Internal Diode Xp testing, Finally!

[Mikewarrior2]

Modified motherboard, of course, since no manufacturer has made a reliable, diode reading mb other than the fujitsu-siemens.

Void Your Warranty Internal Diode Testing.

and, a few quotes to those who don't click the link



<< If you look closely at the two graphs, then you notice that the case temperature, the motherboard's CPU temp, the ambient temperature, and the room temperature are all the same for both tests. However, there is a 7C difference in the internal diode temperature under load. I reproduced this phenomenon by making an excessively thick application of Arctic Alumina, and that produced a temperature 5C higher than a proper application. You can also see in the second graph around 25 minutes where I briefly turned off the fan on the heatsink. The diode temp immediately responded (and cooled down again when I powered the fan back up) but the mobo sensor couldn't see it. >>





<< Second, if the motherboard probe cannot detect a 7C difference in temperature, then if I had to rely only on that probe for notice of CPU over heating then I would make damn sure >>



So, I don't want to sound like i'm bragging, but nah nah nah to all the people who thought that using socket-thermistor's was an accurate way of testing, or that socket-thermistors were good enough, or just plain said I was dumb. There you have results, a 7C internal diode temp with artic alumina had zero effect on the mb socket-thermistor.

The compression is actually even more than i thought it would be... 5-7C diode with no socket-thermistor reading movement. Now come and tell me that you can actually use this method for testing. All statements based on heatsinks/greases can't be made if tested with socket-thermistor due to the tremendously inaccurate nature.


Mike

 

[pHaestus]

In all fairness, Mike, the Asus CPU probe is without a doubt the worst of its type. If you look at the graphs:

http://www.voidyourwarranty.net/review/xpdiode/1utemps.jpg

from the point that the CPU was idled with WPCredit there is a 10 C offest roughly. That means Asus is trying to compensate with adding 10C to the measured CPU temp. If you subtract the 10C offset from the Asus undercore data then you get roughly a 7C delta between idle and load temperature. For the internal diode, that delta is around 30C. In theory then the Asus probe SHOULD detect a difference larger than 5C. In practice, since it isn't even touching the CPU it actually does a little worse when comparing differences in paste application.

Perhaps motherboards that have probes that actually touch the underside of the CPU would do a little better. It is hard to say with the XPs though because they have resistors that will impede temp measurements as well.

I am glad you found the writeup useful

 

[Mikewarrior2]

pHaestus,

yeah, the article was damn useful. Yes, not touching the backside hurts the reading. but most motherboards are just as inaccurate as the a7v133. Most that are as inaccurate do not have the same amount of "temp compensation" that asus uses.

I'm guessing that even with a probe that touches the backside, there is still massive compression. It isn't too hard to imagine a 2C Diode reading showing up as 0C change on a socket-thermistor.



Mike

 

[JellyBaby]

<< , since no manufacturer has made a reliable, diode reading mb other than >>

Why exactly is that the case?

 

[rogue1979]

While slow responding and not nearly as accurate, the Epox 8K7A can detect smaller temp changes. First of all make sure a regular thermistor motherboard has the sensor in the cpu socket. Second of all, carefully adjust it so that as you mount the cpu in the socket it has to gently push the sensor down to insure it is touching the bottom of the cpu. While it can't possibly measure the true internal core temperature, it is alot more sensitive and accurate that you can imagine. Here again, it is tough to do comparisons with different motherboard brands, although I believe that if everybody adjusted the thermistor accordingly that would help. Then we have to take into account Asus's bad attempt at using an offset temp reading.
I have done extensive cpu temp testing, and I could get a consistent temp difference of 2c by using regular thermal grease vs. Artic Silver II, repeating the test several times with the same results. Switching cpu fans several times told me that at idle the Delta cools 3c better, and 5c at full load with my heatsink and cpu, again giving me the exact same results after testing 3 or 4 times. My ambient case temps vary with room temperature, and my cpu temps vary accordingly just like they should.

Yes, the internal thermal diode is clearly a superior tool for accurate cpu temps. But if used properly, the old thermistor set up can give you alot of useful information on your cpu temps. I am not too concerned about my max cpu temps (within reason). I mostly use the information gathered from the thermistor to gauge cooling improvements in order to achieve a maximum overclock. The thermistor is the only way I have been able to check cpu temps since switching to socket A, and I have been very successful in achieving my overclocking goals using this method.

When I am overclocking Pentiums or Celerons, I use the temp info the same way and it doesn't give me any extra overclocking advantage. I have a Celeron800@1120MHz running in my wife's rig right now, with a Tualeron 1GHz arriving at the end of the week. While I cracked the core of my old Tbird750@1017Mhz and killed it, I have yet to destroy a socket A cpu with heat. My old motherboard Abit KT7 (non-A) is running a Duron 750@1064Mhz in my kids system. Of course I use the old thermistor touching the bottom of the cpu and have achieved many successful overclocks on that rig also.

