More questions about T60/p heatsinks...

[LegendaryKA8]

Hey, all. I didn't want to muck up any of the other threads on this topic as I am going to be asking a few different questions here. Firstly, what is considered 'good' operating temps for the FireGL GPUs, both idle and load? I am going to be getting my hands on a V5250 motherboard soon and want to know what to expect.

If this helps anyone, I bent the heat sink down on my X1400-equipped T60, as well as using a combination of a decent amount of AS5, thermal pad sitting on top of the heat sink, and the 'dime mod' to provide a little more pressure to the GPU. Using smart mode on TP fan control, I'm getting idle temps as low as 60C(where the fan turns completely off) which slowly raise up to 65C, where the fan turns on and cools it back down... rinse and repeat. I see load temps in the 80C range, which considering the thermal design in this laptop is pretty decent.

Once I get the new motherboard in, my plans are to use a small copper square(salvaged from a dead T4x long fan; it is almost the exact size as the GPU area on the T6x heatsink) in place of the thermal pad(using Arctic Silver thermal adhesive). I don't have a set of calipers on hand, but judging it against the thickness of a dime the plate is about 0.050" thick. I will be using a new heatsink for this, as I don't want to use my already bent one. Is the stock thermal pad on the heatsink about this thick, does anyone know?

Also, what FRUs should I be looking at for a good performing T60p heatsink?

[JR956678]

Is the stock thermal pad on the heatsink about this thick, does anyone know?

I just installed a new heatsink on my T60p and I measured the thermal pads before I installed. The pad for the GPU is 1mm (.039") uncompressed and the thermal pad for the Northbridge is 2mm (.078").

Also, what FRUs should I be looking at for a good performing T60p heatsink?

The FRU for the T60p heatsink for C2D processors and FireGL 5250 GPU is 41W6407 (you may also see this referred to as 41W6403 which was the IBM part number - 41W6407 is the FRU for the same part). The FRU for the T60p with CD and discrete GPU is 41V9932 (26R9434 is the Lenovo part number). In another thread I described how I modified a T61p heatsink (42X4685 / 42W2028) to fit a T60p (it can be done but it's not a drop in replacement).

As to what is the best performing heatsink that is a $64,000 question and I'm sure opinions will differ. I don't have any good data for the 41W6407 that was original on my T60p as a fan bearing failed on it which is what started me on the path to replace it (a path fraught with too many zigs & zags because I didn't know what the right FRU part numbers are so I bought several "wrong" heatsinks).

What I just put in my T60p this morning and what seems to work better than anything else I have tried is actually a modified 41V9932. From a design view, it has the best heatpipe orientation from the standpoint of cooling the GPU - all the other designs do a better job of cooling the cold end of the heatpipe(s) for the CPU but a poorer job of cooling the cold end of the heatpipe for the GPU - and it's the GPU cooling that's the biggest issue with the T60p.

My observations of temperatures on preliminary testing suggests that the theoretical design benefit of the 41V9932 seems to pan out in slightly cooler GPU temperatures - and I'm using the factory stock thermal pads with the 41V9932 whereas my modified 42W2028 used no thermal pads at all - it sat directly on the GPU and Northbridge using only AS5.

Even though I did it on the modified T61p heatsink, I view the elimination of the stock thermal pads in favor of direct contact and AS5 as a risk / reward proposition. The hoped for reward is better heat transfer and thus cooler temperatures (assuming of course that the heatpipe is capable of transferring all the heat generated from the GPU to the cold end of the heatpipe and getting rid of that heat at the cold end). The risk is damage to the top surface of the GPU and the Northbridge chips if the heatsink plate isn't exactly parallel with the top of those chips and presses harder on one edge of the chip. Think back to all the stories of the early Athlon processors without a metal heatspreader to protect the silicon substrate and fatal damage from improper installation of a heatsink and you get a sense of why I see the risk potential. From what I've seen this morning, even though the concept of thermal pads makes me cringe, they actually work well (if they're new) and there's little risk from them.

[LegendaryKA8]

Wow.... that is a huge amount of useful information. Thanks again!

I just had a couple of questions... firstly, what did you do to modify the 41V9932? And, you said you kept using the stock thermal pads and are using them successfully?

Another thing I'm curious about are your actual temperatures... what were you getting with your T60p with the various heatsink arrangements?

[sktn77a]

"Dime Mod" ???

[LegendaryKA8]

"Dime Mod" ???

