New i5 Macbook Air "incinerating" running Intel Burn Test in Bootcamp. Why?

Very concerned about the heat of the New 13" Air in Bootcamp.
Just tried two i5's and got identical results as follows:
Running Windows 7 Ultimate in Bootcamp,
installed Intel Burn Test and Real Temp,
both latest versions.
Started Real Temp and got about 55 degrees Centigrade idle,
which is high to begin with.
Started Intel Burn Test with the following settings:
Maximum Stress, maximum memory, Auto Threads (or 4),
10 test cycles (or 5 or less).
Real Temp set to default TJ Max of 100 degrees Centigrade.
Start the test, and temps very quickly get to TJ Max of 100 degrees!!
When I go into Real Temp, and adjust TJ Max manually to 110 or
even 120, the temps get up to those maximums very quickly,
yes up to 120 degrees Centigrade!!!
At that point, needless to say, the case of the Air is very hot!!!

I only allowed it to stay at the higher temps for a very brief period,
but at the default 100 degree TJ Max setting, the Intel Burn Test
rebooted without completing 5 test loops at one point.
Other times, it did finish the test without error.

Since I tried several Airs, this is likely going to be the case with any current generation Air. (Please try this and report any different results).

Can anyone provide an explanation for why the Air would run
at the maximum temperature that TJ Max is set to, anywhere
from 100 degrees up to 120 degrees Centigrade while being
stress tested by Intel Burn Test?

The 80 degree range has been (to this point) the maximum temperature I have seen any CPU operate at when stress tested in this manner.

I consider this to be far too hot for such operation in Windows.

 

Have you tried Lubbos?

 

55 isn't hot at all.

the max temp Intel says these processors can run at is 105. Do not let it go over that. Fan control can be a bit off in Windows if you hit it too hard too fast... try a fan control program like Lubbo's which was mentioned and manually make your fans spin up to help control the heat. I try to always keep mine under 90.

 

Do you have a screen shot of the 120c i5? They're supposed to shut off at 100c... ?

 

OK, first off, no... I didn't bother to take screen shots.
You can do the same yourself.
Whatever you set in Real Temp as TJ Max, the Air will climb
to during the Intel Burn Test, up to 120 degrees. Real Temp
will not allow for TJ Max to be set above 120 degrees, for obvious
reasons!

Simply duplicate the test as I ran it and you can see yours go from 100
all the way up to 120 degrees centigrade.
Use default Windows 7 setup with no 3rd party fan utilities.

For the chance that I will keep this particular Air beyond 14 days, I do
not wish to run it beyond 100 degrees.

I think these Airs will shutoff at 100 degrees if you are running OS X, but apparently not in Windows. Test for yourself, again, I am not running mine that hot any more, even if it is going back for refund/exchange.

Why should 3rd party fan utilities have to be used to keep this machine
below 100 degrees centigrade? I have used many a notebook and desktop
with default management without issue. Is it the EFI BIOS at work here?

The i5 in the Air is an Intel i5-2557M. It has many similarities, with slightly
lower wattage than a low-voltage Xeon that I am currently using in a new
storage server I am setting up, an Intel Xeon E3-1220L. It does not have integrated graphics, but, as I said, is a very similar 32nm chip otherwise.
Here is a link to a comparison of the 2 CPU's:
Compare Intel® Products

My Xeon idles in the low 20's, and does not get out of the 60's actively cooled. Passively cooled, it is a pinch higher, but nothing, nothing remotely
close to 100 degrees.

Also, for those saying that TJ Max for these Air CPU's is 105, please note
in my link that the i5-2557M is listed as 100 degrees maximum!

I am seeing posts that these CPU's are ok up to 105, but no idea why that is being said?

Besides, I have stress tested many dozens of CPU's and have never had anything hit close to 100 degrees.

I am not saying that Apple does not have these CPU's adequately cooled,
but I am saying that they are too hot.

 

Because you're using third party software (windows)?

 

they shouldn't be required to do that... some people just like more control, or keep the temps lower than what Apple wants.

You may want to try resetting your SMC, because its not supposed to break 100 on its own... unless the software you are using is interfering with the fan control somehow.

larger units always run their CPUs cooler... its much easier to keep it cooler. This has nothing to do with Apple... Dell, HP, whatever... their laptops always run hotter than their desktops. Thermal limits even from Intel list the Mobile CPUs able to go to higher temps than Desktop/Server ones.

yeah its 100.. I keep forgetting Intel went to 100 on these new ones. Core Duo, Core 2 Duos and the First Gen i3/5/7 mobiles were all 105.

