Unless the adapter was wired wrong to begin with (which likely would've killed the motherboard), there's not much that could go wrong with 'em.

BTW, unless you're really tight on space, you usually don't need those adapters... you can just leave the extra 4 pins hanging over, if they're not already detachable from the main connector.

I suspect what might've happened is that the QDI board didn't recognize the VID setting on mobile chip and fed it too high a voltage. Either that, or you didn't get the heatsink seated as well as you thought.

Re: the PSU 24-pin to 20-pin PSU adapter. Something a bit weird has now happened with the QDI mobo. When I switch on power at the mains, the power + mobo gets switched on immediately - the mobo no longer waits for me to press the little power switch gadget attached to the 2 power pins on the mobo. Another weird thing: I am testing some SDRAM using memtest, and every stick is showing an error.

Unfortunately, on some of my mobos, I am tight on space. (ATM, the QDI mobo does allow me to hang the extra 4 power pins over the edge of the mobo's power socket.) Rather than use the 24 to 20 pin adapter, is it feasible to cut in to the 24 pin PSU plug, in order to seperate the 4 unwanted pins? I think I'd rather do that, than rely on an adapter which could be of dubious quality.

Re: the death of the XP-M CPU. I have put this CPU in the QDI board about a dozen times without issue. And so I am inclined to suspect your suggestion about the heatsink not being properly seated. Say it wasn't seated properly. Would the mobile barton get killed within about 20-30 seconds? Because that's how long I had the screen active, before it went blank and I lost the CPU forever.

Edit: The QDI mobo seems to have gone a bit more mad. It now thinks the XP 2000+ CPU is a Duron 1400 Mhz CPU. About 5 minutes ago, it correctly thought it was an XP 2000+.

Old Thrashbarg wrote:

I suspect what might've happened is that the QDI board didn't recognize the VID setting on mobile chip and fed it too high a voltage. Either that, or you didn't get the heatsink seated as well as you thought.

I agree 100% and second the possiblity if a cracked core 🙁

I guess we'll hear after his inspection, super bummer for RG.

Quick update. I switched off power, but also decided to pull the power plug out of the wall mains socket. (Something to do with "power cycles" ?) I now switch on the mobo again, and it correctly sees the currently installed CPU as an "XP 2000+", and not a Duron 1400 Mhz. I am now running Memtest, and unfortunately it is showing errors on SDRAM sticks I successfully tested on the Abit mobo only this afternoon.

BTW, I am looking at the dead XP-M mobile barton CPU. I cannot see anything obviously wrong with it. However, the light pink head surface has some small areas where this pink color is missing - I am not sure if those missing color areas where there or not before it got killed. Also, the surface does look very slightly wrinkled. But apart from that, it looks OK. BTW, for *all* of my Athlon tests that I have done over these last months, I have never used any "goo" on the CPU head to help out with the heating. I know this is wrong, but I guess that I have been lucky up to now, and I haven't killed any other CPUs. The heatsink(s) I use are really heavy. Perhaps I squashed the mobile barton CPU when I was putting it on?

Edit: I have never seen memtest 86 show error messages before, so I may be misinterpreting them. I think it's telling me that Test 5 fails, with 2 ECC correctable errors. This is the same message that keeps appearing, regardless of which SDRAM sticks I am testing. BTW, now I have corrected the "power cycle" problem, when I switch on power to the mobo, it doesn't immediately start. It waits for me to press the power button on the mobo's 2 power pins.

Hmmm, sounds like your Athlon CPU's have to suffer a lot... 🙄

For testing purposes and quickly trying out CPUs on Socket A boards I recommend using a lightweight aluminium heatsink. With heavy heatsinks the fragile CPU die is easily damaged around the edges.

The holding clip shouldn't be too strong, those grabbing with only one prong on each side are more convenient and less risky to (un)mount.

Without any kind of thermal paste the heat transfer to the heatsink is very inefficient. If the base of your heatsink is just slightly uneven, this could be almost as bad as mounting no heatsink at all!
Even the cheapest, most basic silicon grease is better than nothing and can easily be removed from heatsink and CPU.

BTW, the temperature sensors on old boards are rather inaccurate, mainly because they can only measure surface temperatures. The temperature inside the CPU may be much higher than the board reports.

