Try a lower CPU multi (e.g. 2 or 2.5) and see if the 105 FSB setting gets going...

300 MHz on a 233 cpu is quite impressive!

Most SS7 boards have caps around the cpu sockets I found. If you are a handyman you can "cut" an existing cooler into shape...

For my Pentium 200 testing I went passive and it wasn't an issue. The cooler got hot to the touch but otherwise all was fine

Random thought, but when you tried 3x105, did you check the speed with something like CPU-z, rather than just at the POST screen? It could be that it works fine, but the BIOS just wasn't set up to report odd speeds like that...

Mau1wurf1977 wrote:

300 MHz on a 233 cpu is quite impressive!

I can get it to run at 333 MHz (3.5x multi @ 95 FSB), but Windows 95 / 98 goes insane when it's starting up. I have briefly tried lower multis. I'll retry this again soon...

Old Thrashbarg wrote:

Random thought, but when you tried 3x105, did you check the speed with something like CPU-z, rather than just at the POST screen? It could be that it works fine, but the BIOS just wasn't set up to report odd speeds like that...

Honestly, that was going through my mind when I saw the BIOS POST screen report 300 MHz. Unfortunately, tiredness and laziness got the better of me, and I didn't allow Win98 to boot up to the desktop, for me to then run Sandra 2002 Pro and check this. (BTW, I don't think CPU-z runs in Win98 ?)

Old Thrashbarg wrote:

Random thought, but when you tried 3x105, did you check the speed with something like CPU-z, rather than just at the POST screen? It could be that it works fine, but the BIOS just wasn't set up to report odd speeds like that...

You were right. 😀 I've set the mobo's jumpers back to 105 FSB / 3x CPU multi, and although the BIOS POST screen reports 300 MHz, Sandra 2002 Pro reports the CPU speed as 315 MHz. And the system is stable! Initially, I used 3.2V for the CPU, but that produced occasional graphical glitches on the screen, and also occasional freezing. I increased the VCore to 3.3V, and that has cured those problems.

I ran a few benchies, and compared the results with some older 100 FSB / 3x multi: 300 MHz scores. The new 315 MHz scores were better. Obvious I know, but there's one problem I have ATM. Sandra 2002 Pro oddly reports that the FSB is running at 66 FSB. I don't think this is correct. It must be 105 FSB. Perhaps there is some kind of "communcations failure" between the mobo's BIOS and the Win98 operating system?

I've learnt 2 valuable lessons today. Firstly, the BIOS POST screen may not always tell you the correct speed of the CPU. Secondly, if you've been doing crazy overclocking, and the mobo subsequently appears to have died, remove the CMOS button battery. I'm not sure how long to wait, but I guess a couple of minutes at least. I think sometimes if crazy overclocking goes wrong, it messes up the BIOS. Removing its battery clears this "corrupted state".

I'm going to run some more benchies, and make a screen shot of them. I'll post back today...

The system appears to be stable at 315 MHz. (I used a vcore of 3.4v. I also set the BIOS settings to "fail safe".) There are a couple of odd things. Firstly, I stress tested the CPU by running SuperPi.exe. After the 1M test had successfully completed, the time given was about 8 minutes and 50 seconds. However, as can be seen in the image below, the 1M time is reported as 4m and something.

Ugh, I just noticed in the image that the CPU benchmark has been covered up by another window. Oops. I think the P233 mhz has definitely reached its limit, at 315 mhz. 333 mhz is far too much. I cannot get this speed stable at all. Overall, 300 mhz is a much better option, I think.

BTW, for the heatsink, I'm using the "slimline" copper cooler found on a Powerleap adapter! It just about fits, next to some awkwardly placed caps near the CPU socket.

With the system at 315 Mhz, and the BIOS options set to "fast", DOS seems to work well. Doom, 3dbench2 (144.7 score), and Speedsys all work as they should. Like Sandra 2002 Pro, Speedsys also says that the bus is 66 Mhz.

I ran Quake 1 many times (shareware demo), using the "timedemo demo1" console parameter, and the system is stable. I'm using a Voodoo3 2000 AGP video card, and get these scores:

55.7 full screen
61.7 two notches smaller than full screen

BIOS options set to fast.
61.7 full screen
68.5 two notches smaller than full screen

3dbench2 170.8

Good scores...

Though a K6-2 would kill it 🤣

E.g. at 350 MHz a K6-2 scores 397 in speedsys

I'd be interested to see a Quake timedemo for a K6-2 to compare to the above results. The K6-2 had a weak FPU for its time especially if you weren't using 3DNow which quake doesn't AFAIK. FPU performance is quite important for 3d games like Quake.

I wouldn't be surprised if the Pentium MMX is equal or faster in Quake than a similarly clocked K6-2, even though the K6-2 beats it in synthetic benchmarks.

Here's some interesting reading, albeit for a K6 not a K6-2:
http://www.tomshardware.com/reviews/intel,22-3.html
http://www.tomshardware.com/reviews/intel,22-5.html

I would be really interested to race this "hot rodded" Pentium MMX CPU (233 MHz; OC'd to 300 MHz and also to 315 MHz) against an AMD K6-2 CPU. I don't have any K6-2 300 MHz or 350 MHz CPUs, but I am prepared to buy them. Do I buy the auction listings titled "AMD K6-2 300 AFR" and "AMD K6-2 350 AFR"? Are these the CPUs to get? Thanks very much for any confirmation on this.

