Heat problems on maximum load

Hey people...

Yesterday I got an X9100 cpu (laptop is Acer Aspire 5930G)
The idle temps are quite low... they go below 40 degrees C all the way to 37 actually.
Now, the maximum load temps go as high as 85 to 90 degrees C (90 seems to be maximum) - however, the northbridge heats up at least 7 degrees C higher than that and gets to 97 degrees.
The gpu was giving me some strange temps so I cleaned up, repasted, and then the northbridge went to 105 degrees C under stress test in Orthos and of course the system shut off... dunno why it happened as the time before that, everything was working fine, and I did some measures that reduced the heat emission on the northbridge to the above mentioned temp (I put some thermal paste and a copper coin there).

But the temps seem way too high even still.
While playing ST Online, the cpu goes up to 90 degrees C regulalry, it hovers at 87 or 87, but once it reaches 90, the CPU throttles itself down for a bit and FPS drops (at least momentarily until they go back up - which seems to repeat itself constantly).

This seems to indicate that in games, the system can't tolerate 90 degrees on full load cpu-wise when gaming.

I tried to undervolt, but that went to bust.
Throttlestop is useless because ESI is locked and BIOS doesn't have an option that even displays it, let alone disables IntelSpeedStep.

I also tried CrystalCPUID.
The weird part with that is how CPU-Z reports unchanged voltages on various multipliers, however, the system froze when I gave various multipliers too low a voltage... furthermore, CPU-z seems to report a strange voltage value (0.98 for 6x multiplier and 1.2 for 11.5 multiplier - and when I tried setting the 6x multiplier to that value, the system crashed).

Now, one of those two programs is wrong.
I would wager that the CPU-z is reporting wrong values because it happened before.
But that's not the problem... when I gave the 11.5 multiplier 1.16 Volts, the system worked, but there was 0 decrease in temperatures - which seems to indicate they might not have been applied.
Someone mentioned to tick the 'IDA' slot in the program, but I can't find it (I don't think it exists).

Also, RMclock doesn't work on my system because it would seem x64 OS is incompatible with it.

Anyone has suggestions/solutions?
If I can undervolt the cpu to produce about 10 degrees lower temps, that would be tolerable.
I was also planning on installing some very thin copper shims (1mm) on the cpu, gpu and northbridge to try and get the temps down even further.

 

That should have taught you what 45W TDP means.

There is a reason why people pay *more* and buy T9900. I would recommend you to replace it with that. I have never seen my laptop exceeding 80C under any load (wPrime + Furmark).


--

 

You mean 44W.

Here's an update:
I revisited ThrottleStop and tried undervolting again.
This time it worked.
The minimum voltage I was able to achieve on 11.5 multiplier was 1.1V (reduced from the original 1.2V)... and temps in games saw reduction but only for a short bit.
It held at about 82/85 degrees, but eventually got up to 90 again.

During ORTHOS stress test, I saw temperature reductions yes, and the chipset was at about 90 degrees this time (7 degrees lower than before)... which was more manageable.

I don't get it.
I tried everything I could think of.
Unless the heatsink is not touching down on the components properly (which I don't think is the case because I checked it and it seems to be properly seated)... the undervolting should have affected CPU temps - but it's as if it had no discernible effect at all.

 

What kind of x9100 did you get? A stock x9100 should be running at x11.5 multiplier/1.2625V.

I have had my C0 stepping x9100 overclocked 24/7 at x12.5/1.25V for almost 2 years now, it highest it hits is around 80 C and it only happens during stress testing.

Of course we have different systems, but still I think you should check more carefully about the voltages, forget about CPU-Z, it often reads wrong info. What throttlestop reads is usually correct, I think you got the readings from CPU-Z and throttlestop mixed up.

I have seen a lot people swapping out their C2D for a x9100, I have not heard a single person had problems running them, you should at least be able to run it at 11.5 multiplier with some undervolting if your cooling system is not sufficient.

 

This is a clear indication that your cooling system is inadequate. It can't discharge the heat build-up properly and you see the temps gradually increasing around 90s. Also don't forget: If you play games, your GPU contributes to that heat build-up as well.

BTW, don't expect voltage modification (1.2V -> 1.1V) to drop your temps from 100C to 45C. Looks like your machine can't handle 44W TDP overall.


