How to Overclock Intel xeon x5680 on ASUS P6T Deluxe v1

[MuHcOw]

No I'm not very comfortable 1.4 - I would actually like it to be lower. But the 4.4GHz I do like...
I'm thinking that if it passes at 210 (max BCLK) then I know what the limit is. I'm somewhat sure it won't fail because of lack of vcore The current combination is also pretty close to stock DRAM frequency.
Next test would be with lower vcore.

Temp right now is ~75°C (peak still at 78°C) - The room is getting warmer.

 

[UltraTaco]

Donkey, those temps look pretty good for stressing! Mine at 3.8 with HT and 1.275vcore was about there as well..btw, donkey IMPORTANT!!

CHECK IF CPUZ SHOWS ACTUALLY 4.4 GHZ!! I just remembered, p6t will throttle down to lowest CPU multi if temps reach high temps and will not jump back to normal frequency until restart.

Mine throttle when reaching 80°C, please check if it's still at full frequency and keep an eye as you're approaching 80°. If it drops to lowest multi, then it will be rock stable with 1.4vvore and all stressing will go in vainnif left unnoticed.

If you're locked at wanting 4.4 GHz and that baselock, then once stable try lowering vcore and see if you can get her stable with lower vcore. Because at 1.4, that's getting up there. Any down bumps will be good.

 

[MuHcOw]

Stable at 4,405.5 MHz on all cores - Thermal throttling has NOT been enabled. Minimum distance to TjMAX has been 23°C

9 hours now...

(edit)
I'm not locked at anything yet - But I like 4.4GHz, can't deny that

 

[UltraTaco]

Yeah 4.4 I wouldn't mind that either
How do you disable thermal throttle? My p6t doesn't have that. Pic of bios setting ?

 

[MuHcOw]

Poor choice of words on my part. I don't think you can disable Thermal Throttling.
I'm using HWiNFO and it says that Throttling has not been used (yet).

 

[UltraTaco]

O. Good then. Keep an eye on it, man. You're getting awful close.

 

[MuHcOw]

I will and I am

Coming up on 20 hours...

 

[EarthDog]

24 hours? Geez... overkill. Typically, most users can have a stable system for their uses with 4-8 hour testing and be good most of the time. 24 hours is damn overkill unless its a production system which requires the utmost in stability. Few car that P95 abends at 18 hours when their PC isn't doing anything strenuous for more than a few hours (and few things are as strenuous as stress testing - gaming sure isn't).

 

[UltraTaco]

I would probably agree at this point, since this is on y preliminary testing. This isn't final overclock and not final voltage, perhaps call it done and start lowering vcore until it crashes and then bump like 3 notches up.

Donkey, you have still long way to go testing uncore and ram and then whole system altogether. Better switch to somewhat "fastlane iverclocking"

I ran out of thanks again.
Taco to ED - Thanks

 

[EarthDog]

Not looking for agreement at any point...The final overclock or voltage does not matter. While everyone has their own way to do it, many feel that 24 hour stress testing is overkill unless whatever you are doing on your system runs for quite a long time (production as I said earlier) and or ABSOLUTE stability is of the utmost importance. Most would say a system is stable if it is stress tested for several hours.

Perhaps consider not handing out thanks like candy and you will have them when it matters.

 

[UltraTaco]

Very valid points on candy, but "final oc or voltage does not matter?" Sir?

I had errors on 6 or 8 hours and while it may never show up in real use, it would make me uneasy. Let's rember, ED, prime95 needs to go through all its numbers. 1or 2% instability is still huge when considering that modern cpu makes billions of calculations. Every once in a while 2+2=1 may add up over time.

I prime 20 hours now because it never errorer after that. I did have errors on 16 and 18 hours though..

Someone took my stars away

 

[MuHcOw]

Well I'm planing on installing it in my spaceship - Need absolute stability for that coffeemaker!!
23 hours now - I guess it's stable enough at 210.

So next step is to lower vcore until it starts loosing stability again, right?
I'm testing with Small FFTs now - Should I keep on doing that when testing BCLK stability?

 

[UltraTaco]

Well if you're at 1.4 and you drop it to 1.37 and it crashes fairly soon, then you're in the ballpark and just leave it there if you plan to keeping 4.4ghz, but if she keeps running even at 1.35(unlikely?) Then try raising it even lower until it doesn't run anymore. You know for sure it passes at 1.4, if you find the low spot for vcore, it may be easier to narrow down where you stand on voltage in regards to trying to find lower stable vcore.

Or you could start messing with ram and uncore(those two are sort of tied in a way, because uncore has to be at least 2x ram speed)

 

[EarthDog]

Very valid points on candy, but "final oc or voltage does not matter?" Sir?

I had errors on 6 or 8 hours and while it may never show up in real use, it would make me uneasy. Let's rember, ED, prime95 needs to go through all its numbers. 1or 2% instability is still huge when considering that modern cpu makes billions of calculations. Every once in a while 2+2=1 may add up over time.

I prime 20 hours now because it never errorer after that. I did have errors on 16 and 18 hours though..

Yes, exactly what I said in context is what I meant. I don't see a point in running a 24 hour stress test unless its the parameters I set forth (for myself of course). This means, to me, that once I find a 'stable' voltage and clock after a few hours, there isn't a need to put it through anything else more than 8 hours. I don't know why it would make you "uneasy". If you rig doesn't freeze in your uses, its stable (for your uses). It really is that simple.

I'll take my chances, as I have done for years, in testing for a few hours. Again, this isn't a production system... this isn't something I need to be absolutely stable because I am crunching numbers. I email/web/game/encode for a few hours at most. I really don't care that P95 fails 18 hours in when in MY USES, I do not have any trouble at all. Again, everyone has their methods. Some prefer 24 hours for 'peace of mind' (or out of necessity), others can get by just fine with less.

For RAM an uncore, unless you are benchmarking and need that extra 1-2%, I wouldn't waste my time on that.

 

[UltraTaco]

Okay, you're right.

 

[EarthDog]

It isn't about right or wrong as there are plenty of use cases for such stringent testing, just not for most consumers. If you and this guy want to test for 24 hours, more power to you, but, chances are it isn't needed due to how the PC is used.

 

[UltraTaco]

Thank you, ED

 

[MuHcOw]

Someone took my stars away

Yes ??? - Where did they go?? - You can have mine, if you want...

 

[UltraTaco]

You only have one, brother. Hang on to it my friend, but thank you.

 

[EarthDog]

They were removed. He has the stars he has earned (stars are post count unless staff gets involved - which is what happened to him). But this has nothing to do with your problem. If you would like to chat about the stars, you are of course welcome to PM Taco.