new CPU for my T61

long live T61 ... last of its kind, LOL

the only reason that I went with CPU upgrade from my T7500 is that I love to play Unreal Tournament 3, and this is the only software that makes my CPU hit flat 100% usage on both cores, over half the time while I play.

so there it is, nice T9500 on the way .. in hope of no more FPS drop over CPU


- little info for the geeks, CPUs marked by VID, all socket P

T7500 - current

*SLAF8: 1.075 - 1.250 ; (running at 1.000 - 1.125 with RMClock, SuperLFM at 0.850V)

T9300 - (undervolting with RMClock will result in 2.4GHz max speed as half multipliers are not supported - 12x with 200MHz FSB, instead of 12.5x)

*SLAQG: 1.062 - 1.150
*SLAYY: 1.050 - 1.137
*SLAZB: 1.000 - 1.250
*SLAPV: 1.000 - 1.250

T9500 - significantly more expensive than T9300, like .. ~twice

*SLAQH: 1.062V-1.150
*SLAYX: 1.050V-1.137


as performance goes, here's some PassMark scores

T7500 - 1,272
T9300 - 1,689 or 33% increase; [email protected] ~ 1,525 or 20% increase
T9500 - 1,852 or 45% increase

T7800 - 1,468 or 15% increase, but is still a Merom, not worth the money


So as everybody aims for the SLAYX T9500 (thus keeping the price high), I went with the SLAQH that I found a lot cheaper It was either that or a SLAYY T9300, which I plan to beat with lower voltage from SLAQH T9500 and RMCLock.

as price goes these days, SLAYY T9300 can be found for ~125 and SLAYX T9500 for close to 250. I got this SLAQH for right in between these two

.. expecting the thermal error to pop up at start up, so hacked bios flash will take place as well, enabling SATA2 speeds and more. Speaking of bios, if my laptop works after the flash I'll be happily sending donation to the bios creator, as I've said long time ago

P.S.

the so called defective video card in this T61 still runs at 625/950 MHz core/memory for like 3 years - huge thumbs up for TPFanControl for making it nicely not going past 72 deg C As I remember, 3DMark06 scores are currently around 2,400

 

new CPU in and running already and undervolted quite good, with 0.1V lower than the T7500

x13 (2.6GHz) -> 1.025V.

It ran fine at 1.0125V but gave error at 1.000V on the 4th minute so I left margin for error there. The SuperLFM is 0.925, hither than the one for the T7500

I flashed the BIOS with hacked bios by Middleton so no thermal error on start up. However, I see some differences in temperature readings from different programs ... Here:

When I open HWMonitor:
under ACPI -> TH0 is what TPFanControl reads, i.e. 15 to 17 deg C higher
under ACPI -> TH1 is about same as the temp under the CPU core 0 and 1, but still with about 5 deg C higher than what RMClock reads ... go figure

no biggie though, just my fan may start working harder when not needed as much. Overall the T9500 runs cooler than the T7500, obviously cant tell by how much but I would say like up to 10 deg C cooler.

performance-wise -> Windows loads somewhat quicker, not by much though. The new Super_PI calculation to the 2M finished for like 43 sec ... where as before it used to be like 58 sec (as I remember), so that shows big improvement. As for the Unreal Tournament 3 - going to test it right now

EDIT: I forgot to enable the x14 IDA mode while undervolting, so I just checked that again and realized it needs higher than one_step_over_last_multiplier voltage to operate. At 0.050 the CPU bricked my laptop, working fine at 1.075. Test went with Super_PI as it is single threated application, which BTW finished with 2 sec faster, due to working at 2.8GHz single core and not at 2.6GHz, i.e. 41 sec

 

So, why didn't get the hottest processot you could?

Renee

 

what do you mean by the hottest ? ... I got the T9500 which is best of what I can put. Not talking about the X series as they heat up a lot more, and my CPU and GPU sit on one heatsink.

also, UT3 plays much better, cant see the 10 FPS drop that I had before so all is good. I'm still hitting 100% CPU usage but it doesn't show as flat region, just hits from time to time.

overall - very happy

P.S.
on the side note, the new CPU is smaller than the old one. I got picture of both, may post it later.