 

[pHaestus]

As far as manufacturers not implementing it, I would guess that for VIA chipset motherboards it was primarily because the KT266 came out before desktop Palominos (at least I think that is right) and then for the KT266A mobos they are basically the same motherboard as the KT266 with a new northbridge added. While it is trivial for me to solder two wires onto the back of the socket, it isn't trivial to make all new traces and then change the IC that reads temepratures on the motherboard. For the AMD 761 mobos though, there is really no excuse for the KG7 and the 8K7A not including this feature.

As an aside, I would feel comfortable enough in the performance of this diode to build a circuit that shut the system off at 70C (by blocking the Power_Good line of the ATX connector) and then removing the heatsink from the CPU. Of course, I have been wrong before and it seems like a rather pointless exercise (although H|OCP seems to enjoy frying CPUs, I have to buy mine out of my pocket).

 

[MrThompson]

What pHaestus has done here is provide proof that the HS and thermal grease comparisons done with external temperature probes are a bunch of tripe. Many of us have known this. Nevin is probably too much of a gentleman to say, "I told you so", but this is what he has been saying for a couple of years. Well done pHaestus.

As to retrieving valuable information from a socket thermistor, imagine if an external temperature sensor controlled the thermostat in your refrigerator. By the time the external sensor indicated the inside of your refrigerator was cold enough, you eggs would be frozen and your beer cans would have exploded.

I seriously doubt that many motherboard manufacturers will include the ability of reading the diode in their products. Abit and Asus are the only manufacturers that have a history of doing this and even then it?s been limited to very few boards.

 

[rogue1979]

Thankfully, I have never had to utilize this feature, but both my Epox and Abit have bios controlled emergency temp shut down features. They can be set to activate at 60c. So hopefully even with the slow reaction time to rapid cpu temp increases, and the fact that the real core temp is at least 10c higher, shut down will occur at a true 80c, avoiding core damage. Anyone care to shut down the cpu fan and see if it works? Ha! Ha!

 

[Mikewarrior2]

Jellybaby,

I have no idea why.... Perhaps cost, or perhaps(since most users have retail heatsinks, and these heatsinks showed a 75+C results on the a7vtroubleshooting review) they don't want to reveal higher internal diode temps.


Rogue1979,

even with a backside-contact thermistor, it still isn't all that accurate.... however, the hugest issue with socket-thermistors is a) comparisons of temps that really can't be compared due to different reading location, etc. and b) comparisons at major hardware sites claiming that the rsults are accurate, making judgements on products based on inaccurate testing methods.



Mike

 

[sandorski]

Let's hope that the next generation of SocketA mobos all use the internal diode.

 

[Rand]

It seems it isnt entirely correct to say the Fujitsu-Siemens board is the only board to offer temperature monitoring from the internal diode, Asus claimed in a recent email that their new dual processor 760MPX based mobo the A7M266-D does offer full support for the Athlon MP/XP's new integrated thermal diode and clock throttling features.
Supposedly this will feature will be advertised as "Asus COP (CPU Overheating Protection), and can be enabled/disabled in the BIOS if your using an older Thunderbird processor that doesnt feature the internal diode.

To quote their press release for the board:


<< The A7M266-D incorporates special features to protect the condition of your valuable CPU and motherboard. ASUS COP (CPU Overheating Protection) is a hardware protection circuit that automatically shuts down the system power before temperatures are high enough to permanently damage your CPU. The temperature is directly monitored from the thermal diode on AMD AthlonTM MP and AthlonTM XP processors, providing more accurate readings in real-time. >>



I've not as yet had the opportunity to test the board first hand, so I can't confirm this but I'm inclined to believe Asus is being truthful.\
One might hope that Asus continues this trend in future uniprocessor SocketA mobo's.

 

[Mikewarrior2]

Rand,

I stand corrected... I had not considered the asus dual board since it hasn't been extensively available for purchase, or extensively tested.



Mike

 

[pHaestus]

The Tyan Thunder SMP board also supports this feature (and is $500); not certain about the Tiger (the cheaper one). The nice thing about this hack is that ANY motherboard can be modded though.

 

[Mikewarrior2]

Well, the tyan should have internal diode support, but people can get readings with t-birds in place of athlon mp/xp chips.

So, something may be a bit fishy about its tempr eadings.



Mike

 

[swifty3]

Anyone want to take a guess when the next gen mobo's will be out that utilize the internal diode well?

 

[MrThompson]

The A7M266M is supposed to read the diode. They are having problems with it though. Given the limited number of board that can read Intel diodes, I would not expect a rush to the market.

Joe at Overclockers posted this today. I particularly like his final statement "Why some continue to insist that the in-socket diode is an acceptable test instrument is beyond me."