Odd overclocking results on a Duron.

[Gilby]

Long story short. When I bought my system (Duron 750 w/ KK266) I ended up buying two Duron 750 processors. Sorta happened along a couple of good deals at once and figured I could keep the higher overclocker...or whichever one survived any potential mishaps.

So. Last week I did some simple testing after just luxuriating in the raw speed of a Duron 750 (and coming from a 233mmx, that is an awful lot of luxury.) Staying at the 7.5 multiplier and just playing with the FSB, one Duron refused to post at anything higher than 908 and the other around 950 or so. I resigned myself to not having a gig machine for now, and made plans to sell the slower. (See my FS thread.)

I tried the pencil trick on the faster of the two, and the default Voltage was sucessfully raised, but I didn't get the L1 bridges properly. Still couldn't change the multiplier. No biggie, really. I tried gradually upping the FSB again with the higher default voltage and still crapped out at the same place. Ah well.

Tonight I redid the pencil trick after failing miserably with the defogger kit--that thing's a pain in the ass to work with. I put tape over the penciled bridges to, hopefully, make it a lasting change. And then, at the default FSB, I started to play. For now, I was just hoping to get to BIOS and the System Health/IWILL Smart Settings screens.

900? No problem.

950? No problem.

1000? (Yeah, right.) No problem?

Well hell. Time to head to windows for some Sandra benchmarks (very nice) and burn-in time. No problems.

It even posted at 1050, though I didn't try windows. It failed to post at 1100.

This seemed odd, to suddenly have more speed. Maybe the FSB of this thing just can't handle above 120-something?

Well, next test was 133FSB via the jumper. I set the multiplier to 6 and watched it boot up just fine. Did the same at 7.5*133 for, once again, 1Ghz. That's what it is running at right now, to tell the truth.

I did some more tests, just with FSB, and got the FSB all the way up to 164. It would go no higher, even with the multiplier down to 6, 5.5, and 5.

So my question is...why did it fail to post at a FSB of around 125 (tonight too) but do fine up past 160?

 

[Def]

Your PCI clock was probably still on a 1/3 divider at mid 120Mhz FSB speeds. This results in a pretty high PCI clock that one of your components couldn't handle.

At 133Mhz FSB, or perhaps a bit earlier, your PCI clock goes to a 1/4 divider and you get the default 33Mhz.

This is the reason why people with multiplier locked Intel chips want ones that go to the "next step". As in, you want a 100Mhz chip to get to 133Mhz stable, so you basically have everything "in spec" except for the CPU.

 

[Technonut]

I have a Duron 750 that I use for watercooling experiments and other projects that could be fatal to my T-Birds. It is similar to yours. I have run it set at FSBs such as 133 and below with no problem being Prime95 Torture Test stable at 1GHz. When I attempt to run at a 150 FSB with RAM that I know will take it at CAS2, the 750 is not stable at 950 - 1GHz, no matter if I reduce the RAM timings or not. I'll have to reduce the RAM to CAS3 and try out a 160+ FSB to see what happens.

I have been running Prime95 at a 133FSB with the 750 at 1GHz on a watercooler that I have been testing for over 24 hours now with no crashes. It will be interesting to see if I could get results like yours at a 160+ FSB. What kind of RAM are you using, and at what settings? I am using the KK266/R for the tests.

EDIT: Do you have a fan installed on your NorthBridge HS? I have the stock HS on my KK266/R with Arctic Silver & spring loaded mounting pins from a Blue Orb that holds the HS down much tighter than the original plastic pins.

For the hell of it, I just placed a temp probe on the middle of the NB HS, and with no fan it is at 46.8 Cel. running Prime95 at the 133 FSB/1GHz in an air conditioned room with the case side removed. Pretty damn warm.. I think I will install a fan or waterblock on it and see how much it helps. I have heard of plenty of people running at a 150+ FSB on the KK266 with the stock NB HS with no fan, and had no problem, but like I said, 46.8 Cel. at 133 seems too warm.

 

[Gilby]

Actually, that works out perfectly, Def. At about 126 FSB the PCI bus is running at 42. At 164 it is running at 41. There's the common thread.

Stock HS on the Northbridge. Using a Volcano2 to cool the CPU. The RAM is 256MB of PNY from Staples. It's rated at 133CAS3. I had it at 164 at CAS3, didn't try CAS2. Currently running at 133FSB at CAS2.