Cooling a Q6600

[BonzaiDuck]

When you run two instances of ORTHOS, does one of them seem to be "Paging" all the time? What's going on there?

 

[adairusmc]

Originally posted by: BonzaiDuck
Great! I very much appreciate your help. I don't know if you were following my thread about motherboard ducting, but I almost came unraveled in my enthusiasm at the discovery that I could get the same speeds at lower VCORE on the E6600 -- and with lower temperatures -- for all the arts-and-crafts foam-board tedium.

Now I've got the Q6600 installed, up and running, and SolMeister has me kicking myself, because customer reviews at NewEgg show that they're selling the G0 stepping all of a sudden. I got the B3 -- maybe that's what you have also.

I can't see how Everest Ultimate would report inaccurate temperatures, but this is just a baseline check to see how I'm doing.

Here's the frequency distribution of my 2xORTHOS temperatures for 1 hour, 16 minutes:

Q6600 @ 2.4 Ghz, 1.28V VCORE, baseline temperatures under 2xORTHOS load and 73F room ambient

This is the entire distribution of my temperatures sampled every 8-seconds over that time-period. Your range upper-bound gives me some encouragement for all the little pieces of foam I have to sweep off the carpet now . . . .

I guess I can continue to put water-cooling on the back burner for a while.

Did you notice a wider variation in the load temperatures while you ran ORTHOS at the higher clock settings? I'm going to bump it up to 1,333 Mhz before dinner -- probably without even dropping the multiplier to see how that goes.

Yes, mine is a B3. With my overclock I am running mine at a vcore of 1.26v, and is rock solid so far (and according to 8 hours of dual-orthos).

When I ran orthos at 3Ghz, It would jump up and down maybe 3C max on each core, but for the most part, it stayed pretty close to those values I gave earlier.

 

[SolMiester]

Hi Bonzai, how are you going with your Quad?, didnt mean to get you 2nd guessing your purchase over the stepping, I doubt very much if we have any G0 over here in NZ at the mo, maybe when I have the cash in Oct, they will, but guess I'll be watch for Penryn then...LOL, gotta laugh at your metaphor re the waves, but its very true, though surfing cant be as expensive.
Oh, check out the new improved TJ-06 design....here


Edit - Oh, the last build was Dec 04 for the AMD, the TJ-06 came last year when case temps weren't that good, then a 20"LCD prompted a better card than the 66GT, unfortunatly the TT460w PSU wasnt up to a X1950pro, but after the Zeus 650w upgrade we discovered the card was a dud!....

My poor wife...haha

 

[BonzaiDuck]

LOL!! It looks like we all need to go into the same "12-step" program: for hardware junkies.

I wasn't familiar with the nuances for setting processor affinity with double-ORTHOS, and it's quirky anyway. In my impatience, I bumped up the VCORE voltage to the retail-box maximum of 1.35V. I've got the 6hour, 47min time series to analyze, but peak temperature on the "hottest" core was 59C @ 73F room ambient. The temperature variation, like AdairUSMC said, is narrow, so my intuition tells me that a 7-hour "certification" is probably as reliable as running it twice that long. I'll post a graph soon.

As I understand it now, the G0 is rated to run at 1,333 Mhz FSB. Whether or not that actually means "cooler" -- I can't tell, but it should be worth more stability and lower temps at higher settings.

I think I'll try knocking the voltage down to what AdairUSMC was using. Now the trick will be to get it even within +5C of what I saw as AigoMorla's watercooled load temperature at a 3.6 Ghz clock speed.

 

[BonzaiDuck]

That TJ-06, by my recent "experience," can save you some Xacto-knife blades, a lot of time, and the carpet sweeping!!

I think it basically a single 4" or 5" Acrylic "tube." If it's fairly-well organized into "chambers," I'm not sure how any of my lo-tech mods would improve on the Silverstone's "cooling potential" by that much. But I DO know where you can get more Acrylic tubing -- if you're willing to shell out $51 USD for a 5-foot-long piece of it, and pay freight to New Zealand!!

