VOGONS


386DX40 build

Topic actions

Reply 60 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

Interesting, seems that your board's chipset must have faster memory access than mine. What 3DBENCH score do you get?

If you are squeamish, don't prod the beach rubble.

Reply 61 of 434, by elianda

User metadata
Rank l33t
Rank
l33t

It is already in the Benchlist and scores 15.3 fps.

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 62 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

I've been fiddling with the ISA bus jumper settings and have received some interesting results.

Until now I have been using the async 7.2MHz jumper settings, which are the board's default settings. I tried the /10 divider setting which should give a speed of 4MHz on my system, and my 3Dbench score actually increased.

I tried the /8 divider setting (5MHz ISA bus) and got another small increase. At the /6 setting (6.6MHz) the system would not boot, it beeped 8 times which is supposedly video DRAM read/write failure..... very odd. At /4 setting (10MHz ISA bus) the system would not POST, no beeps at all.

So I can only surmise that running the ISA bus asynchronously incurs some sort of performance hit, compared to running it synchronized to the system bus at a divider. At least on this board/chipset anyway. I would have thought that running the ISA bus at slower clockrates would have slowed it down, not sped it up.
Anyway, here are some results

Changing the ISA bus speed did not affect Norton SI scores more than .1 or .2 of a point, if at all, so I did not bother to put them here. They are basically the same as the previous scores for a similar setup at 7.2MHz async ISA bus.
I didn't run speedsys for every combination either because I was pressed for time and they don't really seem to change that much anyway. Some small boosts to hard disk performance, but CPU and memory performance only has very tiny increases, probably not statistically significant.

Cache enabled, Fast BIOS, Turbo Enabled, /10 divider (4MHz ISA bus)

3Dbench: 15.8
10div%20full.jpg

Cache enabled, Fast BIOS, Turbo Disabled, /10 divider (4MHz ISA bus)
3dBench: 3.7

Cache enabled, Fast BIOS, Turbo Enabled, /8 divider (5MHz ISA bus)

3dBench: 16.1
8div%20full.jpg

Cache enabled, Fast BIOS, Turbo Disabled, /8 divider (5MHz ISA bus)

3dBench: 4.1

Cache disabled, Fast BIOS, Turbo Enabled, /8 divider (5MHz ISA bus)

3dBench: 10.4

Cache disabled, Fast BIOS, Turbo Disabled, /8 divider (5MHz ISA bus)

3dBench: 2.6

Cache disabled, Slow BIOS, Turbo Enabled, /8 divider (5MHz ISA bus)

3dBench: 7.0

Cache disabled, Slow BIOS, Turbo Disabled, /8 divider (5MHz ISA bus)

3dBench: 1.8

Reply 63 of 434, by elianda

User metadata
Rank l33t
Rank
l33t
DonutKing wrote:

I tried the /8 divider setting (5MHz ISA bus) and got another small increase. At the /6 setting (6.6MHz) the system would not boot, it beeped 8 times which is supposedly video DRAM read/write failure..... very odd. At /4 setting (10MHz ISA bus) the system would not POST, no beeps at all.

Well, the default ISA BUS speed at this kind of system should be ~8 MHz. My system features the following CPUCLK dividers:
/5 = 8 MHz
/4 = 10 MHz
/3 = 13.33 MHz
/2 = 20 MHz

I usually run it at 10 MHz, that helps especially video performance.
With a different IDE controller it did run also at 13.33 MHz, but with a few side effects. I think it was a Goldstar Prime 2.
My ET4000 showed at 13.33 MHz here and then wrong pixels and reads from the slave HDD weren't reliable anymore. On the other hand, 13.3 MHz would already be a huge bus overclock.
I haven't checked if I got a card that does 20 MHz reliable (250% overclock), I don't want to burn the old stuff.

So 10 MHz is fair and most of the people ran their systems like this.

8 MHz should run on your machine. If it doesn't something is strange and you should check it.

edit: According to your base clk crystal of 80 MHz, I guess /10 is 8 MHz and /8 is 10 MHz setting on your board. (and /6 13.33 MHz which results in a freeze due to too much overclocking)

Reply 64 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

Yeah the manual isn't very clear. I was under the impression that those dividers were applying to the CPU speed rather than the crystal oscillator.

The manual refers to all settings as CLK2, eg CLK2/10, CLK2/8, etc...

Just noticed that it says underneath 'CLK2 equals the OSC speed of mother board in the 386 system'. So I guess that means that it is going from the crystal oscillator, so the /8 setting is indeed 10MHz.

