Not sure where to go next in OC, PII980

[Nephilim]

Hi,

So as of last night I decided to stop being lazy with my processor sitting at 4ghz just because it was easy to do so and to see the limits.

After messing around a bit I was at 4.2 core 2400 nb 2000 HTlink
With .025 more to the vcore ~1.408-1.425 as reported by cpuz

Running prime though both small fft and blend my third core kept having small failures (no crashes or anything to that extent)

So I dropped my core back to 4.1 and NB to 2200 and everything seems fine.
But the NB seems low for that core clock (It should be more around 2600 right?)

Anyway, Im still a little confused about what affects what when memory is involved, I havent touched any RAM timings or clock speed so is the RAM slacking and causing instability? Or should I just upvolt the NB?

I would really like to keep a 4.2-4.3 core with an appropriate NB to back it up. But I dont want to just start throwing voltage at everything since this will be a 24-7 OC

Here is pics of cpuz and HW mw-->

Spoiler

 

[trents]

When a Prime95 core worker drops out as opposed to say, BSOD, then you usually just need a tad more vcore. Will your temps allow it? And try 1.225v on the CPUNB which should support 2400-2500 mhz on that component.

 

[Nephilim]

Yeah I think the first fail I saw was "Illegal sumout" and the other was "Round was .5 expected .4something or another"

I can handle more temps but my motherboard only increases the vcore in .025
increments, wasnt sure if 1.450ish would be getting to high.

I guess that makes sense though if its a core problem still. On my old mobo It crapped out even hitting 4ghz with the same voltage as now.

 

[BeepBeep2]

If you get the rounding error, increase CPU-NB voltage.
Illegal Sumout, increase vcore...
Any other fatal error increase vcore

0x101 BSOD, increase vcore
0x124 BSOD, increase vcore
(DRIVER_)IRQL_NOT_LESS_OR_EQUAL ..."XXXXX.DLL or XXXX.EXE" increase CPU-NB voltage.
MEMORY_MANAGEMENT same
BAD_POOL_HEADER same

 

[Nephilim]

Spoiler

If you get the rounding error, increase CPU-NB voltage.
Illegal Sumout, increase vcore...
Any other fatal error increase vcore

0x101 BSOD, increase vcore
0x124 BSOD, increase vcore
(DRIVER_)IRQL_NOT_LESS_OR_EQUAL ..."XXXXX.DLL or XXXX.EXE" increase CPU-NB voltage.
MEMORY_MANAGEMENT same
BAD_POOL_HEADER same[/spioler]

Spoiler

^^Thank you that is helpful,

I saw in your sig you say a PII can handle 1600mhz range RAM how is that so?
I tried setting mine manually to 1600mhz and system crashed on startup :X

 

[BeepBeep2]

What is the rated speed of your memory?

 

[Nephilim]

Bump hope someone sees this.
after playing around a bit I realize now that even after raising volts to NB Im still getting some fails in prime.

I think the vcore is fine since I can run small fft's without issue but I seem to fail large fft almost every time on test five. Blend tests willl fail on occasion. Illegal sumout seems to be the same message I get every time.
Ive tested my memory with memtest and its clean, :c

vcore is sitting at 1.440

 

[BeepBeep2]

Try increasing vcore from 1.440 to the next available option. (0.025 or smaller). Make sure temps do not climb too much above 60c though, and if more vcore does not help this issue, make sure to let me know.

Did you sort out the issue with RAM at 1600?

What is the model number of your memory, and/or the stock voltage, frequency and cas latencies as listed on the side of the DIMMs? Your sig only says "Kingston HyperX".
We will get ya sorted out at 1600. Maybe even higher.

 

[Nephilim]

Try increasing vcore from 1.440 to the next available option. (0.025 or smaller). Make sure temps do not climb too much above 60c though, and if more vcore does not help this issue, make sure to let me know.

Did you sort out the issue with RAM at 1600?

What is the model number of your memory, and/or the stock voltage, frequency and cas latencies as listed on the side of the DIMMs? Your sig only says "Kingston HyperX".
We will get ya sorted out at 1600. Maybe even higher.

Actually it appears Ive hit a wall trying to get to 4.2 on the core clock. Im sitting stable 4.1 at 1.408v not sure if I want to go another .05 just to reach 4.2 at this point since at load that will put me at an actual 1.475v ish. :/

As per the memory I dont know the exact model of the ram off the top of my head but it is 1600 native, I havent had the time to fiddle with ram OC yet but Im more interested in hitting tighter timings at this point. Afterall even if I could hit 1600 isnt that going to wear out the memory controller that much faster?

 

[BeepBeep2]

~.05v is normal scaling per 100 MHz from personal and friend's experiences.
No, 1600 will not wear out the memory controller. The only thing that will wear on the IMC is voltage (say you try 1.6v CPU-NB voltage on air...bad) or heat the CPU up to very unsafe temps.

I know a lot of guys that ran 2+ years with ram at 1800+ 24/7, even in mission critical setups, like World Community Grid.
I've been advised to keep things a little more polite in regards to contradicting common belief here in the forum, but as far as I know, the world agrees on 1800 MHz stable being an upper end target for P95 stability, goes for all X2/X3/X4

1600, through just the ratio, should be set and go, in all scenarios, if unstable, then DRAM needs timings loosened or voltage increased.