Turbo switch/jumper

Well, try some more modern program for comparison benchmarks.
Maybe one of these: Speedsys, Norton SI 7, ctcm7r

Check first if the MHz really matches 40 Mhz. (checking the Turbo contact is a good start, try simply with a jumper)
Next things I configured in the BIOS (default 386 Award Bios):
MEMORY READ WAITSTATES - 2
MEMORY WRITE WAITSTATES - 0
(if it doesnt boot beyond the bios switch to higher values)
ISA CLOCK = CLOCKIN/5 <- default, try /4 for 10 MHz or even /3 for 13.3 MHz
If you increase far beyond the default 8.3 MHz check carefully if there are errors on the ISA Bus for certain cards.
f.e. writing wrong bits to the HDD is no fun in the long term...
I could also set to /2 for a ISA Bus at 20 Mhz but this was unreliable 😉.

Next thing check the L2 Cache exclusion blocks. Usually 640 K to 1 MB is excluded. Thats ok. All memory >1 MB should be cached without uncached areas upto your mainboards cacheable size. My cacheable size is f.e. 16 MB.

If you give some bench results, I could compare with my 386DX-40 System. But please no exotic benchmarks.

Well, try some more modern program for comparison benchmarks.

The problem with these programs, that i didn't use them that time. They give me a benchmark result, which means nothing to me.
But of course I'm gonna try them tomorrow, to compare with your results.

Check first if the MHz really matches 40 Mhz. (checking the Turbo contact is a good start, try simply with a jumper)

It's hard to tell. Of course, the turbo switch is on. Without it, ~60mhz would be an awesome result in landmark2.0 😀

Seems to be, i've forgot to link the motherboard. Here it is. The clock gen can be alter within the range of 50-66-80mhz, so the cpu can run at 25-33-40mhz. The problem is, that the jumpers missing from the board, haven't been soldered. Checking the connection between the pads, the top layers seems to be unconnected. Or, it may be hardwired under the top layers. For example aida and other tools i've used determined the clock to 40mhz.

Next things I configured in the BIOS (default 386 Award Bios):

It has an ami bios. One of the reasons why i've chosen this board, because it has almost the same bios as my ex-386 (this one has the awesome hdd autodetect option). Tomorrow morning i'm going to shot some screenshots of the bios.

ISA CLOCK = CLOCKIN/5 <- default, try /4 for 10 MHz or even /3 for 13.3 MHz

I've tried this too. The JP2 on board, determines the isa clock. The info on the link can be confusing, because based on that, it seems, this has effect on the cpu clock. Well, it has to be changed to the actual speed of the cpu. Changing it, there was only a relative huge performance gain on the isa bus (vga for example), but absolutely none to the cpu.

Next thing check the L2 Cache exclusion blocks.

That's the point, where I'm lost the line. 😐 I'm not sure that I understand this "l2 cache / ram cache mambo-jumbo". I mean, cant see how it affects the performance of the cpu. Maybe the best if i do the test with the program you mentioned and do some photo of the bios.

Thx for the reply

I've done some tests.

Speedsys: picture

Norton SI7 summary:

1                            **********************
2                             *   System Summary   *
3                             **********************
4
5              --------------------- Computer ---------------------
6                Computer Name: IBM AT or compatible                          
7                Built-in BIOS: AMI, Saturday, June  6, 1992
8               Main Processor: 80386DX, 40 MHz
9            Math Co-Processor: Non-Intel 80387
10                Video Adapter: VGA, Secondary: None
11                   Mouse Type: Serial Mouse, Version 9.00
12
13                --------------------- Disks ---------------------
14              Hard Disks: 412M
15            Floppy Disks: 1.44M, 1.2M
16
17               --------------------- Memory ---------------------
18                  DOS Memory:    639K
19             Extended Memory:  7,168K
20             Expanded Memory:      0K
21
22             --------------------- Other Info ---------------------
23                    Bus Type: ISA (PC/AT)
24                Serial Ports: 2
25              Parallel Ports: 0
26               Keyboard Type: 101-Key
27            Operating System: DOS 6.22