I used a dime(has a higher actual copper content than a penny) between the heatsink and the clip that normally puts pressure on the GPU to put a little more direct pressure on top of the GPU itself. Since I'm running with the GPU in actual physical contact with the heat sink, I placed the thermal pad on top of the sink to keep the dime from coming loose and falling onto the motherboard.

[crazyfrog]

Probably I can do something to prove LegendaryKA8's theory that 41V9932 (designed for CD) is actually better than 41W6407 (designed for C2D) in cooling GPU.

I have two upgraded T60p about the same age and more importantly, both of them have V5250 and T7400 now. One originally comes with CD T2400, while the other comes with C2D T5600. So I should have both 41V9932 and 41W6407 in hand. The only difference between the two machines is OS. One (originally with T2400) is running windows XP, while the other is running Vista 64-bit. Currently the xp machine is always cooler than the vista machine. But we need to take the difference in OS into acount.

I will use the same hard drive/system to test the two laptops and report the results later.

[JR956678]

I just had a couple of questions... firstly, what did you do to modify the 41V9932?

A couple of (relatively) simple mods:

1 - I epoxied (using Arctic Silver epoxy) two copper plates - removed from the defunct 41W6407 heatsink - that sit adjacent to the CPU heatpipe to provide a bit more thermal mass to help CPU cooling. This seemed like a good idea they had on the 41W6407.
2 - I added two .025 copper plates to either end of the plate which contacts the GPU + Northbridge. This also was an idea borrowed from the 41W6407 and they increase the spring pressure of the hold down clip. I wanted to avoid the penny / dime mod mostly to get some spring action and controllable pressure on that plate.
3 - I epoxied a .100" think copper plate immediately above the GPU and adjacent to that heat sink to get a bit more thermal mass to help the GPU.

I have some pictures of all this but the most recent ones are underexposed so I need to retake them beforeposting and providing some links - it's easier to see what I did than it is for me to describe it.

And, you said you kept using the stock thermal pads and are using them successfully?

Yes. I am actually getting better cooling from this modified 41V9932 than I was from the more radically modified 42X4685 which used no thermal pads at all - direct contact between the GPU and the copper plate assisted only by AS5.

Another thing I'm curious about are your actual temperatures... what were you getting with your T60p with the various heatsink arrangements?

Well - this is sort of realtive because I don't exactly know how transportable readings are from one notebook to another. And as I have no data from the 41W6407 (it died a bit ago and I never replaced it) all I can compare is the heavily modified 42X4685 to the slightly modified 41V9932.

But here goes -

42X4685
CPU at idle - 49
CPU under the max load I could generate - 67
GPU at idle - 64
GPU at max load - 101 was the highest I ever got but I backed off the load at that point

41V9932
CPU at idle - 49
CPU under the max load I could generate - 67
GPU at idle - 63 - 64
GPU at max load - 97/98 and stable

The fact that I am getting better cooling with the 41V9932 than I did with the modified 42X4685 using no thermal pads at all leads me to believe that the thermal pads are not all that bad a thing - especially considering that they are gentle on the chip surfaces and make up for what inevitably can be an imperfect fit.

Subjectively - when under load there's quite a bit of quite hot air exhausting from the rear air discharge vent - it's hotter with the 41V9932 than it was with the 42X4685 - which makes me suspect that the benefits of that design and its better cooling of the GPU heatpipe's cold end is what makes the GPU stay cooler (it's able to extract more heat from the heatpipe).

crazyfrog 's post above and his experiment is going to be interesting.

[LegendaryKA8]

Crazyfrog:[/b ]Hmm.... that would be a very good experiment to run. However, what are your current GPU/CPU temps on both your XP and Vista machines?

I'm not sure about load temps(haven't really tested them out that much), but my GPU idle temps are around 49-53C. Not the best in the world, but still well within specs. I do want to say that I have never had any problems with critical overheating in this system... I just want the better GPU performance of the V5250 versus the X1400.

JR956678: Thanks for the temps. Actually, your idle temps are almost exactly what I'm getting out of my X1400-equipped T60. If I can get similar numbers to yours, I'd be highly pleased with the thermal performance in this notebook. My current heat sink is a 41v9932, but as I did bend the pipes and all that fun stuff in order to get a direct fit. That improved temps drastically, but I will try a new fan with new thermal pads first.

Where did you find the copper plates to do your mods? I already have the AS thermal epoxy, but have had a major time with finding a local supply of copper plates that are actually that thin.

I would also be very interested in seeing these pictures, too. Sounds like you've done a lot of homework regarding thermal designs on these heatsinks.

[JR956678]

Where did you find the copper plates to do your mods? I already have the AS thermal epoxy, but have had a major time with finding a local supply of copper plates that are actually that thin.

