• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

sluggish performance

jstanthr

jstanthr

Limp Gawd
Joined
Nov 11, 2009
Messages
236
on some w/u's i stay at around 600-700 ns/day , but it seems like every 2-3 w/u tht it gets like super sluggish and is the 30 ns/day range, it does that for the entire length of the unit. normally takes me around 45min to an hour to get an average unit done, but the sluggish ones take like 4-5 hours or more. what is causing that? happening to any1 else?
 
Are you processing different units, or are these the same vintage? Some units are worth small amounts of points (i.e. 353 points) and process fast, while others are worth more points (i.e. 1920 points) and can take longer to process. Which units are you referring to as being sluggish?
 
the 1888 point WU's take 5-7 hours depending on your gfx card.. 737 or what ever point they are worth take 3-4 hours and the rest take any where from 1-3 hours..
 
I suggest that you don't use the viewer to monitor your client and instead download a program like HFM.NET or Fahmon to monitor it instead. Those programs will tell you lots of useful info, like percentage progress for the current workunit, the points value of the current workunit, your current points per day average based on the amount of time it takes for you to finish one percent of the unit, etc. I suspect that you're simply running different-size workunits as has already been suggested, and using a monitoring program will tell you for certain whether or not that is the case.
 
Zero82z said:
I suggest that you don't use the viewer to monitor your client and instead download a program like HFM.NET or Fahmon to monitor it instead. Those programs will tell you lots of useful info, like percentage progress for the current workunit, the points value of the current workunit, your current points per day average based on the amount of time it takes for you to finish one percent of the unit, etc. I suspect that you're simply running different-size workunits as has already been suggested, and using a monitoring program will tell you for certain whether or not that is the case.
Click to expand...

Not to mention the default monitor takes a fair bit of CPU power, and has issues displaying anything if your CPU is under load.
 
ok, let me clarify a little bit more, the system is a dedicated folding rig, a c2q q6600 2.4 oc'd to 3.2 on air, never breaks 42-45 c under load on any of the cores, temp from hwmonitor from the chip itself, not the winbond, proc reads like 10c higher than the others. the gpu is a bfg gtx280 oc, it is mildly overclocker above what it came as, and 4gb or corsair ram. running win 7 ult 64bit. i don't know how to tell what type of w/u im getting but i do notice the sizes are different on the regular display window, seems the bigger ones work just fine and on the "performance" line give out around 600-650 ns/day on average. but others even if thery are the same size only get 200-300ns/day and the graphic is slow and jittery, wheras when its at 600-650ns/day its jst flipping around like normal. is there any way to keep it on the faster ones? and another thing, when its on the ones taht are slower ns/day it doesn't use as much resources and it doesn't heat up like normal, and it takes a whole lot longer.
 
its not the process time thats bothering me, its the performance of the w/u itself. reguardless of size, from 8000000 to 20000000 it performs at 600-650 ns/day. and does those in like an hour and the system stays maxxed out and gpu is a balmy 65-66c. BUT on some w/u's they only run at 150-300ns/day system is far from fully loaded, temps go way down to around 50-55c and it takes like 3+ hours.
 
first off.. like zero said.. stop using the built in display window and use a monitoring program like fahmon, fahspy, or HFM.net.. that will give you better answers into why certain WU's are slower then others.. the built in display for F@H GPU effects the performance of the card folding.. and not just by a little bit it can cut speeds on some work units in half where as some it doesnt effect them..

so until you start using a separate F@H monitor theres not much else we can help you with since none of us use the built in display..
 
The built-in viewer is worthless. Stop using it and use a real monitoring program instead.

