Thinkpads Forum

I'm trying to find out how much difference there is in Penryn (T9XXX, T8XXX) and Merom (T7XXX) CPU's temperature/heat dissipation and whether they can operate cool without fan.

Could people with any CPU in their T61 post their idle temperature with TPFanControl and fan off. It is better to let idle temp settle for hour or so to be accurate.

If you do, please post all these:
-CPU idle temperature
-ambient temperature
-which CPU
-integrated or discrete GPU
-OS
-lowest voltage/vid (not necessary)

If there are people who have had both CPU's, was there difference in temperature and did Penryn CPU feel cooler too?

Thanks.

I only let my T61 (integrated graphics) for half an hour idle but the temperature seemed stable... CPU temperature certainly was.
ambient: about 24C
CPU: 55/56C (T7300, both cores running, set at the minimum voltage: 0.9V)
PWR: 49C
PCI: 45C

I'm using it fanless (at a variable voltage going as high as 0.9625) right now and, after a few minutes, the CPU temperature seems to have levelled out at 57/58 with the "pwr" at 50 which I what I'm most uncomfortable with. Of course, I haven't put any heavy load on the laptop or the temperature would shoot up.

EDIT:back on BIOS control, the fan dropped every temperature but the CPU's at or under 42. "bus" was kinda hot fanless as well but it's really "pwr" that dropped the most (-11C so far).

Thanks for the input.

I have T61 with integrated graphics too with T7100. It gives similar idle temp results:
~57C with 0.9V
~55C with 0.85V
...in 25C ambient.

Seems like these Merom temps are normal and there's no bias. Now could at least someone with Penryn CPU post idle temperature so we can see how much cooler it really is in T61?

0.85V? I assume this is a T7100-only feature. It sports the lowest voltage of the regular CPUs in the specs. Or is there a way to get voltage lower than 0.9V on the other CPUs?

HFat wrote:0.85V? I assume this is a T7100-only feature. It sports the lowest voltage of the regular CPUs in the specs. Or is there a way to get voltage lower than 0.9V on the other CPUs?

Lowest voltages usually vary between CPU units of same model, because others are more stable. Not sure if 0.85V is T7100 only. I have read other people always having 0.85V or 0.9V in Merom CPU's. So 0.9V is probably you lowest voltage, unless you want to do some CPU pin modding.

That would be odd. I had always sought these values were model-dependent.

My CPU could evidently run a good bit lower than 0.9V at slow speeds because it's stable at 0.9V up to 1.6Ghz. But perhaps most Meroms can go lower...

I'm using RMClock by the way.

HFat:

in automatic mode (which is set by default), RMClock recognizes my T7500 as a desktop processor. By setting the processor type manualy to "mobile" (under advanced settings), the lowest voltage that can be set becomes 0.8500 V.

RMClock does a wonderful job in keeping the processor temperatures low, its "Performance on demand" setting however leads to a BSOD every once in a while (at least in my case).

George

I already set it to mobile.

As to your bluescreens, I thought "performance on demand" was just a name. You can set up this mode however you like in any case. Are you sure you didn't set some voltages too low? Set them as low as possible and you will get instability...

"Performance on demand" dynamically changes the speed multiplier according to the real time load. Higher load causes the processor to run at a higher frequency. On the other hand, if the load goes under the specified threshold, a lower frequency is set so that the power consumption and heat dissipation is kept low. The other profiles just set the processor at the chosen speed and leave it there, irrespective of load (whether high or low).

The chosen voltage levels are all fine and for each multiplier setting a couple of steps above what would actually have been the lowest possible. The BSOD's seem to occur when there are frequent changes of the p-states. They never occur when the processor is kept at a set frequency, irrespective of the chosen profile (eg. they never occur when performance on demand is set, but no load is applied causing the processor to constantly run at the lowest frequency. The same applies to the other extreme, i.e. applying full load so that the highest frequency is kept constantly or when a static profile like Power saving is chosen).

Unfortunately, I haven't been able to find the root of the cause. It seems though that longer timer intervalls for applying voltage levels between p-state changes mean less frequent BSODs. Unfortunately, Murphy's law has often enough been proven right in that however rare a BSOD happens, it will always be at the worst possible moment...

George

PS.: Apologies to all for the off-topic message, I thought it would be of interest to others as well.

giorgosvam wrote:"Performance on demand" dynamically changes the speed multiplier according to the real time load.

I thought all modes did that but I just checked and they don't. You're right.

giorgosvam wrote:The BSOD's seem to occur when there are frequent changes of the p-states. They never occur when the processor is kept at a set frequency, irrespective of the chosen profile (eg. they never occur when performance on demand is set, but no load is applied causing the processor to constantly run at the lowest frequency. The same applies to the other extreme, i.e. applying full load so that the highest frequency is kept constantly or when a static profile like Power saving is chosen).

I assume you've already tried different transition settings.
I would also try upping the voltages a bit more, especially the lowest voltages. My lowest voltage is a lot higher than it would need to be (because I can't go any lower) and my system is stable. On another CPU, I needed to up the lowest voltage a good bit higher than the fixed-frequency tests would have indicated in order to get rid of occasional crashes. YMMV.

Anymore results?

Not sure these results in such a small sample will mean anything at all. They will be so dependent on the environment the chip is in (computer, motherboard, assembly, software, dust, dirt, etc). A better measure might just be the power dissipation spec of the various Merom and Penryn models from Intel.