Re: Socket 7 upgrade article
I was reading a recent article about upgrading older socket 7 systems using AMD K6-2 processors. I have acquired a K6-2 350 and want to upgrade an ABIT IT5H motherboard currently using a K5. Has anyone tried this combination successfully?
If so, any advice on pitfalls?
If so, any advice on pitfalls?
I had an ABIT TX5 MB with a Intel 233 pushed to 266. Tried to get to 299 but it never made it that far. Then I bought the same CPU, an AMD K6 350. Tried to push the system to 350 and was successful in running it at that speed. However, when running SETI I noticed that my performance decreased! It was now taking approxmiately 35 clock cycles to do a floating point computation, where previously it had completed in 12 cycles. I lowered the speed back to 266 and the performance per clock cycle came back. The only conclusion I can draw from this is that the L2 cache could not keep up with the high processor speed. I did not gain anything by speeding up to 350 on the old MB.
The article was "Socket 7: Fit For Years To Come"
Specifically, I am wanting to use an AMD K6-2/350 (2.2V core/3.3V I/O). According to the article this will work with an ABIT IT5H despite the fact that the IT5H manual specifies a minimum core voltage of 2.5V. I am not sanguine that this great of a voltage disparity is acceptable.
Also according to the article bus multipliers of 5x or greater may be used. Again the IT5H manual specifies a maximum speed of 3.5x and I am not certain how faster multipliers can be attained.
Any experiences with this type of setup would be appreciated. Especially any experiences with the voltage/speed levels the K6-2/350 can handle.
I have a Tyan 1573 ATX board with a TX chip set and had an AMD K6-2/350 running on it today. It usually runs with a P5/233. The board is not supposed to work with K6 chips, but it did. The FSB only goes to 66 and the multipliers only to 3.5, but if you set the multiplier to 2, the 350 will "see" it as 6 and that makes it ~400 (66x6). The first time the computer posted, it showed the chip as a 486 at 66 (2x66), but SiSoft Sandra showed it as a K6 at 401. After the first post, the computer does not show any chip/chip speed, just all the other normal post stuff. Strange huh! I ran benchmarks with the 233 and the 350 (400), the K6 was FAR better. I ran the burn in program in Sandra for an hour and all was good. Also ran some games that work the system (or at least that I like)CFS and Motorhead and both worked great, no crashes and much higher FPS. BUT, then I decided to run the Prime95 program, and it kept saying that there were rounding off errors. This was the first time I had tried the program to "work" the system. So I guess there is something not quite right deep down somewhere.
Bottom line for me is that when my K6-3+/450 comes in this week for my one computer that currently has the 350, the 350 is going in the 233 machine. It runs all the MS Office software and the games I play without locking and Sandra shows that the improvements are real.
If anyone out there knows why Prime95 is finding an error when Sandra is not, please post!
Sorry this was so long, but I hope it helps.