Anonymous Coward wrote:

if you get a chance can you test your M326 motherboard with either 8MB or 32MB of RAM. I am curious to know if you board supports page-interleave mode. You need to fill all 8 slots for it to work (if supported)

Apologies for being back late.

Here are the results. The first image is the result of using 8 x 4M = 32M 70ns SIMMs, the second image is the result after removing 4 SIMMs (=16M, 70ns). I also tried with another set of 16M 60ns SIMMs, but the result was not changed. It seems that page-interleave is not supported or at least not activated on this mobo.

Strange, I'm not able to achieve the same high score as last time. I only get 9.36 but it was 9.40 last time.

Sorry I'm a bit late to this party..

This is a 386 Baby Screamer Series 42 motherboard and 386DX/33. 32 megs of of 70ns 40 meg SIMMs. Its running off a compact flash card, leading to odd hard drive numbers. The memory bandwidth, esp cached, and CPU numbers seem really low, and I'm not sure why. I suspect something isn't running at full speed but can't prove anything.

On first look, I'd say your Turbo is OFF.

You may be right. The machine starts with the turbo light on, but never boots; after a reset, the light goes off and it boots. Maybe it can't run stably at full speed. However all the CPU speed programs say its running at 33 MHz, so I'm not sure thats it. I may have to get an O-scope one of these days and see if the clock or some other signal is dirty.

Also the presence of a maths copro (or NPU as referred to by speedsys) will affect your CPU score. I doubled my 386DX40's CPU score just by removing the 387.

Ailicec wrote:

However all the CPU speed programs say its running at 33 MHz, so I'm not sure thats it.

The frequency doesn't have to change necessarily when switching Turbo. There are as well other implementation with waitstates/internal caches a.s.o.

I think that most 386 boards reduce the frequency when the turbo is off. I've seen many going from 40 to 8, while another one I use right now goes to 16. However, I believe what elianda says is more common in 486 boards. Anyhow, those old dos benchmarks that measure the actual cpu speed, like the Landmark speed test, should tell you what's going on.

The other benchmarks all indicated 33 MHz. The motherboard manual says non-turbo is 8 MHz. The odd turbo light behavior still bugs me, though.

Seems like the cache implementation is crummy or not working right.. it seems to be 33 MHz, but with very little memory bandwidth.

For fun, I'm posting this from the 386, painful as it may be. No more luck, I tried different combos of SIMMs and some jumpers, but could only make it slower. Checked the BS16 pin to make sure it was running as 32 bit, and it was.

Even without cache, the memory results are complete shite. You have some kind of turbo problem. Did you try to software turbo controls? I think it's CTRL + and CTRL -

Yes, I've never got any of the keyboard controls to work. There's supposed to be ones to control turbo, cache, and get in the BIOS, none of them work. Maybe because I had to replace the keyboard BIOS chip? Sad if that killed my performance.. I do have an EEPROM burner, if I ever got the data I could make another one.

Do you still have the original chip? It's certainly possible that you did something by swapping it. I think some KB controllers have special programmable ROM.

I seem to recall somebody else on here with an AMI baby screamer reporting that the system would always boot up in slow mode and would have to use the keyboard shortcuts to bring the system to full speed. It wouldn't surprise me if your system had some extra code in the kb controller to add this functionality. Maybe the kb controller you swapped in is missing this extra functionality.

In theory is should be possible to reprogram a keyboard controller. Though it seems some have eeproms and others don't, so you would have to make sure you get the right model. I have a 386SX board made by AMI with the keyboard turbo controls. Unfortunately I'm in another country at the moment and I don't have the tool to read an i8042. I think VOGONS user "feipoa" has the same motherboard as you, and *may* have a tool that can read and reprogram. I think he's a little busy these days though.

that's a phantastic thread! But might I ask what's so great abot SiS 'Rabbit' chipset? Looks super-cool, very high memory scores, but somewhat low CPU ratings as it seems?
Can it hold against the later compact Cyrix DLC-compatibles...
What are its possible strong features. Rarity?

After reviewing this thread I still can not see what are the best-performer chipsets...

Looks super-cool, very high memory scores, but somewhat low CPU ratings as it seems?

I'm not sure what you mean... the CPU ratings are on par with (and in some cases, a bit better than) every other 386/40.

Edit: Ah, wait, unless maybe you're talking about that quirk in Speedsys related to the coprocessor scores...

Old Thrashbarg wrote:

Looks super-cool, very high memory scores, but somewhat low CPU ratings as it seems?

I'm not sure what you mean... the CPU ratings are on par with (and in some cases, a bit better than) every other 386/40.

Edit: Ah, wait, unless maybe you're talking about that quirk in Speedsys related to the coprocessor scores...