And finally, on mainboards with damaged capacitors, everything is possible. On Socket A designs, most of the problematic caps are used in the voltage converter for the CPU.

About the ECC errors: I might be wrong, but I don't think KT133A even supports ECC.

Thanks a lot for the advice. I will definitely get a lightweight aluminium heatsink for quickly trying out CPUs. I will look for one that has one prong on either side. I might get one with a thin copper base, but not a heavy solid copper design.

Re: the ECC errors - I noticed that the BIOS set up area had an "ECC L2 cache check" (not exact words) option (Enabled/Disabled). I tried both values, but memtest always displayed "ECC errors". I double-checked the SDRAM sticks again on the Abit board, and memtest passed OK.

Briefly returning to the QDI mobo + XP-M CPU "incident" - 1) I didn't do my usual "fail safe" operation of popping out the mobo's button battery when I tried the mobile barton CPU. (I tend to do this now, after I smoked 2 t-breds in an Abit board.) Instead, I just cleared the mobo's CMOS jumper (twice!) before switching on power. 2) When the large heavy copper "boat shaped" heatsink was attached to the CPU, I wondered if I had positioned it correctly. Rather than unclip it, and reattach it, I simply "slid" the heatsink across the CPU in a horizontal movement. Perhaps doing that damaged part of the CPU? I notice that this large "boat shaped" heatsink has 4 little pads on its CPU mounting surface. It's possible that these 4 pads, in conjunction with the CPUs 4 pads, caused the heatsink to "lift off" the surface of the CPU head, and consequently it wasn't touching the CPU head correctly or firmly enough?

It does not, only detects it and can work with it, but does not correct anything.

I took a bit of a risk, and put my only other XP-M mobile barton in to the QDI Kinetiz board. It's a 2200+ rated CPU. It was a bit risky, because unlike the first XP-M CPU which it killed, this one is even lowered powered at just 1.35V. (The first CPU required 1.45V.) I popped out the mobo's button battery, and also cleared the CMOS jumper twice, just to be sure. (I'm sure popping out the button battery helps when swapping over different architecture type CPUs, such as going from a t-bred to a mobile barton, etc.)

The mobo POSTed OK, and the BIOS POST screen reported the CPU as a "Mobile AMD Athlon 800 mhz". Within the BIOS set up area, the core voltage says 1.51v, which is probably just about OK. (Sandra says it's 1.57V, which is a bit high.) I don't think you can adjust (lower) the core voltage on this QDI board. The voltage and OC'ing control is rather limited on this "budget board".

The good news is that the mobo is being powered using that 24 pin to 20 pin PSU adapter gadget. It seems to work OK. I tried cutting the PSU connector plug, in order to "push away" the 21 to 24 (4) unwanted pins, but I couldn't cut through the plastic too well. One thing I did was - not use that awkward "boat shaped" heatsink. I think sometimes it doesn't sit properly on to CPUs. (I used a different heatsink instead, but it's still very heavy, and I need to buy a cheap light replacement ASAP.)

Unfortunately, memtest still errors when I run it on this board. This time, I got it to check 1 stick of fully working 256MB PC133 SDRAM, and it went completely nuts reporting hundreds of thousands of errors.

The Windows 98 desktop seems stable, but 3DMark 2001 s.e. quits back to the desktop during testing. Generally speaking, the BIOS options seem OK to me, and so it's disappointing that 3DMark doesn't work. It's possible that as I am using the same HDD used on the Abit board, when I now attach it to the QDI board, it's best to reinstall things like the VIA driver, etc? When the desktop first appeared using the QDI board, it showed an error message saying that a registry entry was fixed and the machine needed to be rebooted.

Edit: 3DMark 2001 s.e. did complain about a corrupted file, before the first time I ran it on the QDI board. I'm running Scandisk now, and also I will reinstall this benchmarking app. I accidentally knocked the drive yesterday. 😦

Edit 2: The oddness continues. (Hehe, it's a VIA board! Only joking.) I decided to run 3DMark 2001 f.e. (not s.e.). That quits too, but this time it shows an error message:

File \data\gs2\L\WO\TEXTURES\NU4IN_DIFFUSE0030.DDS is corrupted

So, both s.e. and now f.e. are complaining about corrupted files. They can't both have suddenly developed file related errors. Is this one of those infamous VIA chipset problems? I ran Scandisk, and did a thorough scan including the surface scan option. No errors!

