Well, before anything else, I'd replace the capacitors. The bad capacitors are probably less of a factor in this case, since you have the separate VRM on the Powerleap adapter, but they certainly aren't helping anything. And every little bit is going to count for what you're trying to do here.

The fact is, 155mhz on a BX board is really pushing the limits, and it doesn't surprise me at all that it won't work. Actually, it surprises me that 150mhz works. You're right that the cheap RAM probably doesn't much like 150mhz+ operation... usually you need good CL2 PC133, and even then you generally have to loosen the timings a bit to get those kind of speeds. But you're also forgetting the AGP... 155mhz gives you 103mhz AGP speed, and it's very rare for a video card to handle that. And even if you get AGP and RAM worked out, there's still a good chance that the northbridge itself can't handle the speed. Honestly, I think the CPU is the least of your problems... 1.65-1.7ghz is usually achievable on a Tualatin, with a voltage boost.

Like Old Thrashbarg said, you're running the AGP bus at 103MHz.
My question is: what AGP card are you using that survived that frequency and POSTed ? I'm curious...

Wow, I really didn't realise that I was pushing my luck with this overclock attempt. And the AGP part of this overclock equation is much more important than I appreciated! It's an ATI card, with Rage Pro Turbo written on the main chip. Having just looked at this card, I just noticed that the 3 pin "INT VGA" jumper on it does not have a shunt. I'm sure these cards have a shunt on this 3 jumper pins group. My guess is that it either means "Give this card an interupt" or "Don't give this card an interupt".

You should probably just move over to a PCI video card. The 155mhz bus speed would give 38-39mhz on the PCI, which is still within reason for most cards.

Yep, will do. I'll try that first thing tomorrow. Also, I'll see if I can find a CL2 stick of PC-133 SDRAM...

Apparently Nvidia AGP cards are more resilient to overclocking.

Green cap near the PSU header, on the pic it looks like it's bulging.

Apparently Nvidia AGP cards are more resilient to overclocking.

Maybe, but the old Rage cards are typically pretty tolerant too. At 103mhz, though, you're pretty much going to have to hand-pick cards... even amongst the models that are reputed to handle high AGP speeds, you're still probably going to have to try quite a few cards before you find one that will run. PCI just makes things easier, and there are plenty of decently performing options to choose from.

I tried a couple of things to get 155 MHz FSB working. I changed the VGA card, to a PCI Virge. I also found a good stick of CL2 SDRAM. As before, the 150 FSB setting works fine, but the 155 FSB setting freezes during the RAM count, at the BIOS POST stage. I tried increasing the BIOS chipset voltage setting from 3.3, to 3.4 and then to 3.5, but this did not help.

However, changing this value from 3.4 to 3.5 allowed the RAM count to continue for several seconds longer than previous tests. There is one more value for this chipset voltage option, which is 3.6, but I don't want to try it, as getting to 155 FSB simply isn't worth the hassle. If 150 works fine, then that's good enough. 😀

Edit: Re: caps replacement. I'll need to investigate what tools I need for this. ATM, I've got other bits of junk I'd like to get, before I spend any more money on a soldering station and caps, etc.

Introduction
I was going to do this post as a seperate thread, but then I remembered I already talked about this mobo. So, I have appended this post to the end of this thread. I am testing an MSI Microstar MS-6163 Pro motherboard. It uses the popular retro Intel i440BX chipset. It's a slot 1 mobo. I've got a Powerleap adapter in it, with a 1.4 GHz Tualatin server (512K cache) 133 MHz CPU installed. I'm using 1 stick of 256MB SDRAM, and an AGP Voodoo5 5500 VGA. Highlights of the BIOS POST information are:

Award Modular BIOS v4.51PG
W6163MJ V3.10 101100
10/11/2000-i440BX-W977-2A69KM4KC-00

1) Link to the latest BIOS. (Click on Download -> BIOS)
2) Link to the manual. (Click on Download -> Manual)
3) Link to a review of this mobo.
4) Link to the mobo's Intel chipset driver, for Windows 9x.

Notes
1) The latest BIOS download file is called w6163v3a.exe. It is version W6163MJ V3.10 101100. The .10 part is also written as .a, because the letter a in hex is 10 in decimal. The BIOS coding datestamp is 10/11/2000, but it appears that this file was released on MSI's website later, on the 2001-05-11.

2) The file name for this English manual is E6163-20b.exe.

3) They seem to like it. 😉

4) The filename is called infinst_enu.exe. You don't need to install this chipset driver, if you are using Windows 98 SE. This is because the Windows 98 SE CD-ROM has all of the updates on the disc, and will install them automatically. When you visit this webpage, you need to select the operating system you want from the "Select an Operating System" choice box.

When I installed the Intel chipset driver for Windows 95, and after the PC rebooted, I seem to remember getting some kind of fatal error. I reinstalled Windows 95 again, and reinstalled the chipset driver again, and that problem did not happen again. I wonder if it was some kind of temporary "clash" with a BIOS setting?

General testing
I ran WCPUid by H.Oda (wcpu33.zip) in Windows 95, and here are three screenshots from this utility.

Overclocking
The BIOS set up area allows you to enter in an FSB value of 155, but I cannot get this to work. However, I can get 150 to work without any problems. Also, I have set the three memory timings of SDRAM RAS-to-CAS Delay, SDRAM RAS Precharge Time, and SDRAM CAS latency Time to the value of "2", and experienced no problems with the FSB at 133.

Benchmark test - Quake2 (demo version), demomap demo1.dm2. 1280x960, with gl shadows set to "1", 8-bit textures = "no".

Test 1, memory timings (mentioned above) all set to "2". 104 FPS.
Test 2, memory timings all set to "3". 103.9 FPS
Test 3, FSB increased from 133 to 150. Memory timings all on "3". 104 FPS, but Quake2 complains of illogical errors, and so this system is not stable.
Test 4, FSB @ 150, memory timings all on "2". 104 FPS, but Quake2 complains of illogical errors, and so this system is not stable.

I guess these tests show that the V5 card is "maxed out" at 133 FSB on this system, and that increasing the FSB to 150 has no effect. I guess the reason for this is because the V5 cannot do any more processing, because it's already doing 1280x960 with shadows enabled. Also, the system isn't stable enough at 150 FSB. Actually, I should dig up a PCI V5, and rerun these tests. Maybe the OC'd system is failing because of the OC'd AGP bus?

Going back to old thrashbargs post. It is best to replace the capacitors!

Reason why: DC Ripple is possible and causing those illogical errors along the system bus. http://en.wikipedia.org/wiki/Ripple_%28electrical%29

# Within digital circuits, it reduces the threshold, as does any form of supply rail noise, at which logic circuits give incorrect outputs and data is corrupted.
# High amplitude ripple currents reduce the life of electrolytic capacitors.[8]

A PCI V5 should be fine as long as your not doing any AA which is not so good on PCI bandwidth but atleast should have the divider available.

Yep, the caps definitely need replacing. In fact, there's a bad looking one close to the AGP port! Getting caps replacement equipment is on my to do list! 😀