35 replies to this topic

[naapi]

Hi,

Two SMP servers 171.64.65.63 and 171.64.122.76 are down. They normally accept results from SMP clients, but are experiencing issues. I hope I won't lose work due to this development.

M.

[dnamechanic]

Two SMP servers 171.64.65.63 and 171.64.122.76 are down. They normally accept results from SMP clients, but are experiencing issues. I hope I won't lose work due to this development.

Apparently Dr. Pande made an announcement, indicating it would be resolved soon.

Looks like a discussion going about this at the Stanford Folding forums.

Since these servers are not listed here, my guess is that they must provide Windows SMP work units and maybe exclusively quad-core units?

The deadline is pretty tight for SMP WUs (often 3-4 days or less), hopefully the server issue will not cause a deadline miss for you.

[dnamechanic]

Naapi, from the Stats, looks you have taken quite a hit from the server problem.

One thing that helps a bit is to have the computers set up with enough regular Folding clients to provide one client for each core.

In your case that could be several clients

The tips that come to mind are:

- to have each client in a separate folder, and
- to start each client with the local flag (-local).

The local flag can be added to the -advmethods flag, described in this thread. Add the -local right after, or before, the -advmethods with a space between, like this: -advmethods -local

- It is probably best to use the 'Console client' for multiple instances like this. Not sure the Windows Graphical client allows multiple instances to run.
- Probably best to make sure that the MachineID is different for each clent (machineid=1, machineid=2,..). the limit here is eight, but that should be sufficient for a system with eight cores, or less. The numbers, one through eight, can repeat in a different computer because the computer itself has a UserID. The UserID is created by the first install of the client). In Windows, it is usually tucked away in the registry.

When first run the client will go through configuration options. So the machineid can be set when configuring each cleint for the first time.

The client.cfg can be edited with Windows Notepad, but it is probably a 'nix file, so one must be careful not to disturb any non-alphanumeric looking characters. Note: Stanford guys say not to Edit with Notepad! So, using the configuration utility is probably safest.

Then, when the SMP server is known to be down, activate the regular clients. This can help make points on computers that would otherwise be idle, especially with the current batch of Bonus Point WUs (-advmethods).

Edited by dnamechanic, 20 May 2008 - 09:24 PM.

[Johan]

I'm using the F@H 5.03 Windows client, and recently, it doesn't seem to be able to send results. All it says is "Could not connect to Work Server (results)". Does anyone know what's wrong?

[naapi]

I'm using the F@H 5.03 Windows client, and recently, it doesn't seem to be able to send results. All it says is "Could not connect to Work Server (results)". Does anyone know what's wrong?

Johan,

You can always refer to the following page:
http://fah-web.stanf...serverstat.html

You have server statuses there. You can look up the server that your client wants to connect to and check whether it is down or refusing WUs. At this very moment I see that at least 5 servers are down.

Maciek.

[Johan]

I'm using the F@H 5.03 Windows client, and recently, it doesn't seem to be able to send results. All it says is "Could not connect to Work Server (results)". Does anyone know what's wrong?

Johan,

You can always refer to the following page:
http://fah-web.stanf...serverstat.html

You have server statuses there. You can look up the server that your client wants to connect to and check whether it is down or refusing WUs. At this very moment I see that at least 5 servers are down.

Maciek.

Thank you. It seems to be working correctly now, though.

[StrangeAeons]

I installed a client on my parents' computer, and I haven't seen the username on the stats yet (same as the username here).
Also, I've tried installing both GPU clients for Windows Vista and both times it crashed saying that files were missing. :(

EDIT: files missing related to CUDA, installing new driver now. Hopefully that should take care of it.

Edited by PetaKiaRose, 26 September 2008 - 06:26 PM.

[kismet]

I installed a client on my parents' computer, and I haven't seen the username on the stats yet (same as the username here).

I think the computers won't show up in the stats (my 2 computers don't show up in the extreme folding stats) but you should get the points credited under your username.

Also, I've tried installing both GPU clients for Windows Vista and both times it crashed saying that files were missing. :(

EDIT: files missing related to CUDA, installing new driver now. Hopefully that should take care of it.

