Use the best cooling solution you can get - paste, heatsink, fan. From the Cyrix OC threads it is obvious that not even 3.6V is enough for a stable 133 Mhz system. You need a very fine motherboard + cherry-picked CPU combo and at least 3.8V -3.9V by the trimmer mod. Honestly, if you have only one of these rare processors, I would not recommend this extreme OC at all.
(I only have 100GPs, these also need some voltage boost to 120 Mhz.)

Jolaes76 wrote:

Use the best cooling solution you can get - paste, heatsink, fan. From the Cyrix OC threads it is obvious that not even 3.6V is enough for a stable 133 Mhz system. You need a very fine motherboard + cherry-picked CPU combo and at least 3.8V -3.9V by the trimmer mod. Honestly, if you have only one of these rare processors, I would not recommend this extreme OC at all.
(I only have 100GPs, these also need some voltage boost to 120 Mhz.)

I'd be more then happy to run it at stock 120mhz. but if I do that what voltage do I set it to since my MB does not support 3.6V. or am I out of luck? The motherboard supports the 133mhz Cyrix so I don't know why not the 120mhz.

I would try 3.3v first but if its unstable bump it up to 4. It should be fine at 4 if need be as long as your cooling is decent

smeezekitty wrote:

I would try 3.3v first but if its unstable bump it up to 4. It should be fine at 4 if need be as long as your cooling is decent

cool, I'll do that. out of curiosity anyone know what the GP stands for?

http://www.cpu-world.com/CPUs/5x86/Cyrix-5x86-120GP.html

It says here, 3.3V is the required minimum and 3.6V is the suggested maximum.

It has some "power saving" features, so "GP"could be standing for "green power" or something like it.

By the way, during the good old days, I was running my 33x3 Cyrix 5x86 100MHz at 40x3 120MHz (just with the stoc fins/no fan) and had no problems at all.

okay, replaced the amd 133 today with the cyrix 120 but all my benchmark results are scoring slightly lower. am I missing something? I know the cyrix is clocked lower but isn't it supposed to be noticeably faster then a amd 133? and yes I did enable the cyrix functions in BIOS.

also of note. I've had a m919 COAST module for L2 cache for some time now and although it has been reported as detected it never acually worked. figured it out today. I replaced my EDO RAM with FPM RAM and low and behold the L2 cache suddenly worked.

I doubt you really can enable ALL major M1SC enhancements in the BIOS. I would rather use the IBM tools, Peter Moss utils etc. to do that.

Jolaes76 wrote:

I doubt you really can enable ALL major M1SC enhancements in the BIOS. I would rather use the IBM tools, Peter Moss utils etc. to do that.

but even without enabling any of them shouldn't the chip still be faster then an amd 133mhz?

Only in FPU calculations.
With ALU (whetstone bench), it is the other way round: the 3x40 Mhz Cyrix is MARGINALLY better than the 4x33 AMD ONLY when ALL possible M1sc features are ON .
Point of reference is feipoa's benchmark results.

The answers to all of your questions are already in the forums, albiet, perhaps not all in the same place. I'll try to put them into one spot for you.

The M919 does not work well with a 66 MHz FSB. 60 MHz seemed to be the fastest it was stable at. You will not be able to run 66x2 reliably.

Some Cyrix 5x86-120GP chips do not OC to 133 MHz and the one I got to OC to 133 MHz required 3.85 V to run ok-ish. The IBM 5x86-100HF chips will OC to 133 MHz better than the Cyrix-branded chips (though not all will OC). I have an IBM 5x86-100HF running at 133 MHz and 3.7 V.

On the M919, there is a single resistor you need to remove and place it with a 5K variable resistor. This will give you 3.0 - 4.0 V tunability. Let me know if you are interested and I'll pull out the board to find which resistor is needed to be replaced.

I would not run the Cyrix 5x86-120 at 4V for extended periods. It might work at 3.3 V if you are lucky. Your 3.3V setting might be closer to 3.4 V, or they may have made it 3.45 V. You'll need to measure. I've seen some 3.3V markings measured at 3.35V+.

It is likely that you have your cache and memory timings on auto or set too low. Set your cache to 2-1-2 and your memory read/write to 0/0.

