Arctic Silver 5 on T42p: Temperature benchmarks inside

Message

Kenn · #1 Post by **Kenn** » Thu May 19, 2005 9:35 am

15", Pentium-M (Dothan) 745.

Turns out the Arctic Silver site has good pictures and tutorials, so I didn't take any pictures. Thinking back on it, I probably should have, since the installation takes a little bit of disassembly (remove keyboard, palmrest, disconnect LCD ribbon, remove PCMCIA slot, remove left LCD hinge and left-side magensium frame support, remove heatsink/fan). Not ultra-complex, but you need to be comfortable with PC hardware and philips-head screws.

I'm still running benchmarks on Prime95 (torture test, in-place large FFTs), and the AS site says that full effectiveness will require a few thermal cycles (room temperature to load and back again). So for now, this is what I have. I used Prime95, with the system running for 5 minutes at 100% at each setting before taking a reading. Centrino Hardware Control 1.7 was used to adjust voltage.

1.8GHz at 100% load, before applying AS then after:
1.34Volts (default): 65, 61
1.132V: 55, 51
1.116V: 54, 52
My system at 1.084V is "rock solid stable" and has been for over a week, but Prime95 reports a rounding error within a minute. 1.16V errors out anywhere from 5min-6hours. 1.32V seems to be the sweet spot, no errors reported by Prime95 for over 16 hours. Too bad about 1.084V. I'll probably still use it if I need to run at 1.8GHz on battery power, but usually 600MHz is plenty for my needs when I travel.

600MHz at 100% load, before then after:
0.988V (default): 43, 40
0.700V: 41, 40
0.077V, the lowest setting allowed by CHC, is stable for over 3 hours under Prime95. Good enough for my needs, and so far over 50mins more battery life on my 9%-wear-level 6-cell (just over 4 hours on 15" UXGA).

I haven't tested idle temperatures yet, but I imagine there'll only be a degree or so of difference, if at all. So far, this is in-line with my general experience regarding thermal transfer compounds - the major benefits come at load.

Questions welcome.

#2 Post by **K. Eng** » Thu May 19, 2005 8:01 pm

Nice results. Do you have any problems with fan noise, and if so, has the use of arctic silver helped at all?

Kenn · #3 Post by **Kenn** » Thu May 19, 2005 8:54 pm

I've never really noticed the fan after the first week of owning the thinkpad (as my old HP laptop had a 1.4GHz P4 processor that was cool enough where the fan was completely off on battery power). I think using the Centrino Hardware Control application to undervolt the CPU made the biggest difference in heat, though the Arctic Silver did help, too.

From the benchmarks posted above, I don't think that applying Arctic Silver dropped me below any of the fan speed threshold points at full load, but during regular computing I'm sure it will mean the fan comes on later and a bit less frequently.

#4 Post by **K. Eng** » Thu May 19, 2005 11:05 pm

Hmm well it was worth asking about

The fan still stays on in my T4x no matter what the CPU temperature is anyways. I am still sure that there is a thermal sensor near the WiFi card and ICH4-M southbridge that is always tripping.

Kenn · #5 Post by **Kenn** » Fri May 20, 2005 1:20 am

I always assumed there was only one diode (on the CPU), since nothing but CPU temp seemed to trigger the fan.

Btw, I was running all day at 600MHz 0.700V, the system stayed at 33 degrees the whole time, and the fan never came on

UCI2CI · #6 Post by **UCI2CI** » Fri May 20, 2005 3:14 am

good stuff Kenn. I had a question. was you wireless card on or off during the tests.

Kenn · #7 Post by **Kenn** » Fri May 20, 2005 9:30 am

I've been running with wireless off all year pretty much, but during these tests it's been on and connected to the net (but with no appreciable network activity).

Edit: Strike the above, I was getting confused over something

Both before and after tests were run with the laptop connected to the port replicator 2, with internet access through wired ethernet, so the wireless adapter was not on.

beeblebrox · #8 Post by **beeblebrox** » Fri May 20, 2005 10:20 am

So it seems to me that the Dothan can be volted down far more than the Banias (0.15V difference). At least you get the higher performance with still the same battery life like the Banias core. The 90nm to 130nm indeed gives some advantages.

