Q6600 Not stable at stock....

cooter

Gawd
Joined
Dec 2, 2004
Messages
872
So i was able to get my hands on a Q6600 for cheap from a coworker. It was never overclocked in the past. I swapped it out with my E6750 (which is going in my HTPC) and I can't get the Q6600 to make it through a few hours of Prime95 without one of the cores erroring out. I started out with everything in the BIOS on auto and just figured i would try a run for the shits and giggles. I recieved an error on one core. Then after setting the voltage to 1.25 i tried again. same problem. I have now upped the voltage (in steps running prime 95 each increase) to 1.4 and still have not had any luck. This system is running bone stock with water cooling. The water cooling is running really slow right now (fans and water pump on lowest settings so it is nice and quiet) and the temps at 1.4 volts get into the 60's under load via HWmonitor. There is an 8800 GTX in the water look as well so that helps to heat it up a bit. However I don't feel that this should result in an unstable stock system.

Setup:

CPU: Q6600 B3
Mobo: GIGABYTE GA-P35-DS3P
RAM: 4 GB GSkill http://www.newegg.com/Product/Product.aspx?Item=N82E16820231145

All other specs are in sig.

The bios on the motherboard was just updated. The system has never had a problem before. Any ideas on whether i missed something?
 
circuts degrade over time. the more power you run thru them, and the hotter they are alowed to run, the faster this happens. it could very well be that this chip you have has degraded to the point where it no longer runs stabily even at stock speeds.

it could also be that your 520w power supply isnt able to give enough clean power to a quad core cpu and the 8800, plus mobo, ram, hard drives, cd drive, water pump, fans, USB devices, etc.

just a thought, could be something else all together.
 
Are your memory timings set to the spec sheets? You may want to try underclocking your memory to 667 or so and see how that fares.
 
Which Prime95 test are you doing? Don't do Blend as that will stress other things that might be unstable like your ram. Try 8x multi.

Stay with 1 stick of ram, take everything else out of the equation. Give the ram and memory controller (northbridge) a bit of extra voltage in case just to cancel them out as variables in your testing.

Also, don't worry about Prime95 so much. If I can prime for 20 minutes, I pretty much can play for 24 hours of nonstop gaming and have no issues.

BTW, I also have a B3. Needs 1.525v w/ Load Line Calibration for 3.6GHz. vdroop brings it under the 1.5v spec.
 
Do you have the latest BIOS for that board?

Also, did you clear the BIOS via the jumper on the motherboard after installing the Q6600?
 
Which Prime95 test are you doing? Don't do Blend as that will stress other things that might be unstable like your ram. Try 8x multi.

Stay with 1 stick of ram, take everything else out of the equation. Give the ram and memory controller (northbridge) a bit of extra voltage in case just to cancel them out as variables in your testing.

Also, don't worry about Prime95 so much. If I can prime for 20 minutes, I pretty much can play for 24 hours of nonstop gaming and have no issues.

BTW, I also have a B3. Needs 1.525v w/ Load Line Calibration for 3.6GHz. vdroop brings it under the 1.5v spec.

Ha, I am using the Blend test. I'll also look at trying a 8x multi later if other things don't work.

Are your memory timings set to the spec sheets? You may want to try underclocking your memory to 667 or so and see how that fares.

Memory timings are all set to auto. I'll set to either 667 or 800 and take the memory out of the equation.

Do you have the latest BIOS for that board?

Also, did you clear the BIOS via the jumper on the motherboard after installing the Q6600?

I have the latest bios off the Gigabyte website as of this weekend. I did not clear the bios after i updated it via jumper.
 
up the FSB voltage a notch and +.1 to the mch, memory voltage +.3 (for 2.1V total) see if that helps.
 
well you have a low wattage 5 year old psu. I personally thought that psu was really cool back in the day, but they didnt have q6600's and 8800gtx then either. I agree with no blend testing put it back to stock volts clear cmos before install cpu is just basics with me ( i take out battery tho for 20 mins while i do the install etc) Try diff cpu, also just because someone said they didnt overclock it doesnt mean they didnt! maybe person before them did? (id say it should still run stock fine even if overclocked so doubt thats it) also 60C load? or over 60C? id turn up the fan a little or the pump or what ever need to cool it prob take the 8800gtx out of the loop and add diff loop? 55C is the max id let a cpu get to (but im picky?)
 
Well, i went home and made a few changes and now it is running stable.

1. Not running Blend test anymore.
2. Reset BIOS a verified everything was stock.
3. Turned up the fans in case and the water pump.

Running good now. I may start over and try to overclock it up to 3 GHZ, but I need to fix the cooling in my office, the office gets quite hot and I think that is where quite a bit of my cooling problems come from. It must have been like 85-90 in the office yesterday! I need to run an A/C Duct in there.
 