 

[SolMiester]

Dont think I'll need more tubing Bonzai, The hard part(s) will be separating and moving the floppy cage from where I want to put the middle chamber 120mm intake to the upper chamber securely and looking good. Then I guess the middle chamber exhaust in the side plastic window, maybe i'll just mount it by the pci card holders in between the video and audio cards?!

 

[BonzaiDuck]

Actually, SolMeister, I only vaguely understand what you're trying to do, and I'll need to take another look at your side-view photo. If you're attempting to mount either an intake or exhaust fan in a position closer to the center of the case, it will reduce noise unless of course it's a side-panel exhaust fan. But those PCI slot-covers come in handy for that sort of thing. Also, if you check fan-resellers, there is a brass-colored fan-bracket that comes with some fans -- I got it with some cheap $6 Sunon 120x25mm units -- used to mount the fan on the top edge of the PCI slot-covers at the rear-chassis where expansion cards are secured. But you can use it anywhere. I've got it holding a fan hanging off the rear of my big RAID array cage:

case-center drive-cage intake fan with Logisys CCFL light

In fact, you just gave me a great idea for one more addition to my ducting mod.

What do you know about the purpose or impact of the "CPU VTT" voltage setting?

 

[SolMiester]

Hey Bonzai, CPU VTT, knowledge, not that much bud, I have never touch the cpu voltage in order to get a higher clock, so dont have any experience with it.

Basicly, i just want to create 3 separate chambers in the case, the bottom chamber with wind tunnel is already done, just needs windows draft tape on the bottom of the tunnel to stop escaping air slipping up to middle chamber.
I havent had a hot gpu in the case yet, so for the GTS, i want to increase the 80mm intake to 120, but need to relocate the floppy cage 1st. For exhausting the middle chamber air, i was contemplating adding an exhaust 120mm to the side panel, however you think this might cause vibration?, so maybe i will have to settle for a PCI blower next to the cpu.
The top chamber is the drives, however i have yet decided what to use to seal it off from the middle, maybe some rectanglar clear arcylic?, or such. I dont think i will need an intake fan for the top, I already have a 92mm exhaust behind the HDD cage and the PSU pulls air from its front and the optical drive bay.

Whats with the CPU VTT anyway? Are you clocking the shit out of your B3 already?

 

[BonzaiDuck]

CPU VTT is some . . . . reference voltage termination value. I'm having trouble finding an explanation about it, and I've scoured Rojak's BIOS guide and another online BIOS guide.

If the panels you want to seal are not quite so visible from the case-window, use black foam art-board. For instance, a horizontal frame that supports an array of hard disks is only going to present an edge of itself to view from the sidepanel window. If it's something more visible, use 1/8" Lexan plates cut to fit. I've discovered that the high-priced foam-board sold here in the states at a place called Michael's Arts and Crafts store, is high-priced ($6 for 2'x3' versus $2) for a reason. The black-paper covering is a stiffer paper, and the board seems more rigid.

Someone asked me "why all the non-conductive material for your ducting?" by which he meant "foam-art-board." It's non-conductive!! I don't even think it burns very well, or it's made with some fire-retardant additive. It doesn't hold a static charge. It has noise-deadening properties.

Lexan also has many of these properties, but doesn't go a long way toward deadening noise. On the plus side, it looks pretty and you can see through it.

"Are you clocking the s*** out of your Q6600"

AdairUSMC has been very helpful. He has a TT case and model that I'd think has great potential. He was "walking point" on the over-clocking for his B3 stepping Q6600 earlier in this thread -- before I'd made the changeover from the E6600.

With the E6600 and foam-board ducts in my case, I'd managed to hold my "bench-configuration" clock-speed constant at 3.35 Ghz and drop the VCORE from the unducted setup's 1.46875V down to 1.4563V. This resulted in recapturing all the temperature advantage I had when running at 3.285 Ghz -- before I bumped up the VCORE to almost 1.47. The "temperature advantage" was a 4 to 5C improvement over what Anandtech's Ultra 120 Extreme review reported for their clock settings on what I believe was an X6800 processor. To do their tests, they used a higher multiplier because apparently the X6800 is "unlocked." To make a full comparison, I'd need to know the thermal power of that particular processor at those various OC settings.

