AMD 9600 Black OC effort

archae86
archae86
Joined: 6 Dec 05
Posts: 2,712
Credit: 2,708,481,341
RAC: 3,595,621

RE: The highest temp I can

Message 79843 in response to message 79841

Quote:
The highest temp I can find with my Fluke 65 IR temp probe is 56C on the stock Q6600, but only 46 on the 9600 (at 2.639).


Quote:
The Q6600 has:
Asus P5K-VM mobo
Artic cooling 7 freezer pro heatsink/fan

With the Artic Cooling Freezer Pro in place, what are you finding to measure with the IR probe?

Astro
Astro
Joined: 18 Jan 05
Posts: 257
Credit: 1,000,560
RAC: 0

RE: RE: The highest temp

Message 79844 in response to message 79843

Quote:
Quote:
The highest temp I can find with my Fluke 65 IR temp probe is 56C on the stock Q6600, but only 46 on the 9600 (at 2.639).

Quote:
The Q6600 has:
Asus P5K-VM mobo
Artic cooling 7 freezer pro heatsink/fan
With the Artic Cooling Freezer Pro in place, what are you finding to measure with the IR probe?


Both the 7 and 64 freezer pro sit up 1 5/8" from the mobo to first fin(the fins are parallel to the board). Both mobo's are open and laying on my desk, so I can point the IR at pretty much any side of the CPU and copper mating blocks, and even the heat pipes I (except the vid card on the 9600 makes a bottom/up approach difficult). I worry about accuracy given the "shiny" surfaces. I can tell you the Q6600 at stock feels warmer to the touch than my OC'd 9600 (touching the heatsink fins). In all fairness I haven't lapped the Freezer 7 on the Q6600. I examined the nylon push to latch mechanism on all four corners but wasn't sure I could remove them and reuse them safely. There removal appears necessary to get the pad flat on my granite inspection block. The two of them appear identical except for latching methods. Specs say the 7 has 5 more CFM though. Each has 43 aluminum fins and they have the same dimensions (surface area/fin). I.E If I removed the mounting means and label, I would defy anyone to think they were two different models.

Astro
Astro
Joined: 18 Jan 05
Posts: 257
Credit: 1,000,560
RAC: 0

Attempting to employ the same

Attempting to employ the same methodology to the collection of temps. I measured the center of the heatsink from 1 hand width away and at a roughly 20 degree vertical elevation angle but perpendicular to the edge of the fins (so I wasn't shooting between the fins at open air)(these directions are as it's laying there, not as mounted in a standing box). The Q6600 at stock measures 37.7C, the 9600 OCed to 2.639 measures 35.5C. Remember the laser pointer is offset to the left of the IR collector, so to take similar readings of the cpu/heatsink pad, I approached from the rear the board (rear being the rear if it were mounted in a box), placed the bottom of the heatsink on the PS2 connector box with the front of the meter nearly even with the back of the board and pointed the laser at the left corner of the cpu/heatsink(hopefully point the IR collector just left of center of the pads). The q6600 measured 52.8C and the 9600 measured 43.6C. Both running all cores with Simap at 100%.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 480
Credit: 79,701,289
RAC: 66,562

My new Q6600 is running at

My new Q6600 is running at 3.006GHz with temps, as reported by PCWizard, at;
cpu 30C
core1 43C
core2 35C
core3 34C
core4 43C
using Thermalright IFX14 HS, no fan, but 120mm case extraction fan is ~15mm away and in line. Graphics card is reported to be at 47C.
Have had it running at ~3.2Ghz, and no sign of problems when running usual stress tests, but had a one Einstein error a few hours later, so have backed off.
Using Temp meter with probe, not laser, northbridge has warmest mobo component. This is ver 2 of mobo, version 1.x had smaller NB HS, so I think this might be my blockage to further OC'ing as path to RAM is via NB.

