The SiS 496 boards I have tested seemed OK at 60 MHz, but I did not test them extensively in Windows. The Lucky Star LS486E looks like the Zida Tomatoe 4DPS but without the PS/2 mouse capability. My 4DPS worked ok with 60 MHz, but not 66 MHz. 3.30 V could be a bit low for Am5x86-133 chips. You will want to up that to 3.45 V by changing the applicable voltage set resistor near the motherboard's voltage regulator.

I have a very old VL/EISA system using SiS 411/406 chipset. The clockgen supports 66MHz operation, and it seems to work okay with. I also have the option to use a crystal oscillator, which I use for 60MHz and 45MHz bus. I have four am5x86 chips. Two ADZs, one ADW and a 16BGC with 2001 datecode. Only the 16BGC will operate at 180MHz, but it is not stable even with 4V (using either 45MHz or 60MHz FSB). I have four Cyrix 5x86s as well, and none of them work at 2x66 or 3x45 reliably (voltages tested 3.3V to 3.85V in 0.05V increments).

cx5x86-120 at 2x60 is solid though, as were am5x86 at 3x45, and intel dx-50 at 66MHz.

Great overclocker chips were usually hand picked. That meant testing "tons" of chips to find that special one. I don't know exactly how many SK096 I have... but I think around 20 aren't tested yet. And it's not trivial to test them out as many have bent pins, some really bad 🙁

So far I've tested only four SK096 to find the one that boots at 150 and actually work at 133.

IntelDX4 at 133MHz sounds pretty interesting. How does it perform compared to an AMD X5-133 at 160MHz?

From my previous testing:

AMD@160MHz-WB/133MHz-WB

3DBench...98.6/82.1
PCPlayer...27.8/23.1
Doom...1217/1450
Quake...16.1/13.4

The results of the fastest socket 3 chips were tabulated in the Ultimate 686 Benchmark Comparison. Clock-for-clock, the ALU of the Intel 486DX2-133 (66 MHz FSB) is about 10% faster than the AMD Am5x86-133 (33 MHz FSB); the FPU is about 8% faster.

Comparing the Intel 486DX2-133 to the Am5x86-160, the Am5x86-160 is about 8% faster at ALU and 10% faster at FPU. As such, there is probably little interest in setting up a system around the Intel DX2-133.

Kixs, what type of long-term stability tests have you run on your Intel DX2-133? I found with my chip that, while it originally appears stable in DOS, after extended benchmarking in Windows, it starts to choke, particularly with Ziff0Davix 3D Winbench97 and Winbench96.

PCPBench is interesting as Intel 486-133 is around 10% faster than AMD-5x86-160. (31.1 Vs 27.8 ). Otherwise is around 7% slower.

I haven't done any other benchmarks as I've only set everything up a few days ago. Windows 95 is yet to be installed.

What exactly happens with Winbench97 and Winbench96?

They either hang up or spit out an error with the red stop sign.

kixs wrote:

Cyrix 5x86 is a beast in FPU. My 100MHz would do stable 120, but with 133 is very unstable.

Have you tried the IBM branded version? The IBM chips were supposed to be more conservatively rated than the Cyrix ones so overclocking tends to work better on those. They also support a 2x multiplier, so you can jumper the motherboard to 66x2.

I have only one Cyrix 5x86-100 and no IBM. It's an early version and supports 2X multi (if i remember correctly).

Now check stability by linx 😀

This is my UMC board, I found it on Ebay a month ago with an AMD DX4-120 but as the seller was a member here and on Amibay I bought the bundle outside of Ebay as I dont want them earning money from me when I can avoid it. I need to try it out anyhow so I might aswell see if I can get some decent scores 😀.

The board is a MG v1.1A. This model has also been sold as Ability, Amptron, Aristo, Eurone, Matsonic, Houston Tech, PCWare, Alton, Protac, Sybercom, Elpina, V.I.P, SOLAR COMPUTER, Hsin Tech, SYL, ST, TD and as you probably can guess... PC Chips!

Some versions seem to differ slightly but my board seems to be identical to the PC Chips M921 only with real cache. The jumper settings seem to be identical to the ones in the M921 manual. I can not find any settings for 60 and 66 MHz FSB but if the board supports those settings Im sure I will find them. I can not find any jumper settings for the AMD 5x86 either but except for the multiplier they should be identical to the settings for the DX4-120 so I do not think that will be a problem. [edit] I just read that the FSB jumper settings for the M921 are the same as on the Biostar-MB-8433UUD which member feipoa has mapped out. I also think I have identified the multiplier jumper for the enhanced AMD 486/5x86. I will start testing the board tomorrow as its a bit late now.[/edit]

The "MG" v1.1A. I do not know what has happend with the socket, its a "pre existing condition".

