Did you have 12 ns cache to confirm that faster cache would allow for 2-x SRAM settings?

When you say "128 KB", are you using 8 chips + TAG or 4 chips + TAG, whereby 8 chips would be in interleaved mode and 4 chips would not. I have found that when pushing the limits of the memory subsystem, particularly with higher FSB's and greater SRAM quantities, that using double-banked cache (8-chips, hence interleaving) you can obtain faster SRAM timings compared to single-banked (4-chips). For this reason I never run tests single-banked.

I don't have faster chips but kixs does, and I had the exact same issue - see above:

kixs wrote:

Small update... found some 12ns cache chips and swapped them. Now it works fine with fastest 2-1-1-1 cache timings and 256KB cache. Before it would run reliably only with 128KB cache. With 256KB cache it needed 3-1-1-1 and any performance gains with larger cache were negated. I tested different 15ns chips but usually Quake would crash. Average boost is around 5%.

128kB is 4 chips + tag, 256kB is 8 chips + tag. Yes it does seem like it's doing interleaving with 486, since on a single bank I can't even boot any x-1-1-1 setting, but it's perfectly stable on 2 banks. The problem is, this interleaving only kicks in for burst reads. On a 386 it's always full addressing latency penatly even if the next address fetched is n+1 it seems. So 2 banks give you almost zero benefit (386 doesn't seem to gain much from bigger cache) and you loose 1 cycle on wait state which is considerable slowdown.

So basically, as in my summary, it all depends on the CPU you want to use.

Don't know how I managed to miss this post but, I have the exact same motherboard - the only difference is that my motherboard has the OPTI 495XLC chipset.
However, based on the picture you posted and the BIOS ID string of the BIOS image you uploaded, it's definitely the same motherboard, manufactured by A-trend.
I happen to also have the manual of this motherboard.

While going through my backups, it seems that I also managed to obtain a BIOS dump of this motherboard with an OPTI 495SLC chipset (the BIOS ID string of my BIOS image seems to match the BIOS ID string of yours) - I've uploaded both copies.
If you require a scan of the manual, let me know.

Deunan wrote:

I don't have faster chips but kixs does, and I had the exact same issue - see above: […]
Show full quote

I don't have faster chips but kixs does, and I had the exact same issue - see above:

kixs wrote:

Small update... found some 12ns cache chips and swapped them. Now it works fine with fastest 2-1-1-1 cache timings and 256KB cache. Before it would run reliably only with 128KB cache. With 256KB cache it needed 3-1-1-1 and any performance gains with larger cache were negated. I tested different 15ns chips but usually Quake would crash. Average boost is around 5%.

128kB is 4 chips + tag, 256kB is 8 chips + tag. Yes it does seem like it's doing interleaving with 486, since on a single bank I can't even boot any x-1-1-1 setting, but it's perfectly stable on 2 banks. The problem is, this interleaving only kicks in for burst reads. On a 386 it's always full addressing latency penatly even if the next address fetched is n+1 it seems. So 2 banks give you almost zero benefit (386 doesn't seem to gain much from bigger cache) and you loose 1 cycle on wait state which is considerable slowdown.

So basically, as in my summary, it all depends on the CPU you want to use.

Ahh, I had read that then forgot the content later on. When you say "486", does that include the DLC/SXL/DRx2 chips?
Perhaps this is why so few 386 boards had support for 256K. I have a true 386 board (AMI Mark V Baby Screamer) which insists on 12 ns cache for 40 MHz operation. It also supports 256K but uses fairly uncommon cache. Finding the 12 ns TAG in that configuration wasn't easy.

feipoa wrote:

When you say "486", does that include the DLC/SXL/DRx2 chips?

No, all of these chips are using 386 bus protocol that must be compatible with pure 386 chipsets so there's no burst transfers. Sorry for the confusion but my particular mobo doesn't have a PGA socket (or indeed even a place for one) for a 386 CPU - I have a PQFP AMD 386DX-40 soldered in and a combo socket for 387/486. So I forgot about these upgrade chips (even if I use them in other boards) and I when I wrote 486 I only meant "true 486" chips.

jesolo wrote:

While going through my backups, it seems that I also managed to obtain a BIOS dump of this motherboard with an OPTI 495SLC chipset (the BIOS ID string of my BIOS image seems to match the BIOS ID string of yours) - I've uploaded both copies.

I've downloaded and tested both in PCem and curiously that SLC one doesn't boot, and doesn't beep either. What's even more curious is the internal version string says "OPTI-495SLC-3486 BIOS VER:1.1C", but it's dated 08/08/93 and recognizes more CPUs than my BIOS. Mine is "OPTI-495SLC BIOS VER 2.1" and dated 11/11/92. The one kixs dumped is "OPTI-495SLC-3486 BIOS VER:2.2X1", also from 08/08/93 - and this one (after the small repair I mentioned) does work in PCem and my own mobo.