It was more than a bit of a hassle. Some of them were made from a sheet of K&S "hobby" supply copper I got at the local hardware store - it was .025" thick. Other copper plates came from the dead 41W6407 heatsink, and one I made on a milling machine from an old solid copper base heatsink by milling off the fins and thinning the base out to about .085".

Of note - idle temperatures are mostly the result of simple heat conduction - the heat pipes don't really start working as heatpipes at idle (nor should they). Just from watching temperature changes (both how quickly things heat up under load and how quickly they cool off when the load is removed) I would estimate that the CPU heatpipe starts pumping in the low 60s and the GPU heatpipe doesn't start pumping until the mid 70s.

[dr_st]

Lots of good information here.

The only difference seems to be in the max GPU temperature. How big the actual difference is we cannot know from these numbers, since JR956678 stopped the test (to spare the GPU, which is a prudent decision).

I wonder - what tool do you use to stress the CPU/GPU? I've found out that S&M does a "better" job than anything else I tried in achieving high stress over time. It has both CPU-stressing and GPU-stressing tests. ATI Tray Tools and its 3D renderer seems even better for extreme GPU tests.

I just tried running both on my T60 (T7200/X1400, stock 41W6407 replaced by IBM a few months ago).

The highest temperature recorded for the CPU was 94C running multi-threaded CPU/FPU test in S&M.

The highest temperature recorded for the GPU was 90C (and stable) running ATI Tray Tools 3D renderer (for about 10 minutes).

The CPU/GPU idle at about 55-60C. Keep in mind that it is a hot day and the room is not air-conditioned (current temperature inside is 31C), and that the laptop is docked (which is known for some reason to raise temperatures by a few degrees, probably obscures vents somehow).

[crazyfrog]

Today I got some time to do the test. Before I started the games, I opened up the two laptops to give their fans a through clean (dust will significantly influence the fan's performance!). To my surprise, the T60p original coming with CD T2400 has stock 41W6407 too, and I didn't realise that before. So I can't compare 41V9932 with 41W6407 by myself. However, I think it might be useful to produce some performance data of 41W6407 under XP and Vista to compare with JR956678's data.

First, here is the configuration of my two T60p:
T60p No.1: C2D T7400 (original coming with CD T2400), ATI V5250, 15" SXGA+, 3GB RAM, XP 32-bit SP3, PM for XP v1.60 set to ThinkPad Default.
T60p No.2: C2D T7400 (original coming with C2D T5600), ATI V5250, 15" UXGA, 4GB RAM, Vista 64-bit SP1, PM for vista v2.42 set to Power Source Optimized.

When upgrading the two laptop to T7400, I applied AS5 on both processors. The rest things of the two heat sinks were left as they originally were, i.e. stock themal pads on GPU and Northbridge, no bending, no extra copper plates.

Round one - idle test
T60p No.1: CPU 47C, GPU 63C, fan speed 0, room temp 18C;
T60p No.2: CPU 48C, GPU 66C, fan speed 3466, room temp 19C.

Round two - stress both cores to 100% using Prime95 v25.11 (http://www.mersenneforum.org/showthread.php?t=12155)
T60p No.1: CPU 91C, GPU 84C, fan speed 3438, room temp 18C;
T60p No.2: CPU 94C, GPU 83C, fan speed 4096, room temp 18C.

After the temperatures are roughly stable (>15mins) I read several times through TPfancontrol to produce the above average values. I didn't stress gpu as I don't have a proper tool.

So a clean 41W6407 with AS5 performs quite nice under both XP and Vista. With light load and low room temperature, the fan can be completely stopped under XP.

[LegendaryKA8]

Excellent work... this gives a good baseline as to how well the stock heat sinks should be performing. Idle temps on your stock V5250 configurations seem to match my X1400, which does have a modified 41V9932. Load temps are a bit higher, but still within what I would call a safe spec with these notebooks. I've never really heard of anyone with GPU problems on these notebooks, except when they were critically overheating due to dust or some other obstruction of airflow.

I have a brand new 41V9932 on the way, as well as a few dead 41W6407s. JR956678, would you be able to take more pictures of your current setup? I don't really need to see the heat sink taken out of the notebook; I just need enough of a look at the top to determine where all the extra copper plates should go.

[JR956678]

JR956678, would you be able to take more pictures of your current setup? I don't really need to see the heat sink taken out of the notebook; I just need enough of a look at the top to determine where all the extra copper plates should go.