--

 

It's entirely possible my cooling is not great (or at least it's sufficient enough to run it just under overheating as it did the first time around), however, keep in mind that another person who has an identical laptop and didn't change his heatsink managed to get his X9100 running at 70 degrees under load - although, this was likely attributed to a cooling pad he has, which means that without it, his load temps are actually 80 degrees (which is still better than mine).

Regardless of that, reducing the Voltage from 1.2 to 1.1 should have produced a noticeable effect of at least 5 to 10 degrees - and in the end it didn't for some reason (i know the gpu plays a part in games, but not to this kind of an extent because at maximum, my temps before on P7350 for both the gpu and cpu were about 70 degrees for both the cpu and gpu - which means an 80 degrees or 85 should have been attainable with X9100 when taking the higher wattage into account).

Reducing the voltage by a lesser factor on my P7350 produced 10 degrees lower temps - but on the X9100 it's as if it had no effect regardless of how low I went (1.1V seems to be the lowest I can go as 1.088 caused ORTHOS to force stop).

I'll definitely have to check if the heatsink is properly seated.

If I don't stabilize the maximum load temps in the next 48 hours or so, I'll likely return the cpu.

Any other suggestions would be appreciated.

lidowxx
ThrottleStop seems to indicate that without voltage modifications, the 11.5 multiplier runs at 1.2V (just as cpu-z reports).
My X9100 is an ES.

 

Your old p7350 maxed at 70 C already? Dude, I have to say that acer cooling system is really really bad. P7350 is rated at 25w, same as the p8400 I replaced in my g50vt, most 25w C2D cpus run icy cool from what I have seen, mine used to peak at 55 C. Is your x9100 a C0 stepping or E0 stepping?

If undervolting can't lower the temp off the 90 zone, then you might not want to keep it then. Did you properly repaste your CPU with quality thermal grease?

 

Revision says C0.

I think I did properly repaste the cpu. I used a small grain of rice (smaller than that actually) on the center of the cpu chip and allowed the heatsink to spread it. I used Zalman STG2 thermal paste - which seems to be quite good from my previous uses of the stuff.

Lol... my P7350 already peaked at 85 degrees C when I first got it under full load along with the gpu.
I was able to reduce the overall temps by 10 to 15 degrees by undervolting the gpu (and employing the above mentioned cooling paste).

I'll have to recheck the heatsink a bit later today to see if there's anything else to be done.
Aside from that, I will try putting copper shims on CPU, GPU and northbridge of 1mm thickness (and 2x2 cm large) to see if that will make any difference... if it doesn't, then I will probably have to return the cpu.

 

If you told this to me earlier I would definitely advice you not to buy a X9100. It is more than apparent that your cooling is very weak.

Now I think it may not be able to handle T9900 easily as well. Your easiest bet for performance increase is P9600/P9700. If you have the chance return it immediately and look for P9500+.

I am pretty sure it has a tiny aluminum made heatsink, correct? And probably with a low-airflow fan (to make things even worse)...


--

 

Here's the cooling system:



The X9100 is definitely an overkill for this cooling.

 

Ouch, they can hardly be called heatsinks, just some thin layers of aluminum. Skimpy stuffs like that will do no good, no wonder the temp is so bad.

X9100 and T9900 might be too much for that laptop, better to get a p9600 or p9700, or even better never buy an acer again.

 

Lol... better question is how the heck did the guy who has an identical Acer managed to get his X9100 running at 70 (or rather 80 degrees C without the cooling pad) on load - he didn't change his heatsink btw.

Something is amiss... People changed to a T9800 and T9900 in that laptop and temps were fine for them under the same heatsink (even after having comparable temps to mine with P7350 before the upgrade).
It begs to question if mine is seating properly or if I'm missing something.

 

Maybe you can keep the backpanel off and sit it on a cooler? It should give you a drastic drop of temp, of course it's only feasible if your laptop is always stationary, or almost always stationary.

 

lol...
Well, it is feasible because the laptop is for the most part stationary (although I do carry it around with me here and there - and from what I can tell, doing things in Photoshop and 3dsMax doesn't raise the temperature beyond 55/60 degrees C (pretty much similar to my P7350)- the work I do in those programs for the most part isn't long enough to cause drastic rise in temps anyway - and if I do CPU intensive work in them at such lengths, then the laptop is always stationary)... but as I said, I WILL attempt a few alternative options such as checking the heatsink and adding the copper shims before I make a decision if I want to return or keep the cpu.
Doing 'regular' stuff on the laptop doesn't raise the temps too much... watching a movie or browsing the internet, they stay in the 45/50 degrees C zone (again, essentially the same what I was getting on P7350 if not slightly higher) ... it's gaming that seems to cause temps to go up long term.