But when I first installed my TR ULtra 120 Extreme, it brought my temperatures down at 75F from about 53C with the Ultra 120 "original" cooler to about 47C with the Extreme. I think using a diamond-based thermal paste got me another 2C-degree drop. I think that was when the processor was clocked at 3.285Ghz.

So, with the ducting, IC Diamond thermal paste, and the Ultra 120 Extreme, I was able to get load temperatures in the 43 to 45C range. At 70F room ambient, it was definitely between 42 and 43C. And this was with the higher OC regime of 3.35 Ghz.

So now I've made the swap-over to Q6600. I can see the ducting gives me a temperature advantage, but I've only now bumped up my clock to AdairUSMC's proven 3.01 Ghz, 1,333 Mhz FSB, and he quotes a VCORE of about 1.28V.

I certified my system for 7 hours 2xORTHOS at 1,334 and 3 Ghz, but I'd run the VCORE up to about 1.35V -- the retail box printed "Maximum." With that setting, my highest core load temperature was 59C. AdairUSMC said he'd bumped up his CPU VTT voltage to 1.35V to get it stable.

When I dropped my "set" VCORE to 1.30V, the monitor reported voltage shows about 1.28V, and I was only able to keep it running 2xORTHOS for about an hour to an hour, 20 minutes. Bumping up the 1.2HT voltage about 0.05V and raising the NB voltage by 0.05V increased the running time before ORTHOS posted an error.

I've gone ahead and bumped up the CPU VTT setting to 1.35V -- I think that's what AdairUSMC quoted, but I don't know if he meant "set" or "monitored." What I did notice was that raising the VTT voltage in 0.05V increments from lowest 1.2 to 1.25 to 1.3 didn't change the value reported in BIOS monitor, whereas going to 1.35V showed something like a 0.04V increase in the monitored value, so I'm running 2xORTHOS at the 1.35V setting now.

So far, at 73F room ambient, the peak load temperature for the core running the hottest is about 55C. I think this is good. While I'd like to drop the multiplier to 8 and run up the FSB to 1488 as it was with the E6600 processor, I want at least to see how things work out with this setting at 1,333. I've got high-end Crucial Ballistix RAM that was running at 744 Mhz and 3,4,4,8,2T with the E6600 and FSB at 1488. At 1,334 using the E6600, I had them set at 3,3,3,6,2T (and DDR2 = 667). They seem to be running fine that way at the moment, also. Of course, knowing the Ballistix are rated at DDR2-1000, a person wants to push them as far as they'll go without putting an enormous stress on the motherboard or processor. I don't want to implement any other divider than 1:1, though.

 

[BonzaiDuck]

From nVidia Forums:

["CPU VTT Voltage - the FSB Termination Voltage at the CPU - the bus voltage. Intel lists a MAX of 1.55V for the E6000-series, never needed above 1.4V, works well at 1600 MHz 1.3V"]

NOW it makes sense . . .

 

[tracerbullet]

Originally posted by: BonzaiDuck
But even so -- if it's copper and has heatpipes, it would promise to be more effective than the stock cooler.

I went with this theory above all others. I have the AC Freezer 7 myself, and even at full speed the fan is dead quiet. It's a decent sized cooler but not so honking big as to be worrisome of it fitting in the case, it clears everything around it and the case cover closes, and it's been doing a great job thus far.

I only got my new setup together last night though, so I don't have any real #'s with it and have only a few hours even using it, so take that all with a grain of salt (Been seeing 35 degrees reading from the BIOS on a GB P35-DS3R while setitng things up, have no stress tests though). I'm sure I can't set an O/C speed record with it, but I'm more than happy with it especially for the price and how quiet it is.

If I get some real #'s with it I'll chime in on the other threads.

 

[BonzaiDuck]

Tracerbullet --

Your handle suggests that you and AdairUSMC might cross paths in the same circles.