I just tried another experiment. Increased FSB to 110. This should yield a CPU speed of 330. (110 x 3 multi). The BIOS POST display said 300, but I've now learnt to ignore this, as it could be wrong. And it was. Speedsys says 330, and its result is faster than the 300 test. But it's too unstable at this speed. Quake doesn't run, and Doom falls over after a short time. IMHO, 315 is the overclock speed limit for the 233 Pentium chip.

Edit: I'm back to 315 MHz for the CPU. Doom bench = 84 FPS (891 realtics), full screen mode.

Yes those CPU's should be fine.
Here is an interesting site detailing the different K6-2 models.
http://www.cpu-central.com/Articles.asp?artic … 96&decor_int=23

The AFR denotes socket 7, 2.2 volts, 70 degrees C max temp.

There are some 66MHz bus versions which you probably don't want, they are marked AFR-66. If they aren't marked this way they should be 100MHz bus versions.

I'd definately say you've hit the limit of that 233MMX with air cooling especially if you are at 3.3 volts.... any way of telling what sort of temperatures you get?

Re: K6 chips - thanks a lot for the info. I'll order these 2 chips today. Re: volts - actually, I'm using 3.5 volts ATM! Seems OK. 😀 Re: temps. I would love to know what the CPU temp is. There's a sensor inside the CPU socket area. However, the BIOS options set up area contains no temp reading value for me to look at. There is a "PM Temp" field, and I've set this to "Auto". The other choices are Disabled, and then a range of "cut off" values, for example 80C.

When I get to the Win98 desktop, I've tried launching both Sandra 2002 Pro and Everest Ultimate, to look for the temp reading, but I cannot find one.

There are 2 types of K6-2:

K6-2 which is a desktop chip and K6-2+ which is a mobile chip. I have one of the K6-2+ 550. A beauty of a chip. Both my SS7 boards support this CPU and it really brings some life back into Socket 7 🤣

The K6-2+ has on chip L2 cache and also runs on lower voltage. The K6-2 can also be had with up to 550 MHz. I have a 500 MHz K6-2 but never tested it as the K6-2+ is the much better chip.

There is also a K6-3+ which has double the L2 cache and its even harder to find.

There is even a config.sys driver where you can set the CPU multi when you boot the machine 😀

It scores 446.5 in 3dbench 1.0c and here is a speedsys benchmark:

Another interesting Chip is the Cyrix 6x86 and MII. They don't clock very high but are super fast per clock. One of the few chips ever than managed to beat Intel clock per clock.

Very interesting regarding the "plus" + chips. In the GA-5AX manual, it says
"Support Intel Pentium, MMX, Cyrix/IBM, 6x86MX, MII, AMD K6, K6-2,
K6-2+, K6-III , IDT Winchip 2, IDT Winchip 3, RISE MP6 CPUs."
It supports the K6-2+, but doesn't mention the K6-3+. Perhaps it does support them, with a later BIOS?

However for these faster speeds, I'm happy to go Slot 1-P2/P3. My primary interest with Skt7 is with the first generation Pentiums. Vogons user prophase_j's P55C overclock thread really fired up my imagination about the Pentium MMX chips!

Yup the F4 BIOS has support for K6-3 + and also larger HDDs. What board version do you have? As long as it's not one of the early ones you should be fine.

BIOS here: tw2005.giga-byte.com/Motherboard/Suppor ... 205.2).htm

Yea there are so many chips to play with and yea Slot 1 is the way to go if you need more grunt.

I got a bunch of Pentium 200 chips (the NON MMX type). Researching upon it I read that this chip is a bit rare as it cost a lot more than the 166 and most people got a MMX at that stage anyway.

It makes sense but not sure if these rumours are true.

Thanks a lot. I flashed it to F4. (It was on F3 when I got it.) BTW, I think the board is the last one they made: it's Rev 5.2.

Excellent! Good board. It's a keeper 🤣

I have ordered the two K6-2 CPUs: 300 and 350 MHz. But I've just found a K6-2 400 in my attic. I decided to give it a spin (benchies below). Also, I've just received an interesting cooler. It's a copper cooljag socket 370 heatsink. It's heavy (375g), but has a very small "foot print". It just about fits on the GA-5AX, see photo. Curiously, it also fits on to a Powerleap adapter! If you want to have the ultimate Powerleap cooler, look no further. There's 2 of them left on ebay, here. (Looks like it's postage to Europe only however.)

I ran some Quake 1 (shareware demo) benchies in DOS 6.22, using a Voodoo3 2000 video card, for the K6-2 400 and also for the overclocked (to 315 MHz) Pentium MMX 233 CPU. I was incredibly careful when I did these tests, because of the results obtained. They were so surprising, that I redid both tests again by reinstalling the processors again to run all the tests twice.

Edit: I would like to try some more OC'ing fun. What is the upper limit of the K6-2 400 chip? I would like to see what it can handle! 😀