Thinkpads Forum

If anyone here with the 45w processor in a T420, I need to know how much watt can the thermal solution disipate before throttling down the processor.

Any 45w processor is able to keep its base clock easily while using less than 35w while stressed in non-AVX workloads, so temperature measurements are not as important as the watt measurements.

You may need to stress the CPU with a relatively new version of Prime95, plus, stress the iGPU maybe using FurMark.

Use http://www.hwinfo.com/ to get "CPU Package Power".

Please also let me know if you have nvidia or intel fan installed, and please disable nvidia while testing, or, just make sure that the iGPU is the one getting stressed, and nvidia is not running.

And what is the temp and the fan speed? (prefer max normal fan speed of '7' if possible)

Thanks in advance.

I have a _35w_ quad-core in my T430 and under Prime95 small FFTs one of the cores goes up to 96C, the CPU is eating up to 35w, all measured un hwinfo real-time. The fan is running at 64 (>5000rpm) (tpfancontrol). The throttling is minimal, all the cores run pretty stable @2.9GHz and from time to time 1 or 2 cores throttle down to 2.8GHz; I'd prefer some more throttling and 90C as maximum (as does my T420s i5-2520m) though.

AIX wrote:I have a _35w_ quad-core in my T430 and under Prime95 small FFTs one of the cores goes up to 96C, the CPU is eating up to 35w, all measured un hwinfo real-time. The fan is running at 64 (>5000rpm) (tpfancontrol). The throttling is minimal, all the cores run pretty stable @2.9GHz and from time to time 1 or 2 cores throttle down to 2.8GHz; I'd prefer some more throttling and 90C as maximum (as does my T420s i5-2520m) though.

Thank you for the info, this means that 45w might throttle badly, considering you are even using the nvidia heatsink, which is supposed to be better than the integrated.

your signature says that you t430 have nvidia, during this test, what was the nvidia state ? , disabled via bios, disabled by optimus, enable but idle, or something else ?

It's not disabled so I can say it's idling (the Optimus icon is grey, not colored). I'll post a hwinfo log soon, if you're interested. I think 45w is a bit too much for T420/T430's cooling.

AIX wrote:It's not disabled so I can say it's idling (the Optimus icon is grey, not colored). I'll post a hwinfo log soon, if you're interested. I think 45w is a bit too much for T420/T430's cooling.

AFAIK, with the optimus icon gray, the nvidia is temporarily in 'off' state, and not producing any heat.

An extra hwinfo log would be a nice extra to look at some of the other stats displayed.

Thanks.

Yes, you're right, it's more like stand-by state, not idling.

I've run two tests last evening, one Prime95 small FFTs, the other combined - Prime95 small FFTs + Furmark; as you can see, there is throttling both on CPU and nVidia GPU on combined tests.


Prime95 small FFTs
Prime95 small FFTs + Furmark

Thanks for all the info.

I noticed that when you started stressing the GPU in the 2nd file, the battery started to go down, and then when the cpu stress stopped, the battery started charging again.

what power wattage are you using, 90w I suppose ?

to recap, in case someone else is wondering:

3632qm with fan at level 64 (5600rpm)
When stressing the CPU only (3632qm), it reaches maximum multi-core turbo of 29x, at ~34w (temp = 95c), and probably able to hit 35w.
When stressing the CPU+GPU (nvidia), the CPU throtles to 24x~25x, at 27w (temp = 101c), and the GPU is at 92c, and I am guessing that it is throttling a bit too.

I didn't notice the battery behavior, indeed, it goes down once the NVidia GPU is stressed, even the CPU is not, but it seems that it's charged back from times to times during the stress test. I've made another log, battery is at 97%, and started to go down after the Furmark was fired up, but at 96% is charged back in small bursts. I'm using a 90w power adapter and yes, the NVidia GPU is definitely throttling - see the GPU clock. Gaming is not that bad though, I think Prime95 & Furmark put more stress on CPU and GPU than a regular game, but I did the CPU upgrade recently and didn't have the time to test it thoroughly, T420s is still my daily runner.


Furmark + Prime95 small FFTs

AIX wrote: I think Prime95 & Furmark put more stress on CPU and GPU than a regular game

That is true, stress testing like this is nothing like real world application of gaming, it is very rare to find something that does similar load to both CPU and GPU at the same time.

Thanks for the info, I got much more details than what I was originally hoping for

--

This is off-topic, .. there is a mod in the forum here that can allow you to use the T420 keyboard on the T430, if that is of interest to you.

They're working on it.

http://forum.thinkpads.com/viewtopic.php?f=69&t=120776

Physically you might have to cut off some tabs, it plugs in but some keys are wrong or don't work - now there's a firmware mod to fix that.

hhhd1 wrote:This is off-topic, .. there is a mod in the forum here that can allow you to use the T420 keyboard on the T430, if that is of interest to you.

