Talk:VESA Local Bus

VESA on a 386

Someone pointed out a board they found on Google that appears to be an Am386DX40 with a VESA slot [1] After checking the VLB standard signals, it seems there are provisions for use on the 386 also; But obviously only with a 386DX since the SX doesn't have a full 32 bit address and data bus.66.68.118.47 (talk) 21:27, 30 August 2010 (UTC)

More than one or two slots at a time?

the VESA Local Bus didn't have the electrical ability to drive more than 1 or 2 cards at a time.

Is this really so? I've got a 486 sitting next to me with three occupied VESA slots.
This article needs a bit about VESA slots in Pentium motherboards, but I don't know much about that. Crusadeonilliteracy 09:54, 27 Aug 2003 (UTC)

Unlike PCI or ISA, VLB was based heavily on the 486s memory bus and was very difficult to adapt to the Pentium's bus due to the many differences between them.66.68.118.47 (talk) 21:27, 30 August 2010 (UTC)

I agree that this is misleading. Odd that 3 at the most is marked citation-needed. Look over at the picture in the same section - there is a picture of a motherboard with 3 VLB slots ;). I'd say the citation-needed should at least be removed. 68.47.228.146 (talk) 16:21, 15 December 2010 (UTC)

33mhz

i'm sure i read that there were also issues in implementing vlb for 486 chips that used a 33mhz FSB is this true and if so shouldn't it be mentioned in the article? Plugwash 17:01, 21 November 2005 (UTC)

50Mhz

I never had a problem running VLB cards at 50Mhz. The problem was with the L2 cache SRAM chips on many motherboards. 50Mhz required cache chips faster than 10 nanoseconds, preferably 6ns, which were expensive. 7ns would work and 8ns was on the edge.

With a 486 board that was stable at 50Mhz, an AMD 5x86 clocked to 200Mhz (ADW version because it ran the coolest of the PGA type) would run right up there with a non-MMX Pentium.

I never ever had a problem running any 486 board with VLB at 40Mhz. I 'hotwired' several AMD 5x86 CPUs to force them to run at 4x40Mhz (160Mhz) with EIDE multi-IO boards and various high-end video cards in the VLB slots. Plenty fast in their day but would've been better at 4x50Mhz had I been able to find L2 cache chips fast enough.

If the cache chips were fast enough, 5x86 at 200MHz and working VLB was very easy to pull off. I avoided upgrading for a couple of years by hot-rodding a 486 system to run a 5x86 at 200MHz. If you ran the hard disks and video off the VLB slots, a system like that blew the doors off a Pentium 133. You could even run Windows 2000 off it. Jsc1973 (talk) 07:59, 17 February 2008 (UTC)

I used to run mine at 50,and I had tons of problems. 40 usually ran better. Its not a problem with the motherboard. If you had a good quality board that could run a 50mhz chip,it worked. The SRAM in fact was an issue.However,that was not the fault of ANY of the hardware. It was the system builders lack of understanding.If the SRAM on your motherboard wasn't fast enough,then you had the wrong SRAM.Even now I can go out and buy a motherboard and some DDR3 RAM as a combo. But if I buy a board and some slow memory,that would run fine on a lower end chip,but then go buy a really fast chip that just came out,Im going to clock the ram down to make it work,and therefore its going to run slower. Wow,looks that MB is a piece of junk! But of course,we all know,its not the motherboard thats the problem,its that Im a cheapskate and dont want to buy faster RAM.

The real problem was the video cards. Some could go PAST 50. But most would choke if you tried. Video memory was one problem. Board layout was another. One thing I always did was to go over all the boards. You will often see spots where filter capacitors went. Often they were empty. Id make sure to add those. But mostly,I just selected cards I found that worked. In fact you could push the ISA buss faster too. This was a great boon for hard drive speed. I would carefully select my cards to be able to handle a faster ISA buss. My experience was that the hard disk would choke first. Most hard disks could go quite a bit faster than 8mhz,and remember,the original AT spec was just connected right to the AT buss.I had one system with a 16mhz AT buss,although it had some issues. But with windows 3.1,who knows what caused it to hang. And then there was office 2.0. It was so unstable it inspired a song.But as far as 50mhz went on the VESA boards,it was problematic,and if I was working in IT back in the day,having the experience I do now,I would avoid it unless I really needed the performance,and even then I would be very careful with the system build to make sure that I used high end components.

