I have a Cx5x86 120 running at 40mhz fsb x 3x multiplier in my M919 and haven't had a problem at whatever the default voltage is. I didn't have to change anything to get it to work. Maybe some boards and CPU's are just more stable than others and don't need tweaking.

Do you guys have problem running Phil's vga benchmark on your 5x86s with feipoa's default settings ?
For me it quake won't launch the timedemo with those settings. The game will start thus.
And with the CPU default settings, the bench will start. But won't beat a Intel DX4.

Which Cyrix 5x86 features are you setting? Try these:
C:\5x86.exe /LSSER=off /FP_FAST=on /RSTK_EN=on /LOOP_EN=on /BTB_EN=off

Which motherboard, which CPU, what voltage, and what FSB are you running?

feipoa wrote:

Which Cyrix 5x86 features are you setting? Try these:
C:\5x86.exe /LSSER=off /FP_FAST=on /RSTK_EN=on /LOOP_EN=on /BTB_EN=off

Which motherboard, which CPU, what voltage, and what FSB are you running?

Hello,

I'm using those settings :

feipoa wrote:

*My Default Settings* - Cyrix 5x86 Register Bits
[PCR0=5h, CCR1=2h, CCR2=D6h, CCR3=1Ch, CCR4=38h, WBE (CD=0, NW=1)]

I set them with the evergreen utility using the Hex codes provided.
I actually tested with BTB on its own first to train myself using the utilities, all tests worked.
I'm using a MB-8433 rev 2, a 5x86-100GP 1.3, 3.45v, 3*33MHz. I'm under MSDOS 6.22, using the motherboard's integrated IDE controller, a 5GB HDD with a 2GB partition, and DOS starts F5'd so no memory managers.

I can play the game, but "timedemo demo01" won't load. Only "bug" I'v seen.

If you aren't planning on using NT4, it might be easier to use the Peter Moss Cyrix utility instead of the Evergreen utility. I do not know what the issue is with Quake. I haven't tried Phil's bench pack. You might want to try turning each feature on one by one until you find the problem setting.

feipoa wrote:

Which Cyrix 5x86 features are you setting? Try these:
C:\5x86.exe /LSSER=off /FP_FAST=on /RSTK_EN=on /LOOP_EN=on /BTB_EN=off

Which motherboard, which CPU, what voltage, and what FSB are you running?

Ok, those settings work 😀
btb_en=on will stop the bench from starting.
Anyway, with those settings, all bench except pcpbench get a higher score than a DX4 intel.
The DX5 amd can't be touched though.

Thanks for the help 😀

What FSB, CPU, motherboard, and voltage are you running? Cyrix 5x86-120 should be faster than X5-133 and cx5x86-100 should be comparable.

If the recently noted settings were successful, I suspect your motherboard was having trouble with LINBRST or BWRT.
If you want to use BTB in DOS, you must disable LOOP_EN. So, /LOOP_EN=off /BTB=on.

I'm using the settings and system I told you previously, so all stock 3*33MHz 5x86-100GP WB, stock 3*33MHz IntelDX4 WB and stock 4*33MHz DX5 WB.
It took the enhancements you gave me, to take over the Intel with the Cyrix. AMD can't be touched.
I've put that MB back in the drawer now, so I can'"t compare between loop and btb to see which one gives the best results with Phil's benchmarck, sorry.
I'll have to try doing that later.

Sorry, sometimes I get distracted at home and overlook parts of the responses.

Quake timedemo1 320x200
Cyrix 5x86-100-BTB = 14.3 fps
AMD Am5x86-133 = 14.5 fps

Pcpbench VESA Modus 100 640x400 8bpp LFB
Cyrix 5x86-100-BTB = 7.4 fps
Am5x86-133 = 7.7 fps

Doom timedemo3
Cyrix 5x86-100-BTB = 42.8 fps
AMD Am5x86-133 = 47.1 fps

3Dbench v1.0
Cyrix 5x86-100-BTB = 83.3
AMD Am5x86-133 = 83.3

My settings are probably different from Phil's, but you can still get an idea for the comparative results. In tasks which are heavily FPU-dependent, the Cyrix 5x86-100 will excel by about 15%, even with 33 MHz less speed. For ALU-dependent tasks, the Am5x86-133 leads by about 15%.