Reply 20 of 31, by mv_cz
Last week I had some time to play with the first socket7 build. Since I had previously connected mouse header, I could do some real testing.
- at first I wanted to exchange the original S3 Virge (325) 4MB with the Virge/DX (375) previously taken from the super7 build. And because the older virge had more memory, my plan was to move all socketed memory chips to the Virge/DX. That didn't turn well, I positioned the chips correctly, inserted into the DX card, checked again, inserted card into PC, turned it on ......... and nothing, only beeping and smell of burning electronics. I immediatly turned the PC off and the memory chips were hot as hell. So I repositioned them back into the older Virge and tried both cards. They are working, which is fine, but some of the extra memory chips are possibly damaged, since the original Virge with all 4megs now exhibits strange flickering around the mouse cursor. Before moving the chips, I checked that the timing is correct, so what it could be? As I am watching at the card again, it seems that some of the memory sockets on the Virge/DX are soldered in wrong way 😠 I positioned the chips by the angled corner on the memory sockets, but didn't think of situation that these sockets themselves are not positioned correctly on the PCB.
- Anyway, this wasn't the ultimate plan I had with the graphics card in this system. But I had to make some room for another card - first two PCI slots are too close to CPU socket, the third slot is obscured by CPU heatsink and the last PCI slot is blocked by the VRM cooler by few milimeters, damn. So I knew that the "original" CPU heatsink is oriented the wrong way, and is barely touching the CPU which was possible due to special hooks on either side of the plastic mounting/fan (there are several holes, which are good in case you have to move the cooler a bit sideways http://imgur.com/lIecEBo). So I dismounted the heatsink and ........ another surprise, the CPU is not PR166+ but in fact it is 6x86L PR200+.
- So the CPU was running all of its life as underclocked (at 2x 66 MHz) and bit overvolted (vcore was OK, but I/O voltage was jumpered to 3.5V instead of 3.3V, i fixed it immediately). But why it was not running at the real 150 MHz? The motherboard manual list this CPU running only at 3x 50 MHz which would be slower than the previously underclocked configuration of 2x66 MHz. No mention of 75 MHz bus whatsover, no wonder that whoever configured it, couldn't set it right. So with help from fellow vogoners, we identified the clockgen as IMI SC653DYB which in fact supports 75MHz BUS even with fixed PCI bus at 32 MHz 😲 I reconfigured one of the "reserved" jumpers not far from FSB jumpers and bang ....... works at 150 MHz (2x 75 MHz) and the CPU reports correctly as PR200+ 😎 Then I replaced the fan/heatsink with one of my spares, not very larger, because it would be blocked by the HDD cage. But the third PCI slot is now free for larger VGA card.
- Then I replaced 4x 8MB SIMMs for 4 sticks of 16MB SIMMs originally taken from the super7 socket build, so to have 64 MB total.
- And inserted a PCI 3com 3C905b card for smooth access to my NAS storage
- All of this only was preparation for inserting .......
Yes, the voodoo card, 3D Blaster Banshee PCI to be correct. Now I have clear, crisp 2D image and glide compatibility with solid performance. And since the 6x86L was running at it's base frequency I decided to test it a little bit - in GLquake at 640x480x16 the timedemo runs at 26.9 fps. Which is not that bad for a cyrix cpu, which is known not to perform good in games. Sisoft's sandra showed about 50 MB/s memory throughput with EDO SIMMs.
- During my other work (reorganizing and tested my pile of memory modules), I tried to exchange the SIMM for some SDRAM DIMM modules. I know it is said, that i430VX is a bad SDRAM memory performer, but this is not entirely true, in my opinion. I inserted one 32 MB SDRAM DIMM memory stick and one 64 MB SDRAM (reports itself only as 16 MB, but I don't have any other "older" DIMM and I didn't want to have only 32 MB of RAM) to have 48 MB RAM total. The sandra score went up to ~90 MB/s so it nearly doubled my memory performance over EDO SIMMs. GLquake somewhat reflected this and jumped from 26.9 to 29.7, nearly 3 fps more, which represents +10% increase in real world scenario for no big effort 😀 Maybe is it due to cyrix's CPU architecture and performance? Will have to try it with Intel/AMD in near future.
- In order to install larger (and more quiet) seagate HDD, I had previously flashed modified BIOS for my board with removed 8GB HDD limitation. But that didn't work as expected. With larger HDD, my computer doesn't locks up at boot, even BIOS reports correct size, but Win98SE's fdisk is only available to format first 8 gigs and nothing more. I even tried to modify the bios myself by using this cool tool for BIOS modifications http://www.rom.by/ It patched the BIOS and reported several fixed bug, few were HDD' size related, added write allocation support etc. But after flashing it, my board didn't boot UP 🙁 Fortunately this patched BIOS has special failover function - if you boot without KBD connected, it bypasses the added modules, so I was able to flash back another functioning ROM and will solve the HDD with external controller
- Flashed the last firmware for Toshiba XM-6302B CD-ROM - now it reads my CD-RW discs correctly 😎 (thank god for web.archive.org as the original location linked from http://archive.rpc1.org/farzeno/club-internet … ware/cd/cdf.htm is of course dead)