P55 Mobos CPU sensor problem

[bumper]

I wrote to GGTS yesterday. Waiting for reply.

[forumjoe]

Okay, I'll give it a go.

Do bear in mind that my theory about it being caused by a faulty or poorly-chosen sensor is only that - it's a theory. And perhaps "sensor" is the wrong word. What I mean by that is that, if we assume that the CPU itself has a built-in temperature sensor, then it's also just as likely that one or more of those tiny surface-mounted components underneath or to the side of the CPU is a pull-up resistor involved in passing the CPU's temperature level to the BIOS chip. Being of CMOS technology, the BIOS array will undoubtedly connect with high-impedance signals on the board, some of which might require pull-up resistors. But if one of those resistors is faulty or hasn't been successfully soldered in on the board during the solder-flow manufacturing process, the resultant connection into the BIOS chip would flap around, voltage-wise. One possibility is that the solder joint is good when the PC is first switched on but then becomes bad as the nearby CPU heats up. Or it could be due to a faulty or badly-chosen temperature coefficient for the resistor. Either way, with the signal from the CPU flapping around, the poor old BIOS would have a devil of a job then interpreting the input.

I think someone in these forums summised at one stage that this problem was due to a design error with the CPU itself and that there were already murmurings in the Intel camp to that effect. However, I think that that can now be dispelled. There has been a problem with certain Intel CPUs manufactured since about last Summer, and indeed Intel themselves have now published an explanation of this on the Web (I found it yesterday in the news bulletin section at Gigabyte's homepage), but that particular problem is one that affects IDE transfers and has nothing to do with temperature-sensing. It affects only a very limited number of Intel CPUs. Intel have admitted that users with that particular problem will need to return the CPU chip, for replacement. In the meantime, it's important that nobody should confuse the Intel announcememt with the "wrongly-reported CPU temperature" problem that we P55 users are experiencing.  

[forumjoe]

I've now written to GGTS about this, carefully explaining it all.

I must say I'm not very hopeful, as I've found that, generally, the responses from GGTS have been slow and totally unhelpful. They seem to find it impossible to ever comprehend the problem, despite the extreme lengths I go to in order to carefully explain it. And I don't mean just THIS issue, I mean all the OTHER issues I'm having with my particular board. Sometimes, I  wonder if they do it deliberately. They really do wind me up. They don't  even confirm whether the checksum values I've noted when flashing different BIOSs agree with theirs. When such basic bug-tracing is ignored by them, it makes you wonder what the purpose of GGTS really is.

Anyway, they can't really ignore this "misreported CPU temperature" thing now. There are a growing number of people who've now told them about it and we're all using slightly different boards. It most certainly is BIOS-related.

[bumper]

Today, the answer came from GGTS.

At first, we recommend you chekcing H/W installation status. The cooler must be installed on the CPU surface tightly. Then, if there is any thermal glue or plastic membrane on the CPU cooler, please remove it. After that, please kindly apply a thin layer of thermal grease/compound on the surface of CPU die for better heat dissipation then check again later.
In addition, please try to disalbe SMART fan in BIOS to let CPU fan spins at full mode.

 

[Dark Mantis]

I can understand where they are coming from however I would have thought it was obvious that it isn't a straight overheat problem.

[forumjoe]

I'm afraid that's fairly typical of the kind of eventual response you get from GGTS, ie, stating the downright obvious! It doesn't seem to occur to GGTS that someone who has the slkills to build a PC from scratch would know the technique for properly fitting the CPU's heatsink. Personally, I think it's very insulting of GGTS. In any event, they should have had quite a number of identical reports in of this problem by now. It's most certainly nothing to do with how bumper has fitted his heatsink, it's a board problem! It's definitely and irrefutably BIOS-related. It's simply ludicrous in the extreme for GGTS to continue claiming that the board isn't the problem.

[boglav]

As previously stated, I have disabled the C3/C6 state function in BIOS, and I haven't encountered the "sensor" problem since (+4 months). All other settings are default/auto. I monitor my temps from time to time, and I believe I would have spotted if the problem had arisen again (certainly, I cannot state 100% that this is the best solution).
 I would also try to disable CPU overheat warning in BIOS (can't recall the exact name of that function). It works ok this way, with no annoying random beeps (the other issue that seems to be plaguing the P55 boards).
To be on the safe side, I also replaced my cooler with a Mugen 2 rev. B, which has an enormous heatsink that can cope with my I5 even fanless and is quiet even at max RPM (~1300 rotations) at stock settings.
To be on the safe side, especially with OC or a 24/7 working system,  I would reccomend getting a good cooler, with an efficient heatsink (like the Megahalems) and a set of quiet fans for a push-pull config, and disable the automatic fan rotation setting (or use some SW to set a custom value that suits your needs).
GGTS also sent me a beta bios to test (v13 beta something), but I'm not playing lab rat for them, especially since they haven't acknowledged the problem yet, and less likely solved it.

