More K6-266 Results

After posting our initial findings regarding motherboard compatibility with the K6-266, we received several emails claiming that our information was not entirely accurate. We also have received a huge number of questions regarding the 100MHz bus speed and R581A. Some of these questions we have been able to answer, but others will continue to be questions until a few more pieces are in place

We reported that the FIC PA-2007 would not correctly identify the K6-266. This was confirmed by FIC, and they indicated that a new BIOS revision will be available in about two weeks (mid April). The support rep we spoke with indicated that FIC is evaluating their BIOS’s and determining which ones need to be updated, but have been slow to do this.

We received several emails from individuals who took exception to the claims that the PA-2012 would not even POST with the K6-266. A few indicated that they are currently running this combination, though they needed a BIOS update. With this info, we went into the lab and tested 3 PA-2012 boards and 3 K6-266 processors. None of them would work. We tried changing voltages, multipliers and even bus speeds, but no dice. A POST card showed no life whatsoever. When we replaced the K6-266 with a K6-233 (and changed the voltage to 3.2v), the board booted fine. If anyone would care to provide the ‘secret’ for getting this board to run with the K6-266, we would be grateful for the information.

At this point, our list of compabible motherboards is rather short: M Tech R581A, Gigabyte GA586SG, AOpen AX5T-3 & AOpen AP5T-3 (requires 1.60 BIOS) and the FIC PA-2007 (though it is identified as an 80486DX).

100MHz issues

After our initial success with the first attempt at 100MHz, we decided to perform some more in-depth testing to find out whether variations in motherboards or processors would prevent some from achieving this bus speed. Our findings indicate that the short answer is ‘Yes’, but with some questions still to be answered.

We used 5 R581A boards, and 5 K6-266 processors for this test. We also had 5 different manufacturer’s SDRAM modules, and our ‘trusty’ 50ns EDO modules from Crucial Technology. We quickly determined that none of the SDRAM would allow the system to boot at 100MHz. The Crucial SDRAM actually started to initialize Windows, but the others wouldn’t finish POST. The 50ns EDO, however, proved to be the only memory that would allow Windows to fully load. It is fairly obvious that PC100 modules will be required to run at 100MHz with SDRAM.

Armed with several EDO modules, we began to work on getting Winstone results for various 100MHz settings. Unfortunately, we soon found that the 100MHz speed is not a ‘given’. One K6-266 processor would not run more than a few minutes before crashing, regardless of what motherboard it was installed on. Two of the processors would work on a few motherboards, but not on the others. The other two processors seemed to work on all of the motherboards.

Though we tried several dozen times with different combinations of motherboards and CPUs, we were unable to get any of the Winstone runs to complete. Many ran more than halfway through (10 to 15 minutes), but eventually all them would either time out or get some ‘illegal operation’ message. What is very interesting is that we could reformat the drives and install Windows & Winstone without any problems or errors.

It appears at this point that there is no one single ‘solution’ to the 100MHz issue. The motherboard/chipset must be able to run at that speed, the processor must be able to handle the signals coming in at that speed, and the memory must be able to deliver the data without ‘dropping’ bits. We believe that this is what was causing the various timouts, GPFs and illegal operation problems.

Not having any PC100 SDRAM to use prevented us from testing out our theory about why the Winstone tests would not complete. The timing issues are very critical at these bus speeds, and even the 50ns EDO apparently could not keep up with the intensive data transfer requirements of Winstone tests. We believe that the final piece to this puzzle is to use PC100 SDRAM, though the cache speed might have a little to do with this as well.

We have been expecting some Micron PC100 SDRAM modules for a week or so. As soon as these are delivered we can test again to find out if there are still more hurdles to overcome before we can finally say that the 100MHz bus speed is achievable by anyone. Right now, it looks like a hit-or-miss affair, with some being lucky enough to get a ‘matched’ motherboard/CPU pair. At this point it appears as if our very first test just happened to have the correct two components to allow Winstone to complete successfully.

On a related note, it seems that someone has ‘found’ an undocumented 100MHz jumper setting for the GA586SG motherboard. We will be testing that very soon, and reporting the results. Sometime in the next week or two, we are supposed to be receiving a motherboard with the MVP3 chipset and official 100MHz support. This is when the final pieces of the puzzle will fall into place, and all of the 100MHz questions will be answered (and if you believe that…)

Be the first to discuss this article!