AMD 386 DX-40 and VESA LOCAL BUS - Boardreview and Benchmarks

[@EEVblog]

Wow, a combined 486 and 386 co-processor socket!

[@charlesdorval394]

Crazy right?! That has to be one of the most interesting PC motherboard I've seen

[@CPUGalaxy]

yeah, for me as well

[@uwezimmermann5427]

what a great idea!

[@chihhenghuang]

In Taiwan retail market ,some dealer call it "786" main board because this board have 386 and 468 socket ...

[@soniclab-cnc]

I had this board way back.... man I wish I had kept all my retro stuff. No time and no space for boxes and boxes of motherboards, I donated everything to a local computer club.

[@saloparenator]

you made me realise my 386 board actually support 486 cpu !!!! i didnt knew that oversized fpu.socket was a 486/387 hybrid

[@ernestuz]

Ok, now it's clear for me, you are from AUSTRALIA! (Why didn't you add some kangaroos in the snow scene?). Keep up the good work.

[@مقاطعمترجمة-ش8ث]

Or Arnold Shwazniger

[@CPUGalaxy]

Schwarzenegger is the right name

[@مقاطعمترجمة-ش8ث]

@@CPUGalaxy Aha I just used google auto-correction, still I miss spelling it 😬

[@aboodmazen5983]

@@CPUGalaxy Which processors are installed on the socket 7

[@soniclab-cnc]

@@CPUGalaxy Scpuarzenegger

[@TheCarlos206]

Love this kind of videos ! THX for show us this tech!

[@IBM_Museum]

A moment of silence for the 387 FPU, please - hopefully, it wasn't too rare.

[@ThunderClawShocktrix]

dunno why the pin 1 was different direction

[@FOIL_FRESH]

the stack of FPUs had me grinning ear to ear

[@intrinia]

Ikr! Flexing like it's nothing. ^^ :D

[@flymypg]

In the late 1980s and early 1990s I worked at R&D companies developing custom high-speed real-time sensors and instrumentation. One of the key factors was the "cost of compute", where a 'fast' PDP-11 lab computer could dominate the budget of an entire project and wasn't very portable. We switched to DSP arrays, which helped on size but not on cost, since programming them was a PITA. We longed to use PCs, but the I/O bottlenecks were restrictive. Until the 80486DX and VLB kicked the doors wide open. We developed a combined ADC and DAC VLB card that reduced our overall system compute cost by 10x, as even a large stack of 486 motherboards with our card installed cost less than all alternatives, and was much easier to program. We made other versions of our card when PCI arrived, but the VLB solution stayed competitive for the first few PCI generations. The main problem with VLB was that it was much better for writing by the CPU than reading (which makes perfect sense for a video bus). So our DACs ran fast, while the ADCs were bottlenecked. We found some 'tricks' in the VLB spec that allowed us to make our own DMA controller that bypassed the bottlenecks, giving us several advantages in the final cost and performance of our instrumentation. VLB FTW! Edit: We didn't go from DSP arrays to the 486 in a single step: We used lots of VME systems along the way, based on Motorola 680x0 processors generally running OS-9. The VME bus was extremely flexible, but it didn't get truly fast until later generations: VLB really killed it, so long as you were willing to design your own boards.

[@CPUGalaxy]

wow, what a cool story. 😃👍🏻

[@stevef6392]

That's totally the kind of Frankenboard that Asrock would've made had they been around in the early 90s.

[@dawidoszkiewicz5607]

Best comment of the video!

[@Rubikscuberz]

I love just the stack of CPU's in each slot option

[@GadgetUK164]

Love seeing all this old PC hardware - I have a lot of nostalgia for VLB! I found my VLB setup faster than when I eventually move onto early PCI (both top end cards at the time).

[@PixelPipes]

0:43 "We also have technology in Austria" It must take FOREVER to render your videos on a 386!

[@PixelPipes]

Very interesting tests by the way! I guess one might suspect that a 386 doesn't benefit from VLB, but until now I've never seen anyone able to PROVE it.

[@laharl2k]

Well back then people had dedicated cards for encoding video in production environments. Not sure how new they are but it would be interesting to see a comparison between one of those vs a moden cpu encoding mpeg video

[@hgbugalou]

Funny. A guy from Australia got me, an American, following a guy in Austria.

[@CPUGalaxy]

😂👍🏻👍🏻

[@kayakMike1000]

Jones sent me here also! Subbed just yesterday. Merry Christmas!