Einstein units at 0921.15 to 0921.40 in 4 (14347.36s) to 5 (17610.06s) hours, using power users app. Done ~7 units/day over last week, but with three errors, one my fault, two with same errors that others have reported. Einstein gets 25% resources on this computer, one core effectively.

archae86
archae86
Joined: 6 Dec 05
Posts: 2,712
Credit: 2,708,481,341
RAC: 3,595,621

RE: cpu 30C core1

Message 79847 in response to message 79846

Quote:

cpu 30C
core1 43C
core2 35C
core3 34C
core4 43C


Those are rather large core to core differences if real. But they may be due to sensor calibration offsets. If you are interested, I suggest attempting calibration by idling the system for long enough to stabilize with the most airflow you can manage (I used a large box fan placed next to my open case). For even better calibration you can use somewhat under stock clock and lower the CPU voltage below stock (1.1 should be OK at, say 1.8 GHz). The leakage component of CPU power is highly non-linear in voltage, and also in temperature, and these two measures will lower it considerably. With all that, I'd expect the cores to match to less than 1C, more than that is almost certainly sensor offset error. If your Core2 and Core3 report lower than ambient in that calibration test, you'll have an absolute indication of calibration offset.

Otherwise your sink is not well mated, I think.

Quote:
Using Temp meter with probe, not laser, northbridge has warmest mobo component. This is ver 2 of mobo, version 1.x had smaller NB HS, so I think this might be my blockage to further OC'ing as path to RAM is via NB.


It is a very common complaint for northbridge parts to be very hot. Folks with my motherboard commonly mount tiny fans to cool them a bit (I did it too). The funny thing is that very few credible reports of actual OC limitation modified by this matter seem to be posted. Apparently the NB parts are generally shipped with enough speed margin to tolerate a quite high temperature while still working correctly.

Courage.

I run my Q6600 at 3.006, but I consider this a very conservative overclock. I think I currently run at a commanded CPU voltage of about 1.37, though it ran Einstein just fine for months at 1.35.

Based not only on my own experience, but many reports, I'd expect you can get well above 3 if you are willing to use some more CPU voltage and don't have a badly limiting problem from other parts of your system.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 480
Credit: 79,701,289
RAC: 66,562

RE: RE: cpu 30C core1

Message 79848 in response to message 79847

Quote:
Quote:

cpu 30C
core1 43C
core2 35C
core3 34C
core4 43C

Those are rather large core to core differences if real. But they may be due to sensor calibration offsets. If you are interested, I suggest attempting calibration by idling the system for long enough to stabilize with the most airflow you can manage (I used a large box fan placed next to my open case). For even better calibration you can use somewhat under stock clock and lower the CPU voltage below stock (1.1 should be OK at, say 1.8 GHz). The leakage component of CPU power is highly non-linear in voltage, and also in temperature, and these two measures will lower it considerably. With all that, I'd expect the cores to match to less than 1C, more than that is almost certainly sensor offset error. If your Core2 and Core3 report lower than ambient in that calibration test, you'll have an absolute indication of calibration offset.

Otherwise your sink is not well mated, I think.

Quote:
Using Temp meter with probe, not laser, northbridge has warmest mobo component. This is ver 2 of mobo, version 1.x had smaller NB HS, so I think this might be my blockage to further OC'ing as path to RAM is via NB.

It is a very common complaint for northbridge parts to be very hot. Folks with my motherboard commonly mount tiny fans to cool them a bit (I did it too). The funny thing is that very few credible reports of actual OC limitation modified by this matter seem to be posted. Apparently the NB parts are generally shipped with enough speed margin to tolerate a quite high temperature while still working correctly.

Courage.

I run my Q6600 at 3.006, but I consider this a very conservative overclock. I think I currently run at a commanded CPU voltage of about 1.37, though it ran Einstein just fine for months at 1.35.

Based not only on my own experience, but many reports, I'd expect you can get well above 3 if you are willing to use some more CPU voltage and don't have a badly limiting problem from other parts of your system.


The core temp variation, has been observered by many people, I googled when first saw it, within first 20 mins after first switch on. Also it is not consistently those two cores, although core 1 always seems to be high and core 1 & 2 combination comes up frequently. It also doesn't seem to vary according to mix of units from three projects, which I thought might have been cause.

Electronics engineer since apprenticeship started in 1962, so know quite a bit about heatsinks etc. And yes I have checked, but not done any lapping etc. Also checked with HS vertical and horizontal to see if gravity was affecting the mating between cpu and HS, a problem I have had in the past.
So far only voltage adjustments tried, was to see minimum Vcpu it could take a stock speed and still work reliably, 1.126V worked here running prime95 torture test.