Norton SI7 benchmark:

1                                *****************
2                                *   CPU Speed   *
3                                *****************
4
5This       |
6Computer   +****  41.2
7           | 
8INTEL      |
9Pentium-66 +*********************  211.4
10           | 
11INTEL      |
12486DX-33   +*******  71.2
13           | 
14INTEL      |
15386DX-33   +****  35.9
16           | 
17           +----+----+----+----+----+----+----+----+----+----+----+---
18                50   100  150  200  250  300  350  400  450  500  550
19                                      Computing Index
20                                        
21                         Main Processor: 80386DX, 40 MHz

ctcm1.7: picture
ctcm actually got frozen at this point, and beeped non-stop.

And some bios screen. Main, Advanced CMOS Setup, Advanced Chipset Setup.

My System has 64 kB L2 Cache.

My System:
Speedsys
5.98 in Processor benchmark.
Memory values: Read / Write / Move / Average in MB/s
L1 Cache 25.5 / 30 / 37.3 / 31
Memory 15.7 / 30.1 / 19.8 / 21.9

Norton SI7:
42.5

So it seems your system runs well.

So it seems your system runs well.

It's quite interesting, but the facts are facts.

What about the bios options? For me, they seems to be fine. Do you have any idea, which one should i change for a little more performance?

If the BIOS is fine, I would check the ISA cards.
Do you use one of the faster VGA cards like an ET4000AX ?
Do you use a Gravis Ultrasound that does channel mixing in hardware?

Wasn't "bios shadowing" a big BIOS performance tweak back in the day?

Ah yes, could be.
Well I usually enable it for Video BIOS, System BIOS should be enabled by default.
Also if you use QEMM it shadows BIOS usually, but might slow down a bit if you extensivly use DMA.

This applies ofcourse only if the BIOS functions slow down your programs.

Do you use one of the faster VGA cards like an ET4000AX ?

Yes, I've tried. Same cpu performance, same char/sec vga performance as a newer cirrus logic isa vga (the isa bus is the bottleneck).

Do you use a Gravis Ultrasound that does channel mixing in hardware?

Yep. But i didn't noticed a big difference compared to the sbpro2 in cp2, playing 16+ channel modules.

Wasn't "bios shadowing" a big BIOS performance tweak back in the day?

Only the video shadow turned on, but nothing else.

Only the video shadow turned on, but nothing else.

After turning on all shadow option, nothing changed.
Except my One Must Fall 2097 savegame 😉

Here's some info about system and video BIOS shadowing...
http://www.pcguide.com/ref/ram/logic_Shadowing.htm
http://www.techarp.com/showFreeBOG.aspx?lang=0&bogno=187
http://support.microsoft.com/kb/92766

ROM BIOS shadowing is the process of copying the BIOS from ROM into RAM and using either hardware or 386 enhanced mode to remap the RAM into the normal address space of the BIOS. Because reading RAM is much faster than reading ROM, BIOS-intensive operations are substantially faster. For example, MS- DOS uses the BIOS to write to the screen; therefore, with ROM BIOS shadowing, directory listings run more quickly.

Windows NT does not use the BIOS (except during startup); therefore, no performance is gained by shadowing. If ROM BIOS shadowing is not used, more RAM is available. With Windows NT, there is an advantage to disabling the ROM BIOS shadowing option.

Modern OSs don't use the BIOS for much anymore so there's little benefit these days. I have a feeling VESA VBE video modes bypass it too. And I remember Win 3.x bypassing BIOS disk routines with 32-bit file/disk access.

Some BIOSs let you shadow adapter ROMs too.

VESA VBE Video Modes have not much to do with BIOS shadowing, because you access the framebuffer either by the 64 K window at A000:0000 or the VESA LFB. Both is mapped Video RAM. Your only BIOS access is to change modes or pages if you dont use LFB.

As for your 386 system, I still cant see where it should be slower than normal.