BRAD: Only you can know the truth about this...

[gustavo]

I learned of an application wpcredit, it can be downloaded from
viahardware.doc along with 2 .pcr files that help interpret what
the supposed registers mean (guess at this point you know what
I am talking about).

Well, you have a register number 95 composed of 8 bits: by default the KT266 chipset has an 1C hexadecimal or 00011100 on it. The supposed meanings of the bits being set are:

[95:7]=MWQ Time Slot While MWQ is full (units of 8 CPUCLKs)
[95:6]=(same as above)
[95:5]=(same as above)
[95:4]=Write Policy CPU to RAM 0=FIFO>2 or idle 1=disable
[95:3]=PMR Cycle Control 0=Stall if MWQ full
[95:2]=FID Command Detect 0=disable 1=enable
[95:1]=HALT Command Detect 0=disable 1=enable
[95:0]=(Reserved)

If you set the register 95 to 1E or 00011110 (ie set the bit 1)
the CPU temp lowers too much to be true, in my system from 49 to 31 celsius (unbeliable). Even the chassis temp is higher than
the cpu temp (35 celsius). Also there are several people reporting the same at the VIA forums, and saw one thread at the MSI forum reporting this as well.

I smell something is wrong here and want to know your advice here. The cpu is cooler this is a fact i put my hand over the case and can feel it cool but WTF is happening inside ??

Thanks Gustavo.-

[brad]

Yes sir, sounds fishy to me also. I'm no thermodynamics engineer but I don't believe it because:

- according to AMD docs, an Athlon t-bird in full sleep state still draws 5 watts of power and that will still generate some heat.

- the cpu temp sensor sitting in dead airspace under the cpu socket will still sense it, and it will be higher than inside case ambient temp even if only by a little.

- the thermistor sensors used on most boards are hardly accurate, but I haven't found one inaccurate enough to account for the drops being claimed, and I can't juggle the numbers enough to make it believable. Especially since the beginning temp is in the ballpark.

BUT, most importantly of all:

- they're messing with a reserved register. They don't know what changing the value will do without the docs, and they didn't claim to have them.

I would say what I think changing that register does, but I won't because I've been made an arse of here already! <g>

Cheers my friend

Brad

[gustavo]

Brad:

The cpu idle temp is that get so low, if you run cpu intensive applications it goes up to the same level as before (in mi system 50-51 Celsius), but as the cpu intensive application finishes the temp goes down in a few minutes.

Also you can hear the noise of the cpu fan throthling up as you start a cpu intensive application and slowing down when it finishes (maybe it is normal but i had not realized of it before)

The benchmarks are not affected they yield the same as before.

So my bottom line is: you think is dangerous to change the referenced bit ??

Many thanks, dont want to bother you but it seems a as a red and juicy apple to eat even feeling something is wrong.

Gustavo.-

[brad]

Thanks for the additional info that you sent me. Interesting stuff but one has to wonder where it came from because it isn't posted on VIA's site for the public to get, and I won't speculate on wether or not it is safe to use it for that reason. It is for adventurous people who understand the potential risks and are willing to accept any consequences that arise from using it. You must decide because it may have an adverse effect elsewhere and problems do not always appear right away.

I will say with certainly however that based on everything you have said in this thread, the actual temp of your cpu is 5C higher, maybee a little more, than what the mobo is reporting. That is very typical on MSI's socket-a boards, and I've measured cpu temps in accordance with AMD's recommendations enough times to know. Even including that error, your cpu temps are fine, well within AMD spec, and nothing to worry about.

Brad

[jim chase]

It sounds like it's taking advantage of the cpu halt function. Linux and Windows NT have this built in, Windows 9x/ME didn't do this. Software coolers like Rain and Cpuidle do this for Win9x.

On my K6-III/400 overclocked to 450mhz Rain drops the cpu from an idle temperature of 42c to 31c. In win9x the cpu is running full current all the time without Rain.

Jim

[gustavo]

One link here on the subject:
http://www.msi.com.tw/forum/index.php?target=article&tid=4686&page=2
Look for this post:

DanTucny
Normal User
Join Date: Nov 10, 2001
Location:
Posts: 16
I get 35C idle and 41C under load with my XP1800+ and cheap heatsink (system temp 31C), setting register 95 to 1E from 1C in wpcredit the idle temp drops from 35C to 21C (system temp drops to 28C from 31C)...
have a feeling that the thermistors on the Pro2 do not give very accurate results...
_______________
CPU: AMD Athlon XP 1800+
COOLER: Akasa Icicle 765
TEMP: Idle - 35C, Load - 42C
MOBO: Microstar K7T266 Pro2-RU (orig bios 3.1, curr bios 3.3)
RAM: 3x256Mb Samsung DDR PC2100 (2 currently in slots 1,3)
GFX: Elsa Gladiac 920 (GF3,23.11)
IDE1a: Maxtor 27GB HDD
IDE1b: Maxtor 40GB HDD
IDE2a: Toshiba DVD-ROM
IDE2b: Sony CDRW
IDE3: IBM 60GXP 60GB HDD
IDE4: IBM 60GXP 60GB HDD
RAID: Onboard promise raid 0
FDD: Samsung 1.44Mb 3.5inch floppy drive
PSU: Enermax EG365P-VE (350W)
CASE: Coolermaster ATC201
NETWORK: Netgear FA310TX
AUDIO: Onboard
OS: Windows XP Pro (currently)


he also get system temp (28C) higher than cpu temp (21C) ....

Gustavo.-

[brad]

It sure is there Jim. Its just not having the full details of all the registers and the conditions for their use that irks me about the info, which I have to assume is an unknown source at this point. If it was extracted from VIA docs, it should at least be so stated.

Gustavo,

Just read the thread. Nothing surprising there except that a good comparison of thermocouples and thermistors was given. It would have been nice if he would have gone into the many advantages of PTAT type sensors over the best thermocouples for cpu temp reading, mainly accuracy, ease of use, and cost, but he may not be aware of them.

VERY surprised that none of them seemed the least bit suspicous of a cpu temp that reads lower than system temp!!

If you're interested, here is the link to the AD590M sensor and related application data I use to measure cpu temp:

http://products.analog.com/products/info.asp?product=AD590

And here is my display device. (http://www.tm.agilent.com/classes/MasterServlet?view=productdatasheet&pro-ItemID=1000001297&language=eng&locale=US&FamilyTitle=Digital%20Multimeters/Voltmeters&title=1)

Brad

[gustavo]

That´s overkilling to me Brad , I dont even know the difference between thermocouples and thermistors :(

U$D 7500 for the device in the US appears to be too much money anyway, It should cost twice as much (at least) if it were availabe here in Uruguay.

Thanks Gustavo.-

[gustavo]

Another link: http://www.naggelgames.de/vcool/VC_Theory.html

They conclude:

It seems that a simple HLT instruction will halt a Athlon/Duron but it will not put it into a low power mode. Even if you set the Bus disconnect bit in the Northbridge is not sufficient because the Northbridge will only take action if the CPU is in the STPGNT state - a mere HLT in just enough.

To put the CPU into the required STPGNT state you'd have to utilize an ACPI register in the Southbridge.

Apparently some OSs and cooling utilities fall short in one or the other category: Usually the OS has an idle loop and probably uses the ACPI register to put the CPU into STPGNT state but it forgets to set the Bus Disconnect Bit.
This is where a lot of (BIOS-)hacks and tools come into play: They set this infamous Northbridge bit - however if OS doesn't use the APCI register or the tool has it's own idle loop with just the HLT instruction, the CPU just won't cool off.

Brad does it sound reasonable ??