After trying 3 different video cards - all of which had this same issue I dropped the FSB back to 60mhz so the CPU is running at 120mhz but still there is the same issue.

I believe its the bios now. Some setting that I've missed....

Ive followed Feiopa's Updated Manual for the MB4833 and used all his recommended bios settings.

I thought it might be my curse with SCSI so I dropped it and went back to IDE. No change.

I swapped out the Cyrix 120 CPU for a Cyrix 100 CPU and a 120 o/c running at 3.45 volts and I had the same issue with Doom2.

I cant run the system with a PCI Divider at 2/3 ...its definitely not on. I have to use 1/2 when using 60/66mhz FSB

After swapping out all hardware I believe it can only be two things:

1. The S3 and Matrox IS Storm PCI Video Cards I tried are not up to operation above or close to 50mhz. ( I need a G200)
http://vgamuseum.ru/gpu/matrox/matrox-millennium-220/
2. Issue with the motherboard.

Any takers?

Your graphics card is not running at 66 MHz - it is running at 33 MHz. That is what the HOST / PCI clock - 1:1/2 setting is for, to devide the FSB in half for the PCI clock.

Try testing with a single stick of 16 MB to begin with. Leave 256 KB installed and ensure it is jumped correctly for 256 KB. 10% of the 1024 cache from China I found were defective.

Confirm HIMEM (part of DOS, but can be put onto a boot floppy) passes the memory test. If that passes, be sure Memtest v4.0 passes the memory test.

Use chkcpu16.exe to check what speed your system is running at. Cachechk-v7 reports faulse CPU frequencies for 486's. Use Cachechk-v4 if you want a proper CPU frequency reading.

Try testing at 2x60 before 2x66.

You need to run 2x60 at 3.6 V, not 3.45 V. 1:1/2 for the HOST/PCI clock.

For troubleshooting purposes, disable L2 (external) cache in the BIOS FEATURES SETUP.

Try not to run at 4V. The way the CPU regulator is wired, 4V is max, regardless of how far you take the trimmer. Try not to go above 3.94 V.

Not all Cyrix 5x86-120's will operate at 133 MHz. Of the 2 I tested, one would not operate at all, at any voltage. I mean, it would boot into DOS, but crash with Windows. The other 120 required 3.9x V to run.

thanks feiopa.

I believe I have found a slither of stability!

Unfortunately its with the CPU at 120mhz.

2x60mhz

512k cache

host pci clock 1:2/3

My dos programs run.

pcpbench runs

himem.sys passed

memtest86 v4.0 passes twice so far.

Ill install windows 98 next.....

Windows 98 running fine.

It appears I have it stable at 120mhz but its not enough!! arrrgh!

I want 133 and I'm confident now the issue I'm having with the DOOM 2 screenshot is because I'm not hitting the right voltage.

I'll continue to tinker....

Because the plan is to use DOS I've noticed that I can turn branch prediction on without having any issues....

Feiopa I was checking this thread:

486 mobo + 586 chip

Its an old school thread with some good reading.

I can get around 65 on a speedsys score when I run the CPU at 133 however your getting 76.43 on your Worlds Fastest 486 thread.

That's faster than a 200mhz AMD and almost double an intel 486DX4 100.

11 points on speedsys is almost like comparing a Intel 486DX4 100 to a Cyrix 5x86 120

You must have overclocked the 133 somehow to get that result no? But again your screenshots show the cpu clock is working at 133mhz with quite slow cache and ram timings.

Did you patch that Cyrix chip directly into your brain or what? 😜

If you want to run the chip stable at 133 MHz, it is best to use an IBM 5x86c.

You do not really want to run these chips at more than about 3.85 V. If you are taking it up to 4 V, then you will probably damage the chip with prolonged use. Idealy, we are shooting for 3.7 V.

If you are not getting the same speedsys score as me at 133 Mhz, it is because you have not enabled the appropriate Cyrix features. Ensure that,

LSSER = 0
BTB = 1
LOOP = 0 (default setting)
RSTK = 0 (default settin)
FP_FAST = 1

feipoa wrote:

If you want to run the chip stable at 133 MHz, it is best to use an IBM 5x86c. […]
Show full quote

If you want to run the chip stable at 133 MHz, it is best to use an IBM 5x86c.

You do not really want to run these chips at more than about 3.85 V. If you are taking it up to 4 V, then you will probably damage the chip with prolonged use. Idealy, we are shooting for 3.7 V.

