i7-4702 strange behaviour

Hello guys!

I want to ask a question from U, who have more experience in mobile processor world. So I have a Lenovo G510 lappy which originally have had an i5-4200M CPU (37W TDP). Then I replaced it to an i7-4700MQ (47W TDP). With a little undervolt (about -90 mV) and fixed multiplier in ThrottleStop it can run 8 threaded stress tests at 2800 - 2900 Mhz without reaches 100°C. Then I got an i7-4702MQ (37W TDP) yesterday. And here comes the strange things. I started the test with fixed 2600 Mhz and the same undervolt. As soon I started the stress test, the temperature increased to 100°C almost instantly, and the fan kicked in only at over 95°C but it reaches 100 degrees an additional couple of seconds and the machine turn itself off. So here is a weird situation: the 37W CPU heats up much more quickly than the 47W CPU at a given and even a lower frequency? Or I got a faulty 4702? Or is it a BIOS issue maybe?
I hope someone can give me some advice or explanation about this. I thought if I try a lower TDP CPU, I can reach higher frequencies with the same cooling system so I am a lil bit confused now.
Guys, have U got any help or advice about this issue?

Best regards,
Daniel

 

Are you sure the cooling is fitted properly? Is the paste done right?

BTW, a lower TDP CPU won't allow you to reach significantly higher clocks. It will just power throttle earlier.

 

Cooling is OK for sure. I tried switching the two CPUs multiple times. So I was wrong when I thought that using the CPU with the "proper" TDP rating will give me more performance?

 

I still think cooling could be to blame.The 4702 might have a bad slope on its surface resulting in uneven pressure. The other possibilities that came to mind are bad sample (high leakage) and faulty sensor.

Do you get 100C in one core or all the cores?

Unless lower TDP chips are actually binned for higher power efficiency, yes. Even ULV chips don't show systematicly better efficiency against regular voltage ones (under intermediate load), so I don't think the 4702 is anything special.

If a 4700 is strong enough to trigger significant thermal/power throttling in a system, switching to a 4702 might help. But this is no different from lowering power limit via software/firmware.

 

According to ThrottleStop, all core/thread temps moving together. The CPU's die is seems to be normal and flat, at least for the naked eye.
If I dont set a fixed multiplier, the 4702 throttles down way more often than the 4700. So it is because of the lower TDP limit. It seems to be there is difference at the temperature point where the fan kicks in to its fastest level too. The 4702 triggers fan suddenly over 92-94°C, the 4700 triggers it more sooner and gradually. It takes way much time to heating up. The 4702 heats up very quiclky, literally in a few seconds and almost instantly starts throttling down if I dont set a fixed multiplier. If I set it, It just gets hotter very fast. My intention was to find the maximum multiplier where throttling doesnt occur at all. And it is a much lower multiplier in the 4702 than in the 4700. In other words, the lower TDP rated 4702 runs much hotter on any given frequency than the higher rated 4700. What do you think, is it normal, or it might be a sensor fault or bios problem with the fans?

 

Get some good thermal paste... Like ICD or Gelid GC Extreme... Repaste using the cross method! See how it goes then.. Also clean out your fans/vents while your at it!

 

If OP has been switching back and forth between the two CPUs, thermal paste application technique should not be an issue.

 

I have MX-2 and MX-4, which are still not-so-bad pastes, at least for CPUs. Nor much worse than the Gelid that would be make them useless. And the cooling system is clean as f##@

 

They're terrible for laptops due to low mounting pressure.

As for your CPU issue, the simple fact is your 4702 is probably just a bad chip.

Comparing undervolt to undervolt is also pointless, you need to check the effective voltage. Your 4700MQ might be running at 0.9v for 2.8GHz with your undervolt where your 4702 might be at 1v with your undervolt, etc. These are ballparking numbers, but CPUs aren't equal.

As for "TDP", that is a rating. It has nothing to do with how much heat the chip will produce, or how much power it is capable of physically drawing. My "47W" 4800MQ can pull 88W if I so wish it, depending on what test I toss it through. I've set it to be allowed access to such power, mind. If it was at stock it would throttle due to being starved for power in that scenario. But just saying, the rated TDP does not really mean much. It is a guideline for how much TDP a cooler should be capable of dispersing, and it denotes the default power draw limits of the chip as set by intel, but nothing else.

 

Thanky you for your reply, I think I can see much more clearly now.
Indeed, the 4702 runs at higher voltage at the same frequency than the 4700, I checked it. I will run some tests to find out is it really a faulty chip, or not.
I have one more last question: can U recommend me a good paste for the low mounting pressure notebooks?
Thank you again for the explanation!

 

You can try a more aggressive undervolt. Your chip is not "faulty" for having a larger voltage requirement; it is simply binned worse, which is normal. It may also be a leaky chip, which is to say it expels a lot of heat for the load it is put under. But it is not "faulty".

As for paste, if you have a good heatsink fit I and a copper heatsink, your best bet is Coolaboratory Liquid Ultra. If you do not wish to use liquid metal, then IC Diamond gets my vote... just be careful how you remove it or you can scratch the heatsink or die. You need to soak it and gently wipe it off; no scrubbing. It takes a bit of time but will work. Less-long-lasting but just as good, Gelid GC Extreme. Shin Etsu X23-7921 also is pretty good. Thermal Grizzly Kryonaut has had mixed reviews but can also be very good. For a slightly less effective but extremely good value per dollar, Arctic Ceramiqué 2 is also on the table.

Sent from my OnePlus 1 using the force