If files are missing try to find them via windows search and copy them into the gpu-folding folder. That's what I've had to do, I use Win XP though.

[sentinel]

I installed a client on my parents' computer, and I haven't seen the username on the stats yet (same as the username here).
Also, I've tried installing both GPU clients for Windows Vista and both times it crashed saying that files were missing. :(

EDIT: files missing related to CUDA, installing new driver now. Hopefully that should take care of it.

Hmm Vista is a pain in the @rse. Ihad a similar problem and basically gpu client would not record the card because the drivers were not current enough. Is it by anychance a low 8xxx card running on a vista laptop?

[StrangeAeons]

fortunately, after installing the driver everything's up and running right now. It's going at about 1% of a WU every 11 minutes. Is that normal with a GPU client?

[Luna]

Just a question,
I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

[dnamechanic]

I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

News of such a problem, if confirmed, would likely spread quickly. Reading these forums plus many of the large hardware group forums have not revealed that such a problem is common.

The wording "2D graphics card" implies memory damage to regular (non high-end) graphics cards. Certainly this has not been a problem with the classical folding clients, that is the non-GPU clients (list of clients here).

Overclocking (OC) the GPUs in high end cards can certainly raise temperatures of the GPU and the graphics card memory, which in principle, could damage components. Heat is almost always a concern when OC'ing. Usually if components are overheating, the results will be so undesirable (failure to complete work units) that the OC'er will notice and make adjustments before any damage is done.

A common tactic in OC'ing is to raise the Voltage on either the processor or suppporting hardware (such as RAM). Excessive overVolting, in addition to raising temperatures, can directly damage components. OC'ers in efforts to get the last MHz out of RAM have been known to destroy it.

Also, failure to provide adequate cooling in any folding system (overclocked or not) can cause problems. But, usually the problems are remedied by taking measures to reduce temperatures.

Winterbreeze, possibly reports of such damage is uncommon, or isolated, or could be speculation.

[sentinel]

I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

News of such a problem, if confirmed, would likely spread quickly. Reading these forums plus many of the large hardware group forums have not revealed that such a problem is common.

The wording "2D graphics card" implies memory damage to regular (non high-end) graphics cards. Certainly this has not been a problem with the classical folding clients, that is the non-GPU clients (list of clients here).

Overclocking (OC) the GPUs in high end cards can certainly raise temperatures of the GPU and the graphics card memory, which in principle, could damage components. Heat is almost always a concern when OC'ing. Usually if components are overheating, the results will be so undesirable (failure to complete work units) that the OC'er will notice and make adjustments before any damage is done.

A common tactic in OC'ing is to raise the Voltage on either the processor or suppporting hardware (such as RAM). Excessive overVolting, in addition to raising temperatures, can directly damage components. OC'ers in efforts to get the last MHz out of RAM have been known to destroy it.

Also, failure to provide adequate cooling in any folding system (overclocked or not) can cause problems. But, usually the problems are remedied by taking measures to reduce temperatures.

Winterbreeze, possibly reports of such damage is uncommon, or isolated, or could be speculation.

I have an issue semi related to this. I acquired a graphics card in knew to be faulty (snowy graphics) which I put in one of my few PCI express slots. It was worse than I expected and I could only just make out the screens, it then went to a blue screen and I had to reboot. The problem is when i put a working card in the same slot now it also produced snowy graphics, but if I plug the screen in to the low-end-no-use-for-folding card that i sintegrated into the mobo the picture is fine.

I have re-installed the Cuda drivers and re-inserted etc but it seems the slot is now non-funtional. Burnt out? fixable? Any ideas appreciated!

[sentinel]

I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

News of such a problem, if confirmed, would likely spread quickly. Reading these forums plus many of the large hardware group forums have not revealed that such a problem is common.

The wording "2D graphics card" implies memory damage to regular (non high-end) graphics cards. Certainly this has not been a problem with the classical folding clients, that is the non-GPU clients (list of clients here).

Overclocking (OC) the GPUs in high end cards can certainly raise temperatures of the GPU and the graphics card memory, which in principle, could damage components. Heat is almost always a concern when OC'ing. Usually if components are overheating, the results will be so undesirable (failure to complete work units) that the OC'er will notice and make adjustments before any damage is done.