 

omg GPU temp dropped by 6 deg ...

used to hit 72 when my fan would switch to crazy speed ( at >70 deg C), now gets to ~65 and that's it, and this is while running on 625/950 overclock core/memory, just finished playing for like 3 hours

Penryn FTW

.. the coolest laptop I ever had, literally

 

GPU temp drop is probably caused by repasting it while changing CPU

 

not really ... I've put thermal paste twice before on both. It's just that the new CPU works much cooler, and the CPU and the GPU sit on one heat sink, i.e. you cool one and it affects the other

so as I was saying before - Penryn FTW

 

"what do you mean by the hottest ? ..."

Hot = Ghz per second....

Renee

 

OK, here's some usefull information about temperature readings

as most of you know, installing a Penryn CPU on a T61 laptop would result in the famous "thermal sensing error" at boot up, and therefore if you dont press ESC the laptop will reboot ... and going like that in a cycle. I have eliminated this issue with the hacked BIOS provided by middleton (huge thanks for which), and I dont see that error, which lets my laptop starts as normal. However, it's being said that there's an analog type thermal sensor in the Merom CPUs and a digital type sensor in the Penryn CPU, therefore misreadings on the CPU temperature is expected.

I can get temperature readings from couple different programs, but I'll focus on the 3 programs that give me CPU temps: HWMonitor, RMClock, and TPFanControl.

* RMClock - reads temps from both cores, currently showing ~30 deg C for both cores separately

* HWMonitor - read temps from ACPI THM0/1 and cores, currently showing ~35 deg C for both cores separetely

* TPFanControl - reads temperatures for various devices from the thinkpad-acpi module where up to 16 sensors are supported (for my T61 there are about 11 present). Currently the CPU reading shows the same temp as what you'll see under HWMonitor ACPI THM0 reading, or 42 deg C currently showing.

- The thinkpad-acpi module turns out to be quite usefull in our thinkpads

The CPU reading under the thinkpad-ACPI module comes from a sensor that is located very close to the CPU socket and is meant to give separate CPU temp information. Therefore, considering the misreadings that may happen with Penryn CPUs, I would consider that value as the most useful (same shows as THM0 in HWMonitor). The reading obviously would be only one for the CPU and will not show separate core temps, but frankly you dont need those that much. This turns out to be quite good as TPFanControl is the program that controls the speed of the fan, and will kick in fan speed according to your personal settings under the TPFanControl.ini file. Obviously it takes action based on the highest CPU temp reading among the rest of the programs, thus things look good.

also, dont forget that you can adjust the TPFanControl readings if you need to.

here is the explanation for all the sensors (found at thinkwiki.org) from the thinkpad-acpi module in a T61 laptop that you'll see in TPFanControl:

1 - sensor sits near CPU socket
2 - north bridge (?)
3 - Cardbus/ExpressCard Bay slot
4 - separate from the GPU integrated sensor
5 - 9-Cell Battery - Secondary; static @ 50C on 6-cell
6 - Ultrabay - Secondary; available only when device present
7 - 6/9-Cell Battery - Primary
8 - UltraBay - Primary
9 - RAM
10 - Right of RAM, Under Touchpad
11 - Mini-PCI Area (?)

hope this helps answer some if not all of your questions on the subject

EDIT:
I've changed RMClock with ThrottleStop for undervolting purposes, and I can report that it shows same temps as what HWMonitor shows under core 0/1.

- Here's quote of the author of ThrottleStop regarding the digital sensor in Penryn CPUs:
"DTS shows the direct reading of the on chip Digital Thermal Sensor. This sensor is designed to count down towards zero as the CPU heats up. Intel designed their processors so thermal throttling starts when this sensor reaches zero on any core. By clicking on the Temp button, you can convert this data to an approximate core temperature value."