You've stressed tested many mobile CPUs? Most laptops have more room and are more aggressive in their fans. Apple pushes the line sometimes... but your fans are really supposed to be cranking up high once its over 90, so I'm not sure how you can hit those high of temps unless your SMC needs to be reset, your Windows software is messing with the SMC, or there is just something wrong with the machine...

 

have you tried not running the intel burn test? what happens then?

 

***Giggle***

But seriously, outside of applications that are meant to stress your system to the max, what temps are you experiencing during day to day normal use?

 

You couldn't resist that reply?
Seriously, I only run Intel Burn Test when I am testing out a new system or have made a system change, such as an overclock, which is obviously not the case here. Once it has passed, I never run it again.

Mid to high 50's idle, up to high 80's streaming 720p video, which I wouldn't consider all that stressful. Again, the stress testing I was doing was a one-time thing, just to validate the Air for use with Windows 7.

 

Tried to install Lubbos, but first, it complained about a missing .dll from the Microsoft C++ Runtime Library, so I installed that, then it said that the program could not start, press any key to exit. The current version says that it is for the nVidia GPU-based Macs, which this is not. I am running 64 bit Windows 7 Ultimate.

Any ideas?

 

I did a quick search on Real Temp to see what you're talking about and I think you're misinterpreting the results. Setting TJMax in Real Temp doesn't actually change the CPU's TJMax temperature. The setting is an input to tell Real Temp what the value of TJMax is for your CPU.

The readings from the CPU's thermal sensor are not absolute temps, they are relative to either the max or target TJ temp. Real Temp needs to know that reference point to calculate an absolute temp. So if you set it to 70 and then 120 the reported temps will change but the actual temps will not.

The absolute temps won't be accurate unless you've input TJMax correctly and calibrated the idle temp. Instead of focusing on them, look at the "Distance to TJ Max" readings. If these are going to zero, the CPU is reaching its target maximum temp and then throttling to prevent it from going higher.

The original MacBook Air did this a lot because the cooling design didn't have the capacity to keep up with the CPU at sustained high loads. The 2010 MBA models ran cooler and didn't do nearly as much throttling. It's possible that the higher performance of the Sandy Bridge CPUs in Turbo mode is driving the temps up to the point where it's doing heavy throttling again.

Thermally limited behavior should be expected when putting powerful CPUs in a very small package and will happen regardless of what OS you're running. It's not fair to compare with a desktop CPU which has a big honking heat sink surrounded by empty space with lots of airflow (or even active cooling). The Macbook Air has a tiny heatsink and fan.

 

I realize that the TJ Max setting in Real Temp just tells it what the max is supposed to be for the chip. However, after setting it all the way up to 120 degrees Centigrade, I would not expect the temps to then rise to that level!!

Incidentally, I re-tested using CPUID HWMonitor for temps and Prime95 (the tried and true burn-in tester) and temps rose all the way to 99 degrees Centigrade for both cores.

I tested idle and load temps using 720p video streaming in Flash from Amazon in both Lion and Bootcamp, and got virtually identical readings with up to low 70's under load, so I think all is ok.

I don't plan to ever put the CPU at 100 percent load, just wanted to see if the Air could pass the tests.

At this point, I would say that stress-testing an Air is probably not advisable.

 

Why not? You are telling it that the TJ Max is 120C so when the CPU reaches TJ Max, it reports that the temperature is 120C. It makes perfect sense.

Just to be clear, the temperature of the CPU is not actually 120C. There is no easy way to get a CPU to go above the built-in limit regardless of the OS. The limit of that particular CPU is 100C which is consistent with what you get using the other monitoring tools. Obviously, the MacBook Air is not designed for such heavy loads.

 

I say take the kid gloves off and use the machine for what you want to. Intel incorporates throttling into their CPU design to prevent you from harming it. Consequently, CPU failures are quite rare even among overclockers. The real issue here is that maybe the MBA can't extract the full performance potential of the CPU for a sustained period.

 

You basically just proved dmk2's point. You set the temp to 120C and that's what it told you it hit. But your other software told you it never went above 99C, which is just below the thermal limit of the chip. So everything is working fine.

