Re: Driver Pc100 Bxpro
Carmel Kittell
I had a pc100 board super socket 7 (i think) I later found out it was designed to run the cyrix cpu which had a bus speed of 80mhz unlike the bus speed of 100mhz used by amd on the k6-2
so they customised the bios to have a 100mhz setting and to lie that it was running at 100mhz but when you tested it you realised that the board was maxing out at about 86mhz
This is not actually a driver - It is the motherboard manual for the PC100 BX Pro Motherboard. It took a long time to find - eventually found it on the Amptron website listed as M747. Hope it's of some use to someone out there.
If the driver listed is not the right version or operating system, search our driver archive for the correct version. Enter PC100 m747v15n.pdf into the search box above and then submit. In the results, choose the best match for your PC and operating system.
Once you have downloaded your new driver, you'll need to install it. In Windows, use a built-in utility called Device Manager, which allows you to see all of the devices recognized by your system, and the drivers associated with them.
All three PCI slots support bus mastering, with support for up to 4 IDE drives in PIO Mode 0 up to Mode 4 (max. transfer rate of 16.67 MB/second) and master DMA Mode 2 for a maximum of 22 MB/second. Hard disks of between 528 MB and 8.4 GB are supported without device drivers.
Yes, memory bandwidth is important.EDO cards are known to be minimally overclockable and mine confirms this after few extra MHz. A50 can do 15% percent more. However the SGRAM AGP card, with chip clock lower than others, maxed out PLL no sweat. I thought maybe with some crystal mod I could become lucky owner of the fastest 6326 "evah", but the DAC is tied to the same source. I will update this article later if it will be worth it. Interestingly enough SGRAM cards act like their chip clock is significantly below others. Now (Jan 2013) I can finally add results from early versions, starting with C1 featuring 4 MB memory and PCI bus.
Since the beta driver release code named Java with OpenGL ICD was buried in the depths of the internet allow me to save the suffering of other geeks by providing the library here. But guess what, you can play some GlQuake on 6326 without it, because it is compatible with the S3 Quake wrapper for Virge. There is some hud corruption, but I am still surprised it works. What more at 640x480 the Virge wrapper is a bit faster than 6326 ICD. Real gift is Mpact 2 wrapper, properly working and fastest, it became the choice for first and second Quake engines instead of the ICD. Quake 3 is where ICD belongs and to my surprise, the game is rendered correctly with advanced effects. Outside of Id world I tried Star Siege, again succesfully. Why SiS aborted the OpenGL support is beyond me.
Little known brands are stigmatized with bad drivers expectations, but SiS certainly did not left 6326 without proper Direct3d support. Last drivers 1.32 are from the end of 2002, such long lasted support was probably caused by motherboard chipsets integrating 6326. First 3d architecture of SiS exhibits Direct3d compatibility on par with legends of its generation. If only the OpenGL library did not end up with beta release...
The start however was rather rough. First revision C1 seems to be another chip with broken perspective correction engine. Many textures on large polygons are warped, shaking, sometimes even rotated. Even some 2D elements like HUDs can float by few pixels here and there. Four megabytes of video memory are sometimes not enough, C1 did not use texturing from system memory and dropped some textures in Incoming. Most of the cards have a jumper to enable own interrupt handling. I tried both to be sure whether it influences transfers with system memory, but it made no difference. Local memory management is problematic as well, games like Formula 1 or Motoracer 2 requiring minimum of four megabytes can reject the card- perhaps precious kilobytes allocated for Turbo Queue are to blame. Viper Racing scaled down textures even at 512x384, Populous could not run at resolution above 400x300. In Expendable and Resident Evil 6326 C1 rendered utter garbage. Common bug is found in Grim Fandango, where wrong z occlusion makes characters hide behind some backgrounds. While driver reports support for all blending modes, actual rendering has issues in modern engines like Half Life and Quake 3. Here is the gallery to see more.
My C3 card has AGP interface, confirming it already supports AGP 2x with system memory texturing. Image quality issues are the same as with C1, but was memory management improved by AGP? Well, there are few hopes: Incoming did not drop textures and Formula 1 could start, but look at the mess:Since C5 blending is correct and most of perspective problems are fixed. However cases of unstable textures remained, so it still cannot be considered as correct and accurate per pixel division. The image quality of 6326AGP is fine but few things are still lowering overall impression. Major flaw is how some 2d images seem distorted, for example characters in Myth or fonts in Incoming and Unreal. It is very annoying and absolutely inexcusable bug seen in whole 6326 family. Second, smoke effects are still skimpy, sometimes hardly visible. It is just a bit of white shade thrown onto your screen. Third the texturing likes to drop samples at further mip maps, but the amount of shimmering is within expectations of the time. Having started testing with Windows 95 I used last driver for the system- 1.28. Thanks to that I saw properly lighted Unreal which does not happen with last Windows 98 driver. On the other hand Expendable was broken with older drivers. Also Incoming showed some excessive color banding. 1.32 driver on the other hand selects ugly format for space backgrounds in Wing Commander Prophecy. To see screenshots of 6326AGP visit this gallery.Performance3d performance of 6326 is rather low, unlike what theoretical fillrate suggests it does not have enough power for gaming at 640x480. The sweet spot is usually at 512x384. Lowering resolutions further does not help so much like with Virge for example, so the bottleneck is likely somewhere in the early stages. I have to comment on 2d speed as well. DOS performance is lower than others and that does not come from any benchmark, one can tell the difference just by text scrolling. 6326 did not feel very fast in Windows as well, but it does its job. I picked Virge /GX2 for comparison, but now it was obviously an undershoot.