Sorry for the emphasis, but you didn't seem to notice the first time I said it.
 
well, my next ? was going to be on which one to use and how to set it up, this is the first time ive used F@H so i don't know much about it, never thought id have to apologize for being a noob at something, lol
 
Which one you use is largely a matter of preference. I personally prefer HFM.NET as it is the most feature-packed and customizable of the monitoring programs. Fahmon is the most popular, although HFM.NET has a very similar interface with more options.
 
ok, got hfm.net it says the one im doing is 1888 credit, i just don't understand why it doesn't stress the gpu as hard as the lower credit ones. i guess it's not for me to know
 
Even different workunits with the same point values can stress the GPU differently. It depends on the specific workunit and the kinds of calculations being done. More than that I can't say because I don't know. But the real number you have to pay attention to is the PPD, since that is the best measure of overall performance. As long as you aren't seeing any major fluctuations in that number, you're good to go.
 
i take it there aren't any that have a display similar to the default one that has the protein model on it? i know it serves no real purpose, but fun to look at from time to time. i didn't realize until i quit using it tht it was the prog that was fully loading my 4th core to 100% when its not on, my cpu usage is near 0. but the gpu is still doing its thing. can i run more than just the one gpu client?
 
jstanthr said:
i take it there aren't any that have a display similar to the default one that has the protein model on it? i know it serves no real purpose, but fun to look at from time to time. i didn't realize until i quit using it tht it was the prog that was fully loading my 4th core to 100% when its not on, my cpu usage is near 0. but the gpu is still doing its thing. can i run more than just the one gpu client?
Click to expand...
No, there aren't any similar viewer programs. You can run one GPU client per GPU, so if you have another video card you can run it on that as well. You can also run CPU clients if you want to take advantage of that. If you're interested, the most effective way would be to use Linux virtual machines to run the Linux SMP client, which should put out a respectable amount of points since you've got a good CPU.
 
i was going to setup a dual boot ubunti win7 on here, so i guess when i get that done i will, would it be feasible to run the smp client and the gpu client? or the gpu and 2-3 of the uniproc clients (leaving 1 core for random use) if using the smp client is the best, which of them is the better one, i seen there are guides for 2 different smp clients.
 
i was going to setup a dual boot ubunti win7 on here, so i guess when i get that done i will, would it be feasible to run the smp client and the gpu client? or the gpu and 2-3 of the uniproc clients (leaving 1 core for random use) if using the smp client is the best, which of them is the better one, i seen there are guides for 2 different smp clients.
Click to expand...
There is simply no comparison between the Windows SMP client and the Linux SMP client at the moment. For me, Windows gets about 3800ppd. Whereas Linux gives me closer to 10,000ppd. As people here have said, the best idea would be to download VMWare Player 3.0 and run NotFred's in an instance. I wouldn't advise running on a native Linux install, seeing as you have a GPU as well, and that is so much easier to set up in Windows.

Besides which, you only need one SMP client now. It used to be that VMWare player could only support 2 cores per VM, so that's why people used 2 per C2Q. However, VMWare Player 3.0 supports up to 32 cores per VM, so you only need one.
 
One VM also takes up much less RAM and you don't have to worry about affinities.

Set the priority to idle and forget about it.
 
Do you run 1 vm for the Q6600? I have a Q6600 and 2vm's (NF's) and getting 4k at 3.0ghz. Is that right?
The 10k is for the i7's I believe...
 
Do you run 1 vm for the Q6600? I have a Q6600 and 2vm's (NF's) and getting 4k at 3.0ghz. Is that right?
The 10k is for the i7's I believe...
Click to expand...
Because of RAM limitations I got a boost of about 2000ppd moving from 4 VMs to 1 VM. If you have enough memory to adequately feed those 2 VMs, then the difference will be smaller. But your system will feel more responsive, less sluggish, and besides, having so many instances is an eyesore on the taskbar and the desktop.
 
gwarren007 said:
Do you run 1 vm for the Q6600? I have a Q6600 and 2vm's (NF's) and getting 4k at 3.0ghz. Is that right?
The 10k is for the i7's I believe...
Click to expand...

I would also recommend 1 vm. Less RAM used up, and more responsive system. Not to mention faster WU completion.
I get almost 5k from my 1 VM on my Q6600
 
I've just swopped all my Q6600 from 2x VM's in VMWare Server to 1x VM in VMWare Player.
I've found I'm pulling around a couple of hundred PpDpBox more by doing this.
At 3.1Ghz in now getting ~4.7k PpD.
With 5x boxen changed over, that's around another 1k PpD.
Every little helps.