If you are not getting the same speedsys score as me at 133 Mhz, it is because you have not enabled the appropriate Cyrix features. Ensure that,

LSSER = 0
BTB = 1
LOOP = 0 (default setting)
RSTK = 0 (default settin)
FP_FAST = 1

do the other cyrix enhancements not matter?

WT1
BWRT
LINBRST
MEM BYP
DTE_EW
USE_WBAK

I am using this Cyrix chip. I have noticed this Cyrix chip is very temperamental whereas the IBM chip can really take a beating. (variations)

WBAK, DTE, and MEM_BYP should be on by BIOS default. Maybe LINBRST is too, but it is best to test initial overclocking with LINBRST and BWRT disabled.

I do not beleive LINBRST and BWRT did much to the speedsys result. Please refer to the nicely crafted charts and tables in Cyrix Register Enhancements Revealed for which enhancements improve which benchmark.

Did you test your PK-EP586X4 chip with the 2x and 3x motherboard jumper? Did 2x yield a 2x multiplier and 3x yield a 3x multiplier? If so, then the "X4" printed on that label is nothing more than implying DX4 and the chip is not really 4x capable.

EDIT: typo fixed - 2x/2x, 3x/3x.

Hooorah! 😜

Behoooooold!

Cyrix 120GP chip with a 133 o/c and 66mhz FSB with Branch Prediction turned on.

Interesting, a 5x86-120 with stepping 1, revision 3. Usually they are stepping 0, revision 5.

feipoa wrote:

WBAK, DTE, and MEM_BYP should be on by BIOS default. Maybe LINBRST is too, but it is best to test initial overclocking with LINBRST and BWRT disabled.

I do not beleive LINBRST and BWRT did much to the speedsys result. Please refer to the nicely crafted charts and tables in Cyrix Register Enhancements Revealed for which enhancements improve which benchmark.

Did you test your PK-EP586X4 chip with the 2x and 3x motherboard jumper? Did 2x yield a 2x multiplier and 3x yield a 4x multiplier? If so, then the "X4" printed on that label is nothing more than implying DX4 and the chip is not really 4x capable.

I did test the multiplier on this chip.

2x yields 2x multiplier but 3x yields a 3x multiplier.

I believe this is the first known instance of a 120MHz chip that is s1r3.

Did you find out anything about that gainbery 5x86-133 that I backed out on feiopa? I was crazy enough to buy it I just couldn't quite get enough cash together to justify it.

Good to see some interest in these chips, im quite happy with a 120 part but 133 is the holy grail.

No. I think the only way would be to actually test the chip. The cost was not worth the risk. Did the guy ever relist this item?

I have some ideas for making 2x/3x chips work at 4x and have ordered the parts, but have not been in any rush to work on it. I seem to have too many little project ideas with too little time.

My 20 yrs of experience in OCing -or witnessing the results of OCing- whispers NO and holds my hand back above the same motherboard.
Applying almost +0.5V onto such rare a processor to gain stability... Well maybe if I had at least a dozen in stock.
But I am on the practical side: speed of games and apps will never, never touch that of a Pentium in the 133 - 233 Mhz range which is also a nice platform for time travel purposes.

At 3x40 Mhz with the IBM (0 /5) 5x86 my rock solid best results without ANY overvolt are 178,54 / 68, 27 / 48,46 (Speedsys 4.78 L1, L2 and memory) and this is more than enough for a DOS gaming machine which I plan to use for some time...😀

Jolaes76 wrote:

Applying almost +0.5V onto such rare a processor to gain stability... Well maybe if I had at least a dozen in stock.

Amen. This is the reason why I don't want to stress my AMD X5 133 @ 180mhz any longer to gain stability. I know they are plentiful, but I don't see any reason to take any risk since I have much faster machines available.

Feipoa, when I change the voltage jumpers to 3.45 or 5 volts. does it bypass the voltage regulator with the adjustment screw?

I believe the 4 Volt Jumper setting is for the voltage adjustment component correct?

Yes, the 4 V jumper setting is for use with the trimmer on the voltage regulator. The 3.45 V setting still uses the voltage regulator, but it is fixed at 3.45 V. 5 V feeds directly from the PSU, bypassing the onboard voltage regulator.

Jolaes, I was under the impression that rgart was overclocking/overvoltaging the IBM 5x86c, not the rarer s1r3 Cyrix 5x86-120.

I see, but even the IBM 5x86 prices are rising. I have just paid over 46 USD for a perfect specimen. I do not usually care for collectors grade items just for the shiny looks, but I wanted to have an unused or lesser stressed one.