A common tactic in OC'ing is to raise the Voltage on either the processor or suppporting hardware (such as RAM). Excessive overVolting, in addition to raising temperatures, can directly damage components. OC'ers in efforts to get the last MHz out of RAM have been known to destroy it.

Also, failure to provide adequate cooling in any folding system (overclocked or not) can cause problems. But, usually the problems are remedied by taking measures to reduce temperatures.

Winterbreeze, possibly reports of such damage is uncommon, or isolated, or could be speculation.

I have an issue semi related to this. I acquired a graphics card in knew to be faulty (snowy graphics) which I put in one of my few PCI express slots. It was worse than I expected and I could only just make out the screens, it then went to a blue screen and I had to reboot. The problem is when i put a working card in the same slot now it also produced snowy graphics, but if I plug the screen in to the low-end-no-use-for-folding card that i sintegrated into the mobo the picture is fine.

I have re-installed the Cuda drivers and re-inserted etc but it seems the slot is now non-funtional. Burnt out? fixable? Any ideas appreciated!

Come on guys! I can't believe you're as stupid as me, post-mortem or phoenix maker, just talk to me.

[Mind]

Sorry Sentinel, I can't help you here. I am not even sure if the client I installed on my wife's single core XP computer is working correctly.

[Krell]

I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

News of such a problem, if confirmed, would likely spread quickly. Reading these forums plus many of the large hardware group forums have not revealed that such a problem is common.

The wording "2D graphics card" implies memory damage to regular (non high-end) graphics cards. Certainly this has not been a problem with the classical folding clients, that is the non-GPU clients (list of clients here).

Overclocking (OC) the GPUs in high end cards can certainly raise temperatures of the GPU and the graphics card memory, which in principle, could damage components. Heat is almost always a concern when OC'ing. Usually if components are overheating, the results will be so undesirable (failure to complete work units) that the OC'er will notice and make adjustments before any damage is done.

A common tactic in OC'ing is to raise the Voltage on either the processor or suppporting hardware (such as RAM). Excessive overVolting, in addition to raising temperatures, can directly damage components. OC'ers in efforts to get the last MHz out of RAM have been known to destroy it.

Also, failure to provide adequate cooling in any folding system (overclocked or not) can cause problems. But, usually the problems are remedied by taking measures to reduce temperatures.

Winterbreeze, possibly reports of such damage is uncommon, or isolated, or could be speculation.

I have an issue semi related to this. I acquired a graphics card in knew to be faulty (snowy graphics) which I put in one of my few PCI express slots. It was worse than I expected and I could only just make out the screens, it then went to a blue screen and I had to reboot. The problem is when i put a working card in the same slot now it also produced snowy graphics, but if I plug the screen in to the low-end-no-use-for-folding card that i sintegrated into the mobo the picture is fine.

I have re-installed the Cuda drivers and re-inserted etc but it seems the slot is now non-funtional. Burnt out? fixable? Any ideas appreciated!

Come on guys! I can't believe you're as stupid as me, post-mortem or phoenix maker, just talk to me.

Have you checked the power supply? Maybe it can not put out enough juice.

Congrats on your recent ppd increase! I may have to bring in more heavy artillery to keep you from passing me.

[sentinel]

I believe I have read on a forum that some people reported 2D graphics card memory damage after using F@H for few days.
Could anyone confirm the risk with this?

News of such a problem, if confirmed, would likely spread quickly. Reading these forums plus many of the large hardware group forums have not revealed that such a problem is common.

The wording "2D graphics card" implies memory damage to regular (non high-end) graphics cards. Certainly this has not been a problem with the classical folding clients, that is the non-GPU clients (list of clients here).

Overclocking (OC) the GPUs in high end cards can certainly raise temperatures of the GPU and the graphics card memory, which in principle, could damage components. Heat is almost always a concern when OC'ing. Usually if components are overheating, the results will be so undesirable (failure to complete work units) that the OC'er will notice and make adjustments before any damage is done.

