feipoa wrote:

Only way to be sure is to remove the modules from the Biostar board and test them in the UM8810P.

True, without the SRAM tester I'm unable to determine with 100% accuracy which ones are dead and I'm not willing to strip my other board of cache. I do remember reading your thread stating 10% failure rates using the Chinese reprints.

I tested and catalogued the PCI Matrox video cards I have with the Cyrix 5x86 120MHz + UM8810P-AIO.

DOS 7.10
Himem.sys loaded
DOS Quake with -nosound -nomouse -nojoy -nonet -nocdaudio +timedemo demo1

BTB_EN=ON, FP_FAST=ON, LSSER=OFF for Cyrix enhancement gains and all else OFF in the Peter Moss utility for stability throughout all tests.

Name: Matrox Mystique 220

Engine: MGA-1164SG-A
Code: MY220 / 4B / 20
Copyright: 1997
Reivision: A

3Dbench 1.0c : 103.0
Pcpbench /vgamode : 23.1
Pcpbench 100 640x400 : 9.4
Pcpbench 103 800x600 : 6.2
Pcpbench 105 1024x768 : 4.3
Quake 1.06 SW 1st run : 16.5
Quake 1.06 SW 2nd run : 16.5
Quake 1.06 SW 3rd run : 16.5

Name: Matrox Millenium II ( edit: This is a Matrox Millenium II)

Engine: MGA-2164WP-C
Code: MIL2P / 4BN / 20
Copyright: 1997
Reivision: A

3Dbench 1.0c : 103.0
Pcpbench /vgamode : 23.1
Pcpbench 100 640x400 : 9.4
Pcpbench 103 800x600 : 5.4
Pcpbench 105 1024x768 : 4.3
Quake 1.06 SW 1st run : 17.1
Quake 1.06 SW 2nd run : 17.1
Quake 1.06 SW 3rd run : 17.0

Name: Matrox Millenium G200

Engine: MGA-G200
Code: G2+ / MILP / 8D /CPQ
Copyright: 1998
Reivision: A

3Dbench 1.0c : 103.0
Pcpbench /vgamode : 23.1
Pcpbench 100 640x400 : 9.4
Pcpbench 103 800x600 : 6.2
Pcpbench 105 1024x768 : 4.3
Quake 1.06 SW 1st run : 16.2
Quake 1.06 SW 2nd run : 17.1
Quake 1.06 SW 3rd run : 17.1

Name: Matrox Millenium

Engine: MGA-2064W (IS-STORM R2)
Code: MGA-MIL / 2B
Copyright: 1995
Reivision: A

3Dbench 1.0c : 103.0
Pcpbench /vgamode : 23.1
Pcpbench 100 640x400 : 9.4
Pcpbench 103 800x600 : 6.6
Pcpbench 105 1024x768 : 5.0
Quake 1.06 SW 1st run : 17.1
Quake 1.06 SW 2nd run :16.2
Quake 1.06 SW 3rd run : 16.2

Name: Matrox Millenium G200A

Engine:MGA-G200A-D2
Code:: G2+ / QUADP / CPQ
Copyright: 2000
Reivision: A

- Will not work in either UM8810 or MB8433 motherboards (socket 3)

Looks like the G200 comes out on top. What about Millennium II?

I have the G200, Mystique, and G200 Quad. I recall the Quad not working, but this is what I have written about it,

Matrox Millennium G200 MMS Quad 32 MB (Either the motherboard did not turn on, did not POST, or drivers did not install - I don't recall which.)

It would be neat if it was possible to combine the GPUs on the Quad card and make G200 SLI.

I still have a couple more Matrox cards to dig out and I'll add to the list above. I had wished the Quad card had worked in either of my systems. It would be an interesting card to benchmark! My monitor + motherboard turned on but went to a black screen. Nothing more.

I played a bit of Warcraft II, Crusader No Regret and tested Redneck Rampage tonight.

feipoa: Here's another game with a fps counter

Redneck Rampage @ 640x480 using Matrox G200 (no 3DFX) (Demo 4.0 - 18 fps) [rdrate - show frame-rate]

Why is there no GPU acceleration?