Reply 65 of 434, by retro games 100

User metadata
Rank l33t
Rank
l33t

Excellent info, people! @DK, I see that you have tested your mobo @ ~7, 8 and 10 MHz bus speed. Perhaps understandably, there are no documented settings to reduce this bus speed, and also understandably, any increases result in failure. I believe in theory, if you were to replace your CPU with a slower one, it would simply get automatically overclocked, rather than run at its native speed. So, I guess one way to slightly reduce the overall system speed would be to change the oscillator from an 80 MHz unit, to say a 66 MHz unit. The purpose behind this is to try and get some 3DBench mark scores between ~10 and ~15. That might provide some useful flexibility, with some speed sensitive games. The question is, can these oscillators be found any more?

I've got a second idea. You could try undocumented bus speed jumper settings, to see if it's possible to slow down the mobo a bit. The following 3 configurations do not appear on the stason webpage for a 386 chip, for this mobo. Perhaps give 'em a go?!

JP3 JP4 JP6
C O 2-3
O C 2-3
C C 2-3

Go on, I dare ya! 😁

Reply 66 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

I believe in theory, if you were to replace your CPU with a slower one, it would simply get automatically overclocked, rather than run at its native speed. So, I guess one way to slightly reduce the overall system speed would be to change the oscillator from an 80 MHz unit, to say a 66 MHz unit. The purpose behind this is to try and get some 3DBench mark scores between ~10 and ~15. That might provide some useful flexibility, with some speed sensitive games. The question is, can these oscillators be found any more?

I believe that is correct. If you want to install a CPU slower than 40MHz you will need to replace the oscillator.
I've had a look at my local electronics shops but I can't find any crystals in that type of package.

As for jumper settings- my manual says that is JP6 is set to 2-3 the system will run the ISA bus async at 7.2MHz and ignore the other two jumpers- in the table it has an X in the JP3/JP4 settings.

If you are squeamish, don't prod the beach rubble.

Reply 67 of 434, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
retro games 100 wrote:
I've got a second idea. You could try undocumented bus speed jumper settings, to see if it's possible to slow down the mobo a b […]
Show full quote

I've got a second idea. You could try undocumented bus speed jumper settings, to see if it's possible to slow down the mobo a bit. The following 3 configurations do not appear on the stason webpage for a 386 chip, for this mobo. Perhaps give 'em a go?!

JP3 JP4 JP6
C O 2-3
O C 2-3
C C 2-3

Go on, I dare ya! 😁

I will accept your challenge! ;D
Trying out undocumented jumper settings often lead to veeery interesting results!
And at any rate, I don't know the first thing about changing occilators, I don't even own a spare 😜.
I always use my fastest cpu with the most scratches and keep one hand on the power switch at all times! 😜

Reply 68 of 434, by retro games 100

User metadata
Rank l33t
Rank
l33t

OK, great! 😀 Looking at this mobo, the oscillator is socketed, so I think it can be changed fairly easily. I wonder if it's possible to find and then remove a suitable osci from an old defective piece of hardware, to use for these kind of experiments. Also, has the seller shipped your item yet? I haven't heard from him since I bought mine, and I'm still waiting to understand what the shipping charge is.

Reply 69 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

I have just done some quick tests and I can confirm the manual is correct. When JP6 is set to 2-3 the settings for JP3 and JP4 are ignored. I tried the 3 combinations above and they returned identical benchmark scores each time.

As for the crystal- yes it is socketed although I'm not sure where you would get one that fits.

Reply 70 of 434, by retro games 100

User metadata
Rank l33t
Rank
l33t

Thanks very much for those extra tests - really appreciate that. What do you guys think about this: I managed to get 3DBench to produce every result number between 1 and 16, using a 486. I got the different results by changing its bus speed, cache settings, BIOS settings, and turbo status. To me, it seems more flexible than a/this 386 mobo.

Now, please understand that I'm not knocking this mobo, because I've bought one, and I can't wait to own/try it! 😀 But it's just a thought - that a 486 can act like a more flexible 386, compared to a/this 386 mobo. Perhaps I am missing the point. Perhaps it's more important to own a 386, because it's cool to have one, and they are a superb piece of retro history. Also, my slow down tests are just synthetic. Maybe games won't behave 100% perfectly with those slow down "tweaks", and that it's better to run ancient speed sensitive games on a 386.

My 486 slow down results can be seen in the big table here.

Reply 71 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

Depends what you want to do.
If you are happy to own one system and slow it down for all your games then a 486 or even a Pentium system is probably the go, parts are easier to come by - especially Pentium, I find them at the dump all the time.
Mau1wurf seems to have had success with slowing down the Socket 7 Pentiums for older games.