[@wayneholzer4694]

I was sent here by David from eevblog

[@clintthompson4100]

Pretty sweet motherboard that can take not only a 386 processor but your 486 chips, overdrives cpu by kingston(granted it needs to have the voltage regulators)and vesa cards. I like when a manufacturer actually puts in the time to give the person buying their product options and future upgrade paths its details like that that always make some manufacturers always superior over others.

[@uwezimmermann5427]

He tricked me as well, a German living in Sweden!

[@_bacterio]

Hi, good video showing the difference in performance. I'm the FastDoom developer (a Doom port optimized for 386/486 processors), can you try it with this setup? I want to see if there is more benefit using vl-bus cards on 386 processors with my port, as I don't have access to a 386 motherboard with vl-bus. Thanks!

[@CPUGalaxy]

sounds interesting 😃

[@Trancelistic]

@@CPUGalaxy I am intrested as well:)

[@intel386DX]

Wow cool! I want to try this port too :)

[@vapourmile]

@@CPUGalaxy I'm interested too. :)

[@CrassSpektakel]

I guess I know why the VLB IO is slow: You need drivers. Without drivers they are running in some PIO-Fallback-Mode - I remember PIO-0 was limited to ISA/4 or even ISA/5 which is roughly 1500 to 2500kByte/s depending on FSB. And yes, most of these cards do not have DOS-Drivers, only Drivers for Windows. Also most chipsets do not support DMA but at least higher PIO-Transfers. I own a DTC-2278-VLB-Multi-IO-Board which delivers full 33MByte/s by DMA (!!!) using 32Bit-Drivers with Linux, W98 and WNT but only 2MByte/s with plain DOS. I have the original driver disks and manual and it pretty much states "no DOS drivers use a real operating system". Fun-Fact, the W98-drivers do not load before W98 but within W98. No config.sys, just real 32Bit-Drivers just like WNT. Which basically means the system boots with Fallback-PIO and only activates the full drivers after starting the W98 Kernal. The drivers themselves are win32-drivers compatible to WNT and W98 all at once. Pretty mad shit back then.

[@valenrn8657]

Change the CPU to 486. 386DX doesn't have the arithmetic intensity like a 486 or RISC CPU at a similar clock speed. The 386 can't deliver 40 MIPS for 40 Mhz for 40 Mhz VLB.

[@pvc988]

Back then: 1 motherboard supporting 2 different CPU generations (EDIT: even from different manufacturers) Today: Basically the same CPU released 10 times for 3 different sockets

[@nbrown5907]

Yea imagine my surprise back in the day to find out that the SX and DX were the same they just did not solder the Math Coprocessor connection on the SX chip lol.

[@LiLBitsDK]

yeah and those overdrives are like pentiums :D så 3 gens or actually more than 3 gens when you look through all those he stacks up

[@Trancelistic]

Yep, the cpu benchmark truly shows it was totaly worth to ugrade a pc from 386 to 486 (huge jump) and again to a pentium (again insane jump )

[@ResonantBytes]

Wow, I'd never thought I'd see that board again. Awesome video! Please keep making them! :) I had a PC with this board, a TX486DLC and an IIT 4C87DLC (both 40MHz), complete with VLB VGA and controller cards. What a weird combination in hindsight. But as we say in German "Don't inspect the mouth of the horse you just received for free" (non-literal translation) :D Sadly the BIOS ROM got corrupted at some point (or at least that's what I think happened) and without the knowledge or tools to fix it back then, the board went into the trash :(

[@prestigious0]

I love that western digital caviar startup sound! My pair of 2200 drives both still run. It was fun to plug them into a usb to ide cable and browse my digital time capsule from when I was a kid.

[@PyroRob69]

Beautiful stack of chips that bring back a lot of memories.

[@OzzFan1000]

Now THAT is a rare find that I didn't even think existed! A motherboard that accepts 386DX and 486 chips using VLB? Also, a great collection of 386-era x86 chips you have there. Thanks for sharing!

[@perhansson6718]

Great channel! That white mark (I thought it was some polystyrene foam packaging) was bothering me so much before you explained what it was, what a (NOT!) great way to signify pin 1 for the FPU!

[@MrJef06]

Yeah the design is not good to begin with but perhaps they had strong routing constraints. In any case a big fat warning sign on the PCB would have been in order!

[@AndreiNeacsu]

This is one of the most interesting motherboards I've ever seen. Thank you for the video!

