interpretation, the utilization factor tells Boinc that the tasks will use 0.5 of the GPU, ie half of it, so Boinc starts 2 tasks to fully occupy the GPU. Same goes for 0.33 but then we're talking thirds.
Download and run some kind of monitoring software to keep tabs on the temp of both CPU and GPU and you should be fine. I use GPU-Z to check on my GPUs and "Core Temp" for the CPU.
Well Robert yours is a
)
Well Robert yours is a quad-core with GeForce GTX 480 which has 448 CUDA Cores
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-480/specifications
(it says 100C is MAX but you should probably be more in the low 60's C )
And if you set it at .5 you can run 2 GPU tasks at the same time (I pretty much only run the Perseus Arm Survey tasks) which are 3,333 credits each so you can test and see how many tasks X2 it can run in a 24 hour period to see if you get more tasks per 24 hours done running 2 at a time instead of 1 and I can almost guarantee you will get more done running 2 of the Perseus Arm Survey tasks at a time and you could even run a CPU task at the same time (or even 2 or 3 at the same time) And yes most of us run the GPU-Z free program to tell us just what the video card is clocked at and it's running temp.
I try to keep mine running at 60C or better but then the one in my laptop runs 2 GPU's and 8 CPU tasks at the same time with the card temp usually running about 83C and it has been running 24/7 for almost 2 years now.
And as I mentioned my OC'd (actually super-clocked) GeForce 660Ti runs 4 GPU tasks at the same time and one CPU task and I don't use any fancy coolers and my CPU's are not new top of the line as you can see if you check my hosts and it averages 40,000 credits each day by itself (set at .25)
All I have for cooling is what comes with the card (and use PrecisionX to set that fan to run at 50%) and the side panels off with a 20in fan blowing across all 6 pc's and the CPU's are only using stock fans and heat sinks.
Don't worry about that "warning" you see on the preference page.......just use GPU-Z and make sure the temp of your card is in the lower 60's C
The only extra thing I did was all my power supplies are outside the box and run at room temp. and don't take up room or add heat inside the box.
Hi again MAGIC. I think I'll
)
Hi again MAGIC. I think I'll try to run 2 Perseus Arms on the GTX 480 and let Gravitational Wave S6 run on the cpu threads. Turns out that this is a perfect time to start experimenting. I'm having a driver problem on my einstein machine. It will run, at most, 36 hours, until it either blue sceens or stops responding in a dead freeze. (Gruesome analogy, I know.) Since I did not have this problem until I activated the intel graphics, my guess is that NVidia and intel drivers cannot get along. But thanks to my fellow E@H participants, I can effectively deal with this problem. BRP arecibo v1.34 runs on the intel graphics thread yeilding a wopping 62.50 credits per task. Meanhile BRP arecibo v1.39 and BRP perseus arm run on the GTX 480 yeilding 1000 and 3333 credits per task respectively. Obviously, integrated intel graphics are a long way from replacing high end video cards, assuming my experience has been typical. Anyway, I'm done with the intel graphics. About my GTX 480, no way it will run in the low 60c's with stock cooling. I put the 480 online about a year or so ago and can still remember my shock at seeing it running at 92c, and it wasn't even under full load. I knew that computer components were very inefficient users of power, but I wasn't prepared for that. Only a fraction of the power used goes into processing graphics files, the bulk of the power goes into producing unusable heat adding more entropy into an already disordered universe. (Sorry about that, but I do get a little carried away at times.) Anyway I took it apart, cleaned it, added some decent thermal paste, and replaced the casing with a hybrid cooler, and have been running it ever since without temperatures exceeding ~62c. Your GTX 660 is a very good choice. I'm guessing it has more than 1000 cuda cores and probably a central processing core in the 900-1100 MHz range. I'm also guessing that it doesn't require too much support from your power supplies' 12 volt rail (one six plug, or maybe two six plugs). Since you are running around 60c with stock cooling, that's about as effecient as these things get. My GTX 480 is not nearly as efficient as your GTX 660, but you might be pleased to know that I also run two very efficient cards. I run a biology project on a GTX 650. Asus makes a 650 that is completely powered by the PCIE bus. I believe they rate it at 75 watts. It has 384 cuda cores and nets me ~90,000 credits daily and has never exceeded 55c. But the one I'm most proud of is Sapphire's version of the radeon HD 7970. Its an absolute beast. It has 32 compute units containing 2048 stream processors. (I don't know how to compare this to cuda cores, but suffice it to say, it would be equivalent to lots and lots of cuda cores. I run this with stock cooling and have never seen it hotter than 62c. It usually runs in the low to mid 50c's. I run this in an AMD fx-4300 machine which also runs with stock cooling at ~45c. I run milkway@home on this card and the credits come raining down like manna from heaven - about 245,000 per day. One more thing, the new intel icore processors run hot. If you ever think about upgrading to one of these to take advantage of the hperthreading technology, plan on buying an aftermarket cooler. They are very inefficient users of power. (By the way, I know that CPU-Z is very popular, but I've always used 'Open Hardware Monitor'.) Well thanks once again to you and to the others who have posted to this thread, for taking the time to instruct me and for giving me an opportunity to lecture on computer hardware. You have all enlightened me far beyond my expectations and made me very happy that I asked for help.
RE: .... the bulk of the
)
Hey! Entropy rules! Our computers are merely heaters that calculate. :-)
[ However your name is now recorded on the list of those to blame in the far future when the Universe dies of it's Heat Death. ]
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
And by the way MAGIC, forgot
)
And by the way MAGIC, forgot to mention; I just love your setup. Primative, dishevelled, its just very cool!
RE: And by the way MAGIC,
)
Thanks Robert.......that's why I don't have my laptop in the picture since it would look lost there in the Magical Wild Kingdom