Okay, maybe I'm missing something here. Maybe this stuff is common knowledge that I never found out. But both of my computers have really hot northbridges!
Comp 1: Gigabyte K8N nForce4 SLI mobo, A64 3700+ (undervolted)
Comp2: Gigabyte s55 nForce4 SLI, X2 3800+ (undervolt)
CPU temps are fine for both of them, idle around 30-38 C and Load is around 45 or less. But the northbridges are blazing hot, so hot that I can't hold my finger on it. Speedfan reports a couple temps up around 70C, it just seems like that is way too much. Is it? At full idle I still see a couple temps up around 60c, but I can actually touch the HS at that temp.
For my gaming system, I installed the new Thermalright IFX/SLI northbridge cooler, and that has pretty much taken care of that sytem. The other one has a large passive heatsink that it came with stock.
I recently stopped folding on the first one because I was concerned about this heat. When you pull the side of the case off, you can feel the heat from the NB, which heat soaks my passive nVidia 7100GS that is sitting right next to it.
I'm tempted to get another thermalright cooler, but it seems like it shouldn't be necessary?
Blazing hot northbridge?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
The nforce4 integrated MCP chipset is notoriously hot - possibly because it combines the functions of a traditional northbridge + southbridge on a single piece of silicon (at least on AMD boards), together with various fancy features. You could get another Thermalright if it makes you feel more comfortable, but I wouldn't worry too much as long as it's all working OK.
I wonder if any of those screaming 8000rpm fans found on the early NF4 boards are still working...
I wonder if any of those screaming 8000rpm fans found on the early NF4 boards are still working...
Careful! If the Nforce4 MCP hits a certain temperature, it will start corrupting data silently during disk writes. It's not a cataclysmic effect, just a few bytes out of each GB, so it can go unnoticed for quite some time. I started seeing it on large torrent downloads first, during the verification phase, and I didn't diagnose it correctly until it was too late (OS files were corrupted).
The tripping point probably differs from motherboard to motherboard. On my Albatron K8SLI, corruption would start when it hit about 49-50C, and higher temperatures would increase the effect. Monitor your temperatures very closely, and keep an eye out for signs of corruption. I used to test by copying large amounts of data (about 12GB worth of virtual machines) and checking md5 checksums afterwards. Once it hit the threshold, 1 out of 2 or 3 copies would exhibit corruption. Once it went at about +5C, corruption would occur every time.
I combated the problem by adding a slow 8cm fan at the PCI brackets and ducting the northbridge to it, which kept temperatures at 39-45C. The system ran for about a year without showing such problems again (I am now building a fileserver out of it).
Just thought I should warn you
Edit: I wouldn't trust Speedfan temperatures as absolutes, motherboard thermal diodes can be notoriously inaccurate. They are still useful as relative indicators after you find the tripping point.
The tripping point probably differs from motherboard to motherboard. On my Albatron K8SLI, corruption would start when it hit about 49-50C, and higher temperatures would increase the effect. Monitor your temperatures very closely, and keep an eye out for signs of corruption. I used to test by copying large amounts of data (about 12GB worth of virtual machines) and checking md5 checksums afterwards. Once it hit the threshold, 1 out of 2 or 3 copies would exhibit corruption. Once it went at about +5C, corruption would occur every time.
I combated the problem by adding a slow 8cm fan at the PCI brackets and ducting the northbridge to it, which kept temperatures at 39-45C. The system ran for about a year without showing such problems again (I am now building a fileserver out of it).
Just thought I should warn you
Edit: I wouldn't trust Speedfan temperatures as absolutes, motherboard thermal diodes can be notoriously inaccurate. They are still useful as relative indicators after you find the tripping point.
It appears for Gigabyte boards hot NBs are par for the course. You can add GA-MA69GM-S2H to your list.
I have a ATI 690G based GA-MA69GM-S2H and its NB runs very, very hot, even when all the non-essential chipset features are turned off.
Is the NB in the 690G chipset inherently hot? Nope. I also have a Biostar 690G based board. With all the chipset features turned on, the NB on the Biostar still runs very cool. So it's possible to have a cool running 690G NB. But for some reason Gigabyte chose to design their boards to run with a hot NB. Maybe they chose to overvolt it for some reason?
I have a ATI 690G based GA-MA69GM-S2H and its NB runs very, very hot, even when all the non-essential chipset features are turned off.