The M919 is the only board I've seen which has the enable LSSER and Linear burst feature in the CMOS. Linear burst doesn't do much, but LSSER has a good performance boost. If you are using DOS, you'll want to enable branch prediction (BTB) using a 3rd party enabling program. For DOS and Windows, you'll want FP_FAST enabled. This gives a huge FPU performance boost. MEM BYP, RSTK, BWRT, LOOP, and DTE should also be enabled. But even without them, your Cyrix 5x86-120 should be performing better than an X5-133. Please share your BIOS settings.

What are you using to benchmark?

EDIT: Something else you want to keep in mind is that the m919 motherboard sneaks in a 2/3 PCI multiplier when the FSB is run at 40 MHz. So you're PCI bus will be running at 27 MHz. This is part of the reason why even your ALU-based graphic benchmarks may appear slower than the X5-133. The U4BC used the m919, unfortunately. So some 40 MHz benchmark tests have lower scores than they would if the PCI bus was run at 40 MHz. The effected benchmarks are noted in the write-up. They were mainly the heavily graphic-based benchmarks.

thanks for the info feipoa
heres what I have so far. for benching I'm using the standard 3Dbench, PCbench, DOOM, and Quake i'm running 64MB of fpm RAM and using a PCI 2MB ARK 2000MT video card

heres my bench with my old AMD 5x86 133mhz WITHOUT L2 cache (this was before I figured out I needed to swap the EDO for FPM)

3Dbench = 65.3
PCP = 16.2
Doom = 38.46
Quake = 11

now here is the results after I installed the Cyrix 120 enabled LSSR and LBF in BIOS and enabled the L2 cache. now the numbers are mostly higher but I think thats more to do with the L2 cache then the Cyrix

3Dbench = 65.8
PCP = 17.9
Doom = 32.74 (this I don't get, its slower even with the Cyrix AND the l2 cache)
Quake = 11.7

now you were correct about my cache being on auto settings. messing with cache timings and wait states is something I still haven't tried to figure out or delved into yet so I usually leave it alone but following your instruction I got a pretty big performance boost.

3dbench= 71.5
PCP = 22
Doom = 34.59 (still lower?)
Quake = 13.5

so far with the voltage set at 3.3v its been completely stable. I did have some issues using the programs to enable the extra features, Ive tried 3 so far. the Peter Moss utility crashes on me when starting it up. the CyrixGO utility seems to work but after enabling most things I actually get a slightly lower (about 1FPS average) score on the benchs. The IBM program seems to work but I'm not sure how to use it. I go to the features such at BTB which you suggested running, change the value to 1 to enable it and hit f3 to write to registry and everything just resets.

You will be best off with the IBM 5x86 utility for initial troubleshooting. Check my threads.
Your games benchmarks are slower because your motherboard is using a 27 MHz PCI bus instead of a 40 Mhz bus when you have the FSB set at 40 MHz. When you use the AMD X5-133, the FSB and PCI bus are set at 33 MHz. This is a particularlity of this particular motherboard and is not related to a Cyrix CPU issue. There may be some way to hack this m919 to keep the 40 MHz via some hardwiring, but I have not played with this as of yet.

Are you talking about this page?
http://th2chips.freeservers.com/m919/unoff/m919.html
"PCI (hard drive) Speed Boost
For the fastest PCI bus, boot the VIP at 133 MHz and then switch the jumper to overclock to 160 MHz. I do this routinely to avoid problems with my Sound Blaster on the ISA bus. But then I discovered the computer ran faster in this configuration. The way I figure it the VIP's AMI WinBios must set up shop based on the boot speed. It seems the PCI speed is 20 MHz after a 160 boot and 33 MHz when booted in the 133 mode.

HOT switching to 160 MHz must keep the dividers disabled and you end up with a actual 40 MHz PCI bus speed. Connect a switch (the unused keyboard lock works for me) to the "A" jumper on the CPU clock selector JP7 block. Switch on=133 off=160. My hard drive really perks up when I do this: take a look at the dismal 160/160 boot/operation rates."

The only hardwiring the guy is talking of is to make the clock selector available externally, in his case by the keyboard lock.