Concerning Artic silver etc.
It is not necessary, we are not in Prescott times. Stay with factory made stuff or cheap silicone paste.

Actually, some time ago I read a hardware magazine article about various tests and their result was that Artic silver gave a 1-2°C advantage over dirt cheap 70 cents silicon paste (what I am using now on Pentium III systems).

For a highest clocked Pentium IV desktop CPU every °C counts for survival, but on Pentium Ms it is useless to tune the system.

The best part of the article was their finding that Crest Toothpaste was better than Colgate tootpaste. Beating any Artic silver by huge margins.
However, when the toothpaste dried up after some time, the
heat conductivity dropped drastically (water evaporated) and the CPU blew up (yeah, the good AMD CPUs!).

beeblebrox · #9 Post by **beeblebrox** » Fri May 20, 2005 10:23 am

BTW: as soon as you turn on wireless, the fan kicks in after 20 seconds. Always! has nothing to do with the Pentium, it is just another thermal sensor (that is not listed).

you can put the Thinkpad on a frozen gel pad (I tested it) and the fan stays off, even with WLAN turned on.

Kenn · #10 Post by **Kenn** » Fri May 20, 2005 11:24 am

beeblebrox wrote:BTW: as soon as you turn on wireless, the fan kicks in after 20 seconds. Always! has nothing to do with the Pentium, it is just another thermal sensor (that is not listed).

you can put the Thinkpad on a frozen gel pad (I tested it) and the fan stays off, even with WLAN turned on.

Interesting, my experience over the last few months running without wlan may have confirmed that. Too bad that sensor isn't tapped, I like knowing what's going on in my machine.

I also agree for the most part re: arctic silver - the CPU thermal zone is just not critical on a properly-setup thinkpad. However, in my experience benchmarking CPUs, arctic silver has always made enough of a difference to be worth the $9 (especially on a $1500-3000 machine), and I'll take a 4-5 degree drop at load anyday

Torque · #11 Post by **Torque** » Fri May 20, 2005 12:34 pm

[quote="Kenn"][/quote]

Edited by Moderator: Snipped a lot of excessive quoting. Trim your quotes, guys.

I ran a short test today. Had my TP on the desk, connected to AC with WLAN on. After one hour of idling, the temperatures where CPU@42 degrees C and HSS@35 degrees C.
I then turned off the WLAN, at left it alone for another 1½ hours. Temps are now 39 degrees for CPU and 33 degrees for HDD.
So there's a slight difference.
The fan is still running.
I have my CPU set on "adaptive", så it ran the entire test at 600MHz. POWERPLAY was off.

Kenn · #12 Post by **Kenn** » Fri May 20, 2005 11:43 pm

Torque wrote:

Edited by Moderator: Snipped a lot of excessive quoting. Trim your quotes, guys.

I ran a short test today. Had my TP on the desk, connected to AC with WLAN on. After one hour of idling, the temperatures where CPU@42 degrees C and HSS@35 degrees C.
I then turned off the WLAN, at left it alone for another 1½ hours. Temps are now 39 degrees for CPU and 33 degrees for HDD.
So there's a slight difference.
The fan is still running.
I have my CPU set on "adaptive", så it ran the entire test at 600MHz. POWERPLAY was off.[/quote]

42 degrees for idling at 600Mhz seems a bit high, but it's not at all uncommon for 1.7-1.8Ghz - are you sure you were on adaptive for the test period? I'm currently idling at 33-34C at 600MHz, and it wasn't all that much higher before using arctic silver...

UCI2CI · #13 Post by **UCI2CI** » Sat May 21, 2005 2:29 am

i have this theory, but ofcourse i might be wrong. and i want the people who have had repair done to their laptops to chime in:

If you laptop is running hotter than it should and you have had onsite (or even depot) repairs done to your CPU/systemboard/fan, chances are that thermal paste was not properly applied.

If you have had repairs done on your thinkpad and have noticed your laptop is running hotter, (or if it doesn't for that matter) please post a reply.

