Q6600 w/ Tuniq Tower - Added pics

[RussianSensation]

Sold Old System:

E6400 @ 3.4ghz w/ Big Typhoon AS3
Gigabyte P965 DS3 v1.0
2x1GB Corsiar XMS2 PC5400
320GB Seagate 16mb SATAII
PNY 6600 256mb
Viewsonic 19 inch CRT
OCZ PowerStream 420W
Aspire case

New System:

Q6600 VID 1.2500 2.4ghz w/ Tuniq Tower
Gigabyte GA-P35-DS3L v1.0
2x1GB OCZ PC6400 Platinum Rev. 2
400GB Samsung HD403LJ 16mb SATAII
EVGA 8800GTS 320mb
37'' Westinghouse 1080P
Corsair HX520W
Thermaltake Soprano case

I put the new system together with the Tuniq Tower blowing air towards the rear case exhaust fan. I applied Arctic Silver 3 as shown Here. Loaded up the system and got these temperatures at idle:

Q6600 @ 2400mhz
Core 0: 39*C
Core 1: 35*C
Core 2: 37*C
Core 3: 33*C
Idle Core Temp 1st install

What happens at load?

Q6600 @ 2400mhz
Core 0: 60*C
Core 1: 55*C
Core 2: 59*C
Core 3: 54*C
System temperature (northbridge): 47*C
Load Core Temp 1st install stock

Ouch! As a reference point my old system was getting 50-51*C at full load @ 3400mhz! on an inferior Big Typhoon at similar 1400rpm. As I read before, quad cores are arranged as follows:

0 1
2 3

So I knew something was up when cores 1 and 3 were much lower (see above). What happens when I begin to overclock just slightly?

Q6600 @ 2580mhz
Core 0: 63*C
Core 1: 58*C
Core 2: 57*C
Core 3: 63*C
Q6600 @ 2580mhz Load

Wow, Cores 0 and 3 are running already beyond B3's spec on one of the best air coolers...

Still, it's pretty impressive that with only increasing the FSB voltage by +0.1V, this $100 budget board was able to post 430FSB which means it is likely sufficient for Q6600 overclocking (i.e. 9x430 = 3870). I have not tested the top FSB of this board but if there are any requests, I will do so.

<<<<<<<Time to reinstall the Tuniq and fix the northbridge....

1. I already linked the recommended method for AS from the manufacturer. But logically, I thought to myself that drawing a thin line will spread the Arctic Silver but in a way that it won't cover all edges and parts of the IHS. On all my builds before I spread the Arctic silver very thinly with a business card because I want the full base to dissipate heat at all points of the IHS not just where the Quads are located underneath. In the end the heat is dissipated from the surface area of the IHS so all of it should be covered in Arctic silver imo.

2. It turns out the DS3L's northbridge compared to P965 DS3's (which wasn't great to begin with and I before replaced it with a Zalman NB47J) is even more inferior (it weighs less and has a thinner base!!!). Of course I sold my old system but still had the 965 DS3's old northbridge. Time to apply arctic silver here too and swap them. I also pointed an 80 mm Vantec Stealth 2100 rpm fan at the northbridge.

Rebuilding results:

Q6600 @ 2400mhz Load w/ business card spread Arctic Silver 3 and swapped 965 DS3 v1.0 northbridge w/ 80mm fan blowing on it

Core 0: 54*C (-6*C)
Core 1: 54*C (-1*C)
Core 2: 51*C (-8*C)
Core 3: 52*C (-2*C)
System temperatures (northbridge): 39*C (-8*C)
Q6600 2nd install Load

*Note: I already mentioned I most likely installed the Tuniq poorly that is why there are drops of 6 and 8*C. But I am confident additional 1-2*C drop resulted from spreading AS around.

So if you are having similar issues with your Quad build, make sure you have installed the cooler and AS correctly. Dont' forget to add a fan to the northbridge (as long as it's quiet)

-------------------------------------------------------------------------------------------------------

BTW, Prime 95 v2.54 (small FTTs) loads Quads LIKE NOTHING ELSE. In comparison 4 instances of SETI@Home at load on the 2nd rebuild only gave:

Core 0: 49*C
Core 1: 48*C
Core 2: 45*C
Core 3: 46*C
The old method of loading cores with BOINC is out the window with Quads even though it worked fine for me when I had a C2D...

