PDA

View Full Version : x86 VLB Cards



Unknown_K
September 23rd, 2006, 01:07 PM
Besides video/ide/scsi/io cards did anybody ever make anything else for the VLB slots?

fxg
September 25th, 2006, 12:28 AM
Besides those, I just saw a few weird combo cards [like Video + Sound + network on the same card]...

dongfeng
September 25th, 2006, 12:55 AM
I never saw anything else for them, my 486 has a 1MB graphics card that fits VLB and it's the only VLB card I have.

Terry Yager
September 25th, 2006, 06:17 AM
Most cards wouldn't benefit much from being bussed directly to the processor. Cards that provide some kinda input or output gain speed by bypassing the slow AT bus.

--T

Unknown_K
September 25th, 2006, 09:23 PM
Besides those, I just saw a few weird combo cards [like Video + Sound + network on the same card]...

Do you have a picture or any information on that weird combo card? I have a couple video/HD/IO on the same card made by Paradise I think, but never anything network or sound related.

A 10/100 Network card could have taken advantage of a VLB slot, but I have never seen any or heard of any (same goes for a video capture card especialy if it had a SCSI controller built in).

carlsson
September 26th, 2006, 05:31 AM
Had anyone heard of 100 Mbps in the era of the VESA Local Bus? Perhaps on the big server/network/router side, but would those operators use X86 hardware?

fxg
September 26th, 2006, 07:12 AM
@ Unknown_K - i don't have any pics or info. I only saw it a number of years ago. I remember it because it was a weird combination and it was the first VLB card i've seen [16bit ISA + PCI-like slot put together kinda caught my attention :)]. After that, I only saw I/O and video cards...

@ carlsson - I'm not sure but, I've got something interesting: a 5-Port UTP network card on 16bit ISA. I didn't get to test it yet.

Unknown_K
September 26th, 2006, 09:51 AM
Had anyone heard of 100 Mbps in the era of the VESA Local Bus? Perhaps on the big server/network/router side, but would those operators use X86 hardware?

Why not, there are 10/100 ISA network cards out there and that slot is much older then VLB, plus VLB could handle the data rate while ISA would not.

carlsson
September 26th, 2006, 11:06 AM
Certainly. I was just considering that those 100 Mbps ISA cards might've been manufactured post-1995, at which late 486 boards had PCI slots like the newcomer Pentium. I thought the VLB was available much earlier than Intel's PCI bus, but my literature suggests at one point they were competitors side by side.


Today, the VL specification covers 32 bit transfers in 33 MHz. VESA is expected to make an addition for 64 bit transfer. Physically, it requires the connector to be further extended, to meet the demands set by Pentium.

PCI is predicted to be the local bus solution used on more powerful computers, while VL will be used in "regular" personal computers. The PCI specification already contains a description of 64 bit transfers and 3.3V system, which makes it future proof.

Those functions in a computer system that have most to gain from a local bus (either VLB or PCI) are graphics, hard drive, network and video input.

I don't know the reasons why VLB diminished, but perhaps PCI was cheaper to implement than initially thought. It also is more rigorously specified towards motherboard manufacturers and supports Plug'n'Play. Perhaps the latter was the killer feature?

The book also mentions IEEE P1394, referred to as "Serial Bus" and transfer speeds up to 100 Mbit/s, but expected to reach 400 Mbit/s in the future. Would that be an early (1994 !) prototype of FireWire?

Edit: So, FireWire development dates back to 1986, with Apple being the main contributor to the IEEE P1394 Working Group. It was shown to the public at Comdex 1993, and the first version was completed in 1995. Thus my book was agile to pick up the "next big thing".

Unknown_K
September 26th, 2006, 01:04 PM
The problem with VLB is that it was too close to the CPU. Cards had to operage in the 25-50Mhz range and most did not work at 50Mhz at all (I have a few high end SCSI cards that don't like 40Mhz either). From what I read one VLB card on the BUS works great, 2 work decent, and 3 start having problems. I also think too many cards doing DMA mode screws up the CPU getting data from the cache on the motherboard. So with faster chips coming out with 66Mhz or even 100Mhz CPU bus it would have been hell making new cards run at those speeds and the older cards would not work at all.

PCI cards all were 33Mhz, didn't interfere with the CPU and cache, had higher bandwidth, and could also have more slots available foor cards.