I have not tried the method you proposed. I was refering to hardwiring the FSB to 40 MHz just before the line driver chip and bypassing the BIOS. I haven't tried this, but it seems plausible. But if the roundabout method you proposed works, that might be easier. You might be able to get some small microcontroller IC to do the switching for you after some criterion is met.

I think the m919 should be suitable for AMD X5 chips running at 3x60 MHz. I just discovered I have an AMD X5 chip which POSTS at 180 MHz using only 3.7 V.

Hi soviet conscript!

can you tell me what stepping and revision your Cyrix 120 CPU is?
could you post a pic of the top and underside of the CPU?
what operating system are you using?
which cyrix enhancements are you using?

How do you find the PCI 2MB ARK 2000MT video card? Is it a quality card?

Here are some of my results with a Cryix 5x86 120-GP (stepping 1 revision 3)

(Cyrix 120) - 40mhz FSB

Operating System: MSDOS 6.22
speedsys (v4.78): 67.81
3dbench2 (v1.0): 96.0
Pcpbench /vgamode v1.97 (320x200): 20.7
Pcpbench 101 v1.97 (640x480): 7.4
Pcpbench 103 v1.97 (800x600): 5.5
Doom 2 Full Version v1.666 FPS: 39.8
Quake Share Ware v1.06 FPS (320x200): 15.4

(Cyrix 133) - 33mhz FSB

Operating System: Windows 98 SE with Unofficial Windows98 SE Service Pack 3 Update
MSDOS.SYS : GUI = 0
speedsys (v4.78): 75.26
3dbench2 (v1.0): 90.1
Pcpbench /vgamode v1.97 (320x200): 23.7
Pcpbench 101 v1.97 (640x480): 7.2
Pcpbench 103 v1.97 (800x600): 5.3
Doom 2 Full Version v1.666 FPS: 42.1
Quake Share Ware v1.06 FPS (320x200): 16.3

I'll have to hook the computer up to get you some of that info which i can do this Monday.

I took this before I installed it

the PC its in is running 64mb of fpm ram and 256kb L2 cache, I'm running DOS 6.22/Win 3.11
all I'm running ATM is LSSER and Linear burst feature because its supported in the m919 motherboard BIOS. I've tried enabling other features with the IBM utility but I cant figure out how it works.

The ARK card came in the PC when I got it off Craigslist. at first I thought it was some junk card and tossed it aside. later I started reading about how its actually a really fast card for DOS and was in the same league as the much praised Tseng cards so I game it another chance. so far its been a really nice card, fast, decent color and no compatibility problems Ive hit yet.

What voltage did you end up using?

Your running DOS so branch prediction is your friend! once you turn that on your benchmarks will sky rocket. I think It will give you your most significant speed increase. I like the Peter Moss Cyrix utility. Here's how I do it with a simple batch file. There is another official Cyrix utility that will tell you which revision and stepping your chip is which may indicate how unique your chip is but the name escapes me. Below is what it looks like:

I personally use the Evergreen utility. It is a bit of a pain to configure because you need to convert an 8-bit binary format into HEX each time you want to change a single bit. When I glanced through the Peter Moss utility, I don't recall seeing FP_FAST as a usable feature, so I started with the Evergreen utility. Also, the Evergreen utility has an NT4 driver.

I am not terribly familiar with all the nuances of DOS batch files. What is the LH for? Load High memory? What happens if you omit the LH? Do you run this in your autoexec.bat file or in your config.sys using install=5x86 /loop...etc?

Did you confirm that with LSSER=OFF that the LSSER bit is set to 0 instead of 1? This bit is confusing in that setting the bit to 0 is what enables this feature. The M919 BIOS has the "LSSER ON" setting with the bit set to 0, which is correct. I'm not sure if Peter Moss took this aspect into account or not. What speed benefit did you notice with LSSER=OFF instead of =ON? I just want to confirm you are setting this correctly.

Hello feipoa!

I remember you prefer the Evergreen program but one look at it and I went back to the Pete Moss utility 😀 FP_Fast is definitely a variable in the PM utility. LH is for loading high but off memory I don't think I was able to get the program loaded into high memory. My baby is all packed up at the moment so I cant confirm that for you mate. LSSER=OFF gave me a speed increase but I can't recall how much. I will re-post when I can confirm.