-------------------------------------------------------------------------------------------------------
Pics and overclocking to follow.
-------------------------------------------------------------------------------------------------------

Updated Pics: August 20, 2007
Thermaltake Soprano Bare
Tuniq vs. Stock Cooler + Q6600 L724A639
Tuniq vs. Stock Cooler - side view
Tuniq vs. Stock Cooler - top view
Gigabyte DS3L - with inferior northbridge
Tuniq installed on DS3L
OCZ Rev 2
Samsung HD403LJ T166
EVGA 8800GTS 320mb
Half-installed case

 

[cmdrdredd]

So you're saying that spreading the TIM across the IHS was better than using a line and letting the pressure of the HSF "spread" it?

 

[RussianSensation]

Originally posted by: cmdrdredd
So you're saying that spreading the TIM across the IHS was better than using a line and letting the pressure of the HSF "spread" it?

At least for my test system, yes. When I took the Tuniq off, you can clearly see the AS being spread by the heatsink uniformly. However, it isn't going to spread to every corner and side leaving parts of the IHS not covered by the paste. The whole IHS is hot so I thought every single spot that touches the heatsink helps to dissipate heat away and therefore should be covered manually.

I know I can lower temps even more by probably 5-7*C by lapping and changing AS3 to MX-2 for instance. I want to see if spreading the paste produces better results for others.

 

[Markfw]

Duvie said he found a link where TWO lines were done for quads, but you method probably emulates that.

 

[RussianSensation]

Originally posted by: Markfw900
Duvie said he found a link where TWO lines were done for quads, but you method probably emulates that.

Right! In fact when I set it up I "drew" 2 vertical lines where the 2 processors are instead of 1 horizontal line crossing them in the middle. Then I spread the paste around. Gave me a lot better results. Mark, btw, I can't think of many programs that stress the system as much as Prime 95 small FFTs. Seti is full 5*C below at full load for each core. From that point of view, does it make a lot of sense to test our system's stability with Prime 95 for 24 hours? I can hardly imagine anything at all loading the cpus that much (i.e. it largely appears 100% load doesn't mean the processors are actually loaded the same as it depends on the program).

 

[Markfw]

F@H SMP client really pushes them hard, I don't know compared to 4x P95, but I refuse to waste that much time testing, when it can be put to good use !

 

[bryanW1995]

I'm currently running 2xseti on my system, 49c on both cores right now. I just turned on orthos small ffts for comparison, temps immediately jumped to 55c.

 

[RussianSensation]

Originally posted by: bryanW1995
I'm currently running 2xseti on my system, 49c on both cores right now. I just turned on orthos small ffts for comparison, temps immediately jumped to 55c.

Precisely. This of course poses a new problem in stress-testing methology - does Prime 95 then recreate a true real world testing scenario? As Mark pointed above, he states that F@H SMP stresses his system, but it's difficult for us to gauge the extent of the load on the internal transistors and cache of that program compared to Prime95. It would be very interesting to see someone run 4 tasks of gaming, video encoding, and say photoshop and track the temperatures to compare that as well to Prime 95.

You can imagine that a difference of 5-6*C in the real world vs. prime 95 means a lot more overclocking headroom or a cooler running Q6600 for probably 99% of real world tasks

 

[Hauk]

I'm tending to think the "old" way to spread AS5 was better too, especially for a lapped heatsink. The Ultra 120 Extreme I received was concave, as many had reported it was.

If I left it and didn't lap, perhaps the thin line spread would be more appropriate. But for lapped surfaces, edge to edge just seems more logical...

 

[njdevilsfan87]

I prefer just putting a small amount, maybe 2cm radius in the middle of the chip, and letting the heatsink spread it out. I usually get the entire chip covered that way.

When people use the credit card method... how exactly is this done? Do you put a blob in the middle, and then use the credit in a way as if you were tape casting the AS5?

 

[RussianSensation]

Originally posted by: njdevilsfan87

When people use the credit card method... how exactly is this done? Do you put a blob in the middle, and then use the credit in a way as if you were tape casting the AS5?