 

flash middleton modified bios and you'll get rid of that annoying message about thermal sensors at startup. I've checked on mine T61 and it works like a charm with T9300

 

With a small registry mod, it's possible to run RM Clock and have it support the half multipliers.

http://forum.notebookreview.com/6130270-post4855.html

HWMonitor is a good program but I've seen it using the incorrect TJMax value for many 45nm Core 2 mobile CPUs so remember to check your CPU.

Intel® Core?2 Duo Processor T9500 (6M Cache, 2.60 GHz, 800 MHz FSB)with SPEC Code(s)SLAPW, SLAQH, SLAYX, SLAZA, SLB3BW, SLB49, SLB4A, SLB4B

And then edit the hwmonitorw.ini configuration file and make sure it is using the correct value which is 105, not 100 for a T9500.

CPU_0_TJMAX=105.0

You can also check the ThrottleStop.ini configuration file to make sure it is using the correct value. If you just swapped CPUs, it might be using a value based on your previous CPU.

Once that is corrected, both programs should show more or less the same thing. There might be slight timing differences when the sensor is read. RM Clock does not support the 45nm CPUs by default so it will need to be corrected too or it will show core temperatures 5C too low.

If you run an X9000 and a T9500 and you set them to the same MHz and core voltage, their heat output and core temperature should be close to identical. Don't let the TDP rating make you think that these two CPUs are significantly different because they are not. The difference comes when you start cranking up the multiplier in the X9000 and cranking up the VID voltage. Then it can start running significantly hotter but if you keep them the same, there shouldn't be any difference.

The core temperature sensors on my T8100 are excellent. Very accurate from idle to TJMax but its always a crap shoot since the 45nm core sensors were only designed to control thermal throttling and thermal shutdown and were never intended for super accurate core temperature reporting.

 

thanks for the good info on Penryn thermal sensors.

however, what I'm trying to determine is which program reports the "correct" temps for the CPU, as we would get that "thermal sensing error" in our thinkpads with the normal BIOS. Word is that this has been put by Lenovo so that users that want to use Penryn CPUs would have to buy another laptop.


again, these are the readings:

HWMonitor
- ACPI / THM0 - 42 deg C currently (same as what TPFanControl reports)
- CPU / core 0 & 1 - 35 deg C currently (same as what ThrottleStop reports, 5 deg higher than that RMClock reports)
- GPU - 47 deg C currently

I ran some test where I loaded the CPU with Orthos to be working at 100%. Considering that I have changing fan speeds based on different temperatures, I locked the fan speed for accurate results. What I got was this:

HWMonitor
- ACPI / THM0 ~ 62 deg C
- CPU / core 0 & 1 ~ 53 deg C
- GPU ~ 51 deg C

if the CPU core temps were correct (which are in question here), and the GPU temperature is known to be correct as being reported by 3 different programs as same, then how/why would the ACPI module report ~62 deg CPU temp while having two chips heating the heatsink with temps around ~50 deg C (according to the CPU core temps) ??

I would think that this will just make the ~53 deg C temperature more stable along the heatsink/CPU/GPU, rather than increase it to over 60 near the CPU.

what do you all think ?

 

Odd. The T61 I gave to my wife was originally mine and came with a Penryn-core T8300. I would think that with the latest BIOS, there shouldn't be a thermal issue, unless early T61 systems used an older variation of system board.

 

Should be able to support it, Santa Rosa supports 800 FSB Core 2 Duos..

 

Early t61 Thinkpad system boards were apparently designed around the Merom processors, and although the Penryn physically fits in the same socket, it causes a thermal sensing error message that can now, thankfully, be bypassed by using Middleton's bios upgrade. It would be a shame if Lenovo/IBM deliberately made upgrade difficult for the end user, but that may not be the techical reason at all. What makes us wonder is that Lenovo did refuse to "fix" the Bios to allow the newer, faster SATA hard drives to work at more than 1/2 speed - and yet we have seen that they could have done it fairly easily.

My question is, since I already have the fastest factory Merom T7800 rated at 2.6 in my November of 2007 T61p, would I be gaining enough, in performance or heat reduction, to justify putting in the later Penryn processors?