Anyway, "likewise" on the posting of benchtest results. I think everyone here, whether at this thread or the one about choosing a fan for a Q6600, will be surprised with my numbers. I'm going to put it in a separate thread, and try to put a link to the thread -- "here and there."

Do you know the measured thermal resistance for the Freezer? I acquired one from my dentist in a parts-swap, but I don't think I'm inclined to use it when I've got the TR coolers.

 

[poco153]

Hello All -

I have a Q6600 (the B3 stepping) coupled with some AS5 and a Thermalright Ultra-120 Extreme (the version with 6 heatpipes). It's been running fine and all, but I'm still hoping to overclock some. The Ultra-120 has a Scythe S-Flex 120mm fan on it because it seemed like a good balance between quiet and cool. I've been running the box at load for a good 8 hours now, and I'm noticing something odd: the nTune utility included with my motherboard (the eVga 680i SLI board) is reading the load temperature as 57C (which still seems a little high). Intel's Thermal Analysis Tool and CoreTemp both read the cores as 68/61/68/61 (0/1/2/3). I feel like this is way too hot, especially since Anandtech has gotten some serious overclockes on the higher-end C2Q chips with the same cooler. What is the thermal limit of these chips? Does it sound like I'm doing something wrong with my thermal paste application?

Thanks,

 

[aigomorla]

duckie, you know of an easy to way to apply IC DIAMOND.

That crap is the hardest stuff i ever had to apply. It gunks worse then cluber girls looking for free drinks.

:\

 

[adairusmc]

Originally posted by: poco153
Hello All -

I have a Q6600 (the B3 stepping) coupled with some AS5 and a Thermalright Ultra-120 Extreme (the version with 6 heatpipes). It's been running fine and all, but I'm still hoping to overclock some. The Ultra-120 has a Scythe S-Flex 120mm fan on it because it seemed like a good balance between quiet and cool. I've been running the box at load for a good 8 hours now, and I'm noticing something odd: the nTune utility included with my motherboard (the eVga 680i SLI board) is reading the load temperature as 57C (which still seems a little high). Intel's Thermal Analysis Tool and CoreTemp both read the cores as 68/61/68/61 (0/1/2/3). I feel like this is way too hot, especially since Anandtech has gotten some serious overclockes on the higher-end C2Q chips with the same cooler. What is the thermal limit of these chips? Does it sound like I'm doing something wrong with my thermal paste application?

Thanks,

Those temps do not sound too terrible, to me. Is your vcore set to auto in your BIOS?

I fixed mine down to 1.26, and with my overclock of 3Ghz I am seeing about the same temps, with the same cooler and thermal paste in use. I noticed when I first installed the chip that the idle and load temps with the voltage at auto were higher than if I fixed the voltage. Maybe try setting your stock setting at a similar setting and see if that affects your temps at all.

I am using a 680i as well.

 

[poco153]

adairusmc -

I dropped my voltage to 1.26250 in the BIOS, and my temps at load with stock speeds (2.4GHz) are still pretty high - 63/63/60/60 according to CoreTemp. I feel like this is quite toasty for stock speeds. Any thoughts as to what I can do to remedy it?

 

[adairusmc]

Originally posted by: poco153
adairusmc -

I dropped my voltage to 1.26250 in the BIOS, and my temps at load with stock speeds (2.4GHz) are still pretty high - 63/63/60/60 according to CoreTemp. I feel like this is quite toasty for stock speeds. Any thoughts as to what I can do to remedy it?

You can try re-fitting your thermalright to see if you can get a tighter fit I suppose. I saw a think here somewhere where you can take a dime, or a quarter, and put it between the screw of the retention bracket and the block (on the top of the block, in the gap between the heatsink. You might have a loose connection on your thermalright (they are notorious for fitting a little loose sometimes).

Here is an illustration (taken from another thread here on the forum) - pic

As far as stock load temps, those do not look too terrible to me. With my 3Ghz overclock, I am hitting 69,69,67,66 with the same cooler (temps taken after 11 hours of dual Orthos). Maybe re-applying the thermal paste and making the cooler fit a little tighter will help.

