Overclocking on 754 platform... am I missing something?

[FDF12389]

Originally posted by: homestarmy
UPDATE!!!

I realized that with changing my memory settings (just using a cheap 256 stick for testing) that I was able to do this at 1.675v and be stable in prime 95 for overnight, no errors.

Thats a really high voltage, it may kill your CPU.

 

[homestarmy]

It has to do with my Mobo, any setting from around 1.575 to 1.65 give the same 1.536v. If I up it to 1.675, it actually runs at around 1.61-1.63, so it's not really too high.

And many others run it at a full 1.675, so as long as my temps don't go too high, I should be perfectly fine I would think.

 

[selfbuilt]

Originally posted by: homestarmy
And many others run it at a full 1.675, so as long as my temps don't go too high, I should be perfectly fine I would think.

That's one helluva of an oc - 2.7Ghz stable! Congrats, nice job.

However, I would also worry about that voltage. If I recall correctly, the Semprons run at 1.4V default (compared to 1.5V for newcastles like mine) ... so you are effectively giving it a 15% overvolt. :Q That is typically beyond the confort level of most overclockers I know (rough rule of thumb seems to be about 10% max for most of us - but there's no hard data to support that).

In comparison, my newcastle hits its max oc of +30% with only a 5.5% overvolt. My old Barton mobile use to do a +40% oc with only a 3% overvolt (compared to stock bartons) ... ah, those were the days!

Personally, I would probably run it at whatever max oc you can get at 1.536V effective (i.e. a ~10% overvolt). But again, that's just my comfort level. ... like you say, you may be fine if your temps stay under control.

 

[Zap]

Originally posted by: homestarmy
UPDATE!!!

2700 stable!!!

That's great! Someone else was reporting a Sempron 3000+ at 2.7GHz and my own 3300+ hits that speed too. I'm wondering if it's the "typical" upper limit of 128k cache Palermo E3/E6 cores? Yeah, you're being limited by BIOS, but going by what myself and one other person have gotten, you wouldn't get much higher.

 

[homestarmy]

Originally posted by: selfbuilt

Originally posted by: homestarmy
And many others run it at a full 1.675, so as long as my temps don't go too high, I should be perfectly fine I would think.

That's one helluva of an oc - 2.7Ghz stable! Congrats, nice job.

However, I would also worry about that voltage. If I recall correctly, the Semprons run at 1.4V default (compared to 1.5V for newcastles like mine) ... so you are effectively giving it a 15% overvolt. :Q That is typically beyond the confort level of most overclockers I know (rough rule of thumb seems to be about 10% max for most of us - but there's no hard data to support that).

In comparison, my newcastle hits its max oc of +30% with only a 5.5% overvolt. My old Barton mobile use to do a +40% oc with only a 3% overvolt (compared to stock bartons) ... ah, those were the days!

Personally, I would probably run it at whatever max oc you can get at 1.536V effective (i.e. a ~10% overvolt). But again, that's just my comfort level. ... like you say, you may be fine if your temps stay under control.

I'm considering that also. What I really like about where I am however is that nice clean 2/3 divider that should give me a straight 200MHz on my RAM. Oddly enough, it reads as 192 MHz elsewhere.

I did find a way to get the voltage between the 1.53 and 1.63 marks - running Cool N Quiet. I was so happy when it gave me around 1.58 (what I was looking for) and it appeared stable. But when I run Prime 95, it throttles up properly and runs for awhile, but the computer locks up . I suppose running at my actual ~1.63v is pretty high, but I am just so used to those high numbers from Socket A.

Question: Just because these processors do not require the high voltage, is there anything that makes them less able to handle it?

 

[selfbuilt]

Originally posted by: homestarmy
[I'm considering that also. What I really like about where I am however is that nice clean 2/3 divider that should give me a straight 200MHz on my RAM. Oddly enough, it reads as 192 MHz elsewhere.

That's odd ... but it may be due to how the board caculates the decimal value of the divider. There's a real quirk in how boards calculate timings. The 2/3 divider could be 0.6666..., or it could be only 0.665 (i.e. 266/400). You can calculate effective memory speed by the following formula:

(CPU Freq x CPU Multiplier ) / (CPU Multiplier / actual Memory Divider in decimal units)

The key is to round-up the bottom term to nearest 0.5. So, in theory, the 2/3 divider should give you a bottom term of 13.5 (yielding an effective memory of 200). But if the board uses 0.665, then this will give you 13.53, which you must be rounded up to 14 (yielding an effective memory speed of 192.9MHz). Sounds like this is what your board is doing at this setting.

Question: Just because these processors do not require the high voltage, is there anything that makes them less able to handle it?

As a general rule, running processors at higher voltages should shorten their lifespan - I suspect what you are really asking is is this a linear process? I don't know if these newer low-voltage chips scale in the same way as older ones ... it just stands to reason that signficantly higher volts could be a problem. Of course, these chips are so cheap -and I've never seen even an oc'ed processor fail in a regular lifetime of use anyway - so you might be fine if your temps stay low. I think temps are at least a reasonable proxy as to how well the chip is handling the higher voltage ... you could just go by that.

 

[Zap]

Originally posted by: homestarmy
I suppose running at my actual ~1.63v is pretty high, but I am just so used to those high numbers from Socket A.

Question: Just because these processors do not require the high voltage, is there anything that makes them less able to handle it?

My personal rule-of-thumb is to resist the temptation to overvolt by more than 10%. I recall reading CPU whitepapers stating voltages were rated to +-10%.

 

[furballi]

1.55 Vcore should be the upper limit if you plan to keep the CPU for a few years.

 

[Hacp]

Originally posted by: furballi
1.55 Vcore should be the upper limit if you plan to keep the CPU for a few years.

For 90NM chips, 1.55 should be the upper limit. For 130nm, you can safely to go 1.6.