Finally OC'n my PII 920

[repilce]

so far so good at this point.

I need to make a duct, as taking the side off my case creates a 5c or so difference in cpu temp, i removed the stock fan on the true spirit and used two antec tri-cools in push/pull.

 

[trents]

Looks good. All voltages and frequencies displayed look appropriate.

 

[repilce]

Thanks, I evaded a duct mod and expanded the area of incoming air flow to my case, antec 300, Now it inlets through the front from just under the top media bay, to the bottom. testing @ 3.6 now and increased voltage with a max temp of 48C ;D

 

[repilce]

Back to 3.5 going for 24/7 stability test.

Think i may settle here as after two swings at 3.6 i could start to see diminishing returns on voltage increases. But I'll take it !

 

[repilce]

Guess these C2's likes some volts.. I kept get'n errors @ 3.5 on cores 3&4 after 3 hours (every time ) and with an increase of vcore up to 1.5v in bios (1.488 cpuz) and nbcpuV up to 1.3 i still got an error on core 3 after 3 hours, but all other were still running.

If it takes 1.5v "actual" to reach stable at 3.5Ghz, i think i'd be better off on efficiency to just keep it @ 3.3-4 where it will hold more close to stock voltage.

 

[repilce]

Okay, this is frustrating. backed off to 3420mhz , and.. there's that damned 3 hour error on core 4 again.... it's a rounding error. I'm starting to believe it might be ram timing related as in dolks guide he mentions something on 64bit memory stability and loose timings.

I'm re-running my initial 3.5ghz setup 250x14 cpuV 1.44 NBv 1.25
Unfortunately my board is weird and limits the access to certain timings by what divider you have it set to.. and since i'm pretty sure these gskill's wont like 600mhz+ I changed my timings at 1000 DDR2 to 5-6-6-17 26

hopefully this will do the trick.

 

[repilce]

Much to my chagrin, no go..

Even at my old settings.. (kinda) 235x14 1.36vcore (i ran this before realizing the cpuNB could be upped to 9x) and it ran stable, think i had it at like 7x (thinking in terms of HTT and not cpuNB) but with 9x i'm still get'n that damned rounding error. and still after 3 hours.. i mean down to 10min of 3 hrs every time.

So just for jollys, i'm running a 250x14 1.44vcore with 8x on the cpuNB @ 1.2v

 

[QuietIce]

You're kind of running the edge with your load core temps or you could turn the cpuNB volts up a bit.

Have you got the vDIMM at +0.05v stock ...?

 

[repilce]

Under 50c is on the edge? and that was with 1.25+cpuNB even at 1.3cpuNB and 1.5vcore max temp was 50c, so i have room to play with it.. but it didnt fix it. at least at 3.5ghz

My vdimm is at 2.1v

 

[TUK101]

Lol, when I saw the title it took me back about 12-13 years ago when we where actually overclocking the original PII's. But they where far from 920's and hitting 450mhz @ 100fsb was a great OC.

 

[QuietIce]

Under 50c is on the edge? and that was with 1.25+cpuNB even at 1.3cpuNB and 1.5vcore max temp was 50c, so i have room to play with it.. but it didnt fix it. at least at 3.5ghz

My vdimm is at 2.1v

You need to look at the load core temp when talking about OC'ing and stability. You're pushing 52°C and 55°C ±5°C (depending on the CPU) is where instability starts creeping in.


PS
Next step to try would be slowing down the RAM.

You know, you could just be running out of clock speed with that board. Not all boards will run 300+ and many start to sputter about 250 MHz. Sometimes a bump to the NB (chipset) will help that.

PPS
Found this at http://ixbtlabs.com/articles3/mainboard/gigabyte-ma770-ud3-770-p2.html

This motherboard demonstrates standard overclocking results with a processor with the unlocked multiplier. But its results with the locked multiplier are absolutely unimpressive. The motherboard apparently has problems with increasing the reference frequency above 255-260 MHz, because we failed to squeeze more even after reducing the multipliers for HT bus, memory, and CPU NB below the usual level.

 

[sQUiRLy]

You know, you could just be running out of clock speed with that board. Not all boards will run 300+ and many start to sputter about 250 MHz. Sometimes a bump to the NB (chipset) will help that.

PPS
Found this at http://ixbtlabs.com/articles3/mainboard/gigabyte-ma770-ud3-770-p2.html

I agree, 1.25 - 1.3V on the NBchipset if oc'ing HTreference clock.

An updated BIOS may have corrected the clock limitation found in that review. Are you current, replice?