Message boards :
Number crunching :
Head Scratcher???
Message board moderation
Author | Message |
---|---|
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
Wondering if somebody can help me to understand this? I have two computers that I run SETI on during the winter. http://setiathome.berkeley.edu/hosts_user.php Both of these computers have the same motherboard, ASUS M4N78 Pro. And they run the same DDR 3 1600 memory with the same timings. Both running Lunatics and both with 5 and 6 work units respectively. If the above link worked you can see SetiOne has a 100MHz faster processor. Then it is running two GTX 650 TI video cards. I am still running 64 bit Vista with ALL the bells and whistles turned off with that machine. In turn setizero has the 100MHz slower processor and two GTS 450 video cards and I am running 64 bit XP. My question is why is setizero kicking SetiOne's butt as far as rack??? As you can see not by much but still. I would have thought the 2 650 TI's would be stomping the 2 GTS 450's all over the place. And the 100MHz faster processor speed would have just been a tiny bit of icing on the cake. Is Vista, with ALL the bells and whistles turned off, the problem? Any help understanding this would be greatly appreciated!!! |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
I have noticed that on my two rigs that I built. The same case, motherboard, CPU, fans. Ram is that same brand Kingston Hyper x blue BUT the slower machine has 8GB ram the faster has 16 GB, the slower machine also has a EVGA 550 Ti with 1 GB ram the faster one has an Nidia 550 Ti with 1 GB ram. The only thing I can think of is Maybe the ram and or the GPU's. I see as much as a 2,000 gap in RAC between the two. Id like to know why also. [/quote] Old James |
rob smith Send message Joined: 7 Mar 03 Posts: 22202 Credit: 416,307,556 RAC: 380 |
First let's sort out your link so we see your computers not our own: http://setiathome.berkeley.edu/hosts_user.php?userid=8826827 The bulk of the processing will be being done by the GPUs not the CPUs, so CPU speed is all but irrelevant. A few questions: Are you using the two computers in exactly the same way, or is one your daily driver and the other just sitting there crunching. You say running 5/6 tasks - is this a total of 5or six tasks, or are you running 5 (6) tasks on each of your GPU. Have you "freed up" cores to feed the CPUs? (Vista is not the greatest OS that MS have ever released...) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
First let's sort out your link so we see your computers not our own: I think that might be one reason, He has Vista on his newest computer and Win XP on the older one. But still hes only showing a 200 RAC differance. Thats nothing. [/quote] Old James |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
Thank you for fixing the link!! Yes, sorry. I am running 5 tasks on each machine with the CPU's. And 6 tasks on both machines with the GPU's. So that leaves one core of CPU to feed the GPU's. I tried leaving 2 cores free of CPU and didn't see any differance so I went back to just 1 on each machine. Both machines ONLY do Seti work, both running 24/7. Both machines are running 8 gig of RAM @1600 with the same timings. The GTS 450's are running at 850 MHz with 1GB of GDDR5 of memory. And 192 Cuda Cores. The GTX 650 TI's are running at 980 MHz with 1 gb of GDDR5 memory. And 768 Cuda cores. I know the little bit faster processor speed doesn't mean anything but the 650 TI's should be blowing the 450's out of the water,,,,,, shouldn't they?? I also realize the old XP machine is JUST beating the Vista one. But I don't see how. That why I am saying head scratcher, LOL!!! |
bill Send message Joined: 16 Jun 99 Posts: 861 Credit: 29,352,955 RAC: 0 |
Look a the difference in Valids. The difference may be in the number of wingmen reporting. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
His one host shows 273 and the other shows 300 [/quote] Old James |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
Those 650's are also under utilized if you're only doing 1 task on each as they'll produce more work by doing 2 tasks each. Cheers. |
bill Send message Joined: 16 Jun 99 Posts: 861 Credit: 29,352,955 RAC: 0 |
His one host shows 273 and the other shows 300 Which means the 300 host was paid mere credits than the 273 host which means a higher rac, right? |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
I'd try cuda50 version for the 650. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
Fred E. Send message Joined: 22 Jul 99 Posts: 768 Credit: 24,140,697 RAC: 0 |
Philhnnss, I have 3 thoughts: 1) Yes, sorry. I am running 5 tasks on each machine with the CPU's. And 6 tasks on both machines with the GPU's. So that leaves one core of CPU to feed the GPU's. That's a bit ambitious with V7. Most of us have had to cut back from what we ran with v6. And your APR's for S@Hv7 / gpu anon platform are fairly low on both machines. Try 2 at a time per gpu for a week to see what happens. 2) You installed the cuda 32 app on both machines but the 650 ti is capable of running cuda 50. Try that app for a week or so and see if it helps. You can just run the installer again, no need to drain the tasks on hand because the app_info.xml will provide for them. 3) On both machines, you're running gpu tasks at the default priority of below normal. Edit mbcuda.cfg in the project's data directory and change that priority to above normal. (Use Notepad). Another Fred Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop. |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