 

[Noubourne]

Originally posted by: poco153
adairusmc -

I dropped my voltage to 1.26250 in the BIOS, and my temps at load with stock speeds (2.4GHz) are still pretty high - 63/63/60/60 according to CoreTemp. I feel like this is quite toasty for stock speeds. Any thoughts as to what I can do to remedy it?

Unless it is 85 in your house, you have a bad mount.

At 76F ambient, my idle stock speed/volt temps were low 30s, and loaded were mid 40s.

I am at 3.5Ghz with 1.368v, and I get 68-72 load temps now, with mid-40s idle.

And I don't even have the Xtreme version of the Ultra 120.

The Ultra 120 mounts very tight and secure to my board. Unlike the Infinity mount on my old Opty - which I could actually swivel it was so loose - this TR mounting for 775 using the metal backplate and sturdy metal screws is a great mounting system.

 

[BonzaiDuck]

We have to be clear in our understanding of Intel's "TCase" and "TJunction" specs.

There's an excellent article at THG in conjunction with TAT that explains it. I can't show the link here, because it's in my "IE Favorites" on the Kentsfield system currently going through some Orthos paces for a multiplier=8 setting I want to test. I'll try to stick it in here later.

Generally, the "overall" CPU temperature value can go as high as 60C and this is considered by Intel to be "Safe." 65C is "warm" and 70C is "hot." By comparison, my "overall" value for the set of core temperatures shown below is somewhere in the 40's of Celsius.

I used foam-art-board (tentative -- to be replaced by Lexan) to duct the CPU cooler, VGA cooler and motherboard. (Especially -- the motherboard).

I was able to OC the B3 stepping at a VCORE "set" value of 1.375V (2% above the retail box maximum spec of 1.35V) to 3.132 Ghz. Above that, increases in constant Mhz increments required two, three, four or more additional notches of Vcore, so I see 3.1+ as a limit of the processor. I'm guessing that people who eventually experienced hardware failure under extreme over-clocks -- under either water or air-cooling, did so by pushing the VCORE in steep increases too much beyond the maximum spec.

I certified the 3.132 Ghz clock with 2xORTHOS Small FFTs for 13 hrs, 27min, normal manual termination with 0 err, 0 warn. The first four hours were taken under a rigorously controlled room temperature of 76.1F +- 0.5F degrees. My core temperatures didn't vary much other than for different ORTHOS calculation iterations and were 61C, 58C, 54C, 57C.

Room temperature is very important, and I would tell anyone to try and "certify" at room temperatures approaching 80F. You can pretty much normalize CoreTemp log results to 80F by adding the difference (converting to Celsius if needed) in your actual room temperature value and 80F. But you can't certify stability at that temperature unless you can keep your actual room ambients that high for at least a few hours.

 

[BonzaiDuck]

AFTERTHOUGHT: Noubourne's clock speeds -- not sure if he's got the B3 or G0 stepping -- and his VCORE setting may well be indicative of improvements with the P5K's chipset.

I'd be just a little worried (for myself if I were pushing it that high) for what he seems to be reporting as the "overall" or TCase temperature.

Auf meinem System, der Wind mussen Kuhl sein.

. . . . . Like the York's Peppermint Patty . . . .

 

[BonzaiDuck]

ANOTHER AFTERTHOUGHT: Noubourne DOES have the G0, and I'm now guessing he's reporting a range of the individual core values. [Still a bit toasty for my taste, though.]

 

[BonzaiDuck]

Here's the link I promised to post:

Tom's Hardware C2D Temperature Guide

This seems to be written exclusively for the C2D processors, but I'm only guessing that the same parameters apply -- except that there are four TJunction sensors rather than only two, in addition to the TCase sensor.

