[Cooling] Looking for a fan that notifies of failure in Overclocking/Modding http://www.dslreports.com/forum/r22410857 en Sat, 28 Nov 2009 11:15:18 EDT Sat, 28 Nov 2009 11:15:18 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22441373 sempergoofy :
said by DreamCarr See Profile :

If I read you right, you employer makes pci cards of some sort?

If so, get one of your engineers to whip up a custom pci breadboard that outputs 12v, and has a rpm sensor. Then just write your own custom softs to manage notification to the pci bus. Hell build a make-your-eardrums-bleed piezo buzzer onto it too, for good measure!

Do it right, and you've got yourself a new product that a lot of datacenters and enthusiasts will love you for.
I'll pass on the suggestion. Always looking for a good revenue stream. :)
--
nohup rm -fr /&]]> http://www.dslreports.com/forum/remark,22441373 Mon, 25 May 2009 16:10:58 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22438062 DreamCarr : If I read you right, you employer makes pci cards of some sort?

If so, get one of your engineers to whip up a custom pci breadboard that outputs 12v, and has a rpm sensor. Then just write your own custom softs to manage notification to the pci bus. Hell build a make-your-eardrums-bleed piezo buzzer onto it too, for good measure!

Do it right, and you've got yourself a new product that a lot of datacenters and enthusiasts will love you for.
--
never apoligize for being who you are, and never respect anyone that says you should.]]> http://www.dslreports.com/forum/remark,22438062 Sun, 24 May 2009 20:48:27 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22417514 sempergoofy : No problem. Not that I really care if they are blowing full speed 100% of the time. But for some reason this particular server just doesn't seem to be tumbling enough air over the PCI boards with its own factory fans. Not sure if running them at 100% would change that or not.

I don't have personal access to the server as it is currently in some labs in another state. I have only been able to talk to colleagues and pull the IBM manuals.
--
nohup rm -fr /&]]> http://www.dslreports.com/forum/remark,22417514 Wed, 20 May 2009 14:08:09 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22416382 insomx : Alright, I didn't realize you had the server in a place where noise is an issue.

The two tach signals into one controller won't work. I'm not sure what it'd do to report the speed of the fan, but regardless of what that does, if one of the fan fails, it'll just report the speed of the other one.]]> http://www.dslreports.com/forum/remark,22416382 Wed, 20 May 2009 11:06:04 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22416135 sempergoofy :
said by insomx See Profile :

I can't think of a way to do this, (sense a fourth fan).

However another option might be to set the fans to not be variably temperature controlled. I have found sometimes those kind of sensing programs aren't that great and can lead to overheating. There should be an option to make all three fans run at full tilt constantly.
Ever hear some of these IBM servers when you first power them on? You'd think you were working on the tarmac at the airport if they left them on that speed constantly. A toddler might get a heck of ride around the computer room on the server if they blew just a bit harder. :)

The fan we added to blow on the cards is, of course, not variable speed currently. It does the job, but if it fails and we don't know it... :(
--
nohup rm -fr /&]]> http://www.dslreports.com/forum/remark,22416135 Wed, 20 May 2009 10:18:50 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22416093 sempergoofy :
said by Ray422 See Profile :

Maybe this is a dumb question, but why can't you just splice a 4-pin fan into one set of the PWM wires?
Do you mean put a "Y" splitter into one of the current fan connectors? I found there are TX3 splitters available to allow two fans to be plugged into one of the motherboard fan connectors. Makes sense for power, but I don't know how two tachometer signals from two different fans into one circuit would work (if at all).
--
nohup rm -fr /&]]> http://www.dslreports.com/forum/remark,22416093 Wed, 20 May 2009 10:10:51 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22415143 insomx : I can't think of a way to do this, (sense a fourth fan).

However another option might be to set the fans to not be variably temperature controlled. I have found sometimes those kind of sensing programs aren't that great and can lead to overheating. There should be an option to make all three fans run at full tilt constantly. ]]> http://www.dslreports.com/forum/remark,22415143 Wed, 20 May 2009 01:07:30 EDT Re: [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22413541 Ray422 : Maybe this is a dumb question, but why can't you just splice a 4-pin fan into one set of the PWM wires?]]> http://www.dslreports.com/forum/remark,22413541 Tue, 19 May 2009 19:29:58 EDT [Cooling] Looking for a fan that notifies of failure http://www.dslreports.com/forum/remark,22410857 sempergoofy : (Mods: This seemed the best forum in my research. Move if you disagree.)

At my job, we have been evaluating a server with a PCI board (multiples) that my employer makes. In this particular server (unlike others where we never have an issue), we determined during tests that it was thermaling down. When we finally caught it right after it had gone down once, we noticed how hot the PCI boards were. (That is good because until then, we would just find the server powered off with no messages in system bios error log or any o.s. (Linux) logs.)

We bought a small fan, powered with 4-pin molex connector power, that we temporarily mounted in the case to blow more air over the boards. That seems to have worked and it allowed our testing to run the 72+ constant hours at full bore.

The server (an IBM brand server) comes standard with its maximum of three factory fans installed. Those fans plug into the mother board at three fan specific plug points. I believe (per my colleagues) they use four pins, one of which is a tachometer signal to cause variable speed dependent upon temperature. Those fans also can signal failure to the motherboard so that alerts (lights, notices, etc) can be sent. But, since all of the plugs on the motherboard are occupied we can't plug in an additional one.

While our additional simple fan works ok, we would like to have some method of being notified if it fails since the thermal-down we were seeing was so unceremonious. Data is needlessly risked if we can't shutdown our application due to impending thermal issues.

I looked at some of the PCI-slot-occupying cooling fans like the Blitztorm. I was hoping they were drawing power and maybe signaling from the PCI slot. They appear to just use the slot to hold them in place and draw their power using 4-pin molex connectors like DVD drives. No smarts for signaling.

I also looked at USB powered fans since we might drive those off an internal USB connection. Again, none that I found have a failure signaling ability.

Soooo, I'm wondering if anybody knows of any good fan solutions that have a failure-notification capability (any method!). Since we are using Linux, that would help, too.

(P.S. I know about the USB Temperature@lert product. But we would rather have the fan notify of failure than a "I'm too hot, too late alert."
--
nohup rm -fr /&]]> http://www.dslreports.com/forum/remark,22410857 Tue, 19 May 2009 11:33:04 EDT