I will post some pictures this weekend - probably tomorrow as I have to get them off the camera - what you're looking for I already have good ones of (it was the T61p heatsink where they didn't come out well).

In the meantime this was interesting - I ran the same test that crazyfrog did on my T60p with the modified 41V9932 heatsink I have on it. The notebook runs 32 bit Windows XP Professional with 1G of memory. My heatsink was installed with AS5 between the heatsink and the CPU. Using Prime 95 from crazyfrog's link I ran it with 100% CPU utilization on both cores until temps stabilized, and got:

CPU - 87C
GPU - 84C

The fan speed was 3470 rpm when everything stabilized.

It surprises me a bit that my CPU temperatures stabilized at a lower temp than crazyfrog saw - the 41W6407 heatsink does a lot more to help CPU cooling than does the 41V9932 (although a few details on my modified 41V9932 were borrowed from the 41W6407). I don't really have an explanation for that but perhaps all the enhancements they designed into the 41W6407 to get better CPU cooling really aren't necessary.

[crazyfrog]

JR956678, what is the processor in your T60p? I think that could make a little difference. BTW, it is the first time I used AS5 for my T60p. Maybe I am not very good at spreading a thin layer of AS5 on the small core. So far I am happy with my 41W6407 and its performance. I just want it to be a little quieter. So I may try T61/p's heat sink and fan when I have a chance.

[JR956678]

JR956678, what is the processor in your T60p? I think that could make a little difference.

I would think it could and there is a difference - my T60p is a T7200 - although Intel gives the same specs for Thermal Design Power for both the T7200 and the T7400.

I may try T61/p's heat sink and fan when I have a chance.

Be aware that there's a height difference issue - and also a spring clip issue. The T61p heatsink is made to cool an Nvidia GPU which replaced the ATI FireGL of the T60p, and the Nvidia sits lower than the ATI. When I tried to install the T61p heatsink on my T60p I also ran into issues with the clips creating downforce on the GPU + Northbridge plate as there were no embedded screw inserts in the frame for them. I had to modify the T61p heatsink to make it fit (this is actually discussed in another thread on this forum) although I did end up getting it to fit and work.

[bhtooefr]

Of course, there's also this diamond thermal paste...

[LegendaryKA8]

Alright.... I'm going to admit I'm a bit impatient, but I ended up receiving my T60p/V5250 motherboard as well as a brand new 41V9932. I might have to tweak my TPfancontrol settings, but once I first booted up(after a BIOS stint to change a few settings) TPfancontrol reported my new GPU as being at 72C. It has new settled down, and appears to be idling at 65C, although it's dipped down to 64C as I'm typing this.. This is a completely brand new stock heatsink with stock thermal pad, with only my previous 'dime mod' applied. It isn't running quite as cool as my X1400, but I'm not going to complain. CPU temps are a bit lower with the stock thermal paste(47-48C versus 52-53C), so that's an improvement as well.

Benchmarks to come in a while... I'm just going to sit on my laurels before a massive test. BTW, how do I get the 'highest temp' to come up in TPfancontrol? I've never seen that setting before.

Edit: I just ran my machine through 3DMark06(score was 1690, so on par with what I've seen a V5250 do). This is by no means a scientific test, but the highest GPU temp I saw was 94C, and the CPU stayed put at about 76-78C or so. It sounds like these are stable temps for a T60p, so I think I'm going to keep this arrangement as it is. Temps dropped very quickly once the benchmarks ended, so that's a very good sign.

[crazyfrog]

I am thinking of replacing the stock thermal pads on GPU and northbridge with something better, e.g. Arctic MX-2, Arctic Silver Ceramique. has anyone tried them? And what is the expected performance improvement?

[JR956678]

I am thinking of replacing the stock thermal pads on GPU and northbridge with something better, e.g. Arctic MX-2, Arctic Silver Ceramique. has anyone tried them? And what is the expected performance improvement?

You can't use grease like compounds in thick layers - only very thin ones - so to replace the thermal pads with grease you also have to put in some sort of copper plate to make up for the height difference between the GPU and Northbridge chips; on my T60p that's a .050" height difference.

I tried Arctic Silver 5 on the T61p heatsink - but I didn't get as good performance with that as I do with the 41V9932 and the stock thermal pads. I know I don't have an apples to apples comparison but somehow I think any performance improvement is minimal at best but I didn't have thermal pads when I used the T61p heatsink (which was modified considerably - it actually used a plate from a 41W6407 heatsink with an estra copper shim on the bottom to make up this height difference).

LegendaryKA8 - your results seem very close to what I'm getting on my T60p with the same heatsink unit.