I currently run a load of Cyrix systems including a 5x86 100,Media GX 180MHz, 6x86 PR166+(not the L version), MII 333 2.9v. To be honest all of them require a bit more effort with the cooling than say a Pentium MMX or classic. I tend to use a fairly meaty heatsink on the socket 7 chips, the kind of thing that would have been used on a Celeron mendocino or early PIII skt370. That with a thin coat of thermal paste(cheap crap off ebay) and a good fan and no worries. I have to use a fan on my 5x86 100 too or I get crashes, that fancy green heatsink is not good enough, at least for my chip at stock volts and speed, a small 40mm fan screwed on top and all is well.
Hope you have lot's of fun with your new Cyrix build, they are great little chips and it's nice to have something a bit different 😀
Best,
Chris

Tetrium wrote:

There was a lot of (I think) 5 IBM 5x86 from 1996 on Ebay recently and all of them were purchased by at least 2 Vogoners. These may or may not have x4 and may or may not be able to run at 133MHz.

I must have missed that listing. Do you have a url? Not that I need more IBM 5x86c chips. I find that the QFP IBM 5x86c chips from 1996 clock well to 133 MHz at 2x. The cermaic PGA IBM 5x86c chips I have from 1996 were not stable at 133 MHz. Neither the QFP, nor the cermaic PGA IBM 5x86c chips from 1996 that I have contain a 4x multiplier option.

BSA Starfire wrote:

I currently run a load of Cyrix systems including a 5x86 100,Media GX 180MHz, 6x86 PR166+(not the L version), MII 333 2.9v.

Wow, you are extreme! Are these all in cased systems? I thought I was nuts having so many damned Cyrix setups - IBM 5x86c-133/2x, Cyrix 5x86-133/4x, Cyrix MediaGX 266, and Cyrix MII-433GP.

Yep! all cased full systems. I have around a dozen other Cyrix CPU's too,with a few on motherboards. Got a real soft spot for the 6x86. In this hobby I have always been attracted by the more obscure, less well known bits and pieces. I'm a big fan of the AMD K5 too, even have 3 IDT Winchip's floating about 😉

Best,
Chris

How about a RiSE mp6 setup? I don't think we've had anyone on Vogons with a cased system based on the RiSE.

I would but the only MP6 CPU I have is dead. wish you had not brought that up to be honest 😉

Did you try heating it up to re-melt the solder beads? Is LZF not selling the "tested" 266 chips anymore for $25?

I have a working mp6 pr266. I have a pr333 but haven't tested it yet. the real find is the 366 ES.

I still have the chip, how would I go about heating it to re-melt? got to be worth a go as it's useless as is.

Best,
Chris

Perhaps with a hot air soldering gun along the top, around the side edge, and along the bottom. Maybe also one of those hair dryer looking large heat guns. They can get up to 900 F. I'm not sure why so many of these mp6 chips are dead, but another member here hypothesized that maybe it was a BGA contact issue. If that's the case, then reworking the BGA solder joints would help.

feipoa wrote:

Tetrium wrote:

There was a lot of (I think) 5 IBM 5x86 from 1996 on Ebay recently and all of them were purchased by at least 2 Vogoners. These may or may not have x4 and may or may not be able to run at 133MHz.

I must have missed that listing. Do you have a url? Not that I need more IBM 5x86c chips. I find that the QFP IBM 5x86c chips from 1996 clock well to 133 MHz at 2x. The cermaic PGA IBM 5x86c chips I have from 1996 were not stable at 133 MHz. Neither the QFP, nor the cermaic PGA IBM 5x86c chips from 1996 that I have contain a 4x multiplier option.

It was in the What retro hardware did you buy-thread I think. One of the buyers was Artex and the other one was the one who put the listing here on Vogons.

It is here I suppose: Link. IBM 5x86C 100GP.

feipoa wrote:

Neither the QFP, nor the cermaic PGA IBM 5x86c chips from 1996 that I have contain a 4x multiplier option.

Okay, but which ones did you use to test the 4x multi like here?
When it is so difficult to find a Cx5x86 with both a 4x multi and a tolerance for 133MHz, then I better give up. The basic Cx5x86 is very nice already.

feipoa wrote:

If you need a little less voltage drop than a standard 0.7V diode, then a typical schottky diodes drop 0.3V +-0.1V. There are also zener diodes which can be operated reverse bias and have all sorts of voltage drops you can select when ordering, however voltage drops are typically in the 2-40 V range.

