Stress Testing

[wirelessfender]

I have what I thought was a stable 1Ghz overclock on my Q6600. I have been running Prime95 and Orthos daily and most come back with no errors. However only have been running them for 10-20min. But occasionally I do get an error on one of the cores usually after 10min of test. My friend says that all you need to do for stress testing is run orthos for 5min tops and you know you are good. However, I read on forums that it should go overnight or 24 hours?! 24 hours seems kinda extreme if you ask me.

The main reason I ask is because even though the stress test error occasionally, I haven’t had any problems keeping my system clocked the way it is whether it be gaming or normal operation.

Do I have a stable OC?

 

[cnumartyr]

I have what I thought was a stable 1Ghz overclock on my Q6600. I have been running Prime95 and Orthos daily and most come back with no errors. However only have been running them for 10-20min. But occasionally I do get an error on one of the cores usually after 10min of test. My friend says that all you need to do for stress testing is run orthos for 5min tops and you know you are good. However, I read on forums that it should go overnight or 24 hours?! 24 hours seems kinda extreme if you ask me.

The main reason I ask is because even though the stress test error occasionally, I haven’t had any problems keeping my system clocked the way it is whether it be gaming or normal operation.

Do I have a stable OC?

Personally I consider 8 hours fine for CPU.

Odds are if it makes it 1+ hour on Prime95 you won't see errors if you restart daily or something like that and don't do anything overly intensive.

Your friend might think of that as stable, I know plenty of people that wouldn't agree. That isn't even the point of saturation for temps.

 

[wirelessfender]

Im guessing point of saturation means how high your temps go before they stop getting any hotter? My temps (using coretemp) hit 98c at about 5min of 100%load, then I usually run it another 5min and they dont go any higher.

 

[cnumartyr]

Im guessing point of saturation means how high your temps go before they stop getting any hotter? My temps (using coretemp) hit 98c at about 5min of 100%load, then I usually run it another 5min and they dont go any higher.

98C ?!?!

Edit: And yes, Saturation is generally 7-10 minutes.

 

[wirelessfender]

Oh come on dont tell me thats too high???

 

[wirelessfender]

thats under the threshold, and this is a sustained 100% load were talking about. When in real load is it ever going to run at 100% on all cores for more than 30 seconds even?

Dear god please tell me im wrong, lol

 

[cnumartyr]

From Comp's guide:

Scale 2: Quad

Q6x00: Tcase Max 71c, G0 Stepping, Tjunction Max 100c, Vcore Default 1.372, TDP 95w, Delta 10c

-Tcase/Tjunction-
--70--/--80--80--80--80-- Hot
--65--/--75--75--75--75-- Warm
--60--/--70--70--70--70-- Safe
--25--/--35--35--35--35-- Cool

Scale 4: Quad

Q6600: Tcase Max 62c, B3 Stepping, Tjunction Max 100c, Vcore Default 1.372, TDP 105w, Delta 10c

-Tcase/Tjunction-
--60--/--70--70--70--70-- Hot
--55--/--65--65--65--65-- Warm
--50--/--60--60--60--60-- Safe
--25--/--35--35--35--35-- Cool


Not sure if you had a G0 or B3, so I put both.

I have a G0 at 3.6 GHz and I wouldn't let it runs at 100% for 10 hours and never breaks 63C on the hottest core.

I wouldn't OC it at 100% past 70 C just because I consider that "safe."

2 C under the automatic throttle point is (in my opinion) insane.

Tjunction 100 C doesn't mean that's how hot you can run it... That's the point it auto-throttles for protection.

You are right, most likely won't see 100% too often, but I still want mine within thermal specs at 100% as a worst case scenario.

 

[cnumartyr]

What CPU cooler/fan, voltage, and FSB/Multiplier setting?

 

[wirelessfender]

Well what about it being a core temp reading? Core temp is the actual temp if the core. most temp monitoring programs appear hotter than core temp.

I also have the G0, how do you have yours at 3.6 without being as hot as me? You watercool?

I have the antec 900 with all available fans, and a Thermaltake V1 cpu cooler.

 

[wirelessfender]

Vcore is 1.38v but probe and cpuz read it at 1.32. Fsb is at 377 with a multi of 9

 

[cnumartyr]

Well what about it being a core temp reading? Core temp is the actual temp if the core. most temp monitoring programs appear hotter than core temp.

I also have the G0, how do you have yours at 3.6 without being as hot as me? You watercool?

I have the antec 900 with all available fans, and a Thermaltake V1 cpu cooler.