Redneck Rampage needs to be run in straight DOS/DOS 7.10 and it doesn't appear to support 3DFX. The game is made using the same engine as Duke Nukem 3D. Strangely the fps rate is very similar to Duke3d (4-18) but the game feels terribly unresponsive in SVGA mode even with sound off, another game like DOS Quake that is better left to the Pentium class in SVGA mode.

feipoa wrote:

Why is there no GPU acceleration?

Sorry that should probably read (no 3DFX) rather than (no acel) I'll edit it accordingly.

Does anyone know why some Matrox Millenium (IS-STORM)'s have this electrical component and others don't? It looks not dissimilar to the voltage regulator on my motherboard.

(IS-STORM REV B)

(IS-STORM REV A)

Also can someone please Identify this Matrox card? It's the only photo I have and I cant make out the writing on the chip. The date printed on the board (1995) would suggest its a Matrox Millenium card but I'm not 100%.

Tables courtesy of Kresimir @ http://www.512bit.net --------> what an amazing resource! Thank you Kresimir

rgart wrote:

Also can someone please Identify this Matrox card? It's the only photo I have and I cant make out the writing on the chip. The date printed on the board (1995) would suggest its a Matrox Millenium card but I'm not 100%.

Compaq Spare # 137897-001 indicates it's a Compaq QVision 2000+, aka Matrox Impression 2Mb PCI (MGA-I / IS-ATHENA R1). 😀

Thanks Stiletto!

I didn't get any takers on the sound card recommendation so I decided to go with the Yamaha YMF718-S. That should give me good external midi module support without the hanging note bug + Sound Blaster + Adlib + Softmpu support....cant go wrong!

I installed a Waveblaster daughterboard to play around with on the Yamaha card just for a bit of fun.

I also had a spare AHA-2940U2W SCSI controller card, so I spent some time getting that working, installed a Quantum Fireball ST4.3GB, Plextor 12x CDROM, and found a case for this Cyrix 5x86 120MHz PC. Setting the MHz display was as easy as a couple of jumpers!

What I love about SCSI host cards like the AHA-2940U2W besides the speed is that it allows you to boot from CD's.

Yamaha YMF718-S

Wave Blaster Daughterboard

SCSI card, SCSI Quantum Fireball, SCSI Plextor CDROM.

120MHz LED

MHz Display Jumpers (100+16+4)

feipoa: I don't think anyone is going to get close to your benchmarked cache/RAM/Quake speeds without running there FSB @ 66MHz on their socket 3 system. When I was experimenting with the 66MHz FSB on the Biostar MB-8433 my cache speeds came tantalizingly close to yours 😜 Cachechk reported my L2 speed surpassing yours at 101.6MB/sec

These days I'm happy to run the FSB @ 33 x 4

Care to share how you tweaked a whopping 227.36 MB/sec RAM bandwidth? Its extremely unusual.

[My Best Effort]

I do not follow when you say your L2 cache surpassed mine - they are, in fact, identical. Our cachechk results are the same for 133 MHz operation. Not sure why your memory bandwidth is low.

Is your system stable with the Cyrix at 133 MHz?

feipoa wrote:

For comparison, an IBM 5x86-133/2x gets these cachechk results:
L1 = 274 MB/s (read)
L2 = 96 MB/s (read)

I thought your L2 reported 96MB and mine is reporting 101.6MB in Cachechk.

Am I mistaken?

Not just mine, many people who post speedsys results with there Cyrix 5x86 seem to get substantially lower than 227.36MB/sec

I've seen a few memory bandwidth results lower than 100MB/sec, my own were 96MB/sec until I tweaked the BIOS. What have you done differently to your RAM than everyone else?

Thanks for pointing that out. It must have been a typo. I have edited the area in your thread which noted 96 MB/s. Even in my World's Fastest 486 link, I had 101.6 MB/s noted. Re: The World's Fastest 486

Most people don't take the Speedsys memory bandwidth score very seriously. But for starters, you can copy my CMOS settings which are shown in the above-pasted link.

Is your system stable at 133 MHz? Considering that you did not respond to the question, I am guessing not stable.

Hehe damn typo 😀

I haven't played with the 60/66 MHz fsb for some time. I was too scared to do anything other than speedsys and cachechk for fear of damage to the CPU or motherboard. It might turn out to be difficult to replace either of my Cyrix 5x86 CPU.