Interesting. I think I have some exotic diodes like that. But I cannot recall their specifics now.

BSA Starfire wrote:

I currently run a load of Cyrix systems including a 5x86 100,Media GX 180MHz, 6x86 PR166+(not the L version), MII 333 2.9v. ... I have to use a fan on my 5x86 100 too or I get crashes, that fancy green heatsink is not good enough, at least for my chip at stock volts and speed, a small 40mm fan screwed on top and all is well.

An original 6x86 PR166+ system, that is for the purists 😀 I don't even have one of those processors, just L-versions and later ones.
My Cx5x86 has been reliable so far with just the green heatsink. Though in the late 486 / early pentium era they really liked to run things with minimal cooling, and not just the CPU: The VRM on my 486 also got too hot for my liking.

That french seller also had a Cyrix 6x86MX-PR266 (2.5x83MHz), 1997, rated for an odd 2.7V.

Haha, I'm a half purist then - I own the Cyrix 6x86-PR166, but have not based a system around it. Too many cased systems already.

To test the 4x Cyrix 5x86 I used a chip marked Cyrix 5x86 -100/4x -120/4x -133/4x and unmarked multi ST -100.

I too had high hopes for the IBM 5x86c chips from early-to-mid 1996 having the 4x multi, but they do not. I have tried several from early 1996 to mid 1996; all are 2x/3x. Best bet for a 133 MHz Cyrix 5x86 is to get an IBM 5x86c-100, preferably QFP, and run it at 2x66.6 on a Biostar MB-8433UUD. Alternately, there are Cyrix MediaGX systems with 133 MHz CPUs. Do not waste your time with Cyrix 5x86-100/120 4x chips. They are not reliable at 133 MHz. My sample size is about 6. They are marked down in speed for a good reason.

I am looking forward and hoping that the buyers of those IBM 5x86C chips will test them at 2x66. If I remember correctly, there is a Luckystar PCI 486 board, which looks a bit like the 4DPS, which also works with 2x66.

feipoa wrote:

If I remember correctly, there is a Luckystar PCI 486 board, which looks a bit like the 4DPS, which also works with 2x66.

Which board is this? And is there a repository of 486 boards that will work at 66MHz FSB?

It is the LuckyStar LS486E rev.C2 (perhaps more revs as well?)
LuckyStar LS486E rev.C2 and Cyrix 5x86@133

Although the 4DPS is similar (LS486E lacks the PS/2 mouse header though) my 4DPS boards would not function at 66 MHz. I have another 4DPS clone board (w/PS/2), but forget if I tested it at 66 MHz.

I sort of recall there being a 3rd board. Anyone else recall? HOT-433 rev4 board? I forget. Maybe there are only 2 boards in the database thus far.

feipoa wrote:

It is the LuckyStar LS486E rev.C2 (perhaps more revs as well?) LuckyStar LS486E rev.C2 and Cyrix 5x86@133 […]
Show full quote

It is the LuckyStar LS486E rev.C2 (perhaps more revs as well?)
LuckyStar LS486E rev.C2 and Cyrix 5x86@133

Although the 4DPS is similar (LS486E lacks the PS/2 mouse header though) my 4DPS boards would not function at 66 MHz. I have another 4DPS clone board (w/PS/2), but forget if I tested it at 66 MHz.

I sort of recall there being a 3rd board. Anyone else recall? HOT-433 rev4 board? I forget. Maybe there are only 2 boards in the database thus far.

The LuckyStar LS486E does look familiar, but I'd need to check if I have this actual board in storage first (I've been inventorying but progress is slow).

I do know the basics of having cache chips speedy enough and the cache setup and needing faster memory etc, but I'm a bit rusty on info about undocumented FSB frequencies. Wasn't there a thread about undocumented FSB settings on 486 boards? I remember actually having participated there. I'll go see if I can dig it up 😀

My apologies for being rusty on the subject but the subject is an intriguing one, so to say 😀

edit: Test:486 undocumented jumper settings
Though I'm pretty sure there were more threads on this subject.

Most PCI 486 boards have undocumented settings. There are so few jumpers for the FSB that they are simple enough to just try them all and note down the results. I have the combinations written down in my printed manuals. I have found that most "appear" functional with a 60 MHz FSB, but not with 66 MHz. I believe the traces on most boards were impedance matched for 33-40 MHz, and some boards, particularly the larger ones, just don't work well at 66 MHz. The LS486E and MB-8433UUD are smaller boards.