Thank you Fred!!!! OK, on SetiOne, the Vista machine, I upgraded to Cuda 5.035, wouldn't let me go 5.5. Upgraded my Lunatics to Win64 V0.04. Changed my tasks to .50 for the GPU. And changed my CPU to only run 4 tasks. Now I would like to do your last recomendation changing my default priority but I do not have a clue what to do???? I can build a computer. Install Windows. But after that I am lost, LOL!!! Can you give me a "Baby Step" walk through on how to do this? And since you metioned both machines I am guessing I should change this setting on my XP machine as well. So if the steps are differant between XP and Vista could you walk me through that as well?? Thank you for your time!!!! |
rob smith Send message Joined: 7 Mar 03 Posts: 22202 Credit: 416,307,556 RAC: 380 |
See how it runs for a week or so with your new settings - then modify the mbcuda.cfg file. I've highlighted the line you need to change, you need to remove the semi-colon at the start. You might want to see what happens in going from "below normal" "normal" first, just in case it causes any problems. The safe way to do this is copy the whole line, and put it in one line lower, then in the copy remove the semi colon an the "above". Save the file back where it came from and the next time you restart BOINC it will read the file and you should notice GPU tasks take a bit less time - exactly how much depends on a lot of things.... Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Fred E. Send message Joined: 22 Jul 99 Posts: 768 Credit: 24,140,697 RAC: 0 |
Rob beat me to it - only thing I might do different would be to use the priority setting up in the global set rather than bus 1,slot 0 since you have 2 gpu's installed. I think the global setting will apply to both. Do it the same way, just remove the semicolon which makes the line a comment line. Restart BOINC and check the stderr report of some of the first completed tasks to see if it is working on both gpu's. Another Fred Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop. |
rob smith Send message Joined: 7 Mar 03 Posts: 22202 Credit: 416,307,556 RAC: 380 |
Fred, by not "un-semi-coloning" the ;[bus1slot0]line I'm surely working on the global settings? Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Fred E. Send message Joined: 22 Jul 99 Posts: 768 Credit: 24,140,697 RAC: 0 |
Rob, I'm not sure since I only have one gpu and can't test, but I'm talking about the one that is higher in the file: ;;;;; Global applications settings, to apply to all Cuda devices ;;; You can uncomment the processpriority line below, by removing the ';', to engage machine global priority control of x41x ;;; possible options are 'belownormal' (which is the default), 'normal', 'abovenormal', or 'high' ;;; For dedicated crunching machines, 'abovenormal' is recommended ;;; raising global application priorities above the default ;;; may have system dependant usability effects, and can have positive or negative effects on overall throughput processpriority = abovenormal (bold added) Seems like the bottom of the file is an example if you want to get slot specific. Another Fred Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop. |
j tramer Send message Joined: 6 Oct 03 Posts: 242 Credit: 5,412,368 RAC: 0 |
the 650's are slower cards....they have less bandwidth...128 bits versus 256 I think.....makes a big difference |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
Thank you very much guys!!! I will do as you recomend Rob. I'll wait a week or two before I make that final change. My RAC fell considerably when the project was down yesterday, ran out of work on both machines. So I need to wait for it to even out again. If I frack it up I'll be back crying for help!!! |
Fred E. Send message Joined: 22 Jul 99 Posts: 768 Credit: 24,140,697 RAC: 0 |
I use APR (average processing rate) instead of RAC due to the volatility of RAC. That's shown by application on the application details page- from your list of computers, click details and then appliction details. Make a note of the number of tasks completed and the APR for S@H v7 anonymous platform. Wait until that task count has increased about 2 1/2 timeS the number of pending validations you now have so that most validations are new work, and compare the APR's. I see that cuda 50 is now running on the 650ti's so that went okay. If you have a problem with the process priority, yell. Another Fred Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop. |
Philhnnss Send message Joined: 22 Feb 08 Posts: 63 Credit: 30,694,327 RAC: 162 |
the 650's are slower cards....they have less bandwidth...128 bits versus 256 I think.....makes a big difference Well shoot. That's not it in my case. Both of the 650TI's I have; http://www.newegg.com/Product/Product.aspx?Item=N82E16814121686 And the two 450's; http://www.newegg.com/Product/Product.aspx?Item=N82E16814121456 both have the 128 bit memory bandwith. Hopefully what everybody has sugested will help me. But I still would like to know why the 650's were not kicking the 450's butt since they were both at basiclly the same settings? |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.