[forumjoe]

Boglav,

When I first encountered the misreported CPU temperature bug, I played around extensively with the C3/C6/C7 function in my BIOS but found that it didn't make a jot of difference. I think that you and a few others have simply been lucky with your own BIOSs in that regard. Very early on, I also tried playing with CPU Thermal Monitor in that same section of the BIOS, and also with CPU Warning Temperature in PC Health, but again to no avail.

My new PC is in no way geared to gaming or overclocking, as it's used for very serious work instead. The graphics card I use consumes only 30W, even at peak (confirmed with the manufacturer). I constructed the entire machine to be as quiet in operation as possible and so I use specially-selected push-pull fans, together with an Arctic Cooling Freezer 7 Pro Rev 2 that itself incorporates a fan (highly recommended cooler). The entire arrangement runs extremely cool, confirmed by me feeling the finned heatsink with my fingers, and I'm able to run all three fans at around 850 rpm. The CPU temperature, as reported by the BIOS, starts off at around +24 degrees C, then after, say, 10 mins reaches around +28 degrees C. I'd say that's probably about right. But after 30 mins or so, the reported temperature starts going wild, with anything from +20 to -20 being the norm.

I've been building my own PCs for some 15 years or so and am a retired IT engineer, so GGTS cannot accuse me of ignorance or negligence when assembling the CPU, etc. Earlier in the build, I quickly realised that there were some annoying limitations in the onboard fan control, so I ignored Smart Fan Control and the second Sys Fan header and instead incorporated my own manually-controllable fanspeed devices.

I think that probably what's happened here is that a common core BIOS has been used across the P55s and that this error has crept in and it affects different models in slightly different ways. That's to say, the C3/C6/C7 setting will correct the problem on one model but not on another. Put another way, I doubt very much whether your BIOS is exactly the same as mine.

It could still be down to a common hardware manufacturing fault on the board but it remains the case that it's the BIOS that's misinterpreting the data it's being fed. So, in my view, there's no question that it's BIOS-related.

[boglav]

This board is so weird..for example if I switch the cpu fan setting from auto to pwm, my razer mouse stops working (just the tracking, the buttons work fine). I have to umplug the system until all power is out (the lights on the mouse turn off) to make it work again.

[Dark Mantis]

Gremlins!

[forumjoe]

Boglav,

I've just looked again at your screenshot of PC Health and it strikes me that not only is your CPU core voltage on the high side but also your RAM voltage is very high. For instance, my understanding is that DDR3 can withstand up to 1.65v but not more than that, really. Do you not think that you're pushing your CPU and RAM rather too much? Have you tried adjusting the CPU's core voltage and the RAM's voltage downward, to see if that affects your observed problems at all? I know that some brands of DDR3 can tolerate over-volting better than others but I'd personally be alarmed if I ever let my own DDR3 be powered on as quite a high voltage as your BIOS is reporting, namely 1.68v.

Actually, that said, I think you'll find that the BIOS's measurements of voltage and frequency aren't terribly accurate. So, it may be better to check out your various supply voltages at the board's power connector with a digital meter. Do this with extreme care, though; you need to know exactly what you're doing, and the risks involved.

Something else that you, DM, and others will find interesting is that, although earlier I told GGTS that C3/C6/C7 has no affect on the misreported CPU temperature on my particular board, it now transpires that I was wrong about that. I've now discovered that, on mine, it does depend on C3/C6/C7. I'll now have to inform GGTS.

By experiment, I've found that the misreporting starts to happen at a particular CPU core voltage. Although I don't normally attempt to push my CPU voltage at all, I can now appreciate that the whacky CPU temperatures start to get reported at a certain voltage, and then as I raise that voltage further, still well below the CPU's specification for maximum voltage (and cause the CPU to consume more power), the reported temperatures become even more negative. I'm not sure if I'm imagining it, but it also appears to make my PC slightly more unstable, and the random FDD accesses occur more often. By the time that my BIOS is telling me that the voltage is set to about 1.180v, the reported CPU temperature is already negative.

From research I did some time ago at Intel's site, my i5-650 can operate at anywhere between 0.65v and 1.40v, with 1.20v being the nominal value. Obviously, as you go down from 1.20v, the frequency of operation will drop off and also the current taken by the CPU will drop. The opposite happens when you start going above 1.20v. The current will change markedly either side of the nominal voltage, so the amount of power consumed by the chip will quickly increase or decrease.

