Pro/Forums - View Single Post

pHaestus · 10-04-2002, 03:26 PM

Hrm just read that thread on Hardforum, and Steve apparently has little touch with reality. He appears completely oblivious to the writing and discussion that has gone on for a year or so on both diodes and waterblock testing. I am guessing the bulk of the people at that forum never read or go to any other sites and so they will actually swallow the load of crap he is spewing.

Regarding diodes and their inherent accuracy:

I think most people can recognize that I have a fair amount of practical knowledge of internal diode on CPU and the circuits that poll them.

http://www.procooling.com/articles/h...rmal_dio.shtml

http://www.voidyourwarranty.net/revi...ode/index.php3

(sorry for the vanity but apparently with H|OCP it is required to prove you are an "authority")

Neither AMD nor Intel ever stated that the diode was an accurate measurement device suitable for testing. Neither AMD nor Intel ever claimed that the diode reading was without uncertainty. Think that Intel tests thermal performance and dissipation with real chips? ha! There are a ton of factors at the motherboard level that affect the accuracy and the linearity of internal diode readings. I have posted a link regarding such before:

http://www.smsc.com/main/anpdf/an820.pdf

Similar requirements for Winbond and Maxim ics. To think mobo mfgrs follow this is a bit naive. It is not trivial to get good quality data from a diode even with professional monitoring equipment, btw.

Intel and AMD have their internal diodes for overtemperature protection and nothing more. They work tolerably well (=/-1-3C depending on temp range), but ONLY when the reader is designed well. And calibration (not just one point) is a major concern.

One would do well to recall the issues associated with "hacking" the AMD761 chipset to read the diode. The trace length, crossovers, and thicknesses were NOT to spec, and the linearity was not good. Sure you could get it to read room temp at idle with a potentiometer adjustment, but there was NO real guarantee that it was reporting the correct temp at any other time. There is no evidence that mobo mfgrs completely redesigned their boards to fall within spec for accurate diode readings; if they did then their engineer was probably fired for wasting money.

I have played around with verification using a tc under the cpu core as per AMD tech docs; not possible with the organic chips though. In the end, it is not simple to validate and calibrate an internal diode.

Regarding BillA's experience:
Again this floors me. Perhaps Steve has honestly never seen this work?

http://www.thermal-management-testing.com
http://www.overclockers.com/articles608/

Or perhaps he blocked it out because he couldn't follow it?

Anyway, the complete ignorance on the topics that he professes "authority" on does not sit well with me. If Steve wants to posture his results as "real world testing" and "an overview without getting technical" then that is fine. But he is treading into deep waters now; carpel tunnel syndrome is a real concern if he is to ban everyone more knowedgeable than he is.

10-04-2002, 03:26 PM	#117
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	Hrm just read that thread on Hardforum, and Steve apparently has little touch with reality. He appears completely oblivious to the writing and discussion that has gone on for a year or so on both diodes and waterblock testing. I am guessing the bulk of the people at that forum never read or go to any other sites and so they will actually swallow the load of crap he is spewing. Regarding diodes and their inherent accuracy: I think most people can recognize that I have a fair amount of practical knowledge of internal diode on CPU and the circuits that poll them. http://www.procooling.com/articles/h...rmal_dio.shtml http://www.voidyourwarranty.net/revi...ode/index.php3 (sorry for the vanity but apparently with H\|OCP it is required to prove you are an "authority") Neither AMD nor Intel ever stated that the diode was an accurate measurement device suitable for testing. Neither AMD nor Intel ever claimed that the diode reading was without uncertainty. Think that Intel tests thermal performance and dissipation with real chips? ha! There are a ton of factors at the motherboard level that affect the accuracy and the linearity of internal diode readings. I have posted a link regarding such before: http://www.smsc.com/main/anpdf/an820.pdf Similar requirements for Winbond and Maxim ics. To think mobo mfgrs follow this is a bit naive. It is not trivial to get good quality data from a diode even with professional monitoring equipment, btw. Intel and AMD have their internal diodes for overtemperature protection and nothing more. They work tolerably well (=/-1-3C depending on temp range), but ONLY when the reader is designed well. And calibration (not just one point) is a major concern. One would do well to recall the issues associated with "hacking" the AMD761 chipset to read the diode. The trace length, crossovers, and thicknesses were NOT to spec, and the linearity was not good. Sure you could get it to read room temp at idle with a potentiometer adjustment, but there was NO real guarantee that it was reporting the correct temp at any other time. There is no evidence that mobo mfgrs completely redesigned their boards to fall within spec for accurate diode readings; if they did then their engineer was probably fired for wasting money. I have played around with verification using a tc under the cpu core as per AMD tech docs; not possible with the organic chips though. In the end, it is not simple to validate and calibrate an internal diode. Regarding BillA's experience: Again this floors me. Perhaps Steve has honestly never seen this work? http://www.thermal-management-testing.com http://www.overclockers.com/articles608/ Or perhaps he blocked it out because he couldn't follow it? Anyway, the complete ignorance on the topics that he professes "authority" on does not sit well with me. If Steve wants to posture his results as "real world testing" and "an overview without getting technical" then that is fine. But he is treading into deep waters now; carpel tunnel syndrome is a real concern if he is to ban everyone more knowedgeable than he is.