VID, BIOS and TEMP EXPERTS HAVE A LOOK !!

[Leeghoofd]

K today in the P5Q Dlx topic we had an issue that users couldn't downgrade to an earlier bios... I proposed to try out a hot flash in which I succeeded ( for those unaware the P5Q Dlx and co have 2 bios chips) so I downflashed from 1611 to 1406... reset bios to defaults, rebooted and then shut down and popped the battery for half an hour to make sure everything got thoroughly wiped... I rebooted and wanted to verify if everything was working correctly... gamed on the rig for a few hours all looked stable, then again a quick temp check etc...

When I opened Realtemp I noticed a strange anomaly in the temps under load... my quad has always been pretty close in the cores, yet now I got a 7°C difference... I openend coretemp : same temp difference there and ow shocka a VID of 1.125... where did the latter come from ? it has always been a VID 1.250 with this QX on 4 different mobo's...

Look at the screens they tell a zillion words :

CHIP 1 with the 1611 to 1406 downflash ( reflashed it trice now wiuth EZ FLASH and AFUDOS and popped battery each time)



CHIP 2 with 1406 bios ( have also reflashed it today, was 0803 before)




So if someone has an explenation what has exactly happened here : a bad flash like most will suspect ( rig works perfectly passes 15 IBT passes, gamed upon,...) reflashed trice, popped battery etc... and no I'm not a NOOB : all settings are 100% identical

I 've told a few times that on some mobo's the cores on some of my CPU's were stuck. Yet on another mobo they worked perfectly ( same type of mobo )

Again this experience confirms that a bios can bork readouts of 45nm way more than some know... VID seems to be able to change with biosses etc


EDIT : reflashed the borked chip with KODA's utility all is same again as chip 2 Yet it remains fascinating to me that a bad flash can evoke this sort of behaviour on monitoring programs

[unclewebb]

I know CPUs have two VID values. The Minimum VID is used when C1E / SpeedStep are enabled and the processor is idle. The Maximum VID is when you are at full load. I've worked with a Core 2 mobile T7200 chip that could report a range of values between these two values as the CPU moved back and forth between idle and full load.

I've never been able to find in the Intel documentation where the VID information is located within a CPU or exactly how to interpret that data. I looked at some code from CrystalCPUID and went from there. VID information seems to be located in several different MSRs (model specific registers) within Core 2 CPUs.

I've never been 100% convinced that VID was totally independent from the board that the chip was on. Your results would lead me to believe that VID might be a meaningless number.

As for your temperatures, that's an even bigger mystery.

Core Temp and RealTemp are both sending a Rdmsr command directly to the processor to access the temperature information. No change in motherboard or bios should make any difference to the data coming directly from the CPU. Overall, these sensors are fairly consistent. They may do many odd things but are generally pretty consistent from day to day.

Maybe how the load was applied to the IHS might have done something to the readings but it makes absolutely no sense if you switch back to the original bios and the temperatures change back to how they were originally. That's not supposed to happen and there's no reasonable explanation that I can think of.

Download the latest version of RealTemp and see what it says in the Settings window for Min/Max VID. Also try doing a CPU Cool Down Test. This can help show the slope that the sensors are working at and other issues.

The latest version of RealTemp keeps track of the APIC ID in the INI file now after you exit RealTemp. That might shed some light on what's going on here. Typically you will see this:

APICID=0123

Which means that the cores are in their normal order. When a bios or CPU screws the core order up you can get some unusual behavior reported but not what you're seeing.

MSR 0x198 contains multipliers and VID information.
http://www.fileden.com/files/2008/3/3/1794507/MSR.zip

I'm interested in any additional information you can find.

[Leeghoofd]

Well after reflashing with Koda's bios tool ( which also seems to rewrite the read block, unlike Asus's flash utilities) all is fine now... too bad I didn't had the beta yesterday but I'm pretty sure I saw Apic0123 when pushing the little button next to the CPU's name...

I've experienced the core behaviour before where the core readouts were stuck on my Striker board , yet they worked fine on the P5Q Dlx... I also returned a few tray CPU's because of stuck sensors, yet when the shop tested it on their mobo, it seemed to work perfectly...

This will remain a mystery to me, thx for being brave to reply to this... I might try to replicate the bad flash but think it will be a 1 out of 1000 case... if it acts up I will contact you asap...

Kind regards Leegie