As I said, if others were able to get reasonable temps in this laptop with similar/same cpu's, then I likely missed something in mine (unless mine came with an especially poor cooling compared to theirs, but that's unlikely seeing how our temps before upgrades were comparable with same cpu's - mine were even lower).

 

I think I'll have the copper shims sometime tomorrow... I'll make sure to post my progress - I'll make the cpu bow down to my lower temperature thresholds - if I can.

Aside from that... I was contemplating about possibly exchanging that internal fan for something better.
If I'm not mistaken, other posters here indicated that manufacturers often put very cheap stuff inside of not so good quality.
Others were able to replace the internal fans with custom ones and got better cooling in the process.

Any recommendations on that front?

 

Check Triturbo's cooling overhaul topic for some tricks.

 

while the E0 stepping is better than the C0, the x9100 by itself is a good CPU. Maybe that other guy that you're talking about has E0 which runs cooler by design.

dropping from 1.2V to 1.1V should definitely show lower temps, and those should be up to 10 deg C.

I would advice against any copper shims between the chip and the heat sink. Those thin aluminum plates that are over the chips are fine because that's not a device to dissipate heat but only to transfer it to the heat sink, which would be where the fan is. So why not remove any copper mods that you may have done and try to push down the plate so that it has minimum distance to the chips. Then use only thermal paste.

also I'm assuming your bus is 266MHz, so the 11.5 multiplier that you're talking about would equal 3GHz ? And you run that at 1.1000 V ? Sounds normal to me.

I run 2.8GHz on Dual IDA at 1.075V. My temps would max out at about 85 deg C (with LinX after 5 min at above room temp) but that's with my fan running on max speed that (I guess) even the BIOS is not using due to possible fan damage, lol. But we have that utility that can control fan speed, so thank for it I can do that. With Orthos I would max out at about 80 deg C, but it's not as loading as LinX.

speaking of IDA I think none of the extreme C2D CPUs have it, and frankly dont need it.

on the other hand your fan may not be as good as you want it to be ...

 

can you tell me how did you set your X9100 to run on 2.8Ghz?
Did you use Throttlestop or something else, and what were the settings in the program?

Yes, my bus is 266 and the 11.5 multiplier runs at 1.1V.
I agree that the temps should have definitely gone down... it's just odd.

 

Set the multiplier to 10.5.

Default: 11.5 x 266 = 3.06GHz (equivalent to T9900).
With 10.5 multiplier, you get 10.5 x 266 = 2.8GHz. (equivalent to T9600)
You may also try 11.0 which will give 11 x 266 = 2.93GHz (equivalent to T9800).


--

 

I dont have x9100.

but yes, use throttlestop

 

Thank you.

Ok... here's a bit of an update.
I went inside the laptop again, cleaned out the paste and did another application.

Instead of putting a small blop on the chips and allow the heatsink to spread them, I decided to spread it out myself across the chips (gpu, cpu and northbridge).

Currently running ORTHOS stress test (Blend).
26 mins into the test, the temps seem to be holding at 70 to 71 degrees C on the CPU.
Northbridge is at 74 degrees C.
GPU at 63 degrees C.

The temps climbed close to 80 degrees at one fluke point, but I propped the laptops back legs on small stands which raised it's backside and then the temps dropped back to 70/71 (it's possible that 80 degrees was nothing more than a fluke before the cpu 'settled' back on 70/71)
I still need to do a test in games... but thus far, current results are... encouraging.

Oh and, the CPU is running at 11.5 multiplier at 1.1V (3.06Ghz)

EDIT:
I ran a Furmark burn-in test at 1280x 768 and ran Orthos blend stress test alongside it for about 10 mins.
Temperatures on the cpu were consistent at 85 degrees C, while the gpu was at 80 degrees and the northbridge was holding at 90 degrees C.
Oh and bear in mind that my GPU is OC-ed by 20% (from stock 9600m gt levels) via Nvidia Control Panel.

I must say that this is a lot better.
It would appear that my temps have stabilized (and are not going up after prolonged periods of time under stress) on the cpu, gpu and NB to manageable levels.

I think I still might put the copper shims on the cpu and gpu though to see if I can further drop the temps down.