I see what quite a bit was. I read a discussion of filter capacitors in logic circuits by an EE once.He talked about how he got out of school and got a job,and was designing a piece of equipment. When he was adding the filter capacitors to the design (and this was back in the hand drafting days,it was suggested by the older engineers that had been there for years that he simply use one filter cap per ttl IC. Well,being young and just out of school,he though,that seems wasteful. So he spent a week analyzing the circuit and figuring out EXACTLY how many he needed and where. He saved a grand total of 7 our of 100 capacitors,and figured out that most went right by the chips,for a weeks worth of work. So out of the 100 boards they built,he saved 50 bucks. Even then,and even being the new guy,a week of his time was worth a lot more than 50 bucks. But if you built a million,that would be 50 grand. But they know theres an easier way to figure out which ones they dont need,and it also gives them flexibility to put them back if they are wrong. The marginal cost of a through hole is just about 0. So they design them all in ,and then just eliminate the ones they dont need and dont populate those spots. But you can see what happened with the 50Mhz FSB. When those boards were designed,intel told them the spec was 50Mhz. But there were no 50Mhz parts to test with. Which means if your smart,you look at the rise times of the pulses,the propogation times etc,and the amount of skew between the differnt signals thats allowed,make your board,then inject pulses into the cpu socket and make sure the signals all act like they should.If your not that concerned,you throw the fastest CPU in you can and if it works,you send them out. Maybe if your new intern is a real go-getter,he overclocks the CPU to see how it acts. Of course,when he finds out that the new video card has problems at 50mhz,the enginer that built it says its not the card,its the overstressed 33Mhz CPU running at 50mhz,his boss buys it and sends it out,quite possibly ignoring the interns explanation that he added all those caps back in and it works just fine. (That CANT be the case,I spent a whole after noon removing caps until it failed,to figure out how many we really needed. Well,yea I used the 33mhz chip,but thats not the point.) — Preceding unsigned comment added by 172.56.38.200 (talk) 14:57, 10 January 2014 (UTC)

Dates

The article doesn't say anything about when it was developed, by whom, what it replaced etc. The whole article just talks about it being replaced by PCI. Why did this exist in the first place? — Preceding unsigned comment added by 71.114.134.22 (talk)

I'm not sure on the details myself but I belive it was introduced because ISA was no longer sufficiant for the new graphics cards that were coming out and EISA wasn't really that much of an improvement. A quick soloution was needed and the quickest one availible was to put cards on the 486 memory bus. Plugwash 12:53, 4 August 2007 (UTC)

That's a good question, and while I can answer it informally, it should be addressed (and sourced) in the article. Anyhow, IIRC, VL-Bus came about in 1992, and made available 32 bits for the graphics interface (or rarely, other devices), as opposed to the 16 bits available on the ISA bus. This was at a time when hardware-accelerated 3D graphics and their bandwidth demands were just beginning to appear for PCs. There was supposedly a 64-bit extension as well. — VoxLuna ☾ ^orbit_land   20:39, 13 August 2007 (UTC)

The Ron McCabe part at the end seems questionable. He may have helped create the VLB, but I'm sure there was a consortium of companies involved, like with the ISA standard. They should all be mentioned, not just one random guy. —Preceding unsigned comment added by 165.95.15.253 (talk) 14:45, 8 July 2009 (UTC)

The article still doesn't say in what year this bus was introduced.--Srleffler (talk) 14:39, 21 September 2009 (UTC)

Link to Bandwidth page

Although it's mentioned in the "See Also" of this article, in fact I don't see VESA Local Bus listed anywhere in the List of device bandwidths page. And I don't know what its values would be anyway -- AFAIK, it should add up to around 133 MB/s (32 bits at 33MHz nominal), but I'm not sure. — VoxLuna ☾ ^orbit_land   20:39, 13 August 2007 (UTC)

VLB for non-video devices?

Were there ever any VESA Local Bus cards other than Video cards? I never remember seeing any, yet you could get motherboards with more than one VLB slot, so presumably other device types must have existed? 81.135.39.190 (talk) 22:13, 5 July 2010 (UTC)

There were many different devices that could run in VLB slots, video cards were just the most common. Other devices like IDE controllers and I/O cards were also fairly common. 66.68.118.47 (talk) 21:02, 30 August 2010 (UTC)

Some VL-bus based non-VGA cards I saw: 10mbps UTP ethernet with SMC91C92 chip, the "Iwill Side" multi/IO VL card with IDE, floppy, serial, parallel and SCSI controller (Adaptec 6360Q chip) onboard, a noname multi-I/O VL card with IDE, FDD, serial, parallel ports and 4 x SIMM-72 slots for expanding the motherboard RAM. Apparently TV and video recording (MPEG) cards also existed in VL-bus format but they were rare and pricey. 188.143.7.252 (talk) 22:51, 29 April 2021 (UTC)

"While IBM's attempt at producing a successor to ISA"

I think IBM's not attemted, IBM's "MADE" a suucessor to ISA, it worked and worked well for many years and EISA and PCI borrowed some features. The fail at IBM it's not made a succesor bus it's not managing to make it an industry wide standard.

Legacy section

This seems rather biased, and I think it could do with review and better citation, rather than what is probably one person's foggy recollection, that I can doubly refute both by reference to other articles on the same subject, and my own experience.