Do you have any IBM 5x86c chips that you plan to try out 66 MHz with? I don't think that super fast cache is absolutely necessary - you can counter act slower cache by using longer cache wait states.

Plans I have plenty, but if I ever start out actually carrying out any plans I'd like to start well informed 😀
I got enough chips to try out, especially (but not exclusively) the AMD ones. I'm a bit more careful with my Cyrix chips 😜

feipoa wrote:

Some MII-400's will run respectably at 333 MHz, however at 350 MHz, more aggressive benchmarks failed.

The MII-433GP arrived, so I immediately tested it. It runs at the advertised speed of 300 MHz. Runs also at 350MHz, even at just 2.0V. 😀 Strangely I have no settings to push it beyond that. At 350MHz and 1.9V the Quake benchmark exitted with en error near the end.

Cyrix MII-433GP 100MHz BUS 3.0x 2.2V; 85C M2433M100-22 1995-1998 VBSDD928AB […]
Show full quote

Cyrix MII-433GP
100MHz BUS 3.0x
2.2V; 85C
M2433M100-22
1995-1998
VBSDD928AB

DIR0=52 (ID) DIR1=35 (Step ID/Rev ID) DIR2=00 (Sub-Rev ID) DIR3=06 (Family Code) DIR4=01 (Model/Stepping) […]
Show full quote

DIR0=52 (ID)
DIR1=35 (Step ID/Rev ID)
DIR2=00 (Sub-Rev ID)
DIR3=06 (Family Code)
DIR4=01 (Model/Stepping)

Maximum multiplier setting = 3.5x, Minimum multiplier setting = 2.0x. ( I found some MIIs can be put at a 1.0x multiplier, but only on the P5A motherboard, not this one. And only when disabling caches. )

Benchmarks while running at 3.0x100=300MHz at 2.0Volt:
SpeedSys: 215,05 points
Quake1 _VGAtest (320x200): 53,4 FPS
Quake1 SVGAtest (640x480): 20,6 FPS

Benchmarks while running at 3.5x100=350MHz at 2.0Volt:
SpeedSys: 250,83 points
Quake1 _VGAtest (320x200): 60,2 FPS
Quake1 SVGAtest (640x480): 22,9 FPS

_VGAtest = Quake108.exe -nocdaudio -nosound -nojoy -stdvid +timedemo demo2
SVGAtest = Quake108.exe -nocdaudio -nosound -nojoy +vid_mode 12 +timedemo demo2

gerwin, if you look through the Excel sheet from the Ultimate 686 Benchmark Comparison (Excel sheet is now available in that thread), you can see which benchmarks did not fare well at 350 MHz. Specifically, it was Ziff-Davis Winbench96 CPUMark32, Ziff-Davis 3D Winbench97, and Final Reality. Everything else ran fine. Maybe you can run these in Win98SE after the chip has warmed up a bit to see if these will run?

I have not tried my MII-433 yet, so, I'm not sure if it can handle these tests at 350 MHz or not.

What graphics card were you using in those Quake SVGA benchmarks? At 350 Mhz, I scored only 17.8 fps. Used a Matrox G200, Quake v1.06, and timedemo demo1. Sound card was physicaly removed for these tests. Are you able to benchmark under similar test conditions? I know Quake 1.08 gives faster frame rates than Quake 1.06 shareware.

What does -stdvid and +vid_mode 12 do?

It will have to wait a little, currently I don't have time to install windows 98 etc. Also, I don't have a single Matrox graphics card really.

The previous test was at 2.0V. I can still raise 0.2V or maybe 0.3V if need be, so I think it should be able to run anything at 350MHz.

Quake with parameter -stdvid selects the 320x200 VGA mode. '+vid_mode 12' is Quake's mode number for VESA 640x480 8bpp, at least with Voodoo 3 and Nvidia. There is a list modes command in the Quake console: 'vid_describemodes'.

@Glitter: Did you try that new MII-366GP?

OK.

What I have found with testing Cyrix chips is that it is usually one specific feature which does not like to be overclocked by a certain amount. If your intended application does not make use, or does not make sufficient use, of this aspect of the CPU, then the overclock will work fine.

I noticed with the MII CPUs at 300 MHz or higher that increasing the voltage did not help at all.