These levels are nice and stable, but of course lower temps would be even better.


Off to try ST Online now to see how it behaves.

EDIT: ST Online plays just fine on 11.5 multiplier and 1.1V.
CPU was at 70/72 degrees C, gpu was 70 degrees (this is after long term playing - which definitely beats 90 degrees on the cpu and 85 degrees on the gpu I was getting last night).

 

Well if you remove your heatsink and you have almost none paste where the core makes contact, then you don't need copper shims. The only place you'll probably have to put one is on the NB (if it doesn't make good contact). Putting shims may damage your CPU/GPU/whatever if you over-tight it. I do have copper shim between my GPU's core and the heatsink, only because I can't lower it down and there is quite of a gap.

 

of course that one can damage the chips if you over-tight the heatsink with the copper shims... but I'm very careful in that regard.


May I ask by how much did you temperatures drop after doing the copper shim on the gpu core?

 

Glad to see you worked that out, 70/72 C are already very good temp for gaming, and 85 C for furmark and orthos stress test is perfectly acceptable. In real world usage, you may never bring the cpu temp to that level.

I agree with triburbo that you probably don't want to add copper shims, they are only needed if the gap between the heatsink and CPU core is too wide for thermal paste to fill in. If the heatsink and CPU core can make proper contact through thermal paste, the added copper shims will likely give you no benefit at all.

I would personally try to get an effective cooler that works great for your laptop first, see if it could help lowering the temp to your satisfactory level. If it does, voila, you are done. If it doesn't, then you could resort to copper shims, which would probably be the last conventional mean to bring down the temp.

 

85C on cpu and 80C on gpu was for Furmark and Orthos working together at the same time... separately, furmark produces 70 C on the gpu, while orthos does 70-72 on the cpu.

Yeah, the temps are more than ok now, but I'm thinking that a copper shim is the way to go if I do decide to try and reduce the temps further.
I already had a laptop cooler and to be honest, I sold it because I almost never used it - and because I can move from country to country... it's impractical.

 

lol, I should have asked how you put your thermal compound ...

I dont know who came up with the one drop thing but I'm highly against it, as the chance of having area of the chip not covered is big, thus the poor results. Better cover all the chip and squeeze the heat sink tight Never had problems with that method.

 

When I did it, I had HD3650, the drop was whooping 12ºC. BTW that's going from crappy thermal pad to paste-copper shim-paste setup, and I was using MX-2 back then, so subtract another 2~3ºC with MX-4

 

Got it.
Btw...
I'm confused regarding ThrottleStop.

If you recall, I mentioned that my EIST option in ThrottleStop was greyed out (which means it was locked by Acer - my Bios doesn't show any options connected to that) and the creator of ThrottleStop program seems to indicate that in order for ThrottleStop to affect change... Intel Speed Step needs to be disabled in BIOS - which in turns allows for manipulation of EIST.

I'm using Orhtos right now without Throttlestop and temps under cpu are essentially the same with what I was getting when ThrottleStop was running.
It's odd actually.
I tried also raising the multipliers and playing with the voltages, and the system obviously either crashed or ORTHOS crashed because there was insufficient power coming to the cpu.

So, why are my temps essentially identical with 11.5 multiplier running on 1.1V instead of 1.2V?
The difference it would appear is only a few degrees (2 or 3 at most), and without ThrottleStop it seems to cycle back to 70-72 on it's own most of the time.

Is ThrottleStop actually controlling this in the end or what?
Cpu-Z confirms different/lower voltages on 11.5 multiplier, but tems say a bit different story...
And apparently EIST needs to be ticked off in order for ThrottleStop to manipulate them.
So which info is correct?
On one side it seems work, but on the other...

 

you need EIST unlocked for one reason only - to run Dual IDA.

I do believe all extreme C2D CPUs come without IDA because you can increase multiplier to whatever suits your needs, so whether EIST is locked or not is of no interest to you.

ThrottleStop is a wonderful program, but seems to me that you need to read the throttlestop thread to see what each setting mean and how to use it. Unclewebb did a nice job explaining everything there.

 

Ah... so this actually explains why my X9100 correlates with lower voltages (or at least the readings from Orthos and CPu-Z do).
EIST as such in that case I don't really require for undervolting.

 

Hey guys... another problem popped up.

It would appear, that regardless of the Voltages I put into ThrottleStop, they don't seem to affect the temperatures.
They are the same with or without the program running.