Well, i went home and made a few changes and now it is running stable.

1. Not running Blend test anymore.
2. Reset BIOS a verified everything was stock.
3. Turned up the fans in case and the water pump.

Running good now. I may start over and try to overclock it up to 3 GHZ, but I need to fix the cooling in my office, the office gets quite hot and I think that is where quite a bit of my cooling problems come from. It must have been like 85-90 in the office yesterday! I need to run an A/C Duct in there.

Well, you still have a stability issue SOMEWHERE if the Blend test is failing, just it's probably not the CPU. Likely northbridge, memory, something like that.
 
id say reseting the bios solved it? or maybe it would have passed the non-blend test with out those 3 things guess we will never know unless he tries blend test again, ya voltage on chipsets but it should run fine stock/stock
 
Well, you still have a stability issue SOMEWHERE if the Blend test is failing, just it's probably not the CPU. Likely northbridge, memory, something like that.

QFT

Blend also tests memory, so, as he said, the memory or northbridge voltages may not be sufficient.
 
I had a core fail on blend. I ran a memory test and it failed. I had to scale back my timings from 5-5-5-18 to 5-6-6-20 and the memory errors went away and the blend test didn't kill the core 10 seconds in like before.

I have found with my boards, a MCH voltage bump is rather important with 4+GB of RAM on a higher FSB.

My rig sig is as such: vore 1.5, MCH 1.2, and DDR2 voltage is at 2.0 (Rated between 1.8-2.2 depending on how for I wish to push the timings.) The CPU termination is up, but I forget what I set it to. Though I am not sure I needed to raise it at all. (Done before I realized the RAM was the issue.)
 
I'll try the blend test again. I may have to up some of the other voltages as well. I have not had alot of time to work on it, but will work on it while I set up my new media center rig.
 
well, i tried the blend test again, failed. I think i need to mess with some of the other voltages. Not sure how long it lasted, as i left the house and a couple hours came back and one of the cores stopped. I'll assume it is the ram or northbridge voltage that needs adjusted. i'll check it out tonight again. I am just shocked how hot this thing is!
 
I would be good to know some facts...like screen shots or summary of your clock speed, voltages (all, or most), fsb, ram timing, and any motherboard settings related to overclock...like speed step, etc. Any intel cpu should run at stock w/out changing anything (kind of obvious).

.
 
Take a few pics of your M.I.T. section in the BIOS and post them up for us to analyze
 
I would be good to know some facts...like screen shots or summary of your clock speed, voltages (all, or most), fsb, ram timing, and any motherboard settings related to overclock...like speed step, etc. Any intel cpu should run at stock w/out changing anything (kind of obvious).

.

I was just thinking and was going to suggest the same thing before reading your post.

So, yeah, give us all the nitty gritty you can as to how things stand now.
 
I was just thinking and was going to suggest the same thing before reading your post.

So, yeah, give us all the nitty gritty you can as to how things stand now.

Hopefully i can get something on here this weekend, i spent last night getting my new media center up and running, should have more time this weekend.
 
So i was able to get my hands on a Q6600 for cheap from a coworker. It was never overclocked in the past. I swapped it out with my E6750 (which is going in my HTPC) and I can't get the Q6600 to make it through a few hours of Prime95 without one of the cores erroring out. I started out with everything in the BIOS on auto and just figured i would try a run for the shits and giggles. I recieved an error on one core. Then after setting the voltage to 1.25 i tried again. same problem. I have now upped the voltage (in steps running prime 95 each increase) to 1.4 and still have not had any luck. This system is running bone stock with water cooling. The water cooling is running really slow right now (fans and water pump on lowest settings so it is nice and quiet) and the temps at 1.4 volts get into the 60's under load via HWmonitor. There is an 8800 GTX in the water look as well so that helps to heat it up a bit. However I don't feel that this should result in an unstable stock system.

Setup:

CPU: Q6600 B3
Mobo: GIGABYTE GA-P35-DS3P
RAM: 4 GB GSkill http://www.newegg.com/Product/Product.aspx?Item=N82E16820231145

All other specs are in sig.

The bios on the motherboard was just updated. The system has never had a problem before. Any ideas on whether i missed something?
Sounds like your coworking is pulling your leg. Who's going to admit "Yeah I OCed the shit out of it and killed a core, wanna buy it off me?"
 
I guess the reason I believe him is because he has no reason to lie as he. offer edit too give me my money back if it don't work. He dont have the knowledge anyway.
Posted via [H] Mobile Device
 
Ok, so I messed with it again today. I set the voltages up and i am still getting errors. Here are some pics of the BIOS configuration and my desktop that shows prime95 and temps.