[@robertbeighter6336]

Wow, I just found your channel - what a hobby! And what an archive. I have a drawer full of old consumer CPU's I've collected, including my 386dx40... AAAAh the old days, overclocking by swapping crystals. Keep up the good work, definitely gotta sub from me!

[@AncapDude]

Geiles Board und schöne CPU Sammlung. Als ich damals auf meinem 486 von ISA auf VLB wechselte war ich richtig geflasht wie flüssig DOOM auf einmal lief ✌️

[@lactobacillusprime]

I've only heard about these - this is the first time I actually laid eyes on one of those combination boards. Lovely how you actually stacked the supported CPU's on the sockets. Thanks for sharing this.

[@williamambrose3947]

Came here from EEVblog. great and interesting channel looking forward to watching your back catalogue of videos

[@CPUGalaxy]

Thank you for visiting my channel. Hope you like my other videos as well. Peter

[@trnguy6137]

Nice job. I agree that pin 1 orientation is a killer. Also thank you for dropping the chip. It verified gravity is still working

[@christopherlee3977]

Talk about a blast from the past, saw the VLB slots and had to click...then got a bonus section at the end wit the Gravis Pro board mention. :D I think all my antiques are long gone by now though.

[@adriansdigitalbasement]

My thoughts on the lack of speed increase on the hard drive: IDE is just an extension of the 16 bit ISA bus, so it isn't going to go faster even connected to a 32 bit bus as a latch of sorts will be needed to take two 16 bit transfers to convert it to one 32 bit transfer. Probably happens less efficiently than the native ISA bus on that CPU which is doing something similar.

[@hespelf]

Yeah, 386 CPU is not great with VLB. The VESA Local Bus relies heavily on the Intel 80486 CPU's memory bus design. Add on top of that the fact that DOS is not a 32-bit OS natively, and the DOS driver only brings marginal improvement. To me, best use case scenario for this MB using this VLB IDE controller would be below combination : - 486 type microprocessor or Overdrive - an OS with 32-bit native drivers capabilities (like Windows 3.11 did) - optimized jumper settings on VLB controller to maximize performance (HDD mode, IOCHRDY, ...) - a hard drive with PIO4 capability (time-correct with this VLB controller board)

[@seleccionmultiple2905]

That motherboard is the definition of versatile.

[@run4restrun259]

I'm first play in Doom 2 on Am386SX, with small screen and PC speaker...)) And also - I was have 2mb of RAM! In Win 3.11 with swap file it works fine)

[@mimas1996]

ouch, I was pretty worried about that cpu, when you hold the hdd over it :D nice video

[@CPUGalaxy]

😅

[@atheatos]

I am also living in Austria and I have a huge retro collection. Especially Socket 7 stuff and AGP cards. When I find the time I will do some videos. BTW, great M/B I wish I had this too.

[@CPUGalaxy]

supa. schreib mir eine mail mit deiner tel nr. dann kommst in eine whatsapp gruppe solltest du lust dazu haben. ☺️

[@tlafeir]

Amazing channel. I watched a bunch of your videos already.

[@TheAmazingAdventuresOfMiles]

I used to run a 386DX40 back in the day, and I played Doom quite happily on it. In fact I completed Doom 1 and 2 on it. I found that if you switched it to the lo-fi mode (I forget which key did that, F5 maybe) and dropped the render window size a notch, it became really quite playable. It did not look as good of course but that was the trade off. I also played it on my 386SX40 but that wasn't very playable.

[@RobRussellOnline]

I had a very similar motherboard way back when, and I don't remember if I upgraded it to a 486 or skipped to a different 486 board altogether. I think it died and I moved on but I'm not sure. I probably did a Cyrix 586 thing. Anyway, thanks for the clarification of your accent - it's been driving me nuts! Well, not really nuts, but I live in southern Germany and I just couldn't place your accent. Thanks for your videos dude! If I ever get back to the USA I will be sure to bring my collection of CPUs back here. I think I have some goodies that are just sitting in storage.

[@jorisammerlaan3063]

Hahaha a video which brings me along the memory lane! My first PC was a 486DX-66.

[@dennisp.2147]

KZbin algorithm FTW! I wasn't aware of any CPU collectors on KZbin. The stack of 386 processers makes me insanely jealous...

[@Sebastian_L.]