So right, I would agree now that C3/C6/C7 is a factor in this, not just in some P55 boards but also in mine (P55-US3L). Also, it's apparent that as the CPU core voltage increases, so the CPU temperature decreases (the opposite to what you'd normally expect). But it looks like the voltage has to go above a certain breakpoint before the whacky temperatures start to be reported.

Unfortunately, it's not at all easy to say exactly what that breakpoint voltage is, as the BIOS is so inaccurate.

You know, so very much depends on the BIOS. Not only should a BIOS enable an operating system to see the basic system components and the resources but it ought also to measure fairly accurately the things that need monitoring. Clearly, in our cases, our BIOSs aren't doing these two things at all well.

I note that the USB3L, UD3, and now my US3L all use the same basic drivers.

[boglav]

As far as DDR voltage is involved, I'm pretty sure I'm on the safe side. As a matter of fact, it's a XMP profile from my ram (Kingston HyperX 1333 CL7, 1.65 v)  that besides increasing the RAM voltage, also ups the QPI/VTT value accordingly (1.21 v). The CPU voltage is set to auto in BIOS, I tried lowering it but it resulted in an unstable system.
Since the systems works OK atm, I'm done tweaking and testing it. Of course, should I discover anything worth telling about as far as the issue at hand is concerned, I'll promptly report it over here.
On another note, I also watched the date of Bios release for several types of P55 boards. All of them are released on the same day, which makes me think we're talking basically about the same version with minor model-specific tweaks.

[forumjoe]

Boglav,

For what it's worth, I've now reported my findings to GGTS and have uploaded a photo to them of the PC Health screen in the BIOS, showing a whacky CPU temperature of -11 degrees C.

One thing that's very noticeable is that the BIOS cannot read or interpret the voltage and temperature data at all accurately enough. In fact, for the same parameter, you see different values in different parts of the BIOS. The differences between them are greater than, for example, the increments by which you can adjust Vcore. So it means that you can never really know the true value to which you're setting any one parameter.

[forumjoe]

Boglav et al,

I've now had some feedback from Gigabyte on this CPU temperature issue. Judging by a photo they've supplied in the response, they've set up their own US3L but cannot reproduce the problem (If a had a pound for every computer company that told me that sort of thing, I'd be a very rich man indeed!). They've used the same CPU, board and BIOS version as mine. I've noticed, though, that they've used the built-in Smart Fan Control, which I'm not. Well, what can you do? They're of the view that this problem is non-existent. But US3Ls suffer from it, UD3s suffer from it, and USB3s suffer from it. Maybe more. Some of us have even uploaded photos of it, showing the negative temperatures, so it's crazy for Gigabyte to say that there's no problem there.

This is by no means the only problem I have with my US3L, and one other important one is that I get constant HDD accesses and also occasional random FDD accesses. I can see the HDD accesses happening by way of the HDD Activity LED on my PC case and can hear the corresponding constant bump-bumping of the HDD. The Gigabyte techie said this was due to the CD/DVD drives being enabled; if I disabled the CD/DVD drives, it'd go away. Well, I tried that in Device Manager and it's true - if you disable both optical drives, the constant flashing Activity LED all but stops. He referred me to an old Web forum on the issue. The trouble is that disabling your optical drives is hardly a feasible solution, is it?! And nobody in that old forum had a permanent solution for it. Here's the link he gave me:

http://social.technet.microsoft.com:80/Forums/en/w7itprohardware/thread/4f6f63b3-4b58-4154-9298-1566100f9d00

I've now come across this:

http://support.microsoft.com/kb/967715

and Microsoft are offering a fix for KB967715 on the WU website but, as far as I can see, the fix is to prevent Autorun, which is NOT what most of us want. And I don't think this is anything that affects the HDD accessing issue.

The techie couldn't explain why one of my optical drives can't be set to DMA mode; it's permanently greyed out in PIO. Gigabyte have tested with some different drives and found no problem with DMA/PIO settings in Device Manager. I suspect that the constant HDD accesses, PIO problem, etc are linked in some way. And this still doesn't explain why SeaTools cannot even see my HDD! I still maintain that there are faults in the BIOS. But getting Gigabyte to look carefully at the software of the BIOS seems to be beyond the hope of any mortal being. As far as they're concerned, if they can't reproduce the problem, the problem cannot exist!

[Dark Mantis]

There is none so blind as them that won't see!

I agree that it is a stupid idea that just because you can't reproduce something it can't exist and the BIOS writers need to start looking at how this could happen and see if they can come up with a way to stop it. After all that is all we do on the forum as we can't reproduce the faults but we try and look at what circumstances could bring about that condition and then find a way to cure it.