All T61p owners are of course concerned with the longevity of the Nvidia 570 graphics chip, and thus significantly lower temps might be worth more than they first appear. Opinions, anyone? Should I follow suit and upgrade to the Penryn to obtain possible longer system life due to less heat? TPF

 

The penryn performance is noticeable when decoding FHD movies on the T61 as compared to the merom processor. The heat reduction by around 5 to 7 degrees is another benefit of this upgrade, if you move from T7800 to T9xxx.

 

I'm totally sold to the Penryn CPUs after I've experienced their much lower temps (up to 10 deg C lower) . Same possible nVidia issue here. Together with undervolting and TPFanControl my laptop tops out at 65 deg C under full load (100% CPU usage both cores + 100% load over my overclocked GPU [by ~55%]). That's where I've set the crazy speed [64] of the fan to kick in, though I can make it start at lower temp as well.

If you dont want to spend too much on new CPU then aim for T9300, you'll still gain regular performance (aside of the HD decoding) of about 15%. Temperature benefit = priceless in our case

I cant comment on the versions of MB that Lenovo used, but I would assume that mine was one of the first ones (bought my laptop in Q4-2007). Also I didnt experience the thermal error because I flashed my BIOS before the new CPU went it, but I've seen others reporting it not once.

what I'm trying to figure out is which CPU temperature seems correct:
(~35 deg C @ idle) or (~42 deg C @ idle) with fan at the lowest speed [1].
I'm leaning towards the second one though - see explanation above.

 

I have a t9300. 35C at idle would be very very good. I'm in the mid 40's at fan setting 1.

 

what are your specs ? Also, did you change the CPU youself or it came with your laptop ? If the first case, did you get thermal error ?

 

t61p, t9300, 8gb ram, nvidia 140m,500gb seagate XT. All prebuilt from factory except I upgraded the ram and HD. No thermal error as the t9300 was installed from lenovo.

 

t61p with Nvidia 140m GPU?

 

^ probably the 'p' is a mistake.

jedisurfer1, run HWMonitor and see if you get about same temperatures under ACPI THM0 and CPU core 0&1.

if you get about same temp readings then it means that the thermal error is in fact a valid error and is present on early versions of the mother board, and your mobo must be of newer type. This is not BIOS related as I got latest BIOS and I'm getting different temp readings.

thanks

 

yes that is correct.

 

never knew a T61p has a nvidia 140m gpu.

 

Definitely. Penryn is always the way to go when the support is there. SSE 4.1 instructions for better video encoding and work in some rich media programs, cooler running, and up to a 10% increase in runtime off of battery (assuming equal clockspeed).

 

jedisurfer1, can you please run HWMonitor and see if you get about same temperatures under ACPI THM0 and CPU core 0&1 ??

thanks

 

^ too bad jedisurfer1 never came back with HWMonitor readings from his laptop

so far I'm thinking that laptops which got the thermal error probably pop that error for a reason, i.e. all temp reading programs that get their info from the CPU sensors would display very inaccurate temperature.

I would also assume that laptops that do not pop that thermal error at boot up will not have that issue at all, i.e. CPU temp reading programs will display accurate core temperatures.

 

Programs like RealTemp, ThrottleStop, Core Temp, etc., all read temperature data directly from a register in the CPU so the bios version or any issues at the bios level will not have any affect on core temperature readings.

The only problem is that many of Intel's 45nm core temperature sensors are not very accurate temperature monitoring devices. 10C of error is not out of the question.

 