I put 2 thin vertical lines to cover the 2 C2D processors and then started to spread the paste to the left and to the right. Of course you can also put a blob in the middle and start spreading it. The key is to not put a lot. After I spread it as thinly as possible and the whole IHS is covered, I also run it the card a couple of times from one side to the other to remove as much from the layer as I can and to even out the spread. It's not really rocket science - just put as little as you can while covering the whole IHS.

 

[aigomorla]

Originally posted by: Markfw900
Duvie said he found a link where TWO lines were done for quads, but you method probably emulates that.

2 lines is what i use. Guessamate the cores possition.

And i believe i was the first one on this forum to say the cores are alligned

0 1
2 3

instad of how they are on a C2D

0
1


Also, my diagrams represents the chip inside the board mounted inside the case.


Also, lets assume if cores 0 and 2 are lower then 1 and 3, that means you tightened the left side too hard. So you need to losen up on the left side and tighten the right side a bit. Also YOU CAN NEVER GET = CORE TEMPS ACROSS ALL 4 CORES ON IDLE. One will always jump while the others stay. You need to load up the cores to see which area you need to tighten and losen.

 

[Rumple]

I just got my Q6600 G0 today and i am going to order a motherboard later this week. I was thinking of getting the DS3L instead of the DS3R. I do not plan on overclocking really at all right now. Maybe down the road when i can afford a good HSF. What is your impression of the DS3L? I see you dont like the northbridge HS but what about the rest of it? Im figuring if im not overclocking that the northbridge should be ok? Also i can just change that out with a Zalman heatsink later on down the road when i upgrade the CPU cooler from the stock one. Im just wondering if i should jump on the newegg deal for the Abit board with the big rebate the IP5-E i think it is??? Or if i can save myself the cash and just get the DS3L.

Thanks for any input im going crazy trying to choose a motherboard and just when i think i finally have it decided i read your thread!!! Also i have 2 gigs of Crucial Ballistix DDR2 800 if that matters.

 

[RussianSensation]

Rumple,

If you are not going to be overclocking, the DS3L board is going to be just fine. There is also the S3L model without the durable capacitors. However, I figured the durable caps are worth the extra $5 or so when I bought it.

A little bit about the board:
Note that DS3L doesn't have RAID, only has 4 SATA ports vs. 8 on the DS3R, has inferior northbridge cooling (but should be sufficient for stock), older model of on-board sound chip (ALC888 instead of latest 889A on the DS3R). It also has a 4-phase power delivery design instead of 6-phase and as far as I am aware it is only Ultra Durable and not Ultra Durable 2 design as the DS3R. Of course since it is a budget board it lacks Firewire ports (but has 4 USB port connectors for extra expansion and 4 in the I/O module at the back of the case). It has 1 cpu fan header, 1 power head, 2 case fan headers. By upping the FSB voltage by 0.1+ I was able to boot at 430FSB without any issues. This is more than sufficient for Q6600 so I know I won't be limited by the board for overclocking. But of course if you want top notch FSB performance beyond this, I'd suggest DS3R model or another board. The IDE connector hangs fairly low. If you have a DVD drive that's still IDE at the top of the case, it might be hard to reach.

A little bit about the BIOS:
Another pet-peeve of mine is that in PC Health Monitor for voltages of rails and ram are only monitored as "OK" or "Failed." Generally, 99% of boards actually specify the voltages. But I am pretty sure DS3R is no different in this scenario. However, I wouldn't even bother with the DS3R board since Gigabyte has a much better board with same electrical circuitry as their DQ6 top of the line board - DS3P for almost the same price. The bios does have a decent selection for timings for memory and even preset overclocking options (C.I.A. 2). The overclocking options are hidden and require you to press Ctrl+F1 to uncover. The BIOS is very easy to upgrade using Gigabyte BIOS v1.20 software in windows. Gigabyte releases new bios revisions fairly frequently as well and they have Vista drivers on hand along with XP for easy downloading on their site. As with most P35 chipset boards that I am aware of, there are no downward multipliers for CPU:RAM, implying the lowest ratio is 1:1.

But to reiterate your point, if you are not focused on overclocking, since all of these boards will run FSB1333, just purchase the one with the features you *need* and whichever is cheaper (Abit is a great brand too). Even at stock FSB1333 you can still clock that Q6600 to 3.0ghz later on should you choose to.

So far the system has ran completely stable. For stock performance though, $75 Abit board is tough to beat.