Man stelle sich vor ein p24T würde da noch drauf passen! Wieder ein sehr interessantes Video, die Variabilität damals war schon echt aussergewöhnlich

[@PeterMilanovski]

Hello and merry Christmas 🎄 from Melbourne Australia, excellent video! I still have my AMD 386 DX40 with math coprocessor, TSeng Lab's 2mb graphics card and sound blaster 2.0 for audio complete with Maxtor 85mb Hdd lol, oh and there's 4mb ram! I remember that I chose it back then because it had the math coprocessor, the machine did everything well that was available at the time, although I never really knew what the math coprocessor did and I still don't know what it was doing, naturally I was excited to see that someone has made a video based on this processor but I didn't see you install the coprocessor, I don't know if you have done a video showing the pro's and cons of using a coprocessor yet as I'm just going through your videos at the moment! Thank you once again! I'm happy that I still have my machine, I'm one of those people who never really throw anything out! But it seems that I'm going to need to dig it out and give it a full service!

[@PROSTO4Tabal]

Nice motherboard. We need some more videos featuring Gravis Ultrasound :)

[@wayneholzer4694]

Great channel was sent here by Dave from eevblog hidden gem this KZbin channel is cheers from Australia

[@LtSyrek]

For PTI-255W, loading a wbide.exe driver in config.sys increased reading speed of 1GB CF card from 4MB/s to 6MB/s in speedsys in my PC, so that would make a difference. The driver also allows to use 4 drives at once and possibly bypassing bios drive size limit (not tested). Nice video with rare configuration.

[@jacekdobrowolski9115]

A great journey to the past. Tnx. I'll be guest of your channel from now on.

[@CPUGalaxy]

Thank you very much.

[@AntonyTCurtis]

You need to use the DOS driver for the VLB IDE card to make use of its capabilities. The default is for ISA PIO compatibility which is almost always equivalent to ISA 6 MHz performance.

[@CPUGalaxy]

I tried that already. Only slight changes and still the HDD run faster on the IDE controller 🤔

[@AntonyTCurtis]

Would be interesting to examine different VLB IDE performance. Back in the early 90s when I worked at OEM shop, we compared dozens of different IDE hard drive and VLB combinations. We even examined the effect of cable length.

[@mlodzin90]

I have simillar board. Those are interesting boards for testing. Mine was originally with 386DX/33 installed, I upgraded it to 486DX2/66.

[@adriansdigitalbasement]

I recently got a board similar to this that can handle 386 or 486DX chips but the 486 socket can also accept a 387 math coprocessor! And yes, VESA Local bus slots too. I had never seen anything like it before. Edit: your board is the same I see!! Cool! The only difference is my socket has a mark for the pin 1 on the math coprocessor unlike your board. Sorry you chip got killed. The orientation on my board is the same where the math chip is rotated versus the 486.

[@CPUGalaxy]

How cool that you got the same board 😄. Lucky you that the pin1 is proper marked for the FPU. Is there a marking on the silk screen or in the socket? Maybe you can send me a picture. I would be interested to see it. And how cool to see you on my channel 😊

[@adriansdigitalbasement]

@@CPUGalaxy imgur.com/a/vbJ6pvr it's different than your board. This one has 3 VLB slots! It came with the CPU you see in the pics but that CPU was actually dead. I put a 386-DX25 in it and it worked fine with that. Didn't do a lot of testing but I did try the VLB slots with the 386 chip and it was indeed working.

[@CPUGalaxy]

very cool board as well. and of course a huge pin 1 marking for the 387!! 👍🏻

[@SUCRA]

Amazing test, great idea. I wonder how it would come out with a 486 sx in place of the 386. And the pinball with the Gus at the end sounds super cool!!! I can almost feel the ball hitting the table and the flippers. Thanks for another cool video! 👍

[@dorinxtg]

IIRC, they were few VLB IDE cards which used the VLB bus, but there were some compatibility issues (example: many WD disks didn't even boot with those cards), so Winbond and others rolled those cards to work in ISA mode, and that's why you get these speeds. (I don't have any link to what I wrote, I simply remember it from those days)

[@CPUGalaxy]

I recognized now some strange issues with this WD drive. When I put a 486DX-2 into the socket the board went crazy. sometimes boot, sometime crash, then messages like „command.com“ cant get loaded, „wrong version“ and all this things. with a 486DX-33 it was working fine... really strange. Then I took an older Maxtor and everything works fine than. 🤔

[@alain99v6]

wow I had that motherboard ! I bought it when my 386 dx33 board died to restart my system and later upgraded the cpu to 486 dx50, running vesa boards at 50 mhz was tricky the trident video board loved it but not so much the iO board but playing with some settings in the bios made it work nicely

[@piotrludorowski9529]

Nice, i have a collection of vintage cpus. Have also this cpu!

