Disruptor wrote on 2023-12-21, 14:02:

Basically the calculation is: 50 MHz --> 0 VLB slots (just one onboard device) 40 MHz --> 1 VLB slot 33 MHz --> 2 VLB slots 25 M […]
Show full quote

rasz_pl wrote on 2023-12-21, 12:41:

50MHz VLB would be insane. Just pre VLB some manufacturers released 50MHz local bus graphics systems, but everything was soldere […]
Show full quote

50MHz VLB would be insane. Just pre VLB some manufacturers released 50MHz local bus graphics systems, but everything was soldered on the board. Example Re: Pre-VESA Proprietary 32-bit Local Buses?
May 1992 "Swan 486/33 DB, The blinding speed of Swan Direct Bus" with S3 86c911. CPU and graphic chip on a single, separate PCB daughterboard.
PC Mag 12 May 1992 https://books.google.pl/books?id=jcrsPJb76HAC … irect+local+bus
June Swan 486/50 DB running S3 86c924 at 50MHz
PC Mag 30 Jun 1992 https://books.google.pl/books?id=uummfdS7yF8C … local+bus+50mhz

Basically the calculation is:
50 MHz --> 0 VLB slots (just one onboard device)
40 MHz --> 1 VLB slot
33 MHz --> 2 VLB slots
25 MHz --> 3 VLB slots

It may be possible to add one VLB slot, but that is out of spec then. If you want to go borderline you even may try to run 3 VLB slots at 40 MHz.

Jujoba wrote on 2023-12-20, 21:52:

Yes, it exist. And I have one in a socket 3 motherboard and working! I have had many troubles to make it work correctly. Sometim […]
Show full quote

Yes, it exist. And I have one in a socket 3 motherboard and working!
I have had many troubles to make it work correctly.
Sometimes the mouse didn't work, or i have troubles with the video card, i put Windows 3.1, and Dos 6.22 and an adapter from MicroSD to Ide as a harddisk.
This Cyrix work with 5 V

Even an empty VLB slot raises the capacity on the Bus. Perhaps this may cause troubles even if you don't use a VLB card.

Does your Cx486DX-50 run its cache in write through or write back mode?

So I've got a unicorn 😁
My Am5x86 runs at 50x3, with vlb i/o and vlb graphics (Trio32). It is picking in the vlb cards, so i get your point 😉

Honestly i still can't get why Cyrix would bother to bring out a dx50 at all. Lots of boards had proven to be unstable at 50MHz bus, while the DX2-66 was just as fast and perfectly stable with any half decent board.
Intel was the pioneer at this speed so I get why they hoped chipsets would develop along...

I don't do mere insane, more like batshit raving unhinged... VLB at 60Mhz... my system of the 90s, Bektronic BEK V439 motherboard UMC chipset, a cherrypicked GD-5429 gfx card and a cherrypicked UMC i/o card. Both needing 1WS of course. Cherrypicking was a lot cheaper when I was getting them out of computer fair junkboxes for $1 a pop.

Anyway, for VLB eXXXtreme, your selection process should go like, find cards that run at 33Mhz 0ws and play nice with each other. Any that don't do this are crappy, and you should stick them in the box of last resort, they will cause you problems in any setup. Given a large enough random selection you probably have less than 10% that bad, but they are out there. Now you know you have baseline okay cards. Now try running them at 40 0ws, you'll lose half of them here. Maybe they would do 50 at 1ws, but if going for eXXXtreme, you will find these lacking... so 50 1ws should be cake for the remainder right? Maybe, but some will fall out again.... now if you've got 3 pairs that run 50 1ws you might get one lot that run 60 1ws but you might need to swap around between them to find the two that like each other the most. If you've left in the won't do 40 0ws but do 50 1ws this is a lot larger lot to mess with but you'll have so many permutations to try at 60 with it being unlikely that you were wrong to deselect them at 40.

Do not preselect based on your notion of whether a card is slower or faster at stock speeds. You gotta go slower to go faster 🤣 ... nah, what I mean is, that it might appear slow because it has other baked in wait states in the implementation, where these will be an advantage to get to 60mhz, instead of it already being on the raggedy edge of the timing at 33.

waterbeesje wrote on 2023-12-21, 14:37:

Honestly i still can't get why Cyrix would bother to bring out a dx50 at all. Lots of boards had proven to be unstable at 50MHz bus, while the DX2-66 was just as fast and perfectly stable with any half decent board.
Intel was the pioneer at this speed so I get why they hoped chipsets would develop along...

Perhaps marketing.
I'm sure Cyrix wouldn't have wanted to damaged their brand name by releasing processors that will run instable not because of the processor but because of the contemporary fragile chipset at those clock frequencies.

50 MHz FSB even had problems with PCI: Look at the disaster with the PCI 2/3 clock frequency divisor of MB-8433UUD. It's PCI clock was turned on for 1/3 of the period and turned off for 2/3 of the period. Many PCI cards refused to work, even at 40 MHz * 2/3.