Hi
I just upgraded my T61 -7661 from T7300 -> a 2. hand T8100
I flashed middletons bios long ago for some SATA 2 testing, so I have have no idea if the thrmal error would have popped up on boot.
From what I have gathered in other threads, the MB I have is indeed of the kind that would trigger the error.
While changing the cpu I also replaced the fan as it was beginning to make a grinding noise. (I didn't change the entire heatsink, only the fan, which prooved to be the most difficult part of the entire operation.) I live in a rather remote location and I got tired of waiting for the Arctic Silver Ceramic that I ordered, so I ended up using the "stars DRG-33" white thermal paste that shipped with the fan.
I am experiencing some odd readings and I am trying to make heads and tales of it which is a tad difficult due to all the factors at play here.
When I boot the machine, the temperature readings in TPFC (as well as the cpu load reading in the win7 gadget) are very odd as they sometimes jump abruptly and it seems that the cpu temp reading falls out altogether once in a while. Now, after being on for 20min, it seems more stabile indeed. I have been running passive for 15min on default "balanced" powerplan, adaptor not plugged in. This has resulted in a stead temp climb from 47C to the current 58(fell to 57 while typing this)
Hwmonitor: THM0=58, THM1=48, Core0=42, core1=43.
All in all, both temp and cpuload seems to be higher than I had expected (running crome, skype, MSE, TPFC, HWmonitor)
Granted, the thermal paste probably suxx and last time I dabled in apliying it was almost 10years ago, still, I think I did a pretty good job getting the drop of grease spread thin and evenly. Moreover, any problems as to installation of fan or paste should not result in these odd readings.
Just now, the tpfc reading jumped from 60 to 75
UPDATE:
Tried reinstalling TPFC and upon restart it reported 96C (gadget reporting 100%cpuload) while the laptop was cool to the touch underneath and the exhaust air from fan only slightly warmer than my hand. Tried another reboot and changed a adaptive power and speedstep in bios to maximum performance. After this, the tp would turn off just before reaching loginscreen and it would no longer detect the AC being plugged in. Took battery out and back in, tp detects AC again, booting resulting in same behaviour, sudden turnoff before login and loss of AC detection. Changed Bios back to balanced, still the same result. Only got in to windows by booting without AC attached.. this is getting stranger and stranger...
As I entered windows,(gadget reporting 100%load) TPFC reported 85C, then 96, then 85, then 75 declining to 60..
Have I screwed something up? is the 2.hand cpu defective? any ideas at all?
Don't know if some of the events could be random and unlinked.. experienced a similar problem with loss of AC detection a year ago, but the machine has been running smoothly since... I am at a loss
Running stable for a while now. Bios on balanced, Powerplan on balanced fan back to the behaviour of the days before TPFC (((
Still getting odd readings but the laptop seems VERY cool to the touch as well as the exhaust.
Makes my lap feel cold, never had that before.. still.. I want to regain control over my sensorreadings and my fan before I start rendering stuff and playing around with rmclock and such..

 

problem appears to be worse than I thought
just after editing my last post, I ran utorrent, crome, closed and opened skype and bam, tp died, loosing AC detection. removed and replaced AC and bat. Reboot, same thing happened again, and again etc.. as I could not get to login, I let windows run startup repair while doing dishes.. after 30min it finished, froze and rebooted itself, same thing, crash and no AC.
Fiddle with bat and Ac again, this time attempting to boot to Ubuntu, same thing, crash, no AC. At no point has the tp gotten more than lukewarm and it has certainly been very cold compared to my experiences with the old T7300. Rebooted again and again to no avail untill it suddenly accepted booting to Ubuntu from where Iam writing now..
I am getting a little frustrated as the randomness of the issue makes it hard to choose apropriate actions.
The T8100 I bought from ebay was cheap (won it at 45$) but the icelandic seller seemed very nice and I felt no reason to be suspicious.
So far I have only learned one thing: Do NOT replace any vital parts in your only machine while working on your dissertation abroad
Eagerly awaiting any input, might start a thread on tpforums as well..

 

if your fan is working properly (you'll need to doublecheck), then the CPU is not touching the heatsink. That would explain all the symptoms that you're having, i.e. CPU throttling back at high temp, and/or shutting down due to overheat.

also, the default Smart setting on the TPFanControl is for turning the fan mostly off rather than keeping the laptop mostly cool - I had to completely change the settings for my needs (i.e. cooling priority). Here's my temp profile in the ini file:

Level=30 0 // 0 - off
Level=40 1 // - added to prevent fan from turning off @ T<50
Level=50 1 // 1 - 2500 rpm
Level=55 3 // 3 - 3000 rpm
//Level=60 6 // 6 - 3300 rpm
Level=65 64 // 64 - 4600 rpm [MAX]

the above settings let my CPU max out at 65 deg C under continues 100% load and that's it. I'm not using the level 6 of the fan anymore, but only 0,1,3,6, and 64 actually change the fan speeds on my T61.