The VLB design lived barely into the pentium era, I have seen a few motherboards with P60's that had VLB bus (the Nexgen motherboards had it also) but PCI was superior and more usefull to bother messing with VLB after the 486 was phased out. I even have (2) 486 motherboards that are PCI, the last generation of boards that took the 486/133 tended to be PCI (a couple were VLB+PCI and were unstable from what I remember if you use both types of cards).

I guess I collect VLB cards for the same reason I love Nubus/PDS Mac cards, they were just different then PCI (which every system adopted later on) and had great performance for their day.

carlsson
September 26th, 2006, 01:31 PM
Yeah. My book states that all three VLB cards supposedly can be bus masters, but perhaps theory doesn't match reality when the specification is loose. When it comes to data transfer speed (is it equivalent to bandwidth in this case?), the book says VLB could do 125 MB/s in 32 bit, and a PCI card of the day did 132 MB/s. The same comparison table quotes 16-bit ISA at 5 MB/s, MCA and EISA both at 32 MB/s. I suppose those are maximum transfer capacity for the whole bus.

For a bit of reality check, the most recent PCI Express (PCIe) bus can go up to 4 GB/s... so in a matter of a little over 20 years, consumer computer bus capacity (and for that matter, CPU clock frequency) has increased by a factor 800. Hard disk capacity is about 10000++ times larger. The figures are mind numbing.

Unknown_K
September 28th, 2006, 08:39 PM
I won a bunch of EISA cards and a 486 motherboard from ebay so I will be checking out how a caching EISA SCSI controller compared to a caching VLB IDE one.

Trevor07
February 17th, 2007, 01:15 PM
Back in December 1993 I bought a really cool 486 motherboard from a company called "Reply Corporation" (San Jose, CA), to replace the 286 board ("planar", in IBM-speak) in my IBM PS/2 Model 30-286 (10 Mhz 80286). This "Turbo Processor" 486 board came with an intel 486-SX 25 Mhz CPU, as well as 1MB of onboard Vesa Local Bus video (Cirrus Logic CL-GD5428 ). I ordered the mobo with 8 MB of 70ns RAM (single 72-pin SIMM). Only problem was, there was no L2 cache on the board, and no socket for any, either - ouch.

Anyway, soon afterwards, I ordered another 8 MiB SIMM and installed it, for a total of 16 MiB RAM, and over a period of 3 or 4 months I twice upgraded the CPU, first with the Intel 486 DX-2 50 Mhz Overdrive (huge improvement, all around), then the 66 Mhz version - only somewhat faster, but noticeable improvement. Fortunately the motherboard had the 169-pin "OverDrive socket", along with the jumper to disable the original CPU. The DX-2 66 required setting the motherboard bus speed to 33 MHz, via a 3 switch jumper block, which was provided.

A couple years later I wanted to try yet another upgrade, so I bought a Kingston AMD 5x86 133 MHz CPU (about $99 USD), which included voltage regulator (3.3V), heatsink and fan, and I also bought two matched 32 MiB 70ns SIMMS, for a total of 64 MiBs of system memory. The memory went right in, and the BIOS immediately recognized it(!). With the matched pair of 70ns 32 MiB SIMMs, the system's MMU automatically did interleaving and paging, which I later verified using Byte Magazine's Linux benchmarks - the entire 64 MiB of system RAM effectively ran at 31ns, through ~95% of Byte's *very extensive* memory benchmark - almost as if it was a solid, continuous bank of 64 MiB of L2 cache.

Another really good thing about AMD's 5x86 CPU was that it had double the L1 on-die cache relative to Intel's 486 DX-2 OverDrives - 16 KiB, vs 8KiB, which helped further speed things along. With this setup, my little PS/2 sleeper easily ran OS/2 2.1 and Windows95, and after running various video benchmarks (I'm a benchmark fiend), l was getting around 32 megapixels/second from the vesa local bus video, at 1024 x 768 x 8 resolution. The VLB video also did 24-bit TrueColor - 16.7 million colors @ 640x480, as well as 800x600x15 and x16 bit 'HighColor' modes. For a '486, 32 MPixels/sec ain't too shabby. I don't know how many MiB/s of bandwidth that works out to, but I think it's probably in the ballpark of 32 MiB/s (1024x768x8x5?). In any event, around June of '97 it was decent enough bandwidth for me to later play around with Enlightenment (an XWindow manager) on RedHat Linux 5.1, and later Solaris 7 x86, running CDE.