My questions include, which temps should i believe, and why am i still getting these errors.

q6600_1.jpg

q6600_2.jpg

q6600_3.jpg
 
Run memtest. I'm pretty sure it's your ram that's faulty, not your CPU. I used to get the rounding error in P95 with my Q6600 and that was entirely due to ram.

You could try turning on load line calibration as it will minimize CPU vdroop and give you a bit more voltage than 1.3v. If you want to try pushing it, put your vcore up to 1.4 with load line calibration to take issues of CPU voltage out of the question.

+0.3V on the DDR is pushing it a bit. What's your default? 1.8v? You are putting 2.1v into the ram when it might not be able to take it.
 
Run memtest. I'm pretty sure it's your ram that's faulty, not your CPU. I used to get the rounding error in P95 with my Q6600 and that was entirely due to ram.

You could try turning on load line calibration as it will minimize CPU vdroop and give you a bit more voltage than 1.3v. If you want to try pushing it, put your vcore up to 1.4 with load line calibration to take issues of CPU voltage out of the question.

+0.3V on the DDR is pushing it a bit. What's your default? 1.8v? You are putting 2.1v into the ram when it might not be able to take it.

When I ran test last night with Load line calibration enabled i still got an error. I'll run memtest see if it could be the ram. As for the RAM it is made to pull 2.0 thru 2.1 volts.
 
Last edited:
Run memtest. I'm pretty sure it's your ram that's faulty, not your CPU. I used to get the rounding error in P95 with my Q6600 and that was entirely due to ram.

You could try turning on load line calibration as it will minimize CPU vdroop and give you a bit more voltage than 1.3v. If you want to try pushing it, put your vcore up to 1.4 with load line calibration to take issues of CPU voltage out of the question.

+0.3V on the DDR is pushing it a bit. What's your default? 1.8v? You are putting 2.1v into the ram when it might not be able to take it.

Thanks for the info. I tested the RAM with memtest and found that one of my sticks was bad. I have been running these sticks for a couple years with my e6750. I only decided to test it because of the new processor and I kinda wanted to overclock my q6600. And sure enough the ram was bad. I just assumed that because the ram was good when i initially put it in and tested it that it was still in good shape.

Anyway, gotta love gskill, sending RMA now. hopefully in a couple weeks i can mess with this. One question I do have, what temps are the ones that i should pay attention to? the Coretemp ones or the hwmonitor ones? also within that program what is the max temp for the q6600?
 
Direct answer- each cpu is set at the factory and the numbers vary cpu to cup but should be around 100C.

(That is why the "delta to TjMax" discussed below is the most accurate method, you cannot find the actual top number but the cpu knows what its factory calibrated max temp is and reports accordingly. We just cant read it to use it ourselves. To get a "temp" the programmer has to guess at a TjMax and do some math to come up with a number and that is why some programs will vary on the same system, some us 95C some 100C as TjMax and the real story is both numbers are likely incorrect but in the ball park.)

Realtemp (much the same as coretemp just better IMO, but either will work fine) is my favorite now.

At right around 100C the cpu will send out an 'I am hot" signal for the motherboard to spin the fans to 100% speed. At some unknown (likely 10-20C more) higher temp it will shut itself down to prevent damage. Most of us like to maintain a 20C "headroom" or safety (really not a safety margin you could run the cpu at 99C 24x7x365 and it would not be damaged) margin so somewhere around 80C as reported by Coretemp/Realtemp under load is a good place to stop pushing voltage etc.

The really accurate Coretemp/Realtemp measurement is "Delta to TjMax". This is like a countdown timer that uses the actual values reported by the CPU and uses no math to come up with a number. This "distance to TjMax" is like a countdown timer, when it gets to zero the cpu thinks it is hot. So like above as long as under full load your "distance to TjMax is > 20 or so you are good and the higher the number the better.

Max Vcore is 1.55V but the voltage requlation circuitry allows .05V of overshoot. Unless you are into extreame OCing I recommend stopping at 1.45 Vcore and I personally dont like going over 1.4V as I want to keep the CPU for a long time.

I am glad you found the problem. It will be fun when you get that stick back and can actually get that puppy barking like it should.
 
Last edited:
just wanted to send an update. I have some temporary ram installed until i get my stick back. In the meantime I wanted to see what I could get.

I was able to push my B3 to 3.0 on stock volts.

Once I get my ram back I will try to push it a bit further.
 
You should try running orthos and test cores individually and seperately from the ram. Errors might be caused by your ram.
 
You should try running orthos and test cores individually and seperately from the ram. Errors might be caused by your ram.

I was able to figure out that one of the sticks of RAM was bad a couple weeks ago. I literally just put the new stick in 5 minutes ago and am running the blend test on Prime95 on stock settings with the new ram. If it works well, I will run up the OC to see where i can get it.
 
Back
Top