I recently got my 83mhz pod to run at 100mhz stable 2.5 multiplier at 40mhz bus and 4volts to the core, but will not even post at 50mhz no matter what I do even though I have had the board stable at 50mhz with am5x86 mobo is pc chips m919 ver. 3.3b/f not sure what my software fps is in quake

Tip for dark screen pics... hold a sheet of paper over half the screen, half press the shutter to set focus & exposure, then drop the paper and take.

If you allow the camera to do its own thing, it will usually overexpose horribly.

How did you get 4 V running to the POD's core? You must have modified the onboard voltage regulator by altering the voltage set resistors. Could you let us know which of the fat SMD resistor(s) you removed and with what resistance you replaced them with? I have been meaning to get around to this project but have not had the time.

Thank you for asking this question I would like to start a new thread on how to mod this cpu but since im a newb I don't think I have permission. Basically remove voltage regulator solder where the center leg went to the innermost leg then it can be powered by the mobo regulators set them for 4 V. The 5 V+ fan leg and 5 V- leg must be removed or insulated from contact other leg must be left, then run 5 V from psu or fan plug on mobo to the fan, if you do not mod the fan it will run too slow at 4 V and cause it to run multiplier of 1. I did this over a year ago on a POD that had almost every pin on it bent so it was a good victim, after I straitened all the pins of course and made sure it worked haven't had any problems running at 100mhz even with all the latencies set at the lowest and it is crazy fast in the m919 at 40/100mhz. I am sorry this is a little off topic as it gets into modding

stuvize wrote:

Thank you for asking this question I would like to start a new thread on how to mod this cpu but since im a newb I don't think I have permission. Basically remove voltage regulator solder where the center leg went to the innermost leg then it can be powered by the mobo regulators set them for 4 V. The 5 V+ fan leg and 5 V- leg must be removed or insulated from contact other leg must be left, then run 5 V from psu or fan plug on mobo to the fan, if you do not mod the fan it will run too slow at 4 V and cause it to run multiplier of 1. I did this over a year ago on a POD that had almost every pin on it bent so it was a good victim, after I straitened all the pins of course and made sure it worked haven't had any problems running at 100mhz even with all the latencies set at the lowest and it is crazy fast in the m919 at 40/100mhz. I am sorry this is a little off topic as it gets into modding

I started a new thread. Please reply there, Modifying the POD83's voltage regulator for overclocking

I tested the AMD 5x86 and the MG v1.1a UMC board a moment ago with mixed results.

The bad

I could not find any PCI divider settings in the BIOS so it seems I need to flash the PC Chips M921 BIOS to my board which should not be a problem as they are identical. Im just afraid the PC Chips BIOS will disable the cache as the PC Chips M921 usually had fake cache. The jumper I thought changed the multiplier diddnt and I have not had the time to find the correct one so the 5x86 is stuck at 3x.

The good

I did test the 50 MHz FSB setting and it seems to work great. No trouble with the IO-controller or the S3 Trio64 video card, at least not so far. Perhaps there is some auto PCI divider in action...

The first Speedsys result. AMD 5x86 @ 3x50 MHz. Notice the BIOS date. 10/10/1994, this BIOS is old!

Im starting to get too drunk to continue so I will tinker more tomorrow.

Edit

Its now the next day. As I only can attach 4 images to each post I will continue this post as an intro to the next one.

I managed to find the jumper that sets the multiplier for the AMD 5x86 on the MG V1.1A and I guess its probably the same on the PC Chip M921. I will upload the jumper settings here in case someone else needs them.

/Edit

When I benched the MG V1.1A and the AMD 5x86 at 4x40 MHz I got a nasty surprise, the board had disabled both the L1 and L2 cache. I guess its because the BIOS is too old to support the AMD 5x86.

MG V1.1A + AMD 5x86 @ 160 MHz no L1 or L2 cache.

As I couldnt get any high results I aimed for a low one. This is still with the 4x multi so it would probably be even slower running 3x.

Then I switched back to the 3x multiplier and set the FSB to 50 MHz again.

Here is PCP bench VGA mode with the AMD 5x86 at 150 MHz: 25.2 FPS

PCPbench Mode 100, AMD 5x86 at 150 MHz: 10.8 FPS, only 0.2FPS better then the Intel DX4 at 120 MHz*.

*Using the Asus PVI-486SP3 and a Matrox Millenium II

I also tested to see if I could get the 60 MHz FSB setting going using the AMD DX4-120 which has 2x and 3x multipliers but it would not even post at 2x60 MHz with the memory, cache and ISA-bus at the slowest settings. Im now sure this board dosnt use any auto PCI dividers, if Im going to get the PCI-bus to run any other divider than 1/1 I will need a new BIOS.

For reference, here are my unoverclocked (except last line) results, on a MB8433 v2.0, with 32MB FPM RAM, and an S3 Trio 64 :

Of course this table will look better when I've found a stable 133MHz DX4 from intel 😀