[@marcosborges6345]

Cool... a time travel of PC's. I from this time.

[@Shand1982]

Very envious of your chip collection! Cool video!

[@lassekristensen385]

I had a 386 DX40 and I remember it could 100% run Doom smoothly! I know this because I bougjht the PC only for Doom. I clearly remember that it ran Doom smooth and I am certain it was a smooth as it allows in the OG DOS version. At the time I was a kid with no experience in BIOS. But I was the experimenting type. But there was one time where I by mistake disabled the EXTERNAL CACHE and when I did that my 386 ran Doom exactly as your in the video. Very slow and lagging low fps. Because I did not have much experience and at the time did not know what caused this, I was devastated that the pc now ran like this. It actually took me months to suddenly figure it out and reenable external cache in the BIOS and then Doom ran super smooth again and I was SO SO SO happy! :) This is a speciel episode and learning curve in my PC computer history so I remember it clearly that it ran perfect Doom. I cannot remember if I had a Intel or AMD but I know it was the DX40. This is a guess but I dont remember having the INTERNAL CACHE disabled. I am sure both mine was enabled? TBH I dont think ANY CPU has 0kb L1 Cache. TRY to enable that? I remember my friend had a 386 SX 25 and that ran Doom like yours DX. He always had to make the window in Doom the smallest haha! Where I could run smooth full screen!!! This is crystal clear for me. There MUST be something wrong with this Doom performance in the video. Unfortunatly I dont remember my graphic card but I l know I have had a TSENG, MATROX, CIRRUS LOGIC, ATI. Otherwise great video! Did you also use the latest Doom version? dont know if that can make a difference.. Think its 1.9

[@valenrn8657]

Low details vs high details.

[@lemagreengreen]

What a neat motherboard, don't remember seeing anything like this

[@teknoman117]

I looked at the pinout for the 387 and the 486. I'm certain it's rotated the way it is to make the board layout easier. The first three rows of pins on the 486 on the side of the socket facing up from our perspective are all of the data bus pins (D0-D31). The data pins on the 387 are in the same relative position when it's installed correctly. I'm almost certain that the data pins on the 486 pins are connected to the data pins on the 387 pins to make layout simpler.

[@wskinnyodden]

The reason for the pin1 position on the 387 is guaranteedly because it made the wire routing easier for the board plan.

[@marcinkrupinski3782]

Great content, very interesting and entertaining at the same time. Please consider comparing AMD 386/40 and AMD 486/40 using vlb setup.

[@CoreyDeWalt]

I am so happy that you made a joke, and are very intelligent. An instant subscription from me.

[@GGigabiteM]

I would guess the reason the 387 is rotated 90 degrees from the 486 on the socket was because of some board layout issue. It could have been easier to lay the board traces that way, or it may have been impossible to route the traces with both being oriented the same way. If that was the case, they should have just had a separate socket for the 387, there is plenty of open space on that board to do that.

[@duderobi]

I am a simple Austrian I like Austrian content automaticly.

[@RuruFIN]

That was funny how many times you said "we can also put a..." when you showed the compatible processors :D

[@CPUGalaxy]

😅

[@CPUGalaxy]

lol. i had to watch it again and now I can hear also all the lot of „put“ 😂😂. crazy... i was not aware of that.

[@StanislavBruiev]

Oh, that was amazing Edelweiss! By the way, nice reviews :)

[@Tech2C]

Wow you've answered a burning question I had for ages - will a 386DX work with VLB. And I guess from the benchmarks it wasn't worth it on a 386. Seeing no speed improvement in games was disappointing and explains why it's so rare to find 386 VLB boards. Cheers!