Would like a few uninterrupted hours of peace to play, but looks unlikely for next few weeks, busy this w/end then Sister and family, four teenage daughters, visiting area over Easter school/college break, thankfully not staying with me.

Conan
Conan
Joined: 19 Jun 05
Posts: 161
Credit: 5,808,481
RAC: 0

Not wanting to hijack this

Not wanting to hijack this thread seeing as it is about Over clocking an AMD 9600 Quad care chip, I have read with interest about the CPU temperatures you are measuring.
Whilst my computer is older than yours, mine is one of the more earlier attempts at a quad core, namely an AMD Opteron 285.

This has two CPU's both dual cores.
Using Core Temp 0.96.1, I am showing on
CPU 0
Core 0 = 71 degrees C
Core 1 = 69 degrees C
CPU 1
Core 0 = 66 degrees C
Core 1 = 61 degrees C

Now this lop sided temperature across the cores I am sure is due to the way air flows around the ANTEC 650 case.
The coolest core is nearest the exhaust fan and the warmest while it is near the two inlet fans they are below the CPU and so don't direct air directly over the processor.
When QMC is running the temperature on whatever core it is on will go up by 4 to 6 degrees and Cosmology also pushes the temps up a degree or two over the other projects including Einstein.
CPU 0 core 0 is always in the mid 60's to low 70's and CPU 1 core 1 is usually in the mid to high 50's.
The Graphics card also has it's own fan from the side cover and sits between 48 to 60 degrees. As it sits so close to my sound card on this Tyan motherboard I believe it is the reason why this Windows computer just shuts down every now and then, but then it could also be the Power Supply, but I have no way of measuring it.
My computer is not overclocked.

Also I could not get Bikemans temperature command to work on my Fedora systems either and I would love to know what temps they are running.

Thanks guys for an interesting conversation.

My attemps at overclocking an AMD 4800+ ended in failure.
I did try for over a year, and even paid a guy to check my computer, he reformatted it and made no difference.
It seems that my ASUS motherboard, despite having all the overclocking goodies built into it will not overclock but crashes all the time. Placed on stock speeds it has not missed a beat for over a year now.

Astro
Astro
Joined: 18 Jan 05
Posts: 257
Credit: 1,000,560
RAC: 0

Good morning, I see some

Good morning, I see some other Q6600 temps listed and all are much lower than mine. Has anyone and info on removal of those nylon locking pins that secure the heatsink to the mobo? If so I'd really like to lap this heatsink and see if it brings down the temps. Not sure how much "headroom" I have to OC, If I'm already starting at 56C. Anyone know max temp off hand? I also keep hoping one of the linux gurus would point me towards a way to monitor each core temp, or atleast one of them.

Hmmm, perhaps I could put a precision ground steel gauge block on the granite inspection block. A gauge block only a few inches wide, then cut the sandpaper to fit the block? Hmmm.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 446,420,142
RAC: 47,569

RE: Not sure how much

Message 79851 in response to message 79850

Quote:
Not sure how much "headroom" I have to OC, If I'm already starting at 56C. Anyone know max temp off hand?

One has to be a bit careful here, since max temp as per spec depends on th stepping:

The older B3 stepping is spec'ed at 62.2°C ( http://processorfinder.intel.com/Details.aspx?sSpec=SL9UM )

But you seem to have a newer G0 stepping (CPUID : 06FBh), with a "thermal specification" of 71°C . see http://processorfinder.intel.com/Details.aspx?sSpec=SLACR

CU
Bikeman
EDIT: As for a tool to measure temp: Did you try lm-sensors already? http://www.lm-sensors.org/

Astro
Astro
Joined: 18 Jan 05
Posts: 257
Credit: 1,000,560
RAC: 0

RE: But you seem to have a

Message 79852 in response to message 79851

Quote:

But you seem to have a newer G0 stepping (CPUID : 06FBh), with a "thermal specification" of 71°C . see http://processorfinder.intel.com/Details.aspx?sSpec=SLACR

CU
Bikeman
EDIT: As for a tool to measure temp: Did you try lm-sensors already? http://www.lm-sensors.org/


How do I tell? and no to the Im-sensors. I went through all the "main contrib, backdoor(both testing and main)", and just about every other thing on the list. I'll look er up

P.S. It appears I didn't quite have the cache of Boinc Simap wus, I thought I had. All my systems are already out and crunching here (Q6600 excepted)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.