Is the NB in the 690G chipset inherently hot? Nope. I also have a Biostar 690G based board. With all the chipset features turned on, the NB on the Biostar still runs very cool. So it's possible to have a cool running 690G NB. But for some reason Gigabyte chose to design their boards to run with a hot NB. Maybe they chose to overvolt it for some reason?
If this is indeed true it might explain the inexplicable crashes I get with my gaming computer. Temps *seem* to be inline but it just crashes. Also when I overclock it, I can run it OC'd for 3 hours of intense gaming, no problem. Next day, I boot it up cold, and it crashes with the same settings.StApostol wrote:Careful! If the Nforce4 MCP hits a certain temperature, it will start corrupting data silently during disk writes. It's not a cataclysmic effect, just a few bytes out of each GB, so it can go unnoticed for quite some time. I started seeing it on large torrent downloads first, during the verification phase, and I didn't diagnose it correctly until it was too late (OS files were corrupted).
The tripping point probably differs from motherboard to motherboard. On my Albatron K8SLI, corruption would start when it hit about 49-50C, and higher temperatures would increase the effect. Monitor your temperatures very closely, and keep an eye out for signs of corruption. I used to test by copying large amounts of data (about 12GB worth of virtual machines) and checking md5 checksums afterwards. Once it hit the threshold, 1 out of 2 or 3 copies would exhibit corruption. Once it went at about +5C, corruption would occur every time.
I combated the problem by adding a slow 8cm fan at the PCI brackets and ducting the northbridge to it, which kept temperatures at 39-45C. The system ran for about a year without showing such problems again (I am now building a fileserver out of it).
Just thought I should warn you
Edit: I wouldn't trust Speedfan temperatures as absolutes, motherboard thermal diodes can be notoriously inaccurate. They are still useful as relative indicators after you find the tripping point.
I might try a backup and clean install... good way to waste 4 hours.
Any pics (please), I may do the same.StApostol wrote:I combated the problem by adding a slow 8cm fan at the PCI brackets and ducting the northbridge to it, which kept temperatures at 39-45C. The system ran for about a year without showing such problems again (I am now building a fileserver out of it)
Thanks.
This is not the actual construction because I updated to a Core 2, but you can get the idea. Sorry for the crappy balance, I have an exam tomorrow so it's the best I can do.
Front side:
Back side:
You'll need two pieces of string or elastic cord (I used what was left after suspending the disks). The filter is only needed for intake fans - the original construction used the fan for exhaust, so I didn't use any filter.
The duct itself was made from folded A4 paper - nothing spectacular, but it worked wonders
The only problem is that the VGA card will get in the way. I worked around the problem by placing it second PCI-E slot, but the construction is impossible if you don't have an SLI motherboard. In that case, you'll have to - somehow - put the fan inside the case, in front (or on) the VGA card (the HD cage is a good candidate).
Front side:
Back side:
You'll need two pieces of string or elastic cord (I used what was left after suspending the disks). The filter is only needed for intake fans - the original construction used the fan for exhaust, so I didn't use any filter.
The duct itself was made from folded A4 paper - nothing spectacular, but it worked wonders
The only problem is that the VGA card will get in the way. I worked around the problem by placing it second PCI-E slot, but the construction is impossible if you don't have an SLI motherboard. In that case, you'll have to - somehow - put the fan inside the case, in front (or on) the VGA card (the HD cage is a good candidate).
Typically there's a little space between Mobo and whatewver drives-and you cound do a homebuilt bracket to mount a fan. A low rpm (800 rpm?) 120 would be quiet and should give enough added cooling. NF4 was a nice chipset but Nvidia screwed up not doing a hi-tech heatpipe HS for it...so you get the noisy little bitty fans or you spend to get a decent aftermarket
I should probably duct an 80mm fan to blow on the northbridge and also hit the crappy heatsink on that passive video card I have. I just need to go buy some of that sheet plastic and use that, probably be cheaper than buying another Thermalright heatsink. I could also make the duct so it would draw air from the front of the case and not just inside air.. sounds complicated.
Anyone remember what that sheet plastic is called, usually white and bonds very well with super glue?
There's a plastic supply place near my house and I got some from them a couple years ago.
Anyone remember what that sheet plastic is called, usually white and bonds very well with super glue?
There's a plastic supply place near my house and I got some from them a couple years ago.