A common tactic in OC'ing is to raise the Voltage on either the processor or suppporting hardware (such as RAM). Excessive overVolting, in addition to raising temperatures, can directly damage components. OC'ers in efforts to get the last MHz out of RAM have been known to destroy it.

Also, failure to provide adequate cooling in any folding system (overclocked or not) can cause problems. But, usually the problems are remedied by taking measures to reduce temperatures.

Winterbreeze, possibly reports of such damage is uncommon, or isolated, or could be speculation.

I have an issue semi related to this. I acquired a graphics card in knew to be faulty (snowy graphics) which I put in one of my few PCI express slots. It was worse than I expected and I could only just make out the screens, it then went to a blue screen and I had to reboot. The problem is when i put a working card in the same slot now it also produced snowy graphics, but if I plug the screen in to the low-end-no-use-for-folding card that i sintegrated into the mobo the picture is fine.

I have re-installed the Cuda drivers and re-inserted etc but it seems the slot is now non-funtional. Burnt out? fixable? Any ideas appreciated!

Come on guys! I can't believe you're as stupid as me, post-mortem or phoenix maker, just talk to me.

Have you checked the power supply? Maybe it can not put out enough juice.

Congrats on your recent ppd increase! I may have to bring in more heavy artillery to keep you from passing me.

I am the worst of all foes Krell, I have resources, seemingly limitless time and energy and a complete recklessness born of really not knowing quite what you’re doing. I am your nemesis, and I'm messy.

Oh yeah, were a team, sorry .. shot my bullshit bolt early PS is fine, I bought a new 500w one which was happily supplying the MSI 8800GT I had in it, it was only when I introduced the faulty card that the slot appeared to turn an undesirable corner. The only easy thing I haven't tried is to see if the 6pin connector from the PS got damaged by the unfettered heat the dodgey card was giving off - likely?

Thanks for the input guys and for the reassuringly competitive camaraderie and banter, Krell, we need more of that. What's happened to Da Horn after all - I thought that was his thing?


S e n t i n e l (he typed self conciously...)

[sentinel]

[/quote]

I have an issue semi related to this. I acquired a graphics card in knew to be faulty (snowy graphics) which I put in one of my few PCI express slots. It was worse than I expected and I could only just make out the screens, it then went to a blue screen and I had to reboot. The problem is when i put a working card in the same slot now it also produced snowy graphics, but if I plug the screen in to the low-end-no-use-for-folding card that is integrated into the mobo the picture is fine. So it seems to be specific to the PCIe slot

I have re-installed the Cuda drivers and re-inserted working cards etc but it seems the slot is now non-funtional. Burnt out? fixable? Any ideas appreciated!
[/quote]

Bump as if I'm going to get another graphics card online, this is the only PCIe slot I have left, and I want to avoid shelling out for another PC (it's the only reason I bought this one!). Any workarounds, fixes would be great and directly contribute to another 4-5 PPD for the team.

[poser]

Hi. I'm aware that some people buy damaged motherboards but usually it's older boards that had popping capacitors. Usually something as complicated like a motherboard is pretty much impossible to fix.

I have to say that what happened sounds weird but computers can be tough to make out sometimes.

Oh btw, is the adapter your monitor connects (that works) built-in or is it on a card (I had a motherboard that had DVI-connection on a separate PCI-e card)?

[sentinel]

Hi. I'm aware that some people buy damaged motherboards but usually it's older boards that had popping capacitors. Usually something as complicated like a motherboard is pretty much impossible to fix.

I have to say that what happened sounds weird but computers can be tough to make out sometimes.

Oh btw, is the adapter your monitor connects (that works) built-in or is it on a card (I had a motherboard that had DVI-connection on a separate PCI-e card)?

Thanks for your response. Board is fine and all other components including the built in graphics adaptor, so I'm running the SMP client fine, but it's hardly the 5k a day I could be getting out of the machine if i could get the pcie working, or a workaround...

[ajnast4r]

im trying to get f@h (6.2/systray) to run similar to a screensaver... i dont believe there are any actual screensaver options so what i tried was having windows task scheduler open f@h on idle and close it on cease idle.. the problem is WTS wont close f@h on cease idle, even when the option is selected.

also everyone on imminst is working under team# 32461

any ideas?

