Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

and here are the photos of my second try (which I unfortunately forgot to take a pre-mounting photo of):



Am I getting there?

The temps were 1-2 degrees higher with this one (43-44) but obviously not before the thermal cycles which should help cooling

Think I might go and buy some more NT-H1 (one website said they're good for 2-3 years so don't know about my old stuff!)

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

The amount of TIM in post #75 is good, even though the coverage in the corners isn't ideal.

There is too much TIM in post #76 and this increases cpu core temperatures.
I hate to see excess TIM buildup along the edges of the cpu

The picture below is ideal.


In Arctic Silver's picture below, the picture on the right implies that they feel that the coverage in the light red oval is correct.
My approach is that better TIM coverage in each of the corners should help lower cpu temperatures a bit more.



In post #72, Skinneelabs says that he used the "2 grains or rice method".
He probably describes this technique in some detail somewhere on their web site.
My best guess is that he uses two grains that are not too far apart and in the center of the cpu.

Check out Best Thermal Paste Application Methods | BenchmarkReviews again and take another look at the two parallel lines method:



"One of the fears I hear about most with this method is that air may get trapped in the middle. However, one look at the image below and I can safely dismiss this potential problem. After the cooler was lifted, I discovered that the two line pattern was able to reach all four edges of the CPU. It seems that this method could have some real world application if correctly applied.

It took some additional cold tests for me to get the amounts correctly tuned. After the retest results (not pictured), the best ratio I found was using line lengths equal to the distance between the lines; or roughly 1/3 the width of the processor."





and the "X" TIM pattern coverage:

"Next up was the cross method of applying thermal paste to the coolers mounting surface. Since two lines worked well, perhaps overlapping them would have an improved effect? I have not personally used this pattern in my own systems, but other web sites have claimed that it can work."



"After a few test applications, the image above is the best coverage I could get. Plenty of material covers the surface, but there are clear signs that too much TIM has reached the edges. Again, the disclaimer here is that this pattern might work for your setup, and every application is different. In my tests however, I wanted more uniform coverage and the cross method took some practice before it could meet my needs.

In the image above my line length was too long, measuring 2/3 the width of the processor. With additional cold tests (not pictured) I learned that lines of the same thickness but only 1/3 in length did a better job and kept the material amounts in check."



In the above picture I'd like to see less TIM buildup at each edge of the cpu.

Your studies obviously come first, but you might want to take a brief break and try using your older tubes of TIM, who knows, they might work fine.


Below are my RealTemp Sensor test results with the ambient temperature starting at about 16^o(C) with:

• cpu running at 2.00GHz ( 6 x 333MHz)
• cpu voltage = Normal
• memory timings all set to AUTO

If you boot with similar settings and your cpu core temperatures remain the same this probably means that your sensors are stuck in the 40+^o range at the low end.

You could then run Prime95 Small FFT test for about 15 minutes and then see what your maximum cpu core temperatures are.




I wonder if there is something with the 771-mod that is possibly causing higher than "normal" cpu core temperatures.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Hey profJim!

I haven't been home yet to do any more work yet but I posted on a 775 to 771 support thread on overclockers. One guy said my vcore was too high an another said:

"Your Tj max setting in Core Temp is wrong and because of this it displays 15 degrees higher. Set this parameter to 85C".

I thought I would run these by you before going ahead and making any adjustments... I guess lowering vcore and Tj max can't hurt, right?

Lukevega

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

What's that old saying? A little knowledge is a dangerous thing.

As I understand it, cpu core temperatures are computed by how far away you are from the cpu core's TJmax value.
Rather than try to explain it (badly I might add), read the Real Temp - Documentation and you will see that altering the the TJmax value in a software program only changes the displayed value of the computed temperature, and doesn't change the actual cpu temperatures. If you want some additional reading check out the 2008 Intel IDF presentation. Page 16 starts the Xeon section in the document and your X5460 is listed on page 19 with a Target Tj = 85^oC.

Don't jump to conclusions as it is possible for each of our cpu cores to have a different Tj value.

It's time to see if your system is overheating and throttling your cpu with 3.60GHz and higher cpu speeds.
To protect your cpu, Thermal Monitoring should always be enabled in your bios.

Maybe using the 771-mod to run your cpu in an LGA775 motherboard is causing the reported TJmax value to be to high.


Happy reading.

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Here are some scream-shots of two 15 minute tests that I did.

Q9650 running at 2.00GHz:

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Q9650 running at 4.00GHz:

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Right. next time I'm free I'll try and get the TIM paste applied properly and then download Real Temp as well to see if that makes a difference.

I have no idea why my temps are so much higher than they could be :( I even used a small brush and got in-between all the individual fins of the cooler!

I will start adding temperature screenshots to my tests for your critical appraisal too.

[lukevega]

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

No you are not being stupid.

Not make sense?
Give me a break, couldn't you just say "Would it now make sense to change my TJMax setting from 100^o to 85^o?"
I hate double negatives and dealing with RealTemp forces me to think in double negatives because as core temperatures rise, the distance away from TJMax gets smaller!!