[@luheartswarm4573]

Cpu Galaxy is that cool geek foreign uncle we wish we had

[@LiLBitsDK]

would be lovely to see the same test but now with a couple different 486 cpu's on the same motherboard :D and compare those to the 386 results as well :D

[@CPUGalaxy]

This is what I planned 😊

[@LiLBitsDK]

@@CPUGalaxy YAY! :D are we there yet? no? what about now then? :D

[@CPUGalaxy]

😄

[@chouseification]

Seeing that mobo brings back good memories for me. One of my college roommates picked up a Gateway 2000 (complete with cow pattern on the shipping box) that came with VESA local bus, and it was the first time we got to see it. Up until then of course it was ISA 8 and 16 bit slots or the much more expensive EISA, which made more sense for people needing early RAID controllers but didn't get a lot of traction as a video card slot (due to much higher cost). Unless you had an IBM PS/2 with MicroChannel, seeing the VL slot was the first new slot in most users' experience with IBM compatible hardware (before they regularly called it Wintel). Of course years later we got AGP and then the various PCI then PCIExpress slots, but for the era that VL bus was pretty awesome. :D

[@chouseification]

I made that comment before you brought out the stack o' FPUs... ahh so nice. I was stuck on my 286-12 that year my buddy picked up his new Gateway, although I was able to get an 80287 from another buddy for like 20 bucks - and with certain apps and games the presence was very obvious. For those who don't remember, many early CPUs didn't come with a built-in floating point unit - used to quickly do higher precision floating point math than what the math logic unit (ALU) in the CPU could do itself, so if you didn't have one the same math would be done in a much slower fashion using integer operations. This would take many CPU cycles to do the processing, so apps tried to avoid the use of floats where not really needed. Instead if you had a math coprocessor installed and the software knew to use it, it used a simple stack mechanism to push the operands and operators, then a command to run what was on the stack, and then retrieve result... this process was amazingly fast compared to the slow method. Younger folks are just used to CPUs that can do float without an issue, so this history may not make as much sense without the context that a lot of fiddling with memory settings (are you needing Extended or Expanded) on boot were needed to run various things, so you often would have boot floppies (even if you normally booted from hard disk) just to run certain games or apps. You have it so easy these days, so enjoy. :D

[@CPU-Z]

Thanks for that video, this is very interesting. By chance do you have a Ti486SXL in PGA168 format ? I have one of those, and none of my 486 mainboards supports it, and I wonder if this (wonderful) hybrid board would (logically it should since it supports the PGA132 version).

[@Technoid_Mutant]

I was a tech during these days, and it was a little early for VLB. The operating system wasn't mature enough to take advantage of it except in very specific circumstances and with software which knew the hardware.

[@brandonupchurch7628]

While the 387 and 486 don't share pins, they would need the same data bus lines just quickly glancing at the pinouts it looks like the orientation that the 387 socket puts the data pins close to the data pins on the 486 socket so that they could run a shorter trace over to the pin on the 387 socket instead having to route long awkward traces 90 degrees around to put pin 1 in the same position.

[@cornflake75]

Ah, those memories...I mean the past times, not the actual chips 😉

[@tl1024]

Thanks for another awesome video. Please do a full shootout with all the FPUs! I need to know which one was really best for different applications.

[@CPUGalaxy]

Yeah, i definitely will 😉

[@tl1024]

@@CPUGalaxy also, any interest in a AMD K6-2/500afx for the collection? It provided me many many hours of gaming & lan gaming in its day.

[@CPUGalaxy]

I think I have all AMD K6, K6-3 & K6-3. This is an awesome socket 7 CPU.

[@sonyisda1]

Should add #doscember to the title to join in with other channels featuring MS-DOS related videos this month

[@blackbox5141]

im addicted to your vids haha thanks eev vlogs for the recommendation

[@CPUGalaxy]

lol, thank you. I will do my best to satisfy your addiction 😉

[@drpc97]

funny how you don't have more views, it sounds like Arnold Schwarzenegger is talking to me about old computers. its nice!

[@pabloescobar1036]

Please make a video about 386DX-486DLC-DrX2-486SX-486DX-DX2-POD and others at same mainboard.we need this 😎

[@bborkzilla]

All that with a Ni-Cd battery that didn't leak! What a rarity!

[@MaxKoschuh]

super, wie immer

[@necro_ware]

I have a very similar mainboard here and also wanted to make a comparison one day and may be a video about it on my channel :) I guess I don't need to now. I was also interested in VLB performance on a 386, but the benefit seems to be very limited. I guess, it's due to clunky implementation of the VLB here. Actually, it was "invented" exclusively for 486 and I guess, in combination with a 386 on this mainboard, it runs through an additional bridge, so it is not directly connected with the CPU, as it would be the case with a normal 486. However, it would be interesting to see, if it runs significantly better with the Cx486DLC, and a real 486 of the same frequency. Just to see what is the limiting factor....