Yeah, I saw it, but the T430 Chiclet keyboard isn't bothering me too much, is the size that makes me keep my T420s as my daily runner, it just feels right, almost perfect. It's my favorite, as both 600X and T41 were. I'm keeping the T430 especially for gaming and as a backup. Unfortunately Lenovo has yet o put a real discrete GPU in their Thinkpad line, but at least they started to do better in displays area. I'll probably wait until Mass Effect Andromeda is released before I'll buy another Thinkpad, hoping for an adequate GPU for this game (and implicitly for Fallout 4, Witcher 3).

So, are you gonna put a 45W CPU in your laptop? Are there options for a 35W Sandy Bridge?

AIX wrote:So, are you gonna put a 45W CPU in your laptop? Are there options for a 35W Sandy Bridge?

My main laptop now is a T520 dual core, but I am looking to get either a T430 or a T530, the possibility of upgrading to a quad core is a must, even if it will not be immediately.

I am currently debating whether to get T530, and have 45w option, or just stick with the cheaper/lighter/smaller T430, and try to find a good 35w quad.

I am considering the fact that 45w quads are cheaper and a bit faster, and the T530 maybe a bit cooler.
but the 15 inch is also heavier.

I am now looking for good deals on either though, whichever looks good.

AIX wrote:

I wonder if you can get me one more detail ,
I want to know about non-AVX load, specifically, how much watts does the cpu use (CPU package power) to maintain 29x multiplier, this simulates more of usual work and games.

I am guessing it should be in the high 20s .., and hoping it is less.

You need an old version of prime95 to test this, like this one for example, it is an old version before they started using AVX in stress testing:
http://mersenneforum.org/gimps/p64v2511.zip (i think this is the official site for prime95)

Thanks!

Sure, I'll check it tonight.

here you go: t430_i7-3632qm_prime95v25.11b2_noAVX

Thanks for the details, much appreciated.

I think for now, I decided on getting a T430 and not a T530, i think seeing those numbers, a 45w might work with some throttling ...

what worries me about 45w processor is that it can turbo for 30 seconds or more at 50watts, which will push temp high quickly, and will make it also throttle down allot and quickly ...

maybe if 45w with disabled turbo boost should be fine.

I just install i7-2820qm to my T420. Waiting for 90W adapter to do more tests. If you still interested in something, ask me

Python.p wrote:I just install i7-2820qm to my T420. Waiting for 90W adapter to do more tests. If you still interested in something, ask me

Well,

you can get hwinfo https://www.hwinfo.com/download.php
then you can run "sensors only" , then click on the button to start logging (probably the 2nd button from the left)
this will create a csv file with all temps and power used and fan speeds.

now, run the prime95 benchmark, just use the latest version.

keep that for 10 minutes or so, and the stop prime, and end the csv file.


then, let the computer idle, and then repeat all the above (and create a new CSV file) with an older version of prime without AVX, http://mersenneforum.org/gimps/p64v2511.zip


Thanks!.

Hi,

ambient temperature was 24°C

https://db.tt/NcSy7yFB
https://db.tt/O0BZNumZ

i have a lot of ideas to modify the cooling system, but now i dont have time for that.

Python.p wrote:Hi,

ambient temperature was 24°C

https://db.tt/NcSy7yFB
https://db.tt/O0BZNumZ

i have a lot of ideas to modify the cooling system, but now i dont have time for that.

The result is quite great, much better than I expected tbh, it is able to keep 42watts for extended period of time.

Which fan FRU do you have ?

if you do not want to re-open the laptop to find out, do you have pictures of it, or, can you remember how many heat pipe does it have ?

it is either:
1. integrated with 1 heatpipe
2. integrated with 2 heatpipes
3. discrete with 1 heatpipe for cpu + 1 heatpipe for gpu (2 total)
4. discrete with 2 heatpipes for cpu + 1 heatpipe for gpu (3 total)

And, what ideas do you have to modify the cooling system ?

Thanks

I have option 2. Integrated with 2 HP.

I want to use option 4. with 3HP and bend one from discrete GPU and solder it together -> better heat transportation (i hope so).

Or make same pasive colling to absorb some heat to keep turbo for a longer time.

And last one, use Coollaboratory Liquid PRO. In these case we can be able to lower temp for 5°C or more. That is my experience with CLP.

Now i use 3 core from 4. 75% of heat and 150% of i5-2540m performace

Python.p wrote: I want to use option 4. with 3HP and bend one from discrete GPU and solder it together -> better heat transportation (i hope so).

Or make same pasive colling to absorb some heat to keep turbo for a longer time.

And last one, use Coollaboratory Liquid PRO. In these case we can be able to lower temp for 5°C or more. That is my experience with CLP.

Thanks for all the info

if you do any of those methods, please do post the results

how are you disabling one core btw ?

I see the fan is not at full speed, I'm wondering if the throttling would be a bit smaller with the fan at level 64 / 5600rpm or not.

I goning to try. ASAP.