AM3 heatsink facing upwards - VRM no longer gets airflow

[Crayon Shin Chan]

Ever since I've installed my Alpenfoehn "Brocken" heatsink, which, like most coolers, happens to blow upwards instead of backwards on a AM3 board, I've noticed that sometimes the CPU throttles to 7.5x multiplier. At first I was puzzled, but upon touching the chokes with my finger, I quickly ascertained the problem - with AMD's stock cooler, the fan blew down on the motherboard, and the air was directed sideways by the heatsink fans, flowing over the VRMs to the left of the socket. There was a fancy looking heatsink from Asus, but the chokes weren't covered. With this aftermarket cooler that blows upwards, the chokes got no airflow at all. Actually, I don't know the maximum temperature for chokes, maybe the MOSFETs under the heatsink were the ones overheating, the whole area to the left of the socket isn't getting airflow I think.

Have you guys run into this problem before? What have you done to fix it? Personally it's fast enough for me at 2GHz as well - but that does defeat the purpose of me overclocking it if it throttles itself every now and then.

I could reverse the 140mm fan on the top of the case so that it blows air back into the case instead of out of it, but then the CPU fan are blowing in the opposite direction.

[Crayon Shin Chan]

I reversed the rear exhaust fan so now it's an intake, and used a piece of cardboard to prevent the air from going directly to the CPU heatsink and direct it to the MOSFET and chokes instead. A little flap on top prevents the intake air from going to the 140mm top exhaust fan before it's reached the MOSFETs. So far it seems to be working - only 20 minutes in tho.

[just brew it!]

I don't think overheating VRMs would cause CPU throttling; I'd be more inclined to suspect a problem with the HSF.

[Crayon Shin Chan]

Yeah, but HWMonitor said otherwise.
http://ritchan.dasaku.net/wp-content/up ... rottle.png
This was before:


And now:

[just brew it!]

I'm not sure what the screen shot is supposed to be showing me. Did your mainboard temps drop? You've completely changed the airflow in your case, which is going to change the temperature of a lot of things (including the CPU). If it fixed the problem, then great -- but I'm still not convinced that the VRMs were the culprit.

I would also be a little concerned about having anything made of cardboard touching your VRMs (can't tell from the pic whether it is touching or not). If the cardboard is touching, and one of the VRMs overheats, you could end up with a fire inside your case.

[Crayon Shin Chan]

Take a good look at the temps reported by the CPU cores themselves, not the ITE IT8721. In my experience, the difference between the two is not a constant value, so I just go with the CPU reported value. The mainboard temperature can't have anything to do with the VRMs - even with the extra airflow they're 50-70C at least. TMPIN2 is always 128C - never seen it at any other value. CPU-Z shows that it's throttling down to 7.5x when it should be at 14x because Prime95 is loading it. BTW the throttling doesn't happen with F@H. The airflow for the CPU cannot change because the heatsink is blowing upwards to begin with. The rear exhaust would simply be blowing air for nothing - or probably taking away hot air away sideways from the heatsink, which probably wasn't the design anyway. Just to prove it, I'm now running F@H and VMWare with a busy Gentoo compile and it's still 47C for all 6 cores.

A fire wouldn't be nice. But the cardboard is only touching the bracket that holds the fan to the heatsink. It's much too short to touch the VRMs. I wouldn't let it touch the motherboard anyway, much less the VRMs.

I also ran LinX at 17000 54 times with CPU-Z a few hours ago and I haven't noticed any throttling.