[@smunaut]

I'm not surprised DOOM doesn't improve with VLB. The way DOOM works is that it renders all the graphics into main memory first, then once the frame is final, it copies it to the video memory. (AFAIK That's partly because the VGA mode used doesn't support any kind of hw based dual buffering). And that copy isn't the big part of the time. Even if the 386 was able to render 35 frame per second (max doom tick rate), this would still only be using about 2 Mbytes/s of video memory bandwidth.

[@Radek__]

14:53 I know why the result is worst for hdd. On the Winbond 32bit i/o controller you have JP1 jumper, which is divided to sections of 1,2,3,4,5. Section 4 and 5 is for latency delay. I see at 11:50 that you have those jumpers on the default position (all on top), so it will give you slow 600nS. So, set jumpers from section 4 and 5 from top position to bottom to lower latency to 240nS. This will increase data transfer speed but it can produce issues if hdd is from early technology. I believe that slow default 600nS is set for safety to run newest and the oldest drivers in the same time.

[@B24Fox]

wow, good observation!!

[@TheGuruTech]

I recall having to use a DOS driver to enable the VLB IDE controller’s faster performance.

[@CPUGalaxy]

good hint. thanks. i did not consider that.

[@LiLBitsDK]

@@CPUGalaxy yeah I recall that too, it was a pita since omg precious memory :D

[@RobertHancock1]

Yes, otherwise the BIOS doesn't know to use 32-bit transfers to access the drive, plus it likely needs to select a faster PIO mode on the IDE drive itself to speed up the cycle time. The DMA modes were still just a dream in the VLB days unfortunately..

[@death2all79zx]

Did anyone else flinch when he dropped the 386?

[@JorgeCarvalho_web_dev]

Hi Sir, greeeeeeeeeat video!!! I have the same mobo that developed a problem when posting freezes on RTC error. I think that one of the OPTI chips developed a problem

[@georgysb]

Goblins 3, Sea Legends are the perfect ones for 386 era PC.

[@mikegamsjager8944]

Back in the day's there was an dos device driver called WBIDE it was used to improve performance especially in VLB configurations with Winbond ( WB ) Cards :) Maybe you should give the device driver a shot :)

[@mikegamsjager8944]

ifsc2.ifsc.usp.br/drivers/LAN/ISA/PTI-WBIDE/README ifsc2.ifsc.usp.br/drivers/LAN/ISA/PTI-WBIDE/DOS/

[@KJohansson]

Had a Acer Power 386 cirka 1990, it had a soldered 386sx 16, and 486 upgrade socket.. Anyway, have you looked at the Hauppage 4860 boards?

[@CPUGalaxy]

oh man, i would love to have a Hauppauge 4860 board. It includes a intel 860 RISC CPU as well a 486 😍😍

[@KJohansson]

@@CPUGalaxy yes I remember the ads for them "Risc free" was the tagline. 😁

[@wishusknight3009]

The hard drive performance anomalies are part of the reason the OPTi chipsets were considered low end for their day. They had very poor interrupt and bus arbitration performance. VLB will only shine in DMA operations, for example where graphics bios routines can be cached in ram. As far as I know there are no IDE controllers from that era that are DMA outside of a couple of specialized and exotic caching controllers, which would thus require cpu intervention for any bit of data transfer from the IDE controller, and that would require being handled by the interrupt controller which resides on the opposite side of the ISA bus arbiter. So while the ISA controller can service the interrupt controller directly, the VLB controller needs to go through a bridge before it reaches the interrupt controller. You wold probably need to step down to a VLB speed of 33 mhz and see if you can run that controller at 0 waitstate. You would need to run ISA video as well, that may be the only way it will match the ISA card at all. SCSI cards are also typically DMA and you are able to find the occasional VLB based unit, and a drive with async reads and writes is capable of showing that difference. As far as i can remember the SiS chipsets didn't do much better, you needed to play with waitstates and pretty much keep the VLB to a max of 33mhz.

[@ILRBW]

On my AMD 386DX40 / ALI M1429/1431/ CL GD5402 ISA / ESS688 ISA sound, the first DOOM went noticeably faster than on this video. It was possible to normally play full screen in normal graphics. It didn't work out with DOOM 2, the game went up to about the 5th level, and then because of the size and complexity of the level itself, and because of the high number of monsters, the game has already slowed down to 5fps and below. But I went into the "chipset settings" in the BIOS and tweaked the system as much as possible, including raising the ISA frequency to 16 MHz. Moreover, in Windows 3.1, videos of the MS Video 1.1 format were running normally, even on full screen. Unfortunately, I don't remember the resolution of the videos, perhaps 320x240.