Edited by ajnast4r, 07 November 2008 - 09:34 PM.

[sentinel]

im trying to get f@h (6.2/systray) to run similar to a screensaver... i dont believe there are any actual screensaver options so what i tried was having windows task scheduler open f@h on idle and close it on cease idle.. the problem is WTS wont close f@h on cease idle, even when the option is selected.

also everyone on imminst is working under team# 32461

any ideas?

Easy one first, yep 32461 is us.

I don't know if it's necessary to use task scheduler, it may be easier to reduce the priority level of f@h, but best for one of the real techs to wade in here. I have pretty low end machines and TBH don't really find that it interferes with the machine performance, do you use high end apps that demand a lot of resource?

[ajnast4r]

I don't know if it's necessary to use task scheduler, it may be easier to reduce the priority level of f@h, but best for one of the real techs to wade in here. I have pretty low end machines and TBH don't really find that it interferes with the machine performance, do you use high end apps that demand a lot of resource?

the issue for me is f@h pushes my cores to about 50C which kicks the fans in my laptop on high... its annoying to have to listen to them while im working or trying to study. so i want f@h only to run when my computer is idle.

[sentinel]

I don't know if it's necessary to use task scheduler, it may be easier to reduce the priority level of f@h, but best for one of the real techs to wade in here. I have pretty low end machines and TBH don't really find that it interferes with the machine performance, do you use high end apps that demand a lot of resource?

the issue for me is f@h pushes my cores to about 50C which kicks the fans in my laptop on high... its annoying to have to listen to them while im working or trying to study. so i want f@h only to run when my computer is idle.

Understood,my laptop is burning a hole in my thigh as we speak

So Techs, we have 2 issues "hurting" the team. Any takers?

Edited by sentinel, 08 November 2008 - 12:25 AM.

[brokenportal]

I think my FAH just disabled itself some how. Its not in my task bar any more. Its still in programs, so I clicked on it and it sent the icon back to my task bar, but when I mouse over it there is no activity, points being generated on it any more. Has anybody else had this problem?

My internet disconnected itself last night and I turned it back on today by unplugging the router and stuff. That must have been it. But then I think that you have to go and reactivate folding if that happens. It started generating points again just now. If you know anything about this then let me know. Maybe we have to alert people that if their internet goes down that they have to reactivate folding.

[poser]

I think my FAH just disabled itself some how. Its not in my task bar any more. Its still in programs, so I clicked on it and it sent the icon back to my task bar, but when I mouse over it there is no activity, points being generated on it any more. Has anybody else had this problem?

My internet disconnected itself last night and I turned it back on today by unplugging the router and stuff. That must have been it. But then I think that you have to go and reactivate folding if that happens. It started generating points again just now. If you know anything about this then let me know. Maybe we have to alert people that if their internet goes down that they have to reactivate folding.

Never had this problem because of Internet connection. Just shutdown the process in Task manager and start again. FAH will revert to the last save point.

[maestro949]

With multiple GPUs they all seem to be drawing from the same CPU. For some of the slower AMDs I have the one CPU is pegged at a 100% so I'm worried that I'm not squeezing the max out of the cards.

Is there a way to distribute the GPUs' load across CPUs?

Attached Images

• 

[dnamechanic]

...Is there a way to distribute the GPUs' load across CPUs?

Not having experience folding with GPUs, not sure how others manage this.

You could try assigning the specific Fah processes to a core with "Affinity".

In the Task Manager, click the "Processes" tab:



Right-click the process-of-interest, in your case, probably FahCorexxx.exe or something.

Then select from the menu "Set Affinity".

The default for four cores looks like this:



Modify as appropriate, something like this maybe using different cores for each process:



Or however you wish to distribute the load on a multi-core CPU.

Edited by dnamechanic, 11 December 2008 - 06:59 PM.

[maestro949]

I'll try that. Thanks. I was looking for a command line flag and seeing lots of stuff about affinity mgt utilities like this but that seems to be fore CPUs.

[maestro949]

Now both CPUs are pegged at 100%. I'm wondering if the AMD Athlon 5600 2.9GHz and Opteron 185s arn't enough muscle.