The first "test" you should run is the RealTemp Calibration test with:

1. TJMax left at it's current 100^o(C) "default" setting
2. Cpu Host Frequency = 333MHz, the stock FSB
3. Cpu Multiplier = 6.0, the lowest it can go
4. Cpu Core voltage = 1.10 volts
5. With your system at the desktop, your core voltages should be about 7^o(C) higher than your room temperature, but they won't be.
6. Save a screen-shot of the settings about 10 - 15 minutes after your system has been doing nothing but sitting there.
7. After this idle test, you can run the RealTemp Sensor Movement Test and save the screen results.

Now it's time to change your 100^o TJMax setting to 85^o(C):

1. Click the RealTemp Settings button at the bottom of the window.
2. Now that you are at the Advanced Calibration Settings window, click on the Set TJMax button.
3. There should now be four boxes with "100" in each box; change each of these to your new 85 TJMax value.
4. Click the Apply button at the bottom of the window and then click the OK button.

Now RealTemp will use the new 85^o TJMax setting.

It probably isn't necessary, but I would reboot your system and it will be beneficial to repeat the 6.0 x 333MHz (2.0GHz) standard RealTemp sensor tests before you test at 3.0GHz and higher.

The attached screen-shot combines two test results from my system:

1. Running a 15 minute Prime95 In-Place Large FFT test at 4.0GHz where RealTemp was the only monitoring program that was running.
2. The second test repeats the first test, except this time RealTemp and HWiNFO were running at the same time.

Tomorrow I'm going to repeat both of these tests after I lower my TJMax setting from 100^o(C) to 85^o(C) where I expect that RealTemp will now report lower core temperatures but HWiNFO will report normal or near normal core temperatures.



Interesting, the RealTemp maximum temperature readings were more consistent than HWiNFO, even though the minimum and idle temperatures still show a little weirdness. CoreTemp and HWiNFO showed similar but minimally different core temperature results in my previous tests.

Core Temp also lets us change the TJMax value(s) but I didn't find a similar setting in HWiNFO.

Running multiple temperature/voltage monitoring programs might lead to incorrect results.

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

The screen-shot shows the reported temperatures with TJMax = 85^o(C) in RealTemp, but HWiNFO and Core Temp are using the 100^o(C) TJMax setting.

Did RealTemp really lower any cpu core temperatures? No, it's only displaying corrected temperatures.

Can RealTemp (or Core Temp) improve cpu thermal protection by tweaking the displayed results? No, it just "looks" that way.
The RealTemp documentation discusses this.

A web search using: intel cpu thermal protection provides links for more information.

BTW with an air or water cooled cpu heatsink, it is impossible for the any cpu core temperatures to really be lower than the ambient temperature.




Later tonight I will post a screen-shot for the Q9650 15 minute, Prime95 in-place Large FFT temperature results with RealTemp using TJMax = 85^o(C). I predict that the results will look really cool.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Sorry for the double-negative!

I've barely lifted my head from the books to get home and make any progress - I really appreciate all the tests you're doing to help me out!

I'll do a RealTemp calibration test as soon as I get the chance (although I'll have to re-apply TIM and mount my cooler again first) - Do you think it will make a difference if I apply the TIM to the cooler first rather than the CPU? I can hold a ruler next to the cooler easier to get an accurate grain of rice shape.

I will download RealTemp but I think you can also change Tj max in CoreTemp in the CoreTemp.ini file by changing the TjMaxOffset - I could try that too?
Looking forward to making progress once these exams are over!

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

In my P35-DS3L system with an E6300 (R0) cpu, I closed Core Temp, changed TjMaxOffset=0,0; to TjMaxOffset=-10,-10; in the INI file and re-opened Core Temp. Both of my core temperatures displayed 10^o lower, but TjMax still displayed 100. When I clicked on Core Temp's Options and Adjust offsets tabs, the Temperature offset screen displayed -10 in two boxes.

I'll stay with RealTemp.

When I entered -50 for the Core Temp offset, Core Temp showed that my core temperatures are -10 and -12^o(C). Pretty cool!!

I disagree with the TIM experts who say that it's not necessary for TIM to cover the full surface of the cpu's IHS surface. Most of the heat is directly above the physical cores under the IHS, no argument from me. The attached cooling tubes cover the full width of the bottom of the cpu cooler and common sense tells me that more contact area is better than less, and I'm aware that common sense isn't very common and sometimes it doesn't make sense.

I recently reapplied TIM with my AIO cpu water cooling block and used a different technique. I used one finger in a plastic bag to spread the thinnest possible TIM coating on the top of the cpu IHS. I did not spread TIM to each of the top edges of the cpu and left about 3mm uncovered at all four top edges of the cpu's IHS. I noticed an immediate improvement with lower and more even core temperatures. Some TIMs will spread easier than others.

I don't want to touch the cleaned surfaces with my fingers, so a ruler would not work for me. A better solution might be to use 100mm (4 inche) paper that is as wide as your cpu where you can put marks on the top of the paper for the center of the cpu and another mark for how long you want your thin grain of rice line to be. The bottom line is; use what works best for you.

Here's an interesting thermal grease video:

More test results:

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Ok, I don't fully understand what these mean but I'm sure you will...