The system ran like stink, and I always thought that it was really cool that it was all neatly crammed into a tiny little IBM PS/2 "slimline" pizza box case that fit nicely on the desktop. When running RedHat Linux 5.1, this system functioned as our 8-node network's DNS server, as an Apache web server (running 13 httpd daemons), an ftp server, as well as hosting 4 or 5 other users' shell accounts, which they'd ssh into from all around the U.S. and Canada. My roommate and I had a frac T1 coming into our house, and our network was 10-baseT, but that was enough back in '97. At any rate, I remember several times seeing these remote users all logged into this system, brushing up on their Linux skills, reading email (pine), doing perl or shell scripting, etc...all this, while I'd be running Gnome or Enlightenment, browsing the web with Netscape, MUDing at Mercator (cool MUD in Houston, TX, still there! - mercator.mudhaven.com, port 2411, I'm "Trevayne"), or playing around with GIMP or whatever. The system never even broke into a sweat - it was always like, "bring it on!". I'd fire up "top", and there was always plenty of memory and cpu cycles to spare, even with these 4 or 5 users logged on. I remember being awestruck at the power. Simply amazing.

I would have to say that this was the biggest, most massive computing experience I've ever had. Rock solid, powerful, fast, responsive, and stable. I don't recall ever having such a "Massive Computing Experience" ("MCE"?) on the fully loaded VAX 11/780s that we had at my university (U of Maryland), or later on Sun Sparc20s, or Sun Ultra1s, or even on SparcServer 1000s that I played around with at work, for that matter. And those latter two Sun Sparcs were 64 bit machines.

The only slightly weak spot on this system was the hard drive controller, which was straight AT/ISA, and the most I ever got out of it was only about 1.6 MiB/sec, using Quantum Fireball drives (running multi-word/16 sector/cycle xfer mode, hdparm), but with 64 MiB of fast memory, that was usually more than adequate. I had /swap mounted on a second, identical physical drive, which was rarely hit, as I recall. I also ran Toshiba 24-speed and later Teac 52-speed CDROMs as slaves off the hd0 hard drive master, and they ran perfectly.

This system attained [a virtually meaningless] 66.15 BogoMips, when booting up Linux - check out http://www.faqs.org/docs/Linux-mini/BogoMips.html - and do a search on "<paul.doyle@mci.com>" - and there she be!

Overall, the CPU, the memory, and particularly the VLB video (oh yeah, this thread is about Vesa Local Bus -- doh! :P ) ran just great. The ultimate PS/2 sleeper.

IBMMuseum
February 17th, 2007, 03:12 PM
Back in December 1993 I bought a really cool 486 motherboard from a company called "Reply Corporation" (San Jose, CA), which was a replacement board for my PS/2 Model 30-286 (10 Mhz 286)...

Reply made a number of motherboards to replace IBM PS/2 planars, and at least one full system on their own. Quite often there were two different levels ("Powerboard" & "Turbo") for the replacement boards. So successful that in some cases IBM used rebranded Reply motherboards in systems (PS/2 Model 53) or sold them as "Systemboard Upgrades".

I've got the same Reply board you had & was able to get it to 128Mb...

Micom 2000
February 17th, 2007, 03:33 PM
I have a bunch of VLB 486 motherboards as well as VLB video cards and VLB port cards. Anyone interested with playing with VLB they are available for a cheap price. Untested, but I'd imagine they are all working. PM me.

Lawrence

IBMMuseum
February 17th, 2007, 03:58 PM
I have a bunch of VLB 486 motherboards as well as VLB video cards and VLB port cards. Anyone interested with playing with VLB they are available for a cheap price. Untested, but I'd imagine they are all working. PM me.

Not to be any competition, I also will be a good source for the same items. I've tested and marked most of them, and rudimentary testing can be done by me before shipping too. Pretty good supply of the additional (CPUs & SIMMs) parts needed to make them complete too.

As said in another recent post of mine, money would be prefered ahead of most trades at this point in time.

Trevor07
February 17th, 2007, 04:08 PM
Mine was the "Turbo" board - one of the very first ones. The best (flashable) BIOS was 1.07, as I recall.

Anonymous Coward
February 20th, 2007, 04:16 PM
I'm looking for a last generation high-end VLB videocard based around 964 or 968 chipset. Do you have any of those?

IBMMuseum
February 20th, 2007, 05:57 PM
I'm looking for a last generation high-end VLB videocard based around 964 or 968 chipset. Do you have any of those?

I do have some S3 VLB cards I was going through just today, but I was looking for other chipsets & don't remember them now. Tomorrow I'll open the box again. Anything else (VLB or otherwise) you were looking for too?