Case 1: PCI article ... the standard came along in 1992, probably about the same time as VLB itself and not too long after MCA. It had plenty of time to get established even in the early days of the Pentium. Only being established in the soho market by 1996 is nonsense.

Case 2: Reference to Triton chipset ... is very Intel-biased and narrow focus. There were plenty of other manufacturers making IBM-compatible motherboard chipsets at the time, in fact it's arguable that Intel themselves were relatively minor players in the scene they had helped launch, unlike the resurgence they've enjoyed of late. AMD were offering a serious threat to them, and pretty much anyone might be responsible for the ICs providing the low level service on your motherboard. Intel's offerings were actually remarkably conservative compared to the state of the art and certainly through the 420, 430, 440 era it seemed to take them forever to catch up with innovations that everyone else had been enjoying for ages, including SDRAM in DIMMs, large maximum RAM figures, decent large/fast L2 caches, over-8gb HDD support, UltraDMA, low voltage CPU sockets (this one possibly so they could keep selling "Overdrive" CPU adaptors?) and then the ubiquitous Socket 7 (and Super S7 which they avoided completely by going to CPU slots), AGP, ATX, etc. Reasonable enough if you were selling modest-performance but highly-reliable workstations to corporate customers, but unlikely to set a home user's pants on fire. And arguably with the Celeron (the continuing descendant of the 486SX), lower performance examples of their integrated graphics and sound, etc it's a model they've been happy to keep pushing ever since, even alongside the actual market-leading parts.

Case 3: The 486 my family bought in mid-late 1994, which was an entirely normal midmarket example of the breed (but still costing a princely £2000 at the time!). 66mhz 486, 8mb of FPM, 540mb mode-2 HDD (initially PIO-4, but we found out how to twizzle things in the BIOS), 2x CDROM, 16-bit soundcard, 15" SXGA monitor... and a 2mb PCI graphics card, on a motherboard with a decent chunk of L2 cache, a fully integrated drive and serial/parallel port controller, plus two other open PCI slots, three open ISAs, and *NO* EISAs, VLBs or MCAs. And the great majority of new PCs around at the time followed a similar pattern. In fact, amongst a rake of various upgrades we've bought, and other old surplus machines I've owned or used since, from a 286, some 486SXs, 5x86 overdrives, Pentiums (and Cyrixes, AMD K's...) from 120mhz on up, to i5s, I've not actually seen any of those slots up close and for real. There was a single solitary 386 PS/2 at my school which presumably had MCA in it, but I never got to lift the lid.

Honestly, PCI came along and ... that was it. Basically killed EISA, MCA and VLB stone dead overnight for anything other than maintaining compatibility with existing parts and using up leftover ICs/Boards/slot parts. As soon as motherboard manufacturers, system integrators and graphics/storage card designers learned about it and managed to get suitable designs together, they switched all their new models over to it instead of bothering making stuff for what were a trio of instantly obsolete standards. I think I've only ever even seen one advert for a VLB card - the Matrox Mystique - in one of the very first (still 1994) PC magazines we bought, and that was itself also available in PCI form, so you could use it to either upgrade an existing machine, or put it into a new one.

Not in common use until 1996? You've got to be kidding me. We had P200s by then. The beta versions of Quake were already out in the wild. MMX was looming on the horizon. Any computer with an older bus standard was already either relegated to desk-job duties, or trashed. 193.63.174.254 (talk) 17:49, 4 November 2016 (UTC)

Citation work 2019

There are a bunch of dead links and [citation needed] notes all over this article. I'm going to clean them up as much as I can, if anyone else wants to assist drop a mention as help is appreciated. Some of the [citation needed] are going to be hard, like the note on the slang name for VLB was Very Long Bus. I know I remember hearing it and even saying it as a joke years ago but I doubt I'm going to find a scholarly reference explaining that (and it's kind of self evident why IT guys called it that....it's a darn long bus especially compared to what else was common back in the mid 1990s, ISA and PCI). I might just rewrite that sentence to explain it was a slang term and cite a source of it's use. I'll see what I can dig up, let me know if anyone else monitors this page and wants to assist. Kc7txm (talk) 20:19, 27 May 2019 (UTC)

Do you think we can remove the maintenance template alert at the beginning of the article? All four citations needed sections were updated and the four dead links have been sourced. Is there a consensus on removing the alert vs adding more citations? Kc7txm (talk) 22:23, 27 May 2019 (UTC)

What do you think @Zac67:, we have 8 working citations on the page after filling in all four citations needed, should I pull down the maintenance template alert? Kc7txm (talk) 08:08, 28 May 2019 (UTC)

Look good to me – good job! --Zac67 (talk) 17:04, 28 May 2019 (UTC)

Thanks @Zac67:, citations needed alert removed from the VESA Local Bus article. Kc7txm (talk) 19:57, 28 May 2019 (UTC)