[@wskinnyodden]

Oh thank you god, thank you, this one is also missing the 2 PAL chips, for a second I though I may have been screwed with mine :) ATM building it with a DLC 40, 4Mb Cached 80186 based IDE controller, 512Kb Trindent VGA (need to get a 2MB VL one) and a nifty AMD Ethernet Controller. Still missing an FPU... Though I may end up just putting a 486DX on it... Seems my only issue was I had not configured it yet and it was seemingly set for a 486DX 33Mhz, so definitely no go for a DLC 40Mhz (not that the clock would matter in this case)

[@MichaelRusso]

I have a board very similar to yours. I over clock my Intel 386 33 to 40mhz . It runs stable with a 32 bit version of Linux on it, X-windows, Apache...

[@matthewday7565]

On the VLB I/O, JP1/4 & JP1/5 are set to the slowest option - stason.org/TULARC/pc/hard-disk-floppy-controllers/C-D/DTK-COMPUTER-INC-Four-IDE-AT-drives-PTI-255W-VER-1.html - The driver set from archive.org/details/dtkpti255w includes a DOS driver that can enable multisector mode - Appears to be a floppy image maker, but the IMA can be unpacked from the EXE using 7-Zip, and then the contents of the IMA can also be unpacked to get to the readme and WBIDE.EXE (Winbond IDE)

[@ulrichfrank4270]

Maybe VLB-capability is not even active for VLB-IDE-Controller when 386-socket is populated - hence weaker performance. Had that same kind of board once (with"MR-BIOS"). It did not even recognize an Adaptec 2842VL-SCSI-Controller when 386-socket was equipped with a Cyrix 486DLC; it was seen only with a true 486 dropped in the 486-socket.

[@georgH]

I wonder how this motherboard makes the VLB accessible to the 386, as the VLB is tied to the 486 memory bus, with the CPU managing arbitration between VLB devices. According to Wikipedia, OPTI was manufacturing VLB-to-PCI bridges, allowing VLB cards to be used on a Pentium, but this is wildly different than adapting a VLB to work on a 386. How did they manage it? Thank you for showing us this rarity, and that it actually worked! I was surprised to see the increase on performance on the video memory access! I also admire your collection of CPUs and FPUs :)

[@CPUGalaxy]

Thank you 🙏🏻

[@mikoyangurevic8634]

I've always being intrigued with a VLB setup, so I built my own in 2019... For VGA you can use for the 16 bit the WD90C31A-LR 1 MB (Paradise), it's very fast, I believe the fastest for an ISA slot. I have checked ET4000AX, Trident and some Cirrus Logic, all with 1 MB VRAM, but the GUI acceleration for Windows 3.x is very poor than the Paradise. And for 32 bit (VLB) try the SPEA Mirage (Video Seven) S3 864 Vision P64 with 2 MB VRAM. Fastest card (yes faster than ATi Mach64) for VLB bus and with excellent GUI acceleration. On my i486Dx2 66 MHz it runs Quake and it is playable!

[@TecKonstantin]

Ohh that board would be nice with a Aries PGA ZIF socket!

[@CPUGalaxy]

ah yes, I was thinking the same

[@TecKonstantin]

@@CPUGalaxy cant Imaging the horror of taking the CPUs out, maybe a mod for the Future

[@anomaly95]

The ZIF socket could be inserted on top of the soldered-on PGA socket for testing. The ZIF would act like a "wear-out" connector.

[@molletts]

I'd be quite interested to see a comparison between the GD5428 ISA card and the VLB version of it. (I wasn't even aware that there was an ISA version.)

[@CPUGalaxy]

I like your idea 👍🏻. Just need to check if i have a 5428 VLB. Which cpu do you prefer to see in this comparison?

[@molletts]

@@CPUGalaxy I used to use one with an Intel 486DX/33 (which I later upgraded to an Overdrive DX4/100). I think I may still have the graphics card somewhere, although the old motherboard is long-dead.

[@kousakasan7882]

My Orchid Superboard had 32 bit graphics. But I got the 33mhz. But it played Castle Wolfenstein and Descent just fine.

[@marcinkowskitomasz]

My guess for VLB IDE slower transfer is that it indeed is slower BUT it uses DMA transfers instead of polled IO, therefore does not use CPU and it would feel much faster anyway!

[@skyhawk21]

The good ole days of sockets, jumpers and Simms with slow ide hdds