- back on the thermal error issue - I can see that your HWMonitor reading under THM0 is much different than what's under CPU core 0 and 1 ... thus my point that this error actually do exist for a reason. You should go by the THM0 reading as reference for your CPU temp.

 

What do you mean by

?
I have been running on bios mode or manual=7 since the first crash and though the fan exhaust is strong and cool I keep crashing..

If this is the case, what should I do about it? I tightened the screws on the heatsink slightly pretty well, could it have something to do with the "stars DRG-33" paste that came with the ebay fan or could there be a slight difference in size between the T7300 and the T8100? I will try to get some Arctic paste and redo the entire thing I guess.. Any advice on AS5 vs Ceramique vs Mx1/2/3/4 for this particular situation?
As I installed the new unoriginal fan in the heatsink myself, some air might be leaking inside the casing, but if this had ANYthing to do with the fan, then the bottom of the laptop would be heated by the heatsink I assume? And finally, I had 30+ crashes so far, what are the chances that I have damaged something?
Sry for the question spam

 

Update to this thread from my posts in the thinkpad forum
Further observations so far:
No crash has ocurred while idling
Not only the CPU reading in TPFC is affected.
Once in a while, other sensors report 1C and furthermore, "new" sensors (eg. 8 & 12) which are not nomally present suddenly shows up to then dissappear magically
Been running stable in my lap for a while now, running crome, winamp, skype, TPFC(manual=7), HWmonitor and realtemp resulting in the following unstable temps:
TPFC - 1 cpu: 50-58C
Real temp 3.60: 38-48
HWmonitor: THM0/1: 43-57/43-56
Hwmonitor Core o/1: 36-40
Reset the bios to defaults and did a bit of testing, TPfancontrol reported temps as high as 105C while bottom and exhaust remained cool to the touch.

 

just took the machine apart again, and from eyeballing the T8100 I'd say that it is a little thinner than the T7300.. Maybe 0.1mm or less but enough to make a difference?
The thermal paste smeared well onto the heatsink, but I imagine that the quality of the paste would become vastly more important if the T8100 indeed is slightly thinner. I remember reading somewhere about adding a piece of copper sheet to improve contact but firstly this sounds rather.. unpleasant.. and secondly, I have nothing like that... I could try tightening the heatsink further, but I doubt it would help much..

 

I would usually apply more of the thermal paste on the CPU than what others recommend, i.e. I'm putting the paste all around the CPU and not just one drop in the middle. Others may argue about that but I'm pretty happy with the results.

if your fan exhaust is strong and cool then that's saying that it's not pulling heat away, thus suggesting for the heatsink and the cpu not touching enough or at all.

I can not comment about why other temp sensors pop up in TPFanControl and then disappear, this is ACPI module related, i.e. could be BIOS, could be the ACPI module itself, could be driver in Windows, could be the CPU.

anyways, I'd start with repasting the CPU. Bolt the heatsink down as much as possible .. without stripping the bolts though, LOL.

 

"I would usually apply more of the thermal paste on the CPU than what others recommend, i.e. I'm putting the paste all around the CPU and not just one drop in the middle. Others may argue about that but I'm pretty happy with the results."

I have not touched the thermal paste on my T61....and it's been just fine.

Renee

 

Cheers guys I was too tired to report back last night.
I fastened the heatsink better (the first time I was afraid to overtighten the screws and I only tightened them as much as I felt they where when I pulled out the merom.
After that it worked!! No more crazy readings, and so far, no crashes.
I am however, still running a little warmer than optimal I think( while writing this in crome on win balanced on battery, passively cooled in my lap, I seem to stabilize around 48-53C)
As soon as my Artic paste arives I will give it a try, but I am still concerned that I might need to play around with some copper sheet to reach optimal temps. As the paste I used was of the "white paint" kind, I don't know if it has a curing-time or not.

