First post, by feipoa
Here we have my latest build: a Cyrix MII at 300 MHz. This system will soon be used to run the Ultimate 686 Benchmark Comparison.
This style of ATX case was very popular around 2001-2003 for generic and home-built PCs, however this particular case has a manufacturing date of 2006, so it is likely one of the last all white ATX towers. I just so happened to buy it new a few months ago from some old stock seller.
It is running a Cyrix MII-400GP at 300 MHz (4x75 MHz) on an AZZA PT-5IT v2.1 motherboard. The motherboard is based on the Intel 430TX chipset, so the max L2 cache size is 512 KB (pipeline burst), having a cacheable memory range of only 64 MB. For that reason, I only have a single 64MB SDRAM stick installed. The CPU is a bit beaten up as it came from a gold recovery salvage, however I do have another 400GP in a more pristine condition which I paid $5 in 2003.
This AZZA board seems to work fine with FSBs of 50, 55, 60, 66, 75, and 83 MHz. For a regular Socket 7 board, it supports quite a bit of core CPU voltages, 1.95, 2.0, 2.1, 2.5, 2.6, 2.7, 2.8, 2.9, and 3.2 V. It also works with CPU multipliers of 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, and 6x. The BIOS was patched by Jans Steunebrink in 2011 to support K6-3, K6+, and Cyrix 4X CPUs. The motherboard natively works with many of the 6x86 features already enabled, except for NO_LOCK (which is easy enough to enable via software) and Cyrix Linear Burst Mode (which a 430TX does not support).
The Cyrix MII case badge did not come with the case, I sourced it from someone on the web selling badge stickers. The image of the monitor shows some really cheesy Cyrix-made screen saver.
Once the Ultimate 686 Benchmark Comparison is complete, I will replace this TX board with either a 2 MB cache VIA MVP3 board or an ALi5 board which has been proven to work well with a Cyrix MII. By replacing the board with a Super7 ATX board, I gain AGP graphics and 512 MB of cacheable memory range. An added bonus is the ability to run a 100 MHz bus, but I am uncertain if the MII-400GP will be long-term stable with a bus over 95 MHz. I also plan to swap out the graphics card to a Matrox G400 Max, G550, or Parhelia 128 AGP.
I personally made the back panel keyboard/mouse plate; there was some guy on eBay trying to sell thin plates for $20+ delivered, and I wasn't about to pay that. I cut a piece of 1/32" ordinary mild carbon steel to fit inside a common ATX panel cover, then JB welded it to the stock aluminium panel cover. I clamped it in a vice for 24-hrs to dry. I then lined it up with the motherboard and drilled the keyboard and PS/2 mouse port holes.
The ATX case has a momentary (non-latching) SPST power switch. There wasn't a convenient way to mount one of those bulky latching AT switches (10A AC) in it's place. I looked on digikey for a suitable latching switch to replace the case's momentary one, but nothing seemed to just drop in to my satisfaction. While there were some switches I could modify to fit, I decided I also wanted to be able to throw in an ATX motherboard later on without swapping out the switch again.
I set out to build a transistor based momentary-to-latching switch converter. It uses 12 resistors, 3 NPN BJT transistors, 2 PNP BJT transistors, and 1 capacitor. The circuit serves as a single-bit static memory element, looking only for a rising signal to change (output) memory states. This circuit layout is not a new concept, though I have added some transistor/resistors and tuned the resistor/capacitor values for this particular application. The solderboard might look a bit wonky because I decided to use all desoldered components I got off of a dead 1980's stereo receiver. This turned into quite a challenge because the transistors all had different base, emitter, collector pin-outs! The two 1 uF capacitors you see in the image are connected in parallel as to serve as a single 2 uF capacitor. On the schematic, "k" refers to kilo-ohms, and "M" refers to "mega-ohms". The ATX-to-AT latching switch converter has been working fine for a few weeks now.
As for mounting the solderboard in the case, I found it very convenient to replace one of the power supply screws with a stand-off.
My only real complaints with the case are the four 5.25" bays (it is overkill), and the vented side panel. The vented side panel had some plastic exhaust ducts for a Pentium 4 or newer CPU, which I didn't need. I removed the exhaust ducts. The disadvantage of the panel vent is that the case sounds a little louder when everything is powered up than it would otherwise.
Unfortunately the case's supplied USB external connector cable was a few inches too short to reach the motherboard header, so I had to extend it (yellow wire shown on bottom left corner of image). That was quite a time-consuming heatshrink/solder job.
The motherboard currently has a Matrox G200 w/16 MB SDRAM, an Intel Pro/100S network card, a Promise SATA150 TX2plus ATA controller, and a Creative Live! (or Yamaha 192XG) sound card (all are PCI). I had a 200 GB ATA drive laying around, so that is now hooked up to the SATA150 TX2plus with UDMA5 transfer rates. Lastly, I added an ATA Toshiba DVD-RW drive which is connected to the motherboard's IDE port, which is running at UDMA2 speeds. Windows 98SE with the unofficial service pack is currently installed in preparation for benchmarking
I am still trying to source some additional benchmark programs for the Ultimate 686 Benchmark Comparison. Something which specifically tests MMX features would be nice. I am open to suggestions/links for benchmark programs.