Luck ............ :D
 
When I set up the 1st instance and let it run I was getting 1980ppd. and I was using 50% Cpu usage.
I had no affinity set.
 
gwarren007 said:
When I set up the 1st instance and let it run I was getting 1980ppd. and I was using 50% Cpu usage.
I had no affinity set.
Click to expand...
You need to change the number of cpus from 2 to 4 in the file in the notfreds folder.
 
gwarren007 said:
When I set up the 1st instance and let it run I was getting 1980ppd. and I was using 50% Cpu usage.
I had no affinity set.
Click to expand...

Did you set VMWare Player to use 4 CPU's ??
Are you useing WinAFC, and if so is it set for 4 CPU's or 2 ??

Luck ....... :D
 
You also have to be using Player 3 to run a 4-core VM. 2.x will only allow 2 cores per VM.
 
Zero, i'm sure you've heard this before, but ur jst awesome. thanks for all the help, im a multiboxer (World of Warcraft) with 5pcs spinning at all times, i was just wasting power before and remembered about the folding thing here, now that i have the setup down thanks to you, i'll have the other 4 folding for [H] tonight, that will be a total of 10 gtx280's 3 c2q 6600's and 2 i7 920's lets see Evga catch up now!! thanks again for all the help. one last question tho, in HFM.NET how do i set it up so it can get the info from the clients on my slave pc's? and is it ok to have the refresh at 1min? id like for it to be real time, but i don't think it'll do that.
 
Ha!

There is going to be some mowing going on..... See if you can get me ;)
 
jstanthr said:
Zero, i'm sure you've heard this before, but ur jst awesome. thanks for all the help, im a multiboxer (World of Warcraft) with 5pcs spinning at all times, i was just wasting power before and remembered about the folding thing here, now that i have the setup down thanks to you, i'll have the other 4 folding for [H] tonight, that will be a total of 10 gtx280's 3 c2q 6600's and 2 i7 920's lets see Evga catch up now!! thanks again for all the help. one last question tho, in HFM.NET how do i set it up so it can get the info from the clients on my slave pc's? and is it ok to have the refresh at 1min? id like for it to be real time, but i don't think it'll do that.
Click to expand...
I know I'm awesome, but thanks for the reinforcement ;). That's certainly a hell of a farm you've got there; if you overclock those machines, you could get over 120k PPD out of them which would put you right about at the top of the team in terms of per-day production :eek:. You should also consider getting set up with -bigadv workunits on those i7 machines, since you can double your PPD output due to the bonus awarded to people who turn in those large workunits fast enough. The best way to do that would be to set up your own custom VM with a full Linux distro and configure the SMP client to run under that, which is a bit more in-depth but still not too difficult with the guide that is available on the forum. I've personally had the best results with Fedora Linux since current versions of Ubuntu seem to have lower F@H performance for some reason.

As for setting up HFM.NET, all you have to do to monitor network PCs is share the F@H folders on each machine and then point HFM.NET to those directories. It will detect the log file and allow you to monitor all the clients.
 
the servers in the farm are a bit different than what your thinking of i think. they have 2 i7xeon type processors and 2 gpgpu units, i have to find the bare c code for the f@h client in order to make it run via the master terminal. it takes a program similar to what u would use on a tesla card to use the gpgpu's which is what the machines were designed for, i get 45% work load, the other 65% is consumed during business ours for my current contractor. i'll get the exact spec sheet for the farm and get back with you for some ideas. (the contractor is a 3d animation firm doing rendering for a feature length film) initial burn in and finishing up the coolant array will be done mid week, if u are familiar with this type of setup i can set you up with a vpn and admin so we can collaberate on it if you are interested.
 
I think that's a bit out of my area of expertise, but if you need any tips on setting things up I'll do my best.
 
You must log in or register to reply here.
Back
Top