 

^ glad to hear, it was pretty obvious (at least to me) that this was your problem. You can look into undervolting your CPU to reach minimum temperatures. Also, changing the type of the thermal paste used will not make much of a difference, unless you're an extreme overclocker by all means.

^ applying thermal paste on the CPU is required when you are putting new CPU and/or when you have to detach the heatsink from the CPU (even though some just reuse the existing paste). Some of the first notebooks of our series (T61/R61) came with thermal pad instead of thermal paste between the CPU and the heatsink (I've seen pictures on here) - this is generally less effective so people changed it. I had doubts that mine could be using a pad as well so I went to replace it, only to find out that thermal paste was used .. LOL. Chances are your T61p most likely uses paste and there's no need to replace it.

 

"I had doubts that mine could be using a pad as well so I went to replace it, only to find out that thermal paste was used .. LOL. Chances are your T61p most likely uses paste and there's no need to replace it."

I know. I'm a professional. And my T61 does uses paste and am I not an advocate of "paste wearing out". It doesn't and notebook people do not understand that. I recommend Artic Silver.

Renee

 

That kind is often garbage; cheap stuff which is prone to migration over time.

If you have a gap, a bit thicker paste is a better choice. The stuff I use on most people's systems is Arctic Ceramique, as it's inexpensive, non-conducting, and does a good job. For your job, I'd use something more like one of Shin-Etsu's compounds.

Shin-Etsu X23-7783D Silicone Thermal Compound - 1g - FrozenCPU.com

This is what usually ships with one of Dell's warranty mainboard replacement kits. It's a bit heavier, and has better conductivity.

If you use Arctic Silver, remember they claim a cure time of up to 200 hours, so it might read warm at first. I usually only put it on desktops I have on 24/7 for that reason, and because it's messy, which is a bigger issue with small exposed CPU dies than with the large heat-spreaders of desktop processors.

P.S. With respect, I tend to disagree with miro-gt's method. Apple had a number of overheat issues with Macbook Pros a couple of years back; it turned out that the factory and some Apple Store techs were over-applying the thermal compound On a mobile Core 2 Duo, a drop about the size of a grain of rice should do it, and the pressure of the heatsink will spread it out across the CPU.

I'm not sure I get you. Once thermal paste has been used, it usually cures, so if you pull the heatsink, it will need to be re-applied. The old stuff won't usually make a good bond with a new CPU, possibly resulting in the sink not mating uniformly to the CPU and creating temperature anomalies.

 

"Once thermal paste has been used, it usually cures, so if you pull the heatsink, it will need to be re-applied."

I know. I have really been into my t61p. At the same time, I've never been near my CPU. In fact, I've never seen it.

There are people here who think thermal paste 'wears out' which it does not. In fact it's a very bad idea to replace it for a variety of reasons.

Anyone who pulls the heatsink should do so for a reason, not routine maintenance

Renee

 

deleted !

 

I can not comment on what issue apple had, but I would think their story seems like a cover up for something else causing the overheat.

I'm placing thermal paste all over the CPU as this way I'm sure it covers all of it. With the drop in the middle you can not be sure of such coverage as it may or may not cover all area, highly dependent on the viscosity of the compound and on the pressure applied over the heatsink.

here's another idea:
- in general, the more area you have the better the heat transfer will be. The CPU itself is a rectangular chip "glued" to its proprietary CPU board. Usually the heatsink is placed on top of the chip, thus the area of the heat transfer is only the area of the top of the chip. Now putting a bit more thermal paste on the chip will cause small part of the thermal paste to get squeezed out and cover the visible part of the sides of the chip as well (that is not covered by the glue that holds the chip down), thus adding the area of those to the total heat transfer area, thus causing better heat transfer.

the common understanding is that the closer the heatsink is to the CPU, the better. Which is correct, so the argument here is that if more thermal paste is used, the heatsink itself will sit further away from the CPU than if less thermal paste is used. Well I'd say that depends on the viscosity of the thermal paste and again the pressure used to hold the heatsink down towards the chip.

but in fact, those things play small role compared to the fact that with single drop in the middle of the chip you may not cover all top area of the CPU, which would result in much bigger issue with the heat transfer.

hope that make sense somewhat

 

Here's the issue though.

