New: AMD-K6 forum for your informationexchange
First i would like to excuse my poor english. Please send me any suggestions or errors to my mail-address at the bottom of the page.
This web page describes the hacking (modification and upgrading) of a Gigabyte GA-586DX motherboard (dual processor with Adaptec UW SCSI onboard) with a AMD K6-2 300 mc/s processor. The improvement is dependent on application, in some case it can be up to 80% (compared with K6-200). Rumors are that it can be made to run even at 400 MHz. (armed with a stronger (higher capacity) voltage regulator or 2 paralell voltage regulators). The K6 is unfortunately not able to do multiprocessing. The board runs only with one CPU!
Disclaimer: I do not take any responsibility for the accuracy or correctness of the following information. It is assured only that with me the described method was tried and had been successful. Your system may suffer damage beyond repair!
Nevertheless it is possible spend little money to enable this board to operate the K6-2 300 successfully.
The
DX-Board has an autoamtic Dual-Voltage recognition. But it works only with 2.8V
CPUs reliable. You can the put in the JP2 Jumper, or, even simplier, put in
a bridge between Pin 1 and 2 of the Jumper. (Who wants to use 3.3V CPUs in these
days) To get the 2.2 Volt core-voltage, you have to use a resistor of about
180 Ohm between pin 1 of the voltage regulator and ground ( Gnd-Pin of the cooler).
This gives together with the parallel R32 (182 Ohm) and R148 (422 Ohm) a Regulator
R1 of about 75 Ohm. Together with the resistor between pin 1 and 2 R33=100 Ohm
you get the 2.2 Volt with the formula (75/100+1)*1.25 = 2.2V


The changes you have to do are marked green in the picture. red marks are for
information only.
If
you use a load resitor of 100 Ohm between B16 and A17 in the CPU-Socket (in the
pic to the upper right), you can measure the core voltage of 2.2V. At the resistor
between B18 and A19 (lower right) one can measure the 3.3V supply voltage. (Don't
regard the red and black in the picture, you measure between the ends of your
resistor, of course). The third 100 Ohm resistor in the upper left is for switching
to dual voltage.
CAUTION: before power on close the Socket! The voltage regulators should not be used without load (empty socket)!
The core voltage can be measured between pin 2 (or the housing of the VR) and ground.
The LX8383A on the GA-586DX is specified for 7.5A. This is normally enough for the K6-2. Peak-loads of up to 8.45A can be done with good cooling. If you get stability problems anyhow, you can use an LT1083CP (about 15 Euro) instead of the LX8383A and/or use an additional CPU-cooler for the voltage regulator.
| Bus frequency (66MHz) | ||||
| CPU [MHz] | factor | BF0 | BF1 | BF2 |
| 233 | 3.5 | 1 | 1 | 1 |
| 133 | 2.0 | 0 | 1 | 1 |
| 200 | 3.0 | 1 | 0 | 1 |
| 166 | 2.5 | 0 | 0 | 1 |
| 366 | 5.5 | 1 | 1 | 0 |
| 266 | 4.0 | 0 | 1 | 0 |
| 333 | 5.0 | 1 | 0 | 0 |
| 300 | 4.5 | 0 | 0 | 0 |
The K6-2 300 is operated with the factor 4.5 (4.5x66=300) in 66Mhz Boards. As there is no BF2-Jumper on the DX-Board, you can use a little piece of loudspeacer cable between W35 and V36 in the CPU-Socket. Since you can use only one CPU, you can put the cable link in the other one.
You can put W35 to ground instead, if you don't like small wires on your board. BF0 and BF1 (Schalter 3+4) are both set (see pic):
Newer K6-2 starting from 400 MHz interpret x2 as x6, so you don't need to change anything regarding BF2. However, AFAIK still nobody suceeded in operating a 400MHz CPU or an K6-III with this frequency on the DX-Board. The maximum reported frequencies are 333 or 366 MHz. The problem seems to be the BIOS.
My DX board announces 486 DX60: -) I have the board revision 3b and the newest officially BIOS available with gigabyte (version 3.43).
To get the max. possible performance, get the DOS utility setk6 of german c't magazine. Thus Write Allocation can be activated at boot-time, which should bring approx. 5% performance gain.
For Windows the Tool CPUIdle is absolutely recommended. Hereby on the one hand the CPU temperature is lowered around approx. 10 degrees. Additionally the voltage regulator is significantly less under stress (approx. 20 degrees cooler) and of course Write Allocation is activated.
Finally use the tool ctcm from the same source, to do several performance tests with the RAM Interface and get an idea how much performace-gain you got from the whole operation.
Have much Fun! I would be pleased about positive acknowledgements!