After 66MHz FSB operation I was able to return the board to a 33MHz FSB without incident. I think I read in your thread one of your boards refused to return to 33MHz FSB operation. ( something else I worried about)

I may have jumped the gun on reasoning for the thread you speak of. I think I forgot that I upgraded the cache to 512 KB single-banked, in which case, 40 MHz with fastest cache timings is less stable than 256 K double-banked. Or I might have been having issues with the PSU connector at that time.

Cyrix 5x86-120 chips are pretty easy to replace, however, I haven't really found any which were stable at 133 MHz and 3.7X volts. I have had the best success with IBM 5x86c chips with the QFP package.

True. Cyrix 5x86 120MHz CPU ( S0 R5 ) are not too hard to come by but have you found a source for Cyrix 5x86 120MHz ( S1 R3 ) CPU?

Do you own any ( S1 R3 ) Cyrix 5x86 marked above 100MHz?

I have some S1R3 5x86-100 CPUs, but all my 5x86-120's are S0R5. It appears as if Cyrix abandoned S1R3 later in production, in favour of S0R5. I'm not sure why, but perhaps S1R3 chips didn't scale well beyond 120 MHz. I do have a Cyrix 5x86-120 with factory heatsink, which aren't very common. Also have a Cyrix 5x86-120/4x.

There have been reports that S1R3, 100 MHz CPUs run fine at 120 MHz. If that is true, then I suppose it isn't so important to have a specifically marked 5x86-120 which is S1R3.

So a couple of updates, after installing the SCSI card, hard disk and cdrom on the Cyrix 120MHz system and installing Windows 95 I was suddenly having this weird mouse problem. Basically the mouse cursor appears all squashed and wont go any further down the screen than about half way. I had the exact same problem a few years ago messing about with the same Cyrix CPU and the same SCSI card in the Biostar MB8433, at the time I blamed it on branch prediction but it seems I jumped to the wrong conclusion.

Cursor miniaturization(squashed)

Right click where the cursor is located and the menu appears below.

I figured it MUST be the last change I made to the system! I ripped out the SCSI card, cdrom and hard disk and replaced it with IDE components and I couldn't believe it when I still had the same mouse problem!

NOTE: My 2 extra VOODOO2 cards arrived from Russia so I tested SLI mode while I had the system turned over to IDE: No FPS increase at all. I guess the CPU is completely bottlenecked.

I returned the SCSI components and turned off all Cyrix enhancements and the mouse cursor went back to normal, so one by one I tested each Cyrix enhancement and found it was LSSER=OFF causing my problems. I know from feipoa's thread that LSSER=OFF gives a substantial performance increase up to 8% in some applications and games. *frustrating*

I tested a few different serial mouse and still continued to have the problem.

Strangely Quake and everything ran fine, everything looked fine it was just this damn mouse cursor! Not being able to move the cursor past the half way point made things difficult.

Feipoa suggested I drop the CPU speed down to 100 MHz to make sure its not a frequency related stability issue or alternatively I could have popped a Cyrix 120MHz Stepping 0 Revision 5 CPU in but I did neither of those things. Somewhat fed up with tweaking after such a long time of stability I removed the Cyrix 120MHz CPU and installed a Pentium Overdrive at 83MHz.

I tested the POD at 100MHz in the UM8810 and I had a couple of crashes getting to the DOS prompt and trying to run speedsys.

I increased wait states, changed timings to 3,2,2,2 and tried again and got a little further but still had some crashes at the dos prompt.

POD back @ 83MHz 😀

I admit I didnt spend much time on attempting to OC the POD and I'm satisfied with that.

POD 83MHz + UM8810 + VOODOO 2

3Dbench 1.0c : 84.1
Pcpbench /vgamode : 22.5
Pcpbench 100 640x400 : 8.5
Pcpbench 103 800x600 : 5.3
Pcpbench 105 1024x768 : 3.8
Quake 1.06 SW : 21.3
GLQuake 0.98 800x600 : 33.9

The Quake + GLQuake score's are excellent but the DOS VGA and SVGA modes are lacking. Any application or game that depends on the ALU heavily is probably somewhat crippled by the POD @ 83MHz when comparing it to the Cyrix 120MHz and Cyrix 133MHz 5x86 CPU's.