You and I can probably agree that the main reason for using CPU thermal compound is that both the CPU, and the heatsink, are not only not 100% flat, they are not 100% smooth. The surfaces are porous; there are microscopic gaps. Thermal compound is meant for one purpose: to fill those porous gaps and make a smooth, flat surface between the CPU and the heatsink.

The heatsink itself, is actually a better conductor than the compound. Over-use of thermal compound can actually increase the gap between the CPU and the heatsink (seeing as thermal compound isn't infinitely compressible), and if this happens, the compound starts to become more of an insulator than a conductor, increasing heat. That's why applying a lot of goop is bad.

The advantage of applying a small amount in the center means that when you tighten the heatsink down, the compound is forced (via compression) to spread across the CPU die, creating an even surface, as well as filling all the gaps. Due to the compression, there will be no problem with the thermal compound spreading across the CPU; because the gap between CPU and heatsink is so small. Once the heatsink is tightened down, it is guaranteed, provided you tighten the heatsink down evenly.

The problem with this is because the thermal compound is supposed to be a transfer between the CPU and the heatsink. Once the paste is squeezed out on the sides, it's not really touching/transferring to the heatsink any more; it's just insulating, trapping the heat close to the die.

The concept of surface area is extremely important as it pertains to the heatsink. However, themal compound isn't there to increase surface area (since it's for heat transfer to the far more efficient heatsink); it's only there to get the heat transferred to the heatsink. That only happens at the point of contact from the CPU die to the heatsink, not at the sides of the die.

 

while I mostly agree with you I would like to point out that a drop in the middle would work best if the chip was actually in a circular form, and not rectangular.

for rectangular form of the CPU (~100% of CPUs out there) a drop in the middle will result either in not enough coverage at the corners of the chip, or in squeezed thermal compound mostly at the narrow sides of the chip.

here's a good article:
80-way Thermal Interface Material Performance Test | Thermal Interface Material,Thermal Paste,Heatsink Compound,80-Way Thermal Interface Material Best Thermal Paste Heatsink Compound Cooling Performance Comparison Benchmark Tests

^ pages 4 and 5 in that article have pictures

the thickness of the compound layer that gets left between the CPU and the heatsink depends to the high extent on the viscosity of the last and the pressure applied on the heatsink, and not as much on the amount applied as the CPU itself has rather small surface (today's laptop CPUs).

 

My experience with thermal paste is similar to Lonewolf's. I don't think that vendors use the best thermal paste. I also don't think they put it on in the best fashion possible. I prefer my mode of putting it on since I have always built destops. I put on Artic Silver in a limited and controlled fashion. Too much or too little is not desireable. I put it on sparingly.

Renee

 

I've tried a few methods and I found that when applying a thin paper-thick layer of AS5 does not work as well as being more generous. I think the design on laptops are different, and some does not allow you to force the hsf on the cpu enough. It's not like the old AMD k7 days.

 

I must admit, I have not put a thermal paste on my notebook. I have applied it to many desktops.

Renee

 

Course not. Manufacturers already hardly make money on computer components, spending thousands of dollar on decent thermal paste is adding insult to injury. That and the tech who puts it together probably brainlessly applies any old glob of thermal paste.

You really only need like a raindrop sized glob and spread it evenly.

 

Arctic Silver 5 isn't nearly as viscous a paste as most; for this reason, a bit more may be required than some other types.

Laptop designs have to allow high enough pressure; if they do not, CPU-to-heatsink contact is poor, with the resulting problems we all know about. To do this, they usually have spring-backed (for lack of better terminology) screws that have a maximum tightness. They allow pressure against the CPU, but won't allow you to be so tight as to crack the CPU die.

As for most computer vendors, most of the compound I have seen is from Shin-Etsu, who is a major OEM provider. SE has several blends, but most of it is decent stuff. Vendors do of course, get it a lot cheaper direct than the end-user can, and skill in application at the factory may vary.