Yes, this bit of uncertainty 😀 There are lots of scores, of CPU rating of around 10 or 9. And SpeedSys says that NPU is also present in most of these cases.
Indeed I took notice of the 'paradox' found by DonutKing (removed FPU/NPU increases CPU rating).
I just could not imagine so many people had systems without FPU, to have scores around 10.

So, 'Rabbit' SiS simply has an "honest" rating of almost about 7, and many others around 10 are inflated? 😀
It can be, testing programs are like that. Bit a solid performance competition/comparison of SiS with a pair CyrixDLC+Cyrix FPU could be useful.

What I'm thinking about, the later boards can 'enable internal cache' on Cyrix DLC and it could give them the edge in performance.

There are lots of scores, of CPU rating of around 10 or 9. And SpeedSys says that NPU is also present in most these cases.

In those results, they're also using a 486DLC. That's not always apparent in the screenshots, since Speedsys doesn't seem to properly recognize the DLC chips. And it makes sense that 486DLC+NPU would be faster than 386DX+NPU, because the DLC is a faster chip.

A 486DLC with an NPU scores ~10 (as compared to 6-ish for a 386DX+NPU), but if you pull the copro it'll score ~20 (versus 9-ish for a 386DX without NPU). I suppose you could say that the non-NPU scores are inflated, but I don't think anybody really knows what metric they use to come up with the numbers, so there could be some sort of reasonable logic behind it, I dunno.

I just could not imagine so many people had systems without FPU, to have scores around 10.

Really the only reason to have an NPU in a 386 is just for shits and giggles, to fill the socket. There's not much software that actually takes advantage of it... I think 2 or 3 games, and then CAD and business applications. That's about all. So it's not at all uncommon to have a 386 system without a coprocessor.

What I'm thinking about, the later boards can 'enable internal cache' on Cyrix DLC and it could give them the edge in performance.

The SiS boards are later boards. I don't know if all boards with that chipset support the DLC's internal cache, but I know mine does. And even if a board doesn't have native support for the cache, you can still enable it with a software utility.

Old Thrashbarg wrote:

...And even if a board doesn't have native support for the cache, you can still enable it with a software utility.

Yes, thanks, I'm just meditating over these pages of wisdom 😀
http://ps-2.kev009.com/ohlandl/8570_Tim_OConnor/cyrx.htm
It's surely worth a try...

Concerning FPU (being next to useless), I've seen a youtube video by Elianda...
https://www.youtube.com/watch?v=E2zsV-kXMIk

This should use the FPU! unfortunately it's not too fast to shine... 😀 p.s. a beefed-up 486 +Voodoo2 does it much better.

Mithloraite wrote:

This should use the FPU! unfortunately it's not too fast to shine... 😀 p.s. a beefed-up 486 +Voodoo2 does it much better.

The Quake on 386 video is more or less a Tech-Demo since Quake was targeted for a Pentium CPU. It was never intended to run on a 386.
The video was primarily related to another video on youtube for running Quake on 386 that was a fake.

I come to the conclusion that finding a good 386 motherboard is really hard. Many boards don't do DMA properly, some older boards don't like DLCs, and most newer boards don't have interleaved DRAM. My own readings and observations tell me that SiS Rabbit, C&T, and VLSI made the best 386 chipsets (don't have exact model numbers). But as 386 stuff is so hard to get now, I haven't been able to test any of this stuff myself. As far as I know Feipoa has boards with all 3 of these chipsets, but I remember he didn't have good luck getting his VLSI based AMI Baby Screamer going. I seem to recall his C&T board was picky about RAM upgrades, and he seemed to like the SiS Rabbit somewhat.

elianda wrote:

Mithloraite wrote:

This should use the FPU! unfortunately it's not too fast to shine... 😀 p.s. a beefed-up 486 +Voodoo2 does it much better.

The Quake on 386 video is more or less a Tech-Demo since Quake was targeted for a Pentium CPU. It was never intended to run on a 386.
The video was primarily related to another video on youtube for running Quake on 386 that was a fake.

in no way I'm blaming this experiment... It's actually adorable! 😀 Of course it's way too much for the poor old 386. It's more of a joking instance that here we ~do~ need an FPU... as if it would help 😦

What should be pretty real is the idea that before the Big Quake no one actually required the FPU. So having none in one's 386 is so much OK...
AFAIR only the users of specialized programs not games had the real need for FPU.

That was a great try. I wonder if using a VLB video on "3/486" board might help to process this monster-heavy video (for a 386 CPU.)
Could a EISA system do something about it too...

afair yours was a Cyrix FastMath FPU for this, but there's a special "pair" FPU for CyrixDLC, a 487DLC... Could that help a bit?