What is strange though is that if I apply a voltage of say 1.088V on the 11.5 multiplier... ORTHOS will produce an error.
Voltages lower than that produce BSOD and/or freeze the OS (requiring a restart).

So, why do the low voltage changes correlate with OS freeze and BSOD, but don't affect temperatures?

I tried running ST:Online with and without ThrottleStop, and results are the same temperature-wise.

Any of you heard of something like this happening?
Any idea, WHY it's happening?

EDIT: Or it simply might be that lowering the voltage produces a marginal effect in my particular case which doesn't reflect on the temperatures.

 

Maybe because you shoved a 44W TDP CPU into a laptop that had a marginal cooling solution that was barely adequate to cool the previous 25W TDP CPU.

That might have something to do with it.

Laptop fan speeds can be controlled by the temperature of your CPU. Fans can kick in and run at a higher speed for a while and then slow down after the CPU has cooled down to the required level and this cycle will go on indefinitely. When you are running extra voltage, if this causes your CPU to heat up more, it might trigger your fan to run faster which cools your CPU down. You can't do any sort of testing on a laptop unless you can control the fan speed and fix it at a constant speed. Without that, your temperature test results are meaningless.

ThrottleStop does work on your X9100 and it seems to work for all the other happy X9100 owners too.

 

Actually, the odd thing is that I was getting 70 degrees C under load with an UV-ed P7350.
75 degrees in Orthos with the X9100 is perfectly acceptable for a CPU of this magnitude at stock clocks (I mean, the difference compared to P7350 is marginal)
I seem to recall cases where people's laptops heated to about 95 degrees C and they considered it 'fine' (granted, mine was too up until recently, though that was my fault for improperly applying the thermal compound).

Yeah, I'd have to conduct proper testing by locking the cpu fan speed to max at all times and see what the results are with 1.2V and 1.1V.
Current fluctuations are unreliable for a proper test as you said.

Which software should I use to lock fan speed to max?
I found reference to SpeedFan as a candidate but upon opening it I don't have the option of modifying the fan speeds.
RivaTuner on the other hand has no option for fan speeds (probably due to WHQL drivers - and I need them).
Any recommendations?

 

If you want to drop temps further with your X9100 then consider a swappout for a E0 stepping X9100. RickiBerlin did a comprehensive comparison titled temperature comparison between X9100 "QHBQ" and "SLB48" on his HDX9000. He's currently running his at 4.0Ghz where it runs 26 degrees cooler than the C0 at full load and is slightly faster. See his comments.

 

^^ Indeed. That's why I asked you which stepping you got, Deks

 

Well, there are no X9100's with E0 stepping. At least none that I was able to find, and one or two that I came across were ridiculously priced.

Anyway... I'm gonna do a copper shim mod on my CPU and GPU over the next few days.
That should hopefully lower the temps further into the 60-ies.

 

E0 X9100 at 4.0GHz and below 80 deg C ?!?!

D-d-d-d-d-amn

EDIT: but then you call that HP a laptop, lol. Sorry, talking too much about that link that nando4 posted

 

Where did you find it?

Well, good luck and be careful

And it's running on PM965 like ours

 

The user no-tech is running his at X9100-E0 at 4.15Ghz, adding a GTX280M MXM module plus grafted in a LCD module and routed it back to the LCD. The latter because they couldn't manage to mod the vbios to output to the internal LCD. See vid.

They are also trying to get quad-core multiplier unlocked QX9200 ES or QX9300 working as well. It's an impressive set of upgrades to those avid 20" HDX9000 (Dragon) fans, a system like no other.

 

how ?

did they modified their BIOS to accept that CPU? I'm sure if I plug it into mine it would not work :|

sorry OP, it's still about x9100 discussion though

 

Pin-mod for FSB "down-clocking" (since it's 1066MHz FSB CPU), it would boot @2,3~2,4GHz and then with ThrottleStop - UP AND AWAY If I manage to find E0 one, I'm SOOOOO going to try it! The temp difference would worth every cent. I hope that it would boot, of course

BTW No BIOS mod is required since it's Penryn as the ones we're using at the moment

@nando4 - The quad is VERY interesting project, but I guess they'll need completely new BIOS for the purpose.

 

Well, the C0 stepping one can be cooled down to nice levels... albeit that doesn't allow for a proper OC like the E0 can apparently give.

Anyone can find reasonably priced E0?