I think I will uninstall -

nVidia graphics driver
Both Breese patches
VIA 4-in-1 driver (v. 4.43)

Then reinstall -

VIA 4-in-1 driver (v. 4.43)
Both Breese patches
nVidia graphics driver

Edit 3: I am beginning to suspect that the HDD and/or mobo is damaged. When I boot in to Windows 98, it says that the registry cannot be accessed, and the machine needs to be rebooted to fix the problem. This has occured about 10 times today. I run scanreg in DOS, and it can't find any errors. Also, Scandisk passes OK. If I try and access the Control Panel -> System - > Device Manager, I get a blue screen error message. The System Device screen then appears, but appears corrupted. The machine then shuts down.

Check for visible problems, mainly caps and if you are able, measure the ESR - that would narrow down the HW failure - also dust may be a problem in this, and of course use good power supply, wich I believe you have.

The almost continuous boot up error message of "windows cannot access your registry - and needs to restart your computer" has now stopped. What's interesting is that inside the System Devices area, I now cannot see the special LLA driver anymore. It's possible that the registry scan/fixing operation has removed it.

The caps on this QDI board look in very good condition. Also, there is only a bit of dust on the board. Generally, it looks in good shape. And yes, I am fairly certain that power supply is OK, because I'm using two quite good ones. Also, how can I measure ESR? What is ESR? Thanks.

I reran 3DMark 2001 s.e., and I noticed on the "nature test", that some of the background "mountain graphics" were flickering strangely. I've seen weird stuff like that before, when experimenting with the Abit board. I think it's either an imcompatible BIOS setting, or it's the CPU being fed too much juice. If the CPU wants 1.35V, and it's being fed about 1.55V at the moment, then that could be the problem. (Unfortunately, I can't reduce the core voltage in the BIOS!) What I could do now is abandon this XP-M CPU, and just put in a t-bred instead..

Edit: I tried the t-bred. It's no good. I still keep getting strange file corruption messages from 3DMark 2001. I'll dig out the Abit board again, and make sure that all components (CPU, HDD, DVD-ROM drive, RAM, FDD, graphics card) are working OK. If they are, then I think something has gone wrong with the QDI board.

Edit 2: All I did was remove everything (eg HDD, graphics card, etc) from the QDI board, and attach it to the Abit board. I switched on power, and when the Windows 98 desktop appeared, a message said that a VIA PCI bridge? device was being installed. Sorry, I forget the exact wording. I wonder why that wasn't automatically installed on the QDI board? I ran 3DMark 2001 s.e., and it worked. No quitting back to the desktop. No strange file corruption error messages. Also, I ran memtest on the SDRAM stick taken from the QDI board. That passed OK. I really think that when the QDI board killed the XP-M mobile barton, it also fried something on the mobo which is now causing strange instability problems with it.

ESR states for Equivalent Series Resistance - it basically defines the impendance of certain electronic devices, mainly inductors and capacitros. I really don't know what level should be reached to know when the cap is going bad, just google it. But basically, this is a random behavior and it may be caps with higher Ω values than the others (maybe the whole lot), you will easily find out yourself...also you need an ESR meter 😀

I have finished testing the QDI Kinetiz board, and I have come to the conclusion that when it fried the mobile barton CPU, it also damaged part of its RAM processing functionality. Every fully tested + working stick of SDRAM that I have tried on this board in conjunction with Memtest results in Memtest reporting errors. Also, when I run 3DMark it complains about corrupted files. This 3DMark FAQ comment is interesting -

Q: 3DMark2001 SE complains about Corrupted DDS files!
A: This might happen if you have overclocked your CPU or have faulty memory. Try to lower your CPU's MHz and/or change your memory. (FAQ found at: http://www.futuremark.com/support/3dmark2001/)

The SDRAM is not faulty, but I think the way in which the mobo "deals with it" has now become faulty.
(Also, I ran Prime95 for about 20 minutes, and it didn't fail.)

5u3 wrote:

Hmmm, sounds like your Athlon CPU's have to suffer a lot... 🙄

Here are three of my "best" Athlon squashers! 😉

Without any kind of thermal paste the heat transfer to the heatsink is very inefficient. If the base of your heatsink is just slightly uneven, this could be almost as bad as mounting no heatsink at all!