I now what Core Temp is. 75 C is the general limit here. 71C Tcase (What most programs read as "CPU" temp) is the max specified by Intel. Generally cores run +10 C over the Tcase +/- 2 C. That makes anything over 80 C pretty darn hot.

I have a Thermalright Ultra 120 Extreme.

It really sounds like you might not have it mounted properly or something or use way too much TIM. The V1 isn't a horrible cooler... I'd try remounting it if you decide that it is too hot for you.

 

[cnumartyr]

Vcore is 1.38v but probe and cpuz read it at 1.32. Fsb is at 377 with a multi of 9

Probe/CPU-Z are reading the actual Vcore. Vcore setting "droops" from what you set it at. I set mine 111mV higher than what I actually get in Windows. From the errors it sounds like you need more voltage to get it stable for more than an hour.

Something is wrong with that V1. It shouldn't be letting you get 98 C Core Temps with a 3.4 GHz clock with 1.32 vCore (which btw is low for the "average" Q6600 at that speed and why you are getting errors).

 

[wirelessfender]

I have remounted it already, But im using zalman thermal grease. I wonder if thereisnt enough on it...

 

[wirelessfender]

hmmmm... more voltage would mean more heat... but if your right, with my cooler I shouldnt be getting this hot. I have to say though, its mounted on there really good, I dont think its not mounted properly.

 

[cnumartyr]

I have remounted it already, But im using zalman thermal grease. I wonder if thereisnt enough on it...

It's entirely possible you have a horribly bum chip.. What does CoreTemp display as the VID of the processor?

With a TIM less is normally better. Zalman isn't a great performer... I'm not sure man, something is up with how it's mounted or something somewhere.. But yea.. I'd back off on the overclock for now. Or you used too much... or maybe too little.. but I doubt too little. Did you follow manufacturer's recommendations for application of the paste?

 

[wirelessfender]

VID is 1.2250v. and what is TIM?

 

[cnumartyr]

VID is 1.2250v. and what is TIM?

Thermal Interface Material.

1.225 is a great VID (mine is 1.2875), so I think it's the cooler: either the TIM under it or the mount.

 

[Evilonigiri]

Ouch...98C. You can boil water on a small mountain with that amount of heat. Say cnumartyr, was your Thermalright 120 extreme concaved on the bottom? I heard nearly all of them are.

 

[wirelessfender]

I have some thermal grease around somewhere... I should give that a try. Would checking to see if the bottom of the V1 is completly flat be an idea? I could sand it down if its not.

 

[cnumartyr]

I have some thermal grease around somewhere... I should give that a try. Would checking to see if the bottom of the V1 is completly flat be an idea? I could sand it down if its not.

In the long run, yes. However it still shouldn't perform that horribly even if it was shaped like a pyramid (exagerration, but you get my point).

Also yes, my TRUE was convex. The IHS on the CPU was concave. They didn't mesh perfect as they weren't machine for each other...

Here is my old temp log from before I lapped.

Clock.....vCore......Load
2.4 1.248 42
2.6 1.248 46
2.8 1.248 47
3.0 1.248 48
3.2 1.264 51
3.3 1.264 52
3.4 1.312 55

Those were all measured with 10 minutes of 100% Prime95 max core temp (hottest the hottest core got). All were taken at a 24/25 C Ambient.

 

[wirelessfender]

cnumartyr, may i ask what your OC settings are for your cpu and memory?

 

[cnumartyr]

cnumartyr, may i ask what your OC settings are for your cpu and memory?

400 MHz FSB @ Stock NB Voltage

Memory is stock at DDR2800 (1:1 with FSB) @ 1.8v 4-4-4-12 (Transcend)

CPU is 400x9 with 1.400 vCore (actual) under load.

 

[wirelessfender]

Right on. Im thinking I might try something like that. I gotta work on my memory though, with ballistix tracer 1066 its running at 980ish and the timings are 5-5-5-15(bios) 5-5-5-18(cpuz). I would like to figure out why the timeings differ like that, and try to get them closer to 1066 or higher. but right now I got to get to the bottom of this heat/stability issue.

 

[wirelessfender]

You know what else is wierd right now.. PC Probe goes wankers every now and then at reads 4.08v across the board. Think thats an unstable OC talking? Probes freaking me out.

 

[cnumartyr]

You know what else is wierd right now.. PC Probe goes wankers every now and then at reads 4.08v across the board. Think thats an unstable OC talking? Probes freaking me out.

Yea, I'd take the OC off for now and see if it continued while you try to figure it out.