 

[poco153]

Originally posted by: adairusmc
You can try re-fitting your thermalright to see if you can get a tighter fit I suppose. I saw a think here somewhere where you can take a dime, or a quarter, and put it between the screw of the retention bracket and the block (on the top of the block, in the gap between the heatsink. You might have a loose connection on your thermalright (they are notorious for fitting a little loose sometimes).

Here is an illustration (taken from another thread here on the forum) - pic

As far as stock load temps, those do not look too terrible to me. With my 3Ghz overclock, I am hitting 69,69,67,66 with the same cooler (temps taken after 11 hours of dual Orthos). Maybe re-applying the thermal paste and making the cooler fit a little tighter will help.

I'm up for trying this out, and I greatly appreciate the help! One thing I must ask, though, is how likely am I to destroy something (I'm most worried about cracking the chip or the motherboard) by putting a coin under the bracket? If you could link me to the forum topic, I'd appreciate it (although I suppose I can google it too). The mount is pretty loose, I can rotate it very easily.

Another new-to-overclocking question - do I need to clean off and re-apply AS5 every time I pull the heatsink off?

Thanks,

 

[adairusmc]

Originally posted by: poco153

Originally posted by: adairusmc
You can try re-fitting your thermalright to see if you can get a tighter fit I suppose. I saw a think here somewhere where you can take a dime, or a quarter, and put it between the screw of the retention bracket and the block (on the top of the block, in the gap between the heatsink. You might have a loose connection on your thermalright (they are notorious for fitting a little loose sometimes).

Here is an illustration (taken from another thread here on the forum) - pic

As far as stock load temps, those do not look too terrible to me. With my 3Ghz overclock, I am hitting 69,69,67,66 with the same cooler (temps taken after 11 hours of dual Orthos). Maybe re-applying the thermal paste and making the cooler fit a little tighter will help.

I'm up for trying this out, and I greatly appreciate the help! One thing I must ask, though, is how likely am I to destroy something (I'm most worried about cracking the chip or the motherboard) by putting a coin under the bracket? If you could link me to the forum topic, I'd appreciate it (although I suppose I can google it too). The mount is pretty loose, I can rotate it very easily.

Another new-to-overclocking question - do I need to clean off and re-apply AS5 every time I pull the heatsink off?

Thanks,

the likelihood that you are going to break something is nil, as long as you are careful. The biggest risk (and this applies to mounting it without the coin) is that when you are pushing down one of the last screws to get it to screw into the backplate, and your screwdriver slips out and hits the motherboard (I did that without the coin. Though it didn't damage anything, my blood pressure went up for quite a while).

I have a dime under mine to help with the loose fit, and it works much better than before. There is still a little play in the heatsink, but it is not nearly as bad as it was before. As long as you get the coin under that screw like it shows in the pic, and you do the mounting with the motherboard out of your case.

Here is a link where I found how to do the dime trick, posted by the member "wfp"

And yes, you should re-appy the arctic silver 5 every time. You want a thin and even layer (as thin as possible). I use a razer blade to spread out the paste on both the CPU and the heatsink, as thin and even as I can get it.

 

[BonzaiDuck]

I'd heard much over the last couple months when the Xtreme first hit the street -- about this "loose-fitting" problem, but I didn't seem to have it.

I'm soo tempted to use the idea for another reason, though. I lapped a 2007 silver dollar smooth hoping that putting it between the heatspreader and the heatsink base would enhance cooling, but on an efficient cooler, no matter how it works, the fins and pipes will rule the day. There may have been air pockets -- any number of things -- but THAT coin trick didn't do anything but increase temperatures -- possibly just for adding an additional thermal interface.

But sticking it on top of the heatsink base -- that's another thought. There's room between the scissors to epoxy some heatsinks -- maybe four of them.

I keep . . . . dreaming up. . . . . these air-cooling Rube-Goldberg ideas . . . and it's hard to convince oneself that adding more heatsinks to the cooler changes the simplicity of the cooler. . . .

I'm at 3.132 Ghz with a peak load temperature of 61C @ 76F-room-ambient. I think I'm going to shoot for 3.2. I'd still rather drop the multiplier and get a higher memory speed, but -- what the heck . . .