LinkBack URL
About LinkBacks
I've got a frustrating problem. I finally got fed up with my old rig which was a modified Duron 1400 @ XP2600ish (11x189 for 2079 MHz), on an Epox 8RDA+ V2.1 with 2x256MB TwinMOS PC3200 sticks. It's run stable for a year at 189MHz FSB but I could never get it to Prime any higher, which I assumed was the RAM's fault.
From some impulse purchases on these forums I ended up with another Tbred-B and an SLK800; I needed a new HD recently and couldn't resist adding two sticks of 512MB TwinMOS with Winbond CH-5 to my Komplett order. I shoved them in in place of the 2x256 sticks but found that my limit was still 189MHz, which I figured meant that the motherboard must be at fault for the low maximum FSB.
I had an 8RGA+ lying around which I was given by OiD some time ago to use for hotflashing a bios chip. I set up a little test rig by shoving the new Tbred and the 2x256 sticks on it, and booting up UBCD from an old CD-Rom. I found that even with an atrocious ebuyer PSU the memory could pass memtest at 200MHz 11-3-3-Cas 2, and it primed for over four hours at 11.5x200 with 1.6V Vcore. So, on Sunday I finally got round to ripping the 8RDA+ and Duron out of my main PC and replacing it with the 8RGA+ and XP combo. The only difference to the test rig setup is that I'm using a much better PSU, I put the 2x512 sticks in in place of the 2x256, and I put it all into a slightly better case on which I've cut out the fan grills and put two 80mm fans as intake and exhaust. Oh, and I pulled the 8RGA+'s cruddy original circular NB heatsink off and replaced it with the better one from the 8RDA+, having cleaned off the pink gum with lighter fluid and replaced it with thermal paste.
And here's the point of this long and tedious tale: I still can't get the damn thing prime stable at 200MHz. In fact I left it on overnight at 12.5x187 and it had failed by this morning. CPU temp never got over 50C and in my experience that's fine, I'm 99% certain the processor is stable at that speed. When I get home tonight I'm going to try bunging the 2x256 sticks back in (if that works I guess I'll have to RMA the 512 sticks). I wonder if the NForce2 controller struggles to cope with two double sided DIMMS in Dual Channel mode? Is it worth my while trying it in single channel?
The only other thing I can think of is that I put too much goop between the NB and the new heatsink- I spread in on thinly at first but I couldn't get a good contact when I test fitted it until I put on a layer maybe 1/3 to 1/2mm thick. Certainly increasing the VDD seems to make the computer less stable if anything.
If anyone has any advice (or even just sympathy
Rich :¬)
Edit: it did manage nearly two hours of Prime at 12.5x189, so I must be close to getting it stable. And please don't tell me not to worry about Prime stability- this computer will be running find-a-drug 24/7 so anything less than total stability is not acceptable.
Reply With Quote
).
. I thought I'd try the L12 mod on the old rig since it'd involve a lot less pulling stuff apart. I slapped it back together without even bothering with new thermal grease on either CPU or NB, I just bunged the old 8RGA+ HS on. I started off without doing the mod to get a benchmark figure- and it's now priming away seemingly quite happily at 9x220 Cas2-3-3-11 on the old TwinMOS memory. It never got anywhere close to that when I had it in my case. The only difference is that it's using the crap ebuyer cpu and a PCI vidcard instead of my softmod 9500.
Originally Posted by Rave
.
