How temperature is affecting my rig.

[Jeaze_Mcleod]

I am begining to think that temperature in my rig eventhough not at overheat levels is playing a major role in the stability of the system. I am curretnly running at 3.4 Ghz and I have thrown everything (Prime 95, Orthos, Super Pi, 3DMarks, etc) except the kitchen sink and it seems to be stable for hours on end.

 

So one day I decide to install NFS Underground 2 and set everything up to highest detail and then start to play. The temperature on average of the cpu at idle was in the 43c degrees according to Coretemp. That looked fine to me considering that I am running overclocked. So I start to play and everything is fine and after 4 or 5 hours of play NFS just quits and it dumps me to the desktop. :confused:

 

I tried to load windows task manager and it would not load and the desktop started to act kinda funny and it started to jitter. I restarted the system and everything has been fine ever since. Also the temperatures were in the 55c degrees as soon as I took a glimpse at the desktop upon getting kicked out of the game and also checked the video card temp and it was in the 51c degrees which should be ok as well.

 

NFS Underground is really not that groundbreaking in graphics but it definately did a number on my rig in the long run. I kinda noticed a very slight slowdown almost down to a stutter but I dismissed it and didnt think much about it. I am going to try and use the game again after lowering the room temperature to about 65c degrees and see what happends from then on.

[Thasp]

I passed 25 hours of prime95 on my old 3000+ venice system, with a gig of RAM, at 2.4 GHz. blend test. I passed memtest for about seven hours.

 

I got errors 7 minutes into FEAR.

 

I up the CPU voltage from 1.475v, to 1.5v, and it's gone.

 

There is no such thing as a stable overclock, just what's stable for you. Weird things happen.

 

Up the CPU voltage 0.025v and see what happens. Is that RAM elpida ICs, or micron D9 ICs? What voltage is it at right now, and what speeds are you running the memory at?

[Jeaze_Mcleod]

I passed 25 hours of prime95 on my old 3000+ venice system, with a gig of RAM, at 2.4 GHz. blend test. I passed memtest for about seven hours.

 

I got errors 7 minutes into FEAR.

 

I up the CPU voltage from 1.475v, to 1.5v, and it's gone.

 

There is no such thing as a stable overclock, just what's stable for you. Weird things happen.

 

Up the CPU voltage 0.025v and see what happens. Is that RAM elpida ICs, or micron D9 ICs? What voltage is it at right now, and what speeds are you running the memory at?

 

From what I understand the chips on my RAM should be Microns, but is there a sure way to find out? I am running the RAM at 2.25V with 4-4-4-12 timings.

On a side note so far upping the voltage to the CPU is just increasing temperature and not much stability wise... not sure if temps above 55c degrees just wont let it work any better.

 

Might go with water cooling when I have the cash to spend.

[Thasp]

Ok, if it is micron, then up the voltage to 2.3. You can do this safely. If you were using elpida ICs at 2.25v, they would either not be working, or fail memtest in five seconds, so my best guess is they are micron.

 

Is that 55c coretemp, or 55c motherboard socket temp? If socket temp, perhaps you are pushing it. If coretemp, you are well within acceptable limits.

 

It is said that 60c is what you should stay at if you're looking for a decent overclock and you're an average overclocker. However, it won't begin to throttle until both cores hit the 82-83c point. With best case scenario air cooling(tuniq tower 120, high end scythe, high end thermalright), won't happen until you get to 1.535v-1.55v in a 77F room. This is complete approximation, I am just trying to give you an idea from my experiences and what I have read. This is with a 2000 RPM 120mm fan.

 

For me, I have done lots in the above 60 range, as long as I stay below 75, this is ok fo rme, but others may disagree.

 

Adding a tiny bit of voltage to the CPU won't add that much heat. These boards have smaller voltage steps than the previous nforce4 models I believe, 0.015 instead of 0.25 I think.. so adding a tiny bit won't hurt. Even with the nforce4 boards, going from 1.475v to 1.5v added a barely noticeable amount of heat. If you add a tiny bit, at the same speed, it can solve the stability issue. If it doesn't, then you know it isn't the CPU(or the CPU needs a lot more voltage), but in my experience I always add a tiny bit of CPU voltage, if it fails in a game after passing orthos and 3dmark.

 

As for watercooling, if you are looking at setting a rig up, post a budget, and whether you insist on all the parts fitting inside the case and I(and others with more experience) would be happy to help you pick out some parts. I have little/no common sense, I learn from making mistakes.. and I made mistake after mistake with watercooling, so I could perhaps keep you from making some of the mistakes that I made. The Corsair Nautilus is a nice simple beginner's watercooling unit, and for the $150 price it does an awesome job. If you're willing to spend higher($300 range), you can get something that'll do about 7-10c better and cool the chipset and GPU.. but would leave you learning as you go opposed to following a manual. If you decide to spend more than what the corsair nautilus costs, I recommend going the custom route so you get the best possible setup. Either way you go, if you decide to leave aircooling, post your ideas here.

[wevsspot]

Jeaze,

 

If your E6600/975 INF combo acts anything like my stepping you will need at least 1.46 vcore at 3.4ghz to get you 3D stable for hours on end. I played around a little bit at a lower vcore but had a similar experience to yours, only mine was a marathon session of BF2. I crashed to desktop and then everything went flaky. Auto-hide taskbar icons quit working, my anti-virus failed to load at boot and other weird behavior. A bump up to 1.46 vcore completely cleared up my problems. And thankfully it doesn't appear that I boogered up my Windows install in the process.

 

In addition to cpu heat concerns (of which I don't think you have any as mine has hit 63C (dual core temp utility) after running eight hours of Orthos without a crash, you should keep an eye out on your n/b temps. When that gets too hot it will affect your stability alot also.

[Jeaze_Mcleod]

Is that 55c coretemp, or 55c motherboard socket temp? If socket temp, perhaps you are pushing it. If coretemp, you are well within acceptable limits.

 

I am using CoreTemp to read both core temperatures.

 

In addition to cpu heat concerns (of which I don't think you have any as mine has hit 63C (dual core temp utility) after running eight hours of Orthos without a crash, you should keep an eye out on your n/b temps. When that gets too hot it will affect your stability alot also.

 

The temp with the NB I think I have under control and keep it at around 45c most of the time even with the current overclock and full 1.75v mod. It used to climb up to 51c with no active cooling.

 

I was getting problems with NFSU 2 even at stock speeds so I decided to uninstall and then reinstall the game and so far so good. It seems I had been misled by the game crashing on me for no reason. I have gone thru one xp recovery already when trying to get above 380 fsb so maybe that messed up the whole thing. I will try to keep on experimenting with at least a room temperature of 70f so I can at least achieve 400fsb.

 

No my question to you guys is how far can I up the voltage to the CPU? So far the largest gain has been to 1.58V. How much is too much so I dont fry the CPU?