This strikes me as a great understatement, so I'll emphasize it. Running an Athlon without thermal paste will kill it. It's not a matter of 'if' but 'when'. If you've been doing it that way, I can only say that you are extremely lucky to any working CPUs left.

Those Athlons put out quite a bit of heat, even the mobile chips are 35+ watts, and moreover they are quite sensitive to heat. Many of the Socket A coolers were barely adequate even with thermal paste.

The sensitivity to heat is a combination of effectively zero thermal protection (the BIOS stuff is nigh useless), no significant power management, and the lack of a heatspreader with paste already underneath. In retrospect, it is astonishing that they didn't have a heatspreader. The no-heatspreader era for Intel and AMD was a pretty strange time considering they had die protection on older CPUs.

Whether or not you'll toast one without paste is dependent on the quality of the heatsink though..... The small die may have significant metal contact if the heatsink is flat enough. I'm sure it will run hot and have worrisome hot spots across the die, too.

As long as you're under ~80 C you actually will probably be fine. I've seen this out in the world 🤣.

The thing is, most heatsinks aren't nearly flat enough. Plus, a lot of the CPUs themselves had a little bit of a curve to the die surface. While that's all well and fine if there's paste there to take up the difference, you're asking for trouble when there isn't.

Some Athlons can survive just fine at a reported 80C (which may actually be as low as 70 given the variances in the temp sensors), especially the Palominos which were hot no matter what you did. Technically all of the series should be able to handle up to 90 or 95C. But... there are also a great many of them aren't nearly so hardy.

Absolutely. Running without paste is stupid. Everyone should have a tube of cheap gunk around that can be wastefully used for our crazy testing stuff. You can get a huge tube of Arctic Silver Ceramique for very cheap if you shop around. (for example)

But really, Palomino heat, like so many other "hot" CPUs, is exaggerated. They are cooler than much of the P4 & A64 lines and definitely cooler than some multi core chips. I've had to work on a few Pallys rather recently and found I could keep them under 60C with even a typical AMD OEM cooler. Sure they are not at all like a luke warm Tualatin or Coppermine, but really now.

I actually have an 1800+ at home running with a 7-volted AMD OEM cooler. 😀

What's disappointing about these old chips is their lack of PowerNow/CnQ so they sit there sucking down lots of power doing nothing. I found a nice program that enables idle S2K bus disconnect though. Not many boards bothered to implement it natively. This drops temps rather dramatically actually.
http://www.silentpcreview.com/forums/viewtopi … oftware+cooling
(specifically S2kCtl http://www.stargaz0r.nm.ru/)

I've got an idea - why don't I try using some thermal paste?! 😉 How about these tests:

Test "ingredients" -

• QDI board
• Palomino 2000+ CPU
• Noctua NT-H1 thermal paste ( http://www.noctua.at/main.php?show=productvie … ts_id=13&lng=en ). I've had this tube, unopened, for over a year now. I think it's about time I crack it open.

Tests -

1) No paste.

a) Measure temperature after 5 minutes of Windows 98 idle desktop usage.
b) Run Prime95 for 5 minutes, then measure temperature.
c) Run coolon.exe for 5 minutes, then measure temperature.

Allow system to cool.

2) Apply paste.

Repeat tests a) to c)

Any comments or suggestions please? I'll do these tests tomorrow morning..

The biggest concern with not using paste is that there will be areas of the core that are completely open to the air. It is unavoidable. This will create hot spots compared to other areas. And as physics goes, areas that are significantly hotter than others will expand more and other bad things compared to cooler areas. The crap mobo thermal sensor will be oblivious to this kind of thing. 😀

I've got an idea - why don't I try using some thermal paste?!

Better idea: rather than 'trying' it, just start using it as a matter of course. You already have some, and it's easy to get more for cheap, so there's really no excuse for taking the risk without it. There's no need for tests about it... there have already been a great many tests showing that the stuff is necessary, proven by a great many people with much greater expertise and much more expensive measurement equipment than you have.

Seriously... I'm trying not to be harsh here, but with as many dead CPUs as you have, and all the random instabilities that seem to befuddle you... logic would normally suggest that a person take a step back and consider whether he's doing something wrong, because nobody has such consistently bad luck. While thermal paste probably isn't the cause of all your issues, you really can't even pretend to do any troubleshooting when you don't bother to include such a basic step as that.