I just wanted an authentic-to-the-era 386 system for my menagerie 😀

Reply 72 of 434, by elianda

User metadata
Rank l33t
Rank
l33t

I don't see the point here. With the 386 you already have the Turbo / Cache / ISA CLK / Waitstates options and the speed range you can work in is already very small, since the system already caps at ~16 fps in 3dbench. So the range is negligible compared to a Pentium f.e.

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 73 of 434, by Mau1wurf1977

User metadata
Rank l33t++
Rank
l33t++
DonutKing wrote:

Mau1wurf seems to have had success with slowing down the Socket 7 Pentiums for older games.

While that is true, I would much rather have your 386 machine 😀

I do get RG100s point (486 being more flexible) but there is flexibility and then there is authenticity. A 386 is cool and as authentic as it gets. My slowdown projects are more about the journey than the destination...

Your 386 machine is definitely a keeper. I wouldn't bother soldering oscillators or anything like that. It would be a shame. It's perfect for the games from that era with plenty of options to slow it down for speed sensitive 286 or XT era games.

A pentium doesn't allow granular speed settings at all. The steps are harsh and huge. Apart from playing with the cashes there isn't much you can do.

With a 486 you do get the Turbo button thrown into the mix, which together with the caches gives you a lot more options.

However there is also the issue of having a video and memory subsystem that's much faster than a 386. So while you can emulate the CPU correctly, you would also need to insert an ISA card into your 486 or Pentium to slow down the graphisc subsystem. Some games don't scroll correctly otherwise or run too fast when there isn't much going on (on the screen).

So yea ket me know if you ever sell your machine. I'll grab it right away

Last edited by Mau1wurf1977 on 2011-01-03, 13:33. Edited 1 time in total.

Reply 74 of 434, by retro games 100

User metadata
Rank l33t
Rank
l33t

Aren't there some speed sensitive games, around the 386 / 1990s era? Some of these games run on a "maxed out" 386 DX might run a bit too fast. But if you drop the speed slightly, to act more like a low to mid range 386 DX, then these games might run at a more favourable speed rate. That's why I am interested in tweaking a 386's speed, around the 3DBench 10-16 score range. Some of DonutKing's results show 3DBench to be either ~10, or ~15/16. There's nothing in the middle, between 10 and 15. I just thought it would be useful to get some of these "missing" scores.

Reply 75 of 434, by Mau1wurf1977

User metadata
Rank l33t++
Rank
l33t++
retro games 100 wrote:

*RG100 braindump* 😉

To a certain degree I would agree with you.

Disabling Cache turns his machine into a 386DX-25 (10 fps) and some of the other settings give you a 386SX-20 and 386SX-16.

The obvious "missing setting" would be a 386DX-33...

So all up he can turn his machine into:

386DX-40
386DX-25
386SX-20
386SX-16

A crappy video card (Oak or something like that) might work here?

Reply 76 of 434, by retro games 100

User metadata
Rank l33t
Rank
l33t

Yes, that's a good point - changing the ISA video card might help. I would really like to run some tests, like that. When I ordered the 386 mobo, I also bought some ISA video cards from a seller in the USA. (I don't own a single ISA video card, but that's just changed!) Very luckily, there was this one seller who has got a whole pile of them, so I saved on the shipping bill, by getting them all from the same seller. I got a mixture - Tseng, WD, ATI, Headland, etc.

Reply 77 of 434, by Mau1wurf1977

User metadata
Rank l33t++
Rank
l33t++

Back in the days the Tseng ET4000 was the fastest card. It was mentioned in the "Wing Commander" book. It was a book about the Game Wing Commander. Had game tips, explained the memory requirements and listed ideal hardware configuration. I believe a 386DX33 with a Tseng ET4000 was the recommended system.

OAK had a bad reputation I believe. My first PC (A High Screen 386DX-33) had issues with the video cardo one day. So we took it back for warranty and they swapped the video card. And I now had an OAK and all my benchmarks scored lower 😢

I got an ISA Tsend ET4000 later and was very impressed with the performance. My next machine was VLB and I also got an ET 4000, just the VLB version. So I was Tseng loyal by that time 🤣

Cirrus Logic was good AFAIK. Not as fast as the ET4000 but good. I didn't have much experience with other cards though.

Reply 78 of 434, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

Nah I'm not going to be selling this.... at least not in the near future 😀

My goals are different to RG100's. I like having the actual hardware, for collectors value, to compare to modern systems, etc. And I get some sort of perverse pleasure from fiddling with and configuring DOS 😜
I'm not really concerned with having a multitude of granular speed options. It's played every game I've tried with it pretty well so far. If I come across one it doesn't I'm sure I'll be able to get it running 'close enough'.

If you are squeamish, don't prod the beach rubble.