Posted by
Rampage Extreme X48 Northbridge cooler fix
Despite the age, Rampage Extreme is still very popular board among the overclockers. About 22000 submisions till this day on hwbot, when this article is posted is an impressive number. Overclocking on 775 platform changed during the years – at the beginning everybody used single rank 2x1GB Micron D9GTR/GTS kits. After Elpida Hyper release, overclockers started using RAM with this IC – dual rank 2GB modules give better performance than single rank, also the IC allows to go higher with the frequency with lower latencies. Binned Elpida Hyper sticks can do 2000MHz+ clocks on very low timings like 6-7-6. Unfortunately not every board is able to run 2000MHz on dual rank modules – memory controller load is much higher in comparison to single rank.
During those years people also figured out, that board must be binned too and very often boards from the beginning of the production provide the best results in terms of FSB and memory overclocking – pieces with S/N starting from 87-8A are usually the best and the most valuable. Other major issue when pushing the whole platform to the limits is the NB temperature. X38 and X48 chipset used the IHS with the thermal paste below it. Unfortunately after 10+ years this thermal paste is dry, which increases the temperature of the northbridge and making some stability issues. Extreme overclockers usually remove the whole cooler and use water block to cool NB (with chilled liquid) or use LN2 cooling on chipset. Unfortunately a lot of binning and testing is done on standard cooling, to avoid any condensation issues. Northbridge running in 70C+ temperatures doesn’t help with that. I was cleaning one Rampage board recently and I’ve found a contact issue between the cooler and NB. I made a short guide how to fix it.
-
Checking the contact
Board cooler is attached not only to the NB, it cools also Fujitsu cap, several mosfets and 3 chokes near the NB area. I removed all the thermopads and thermal compounds and applied a very thin layer of thermal paste. If you never did it, it’s very good method to check the contact between multiple parts with different heights. By looking at the paste spread you can also get some information about the pressure.
Here we go:
Then we put the cooler back. It’s enough to use only 3 screws near this area – two on NB and one on mosfet heatsink.
And now we can remove the heatsink again:
As you can see contact is… very bad. Notice the fact, that I removed the layer of the thermal compound from the place where are the chokes. It was about 0,5mm thick and cooler is still touching chokes at this place, lifting the whole cooler and provides almost no contact with the NB.
Ok, now we know the place which is responsible for the bad contact. I noticed also that, if the board is bend near the CPU area then it’s even worse (very common issue due to pushpin coolers, which makes in my opinion way too big pressure on the CPU).
-
Fix
I took the cooler to one of my milling machines. That was the fastest method for me. In fact it can be done using a flat fille (medium grid) – surface don’t have to be extra flat, because those are only the chokes and soft thermopad can be placed here. I took about 0,4mm of material.
MUCH better. You can notice also a small contact with the Fujitsu at this moment. Unfortunately due to old thermal paste below the IHS there is almost no improvement in temperatures – maybe 2-3C in total. In fact X48 chipset is not that hot, it just need a proper cooler with good heat transfer. Delid is the thing which should be necessary in this case, especially when you consider the fact, that thermal past below the IHS is 12 years old.
During the later tests I figured out also, that due to bad heat transfer to the cooler, back of the board is getting hot. During the tests without the case, I placed the board about 8-10cm from desk and added a fan below it to cool the pcb. This have me about 5-6C improvement, which is very good result.