Sure, it runs hot. It's a Mac laptop, and a very thin one at that. Nothing new.

 

How long have u been running under 100% load?.
I was runing stress test in AIDA and it seems that the fan speed up relatively late (at around the 95°C mark), until then running at the minimum level of 2000 rpm. Apple want to keep the lappy silent?.
Anyway once the fan speeds up starts managing the heat pretty good, and temps go down bellow 90 though sustaining the CPU at load.
Therefore in my opinion those extremely high temp (u see also the cpu starting throttling) are only transient and caused by the fan control system, which only intervenes once extremely high temperatures have been reached.

 

Temps of 100C in IntelBurn test are not unusual. My Acer has a great dual fan cooling system and, with factory paste, I have seen temps over 100C in IntelBurn test. I would expect an I5 in the Air to easily get to 100C and throttle back, given the metal case and rather small sink/fan. I don't know why you're so surprised.

The real question I would be asking myself is how did the temps get above the I5's max temp.? How come the CPU did not throttle at the max temp?

 

Use HWMonitor 1.16 to monitor temps:

CPUID - System & hardware benchmark, monitoring, reporting

 

That's right, HWMonitor !

I'm using my MBA 11 for 3 months now and I only once reached 100 C on one of the cores. Generally it wouldn't let it go above it, coz it will throttle.

But I must say some other ultrabooks are doing a bit better - I think I'm gonna open my MBA after some time to clean the dust and I will install a good thermal grease.

 

i'm sorry, but what is the OP's point here?

is it that he believed that he could change the tjmax of the cpu and therefore permanently damage the cpu if he ever ran 100% cpu usage for a long period of time?

 

I would not adjust the TJMax manually. It is the MAX for a reason.

My MBP also runs unnaturally hot in Windows. I don't believe the hardware is entirely to blame. At idle, it runs at 35-45 C in OSX with the fans at 2900rpm. In Windows, it idles between 65-70 C with the fans at 6000rpm. Ridiculous.

 

6000RMP in windows? Something must be wrong. I have a boot camp partition win Win7 and runs in iddle at 30-40 C degrees most of the time (even with web browsing, office, etc).
Only with a high load (gaming, Crysis) the temperatures rises and fan starts to speed up.

 

Are you running the latest bootcamp drivers and when was the last time you cleaned out the fans?

 

when using a mac under windows gaming, always set fan speed to MAX RPM!
and yes, the SMC is flawed, it's not new. but strange thing is my 2008 mb responds to 80c instead of 95

mba 2010+2011= non standard screw



get the screwdrivers from ifixit.com if you really need to clean the fans


and look at the small fan! combined with the HD3000 in the same chip, temps above 95 is expected

 

does it only run that hot when running windows? or will it run hot like that even on lion?

 

Can running IntelBurnTest damage an Air? I ran it on "high" setting for 50 passes and it got very hot (I didn't check temps), it failed at pass 45 but didn't give an exact reason, the system never froze or blue screened.

It's a new Air mid-2012. I'm worried I may have shortened it's lifetime. If everything still appears fine, should I be ok?

 

no. you're fine. just don't do it again kind of thing.

while the CPU itself is relatively safe, the fan is going through extra wear and tear for no reason. As well, you risk corrupting the OS/data.


in general, running stability test programs on a laptop is a bad idea. why someone does it...i don't know, because they're there for OVERCLOCKING purposes.

 

Stability programs are not just for stability testing. These programs can shed light on the construction quality of your machine. You may not see quality issues arise through day to day, but again it is an underlying issue and can lead to other issues that may crop up. For example, the Nvidia 8400/8600 mobile GPUs suffered from a bad substrate material mixture. This lead to excessive thermal wear. Every time the GPU went through a thermal cycle (heat up and cool down), the GPU suffered sub-surface fractures. At normal 2D frequencies, the internal temperature was not high enough to cause separation of material at the fracture points. However, 3D operation suffered from artifacts. Even if you never played a game on the 8400/8600 mobile GPUs, your GPU was still bad and getting even worse. Numerous chips failed very very early and Nvidia was sued.

Or how about this?

Did you know that a bad power circuitry can cause silent data corruption? If your power circuit is not able to precisely modulate voltage and current, you can actually flip bits while they are on the fly. You won't see this issue until it's too late. However, stress/stability testing your machine runs your system components at their rated levels and if there is an issue and you'll notice it.