There are several variants of the OPTi 495 chipset boards, do you still have the one you got this BIOS from? Or photos?
BTW, there are some dumps of the 495SLC BIOS on russian sites but all of them are corrupted beyond repair. Your's however looks good.

In which "true 486" is referring to PGA-168, but excluding the SXL2 in the PGA-168 flavour?

Correct. Frankly I'm not even sure why TI bothered. Probably it was faster/cheaper to package the SXL into PGA-168 and try to convince mobo manufacturers to support it rather than make a feature-complete 486. We know how that turned out.
In general the question as to what is a "true 486" should then be - does it support burst transfers? Other than the improved, pipelined core this was the second fundamental change that made 486 so much faster. But obviously this only results in a speed-up if there is internal cache, otherwise you'd be wasting tons of memory bandwidth always pushing 16 bytes to the CPU for any kind of memory read request.

Deunan wrote:

I've downloaded and tested both in PCem and curiously that SLC one doesn't boot, and doesn't beep either. What's even more curio […]
Show full quote

jesolo wrote:

While going through my backups, it seems that I also managed to obtain a BIOS dump of this motherboard with an OPTI 495SLC chipset (the BIOS ID string of my BIOS image seems to match the BIOS ID string of yours) - I've uploaded both copies.

I've downloaded and tested both in PCem and curiously that SLC one doesn't boot, and doesn't beep either. What's even more curious is the internal version string says "OPTI-495SLC-3486 BIOS VER:1.1C", but it's dated 08/08/93 and recognizes more CPUs than my BIOS. Mine is "OPTI-495SLC BIOS VER 2.1" and dated 11/11/92. The one kixs dumped is "OPTI-495SLC-3486 BIOS VER:2.2X1", also from 08/08/93 - and this one (after the small repair I mentioned) does work in PCem and my own mobo.

There are several variants of the OPTi 495 chipset boards, do you still have the one you got this BIOS from? Or photos?
BTW, there are some dumps of the 495SLC BIOS on russian sites but all of them are corrupted beyond repair. Your's however looks good.

To be honest, I cannot remember where I obtained the OPTI 495SLC BIOS ROM that I included in my previous post (it's definitely not from a motherboard that I own).

However, I do have another 3/486 OPTI 495XLC motherboard (manufactured by Jetway, the J-402B) where the BIOS became corrupted and through my various searches and posts over the years, I suspect that this is where that OPTI 495SLC BIOS came from.
I did find some of the BIOS images on those Russian sites but, I managed to track down two main sites where I found some old BIOS images for OPTI chipsets:
http://chukaev.ru54.com/bios_en.htm
http://sannata.ru/bios/

The OPTI 495XLC BIOS image that I posted was from the motherboard in the picture

Yeah, that's the sites I've tried. Both OPT495SL.ZIP and 3opm008.zip files contain the same broken image. First 8192 bytes are identical to my 11/11/92 BIOS, the rest sort of looks like a part of another 08/08/93 495SLC BIOS but there's no version string and the data doesn't match any other BIOSes for that chipset that I've found.

As for PCem, it could be a bug/limitation of the emulator that it doesn't boot the image. For example I've just noticed it won't boot the 2.2X1 image as well when the CPU is set to 486DLC rather than plain 386. It doesn't have this issue with my 2.1 though so clearly there's some difference in the init code. I'll try to test it on a real mobo when I have some free time again, but it won't be soon.

What happens if you try it with the OPTI 495XLC BIOS image that I posted?

jesolo wrote:

If you require a scan of the manual, let me know.

Actually, I happen to have the same, or very similar mainboard and scanned manual would be great. The publicly available list of jumpers is rather brief, and when I received the board, one jumper fell off... 😀
Thanks 1E6!
Jakub

Here's a scan of the manual of the above-mentioned A-Trend 3/486 motherboard.
It doesn't include the AMIBIOS Setup section, since this doesn't provide much useful information (compared to what you will find in other 486 motherboard manuals) and, it would have increased the size of the file over the limit.

Does VLB work with 386 CPU?

Yes it does.

G'day! I'm super glad I found this thread, since I just picked up one of these boards, and couldn't find any info on it anywhere else!

I haven't managed to check all the jumper settings yet, so I haven't tested it, but I have dumped the BIOS ROM!

http://uplategeek.com/wp-content/uploads/2020 … /opti495xlc.zip

The chipset on my board is the XLC version. I have no idea if that makes a difference, but it was marked AA3535949, and is different from the other ROM images in this thread.

EDIT: I just noticed the XLC ROM, which it is identical to. As AvE says, you don't know till you know and now you know.