Torque · #14 Post by **Torque** » Sat May 21, 2005 3:47 am

Kenn wrote:

Edited by Moderator: Snipped a lot of excessive quoting. Trim your quotes, guys.

42 degrees for idling at 600Mhz seems a bit high, but it's not at all uncommon for 1.7-1.8Ghz - are you sure you were on adaptive for the test period? I'm currently idling at 33-34C at 600MHz, and it wasn't all that much higher before using arctic silver...[/quote]

Yes, I am positive than it ran adaptive. Had Mobilemeter running in the background, and it showed 600MHz all the way through. When idling on battery, im typically at the temperatures you mention. 33-35C normally. I think the main issue is POWERPLAY. But of course, that doesn't change the comparison.

Kenn · #15 Post by **Kenn** » Sat May 21, 2005 7:57 pm

UCI2I: I had a system board replacement within the last 2 weeks and the tech used no new thermal paste at all, and I had no changes in CPU temp afterwards (and I had mobilemeter running before and after for days, specifically checking this).

But I believe Torque's running pretty hot for 42 degrees at 600MHz idle. For comparison, my system with no new thermal paste applied or redistributed after servicing, at 100% load and default voltage, topped out alternating between 42-43 degrees C. Idle is between 33-35 degrees.

UCI2CI · #16 Post by **UCI2CI** » Sun May 22, 2005 3:38 am

i find it odd that the tech would not use new thermal paste on a replaced system board. In the HMM is states that thermal paste has to be applied to the fan assembly where it contacts the CPU.

As for Torque, yes temps are kind of high. At 600 mhz, you should be running in the mid 30's whe idle. But what really matters is what temps you get at full speed under heavy load (as far as overheating is concerned). Torque, try running prime95 torture test at full speed for 15 min and see what temperatures you hit. if it hits 80, i would call IBM sometime before your warranty runs out. Below that, i wouldn't worry about it. I was hitting 85 before arctic silver, now it tops off at 70 with wireless on and the power plugged in (so powerplay inactive).

Kenn · #17 Post by **Kenn** » Sun May 22, 2005 11:28 pm

Yeah, as with any service call, I think ymmv with regards to what techs do/know/etc. In my case, I was comfortable taking apart my own system and knew to look out for temperature changes after something like this, so I didn't raise the issue...but it's always a good idea to make sure thermal paste is properly applied whenever you remove a HSF!

cchsiao · #18 Post by **cchsiao** » Fri May 27, 2005 1:06 am

Does anyone have their results for 533FSB Pentium M? I try Arctic Silver 5 on my Thinkpad (with a 770 in it), and the temperature still stays high. It's at least 45C when idle (with or without wireless on). This is after I volt down the CPU to 0.700v. When it runs at full speed under 1.116v, the temperature will reach 65C or so. I don't know if this is normal, but it looks like a 760 will be much cooler than a 770 since my friend's 760 will only reach 57-58C at full speed under 1.132v. Any ideas?

roast · #19 Post by **roast** » Thu Jun 16, 2005 11:03 am

cchsiao wrote:It's at least 45C when idle (with or without wireless on). This is after I volt down the CPU to 0.700v. When it runs at full speed under 1.116v, the temperature will reach 65C or so.

I've been getting the exact same thing as you do, without Arctic Silver applied. I don't even know if the undervolting is affecting it at all.

Scorpiontico · #20 Post by **Scorpiontico** » Fri Jun 17, 2005 12:37 am

cchsiao wrote:Does anyone have their results for 533FSB Pentium M? I try Arctic Silver 5 on my Thinkpad (with a 770 in it), and the temperature still stays high. It's at least 45C when idle (with or without wireless on). This is after I volt down the CPU to 0.700v. When it runs at full speed under 1.116v, the temperature will reach 65C or so. I don't know if this is normal, but it looks like a 760 will be much cooler than a 770 since my friend's 760 will only reach 57-58C at full speed under 1.132v. Any ideas?

could you please write or post a tutorial about how you applied the arctic silver 5 in your system?
steps and all that...

thanks alot.

Arctic Silver 5 on T42p: Temperature benchmarks inside

Arctic Silver 5 on T42p: Temperature benchmarks inside

Who is online