PCI vga right? FIC VIP IO have medium PCI performance. Not the best, due to the chipset.
What you get on this motherboard, is a feature rich board.
If you can get one, then try with a VLB VGA card instead.

Thanks, i will try with a VLB one...i have many..

Check your ram and cache settings - they're most likely below what's possible.
Also change all your srams to 15ns to harness the possibility for tight cache timings.

And just pointing out the obvious that you’ve probably already thought of - is the turbo jumper closed?

It still looks ok to me. I found this from another Vogons user, that have included some benchmarks, using a 5x86-133.
Focus on the memory speed. Looks like you have it set up correctly, and only need fine tuning.

algernon wrote:

Thanks, i will try with a VLB one...i have many..

When dealing with VLB, my goto cards are something like this below here. They are not the fastest, yet they are above the middle, performance wise. Yet it gives a wonderfull image and nearly no spaghetti lines....

download/file.php?id=38677&mode=view

For years I have always been warned about the VIA PCI boards. Although I have never owned one, they have a reputation for being less than fast.

For comparison, my 486DX2-66 VLB based system gives the following scores:
a) 3dBench 1.0c - 49.5 fps
b) Quake demo - 7.5 fps
c) PC player - 10.5fps
d) Chris's 3dbench - 48.2 / 28.9 fps
e) Landmark/Supersoft - 316mhz AT, 487mhz 80287
f) Speedsys 25.03 CPU bench, 37.63meg/sec memory speed

Specs:
* Intel 486DX2-66 S Processor
* Asus VL/I-486SVGO VLB motherboard - 15ns cache
* 32meg 72pin Memory, FPM and 60ns.
* S3 805 VLB card
* Winbond VLB I/O card
* NE2000 clone with XT-IDE controller.
* CT2940 - SB16 card.

In comparison to my computer, it looks like the FIC board is really struggling with I/O performance. Pure CPU performance looks fine though, it just can't get that performance down onto the road.

I'll power up my 5x86 133mhz later to compare PCI performance on a differnet chipset to the OPs computer.

My dx2-80 is a tad faster. However that is 40 and not 33 mhz fsb and 80 not 66 in core speed.

https://to9xct.blogspot.com/2018/09/dextor-ta … -build.html?m=1

My 5x86-133 at 133mhz (4 x 33mhz)

a) 3dBench 1.0c - 72.9 fps
b) Quake demo - 12.7 fps
c) PC player - 17.1fps
d) Chris's 3dbench - 86.6 / 51.1 fps
e) Landmark/Supersoft - 634mhz AT, 976mhz 80287
f) Speedsys 50.04 CPU bench, 37.82meg/sec memory speed

My 5x86-133 at 120mhz (3 x 40mhz)

a) 3dBench 1.0c - 75.2 fps
b) Quake demo - 12.4 fps
c) PC player - 17.4 fps
d) Chris's 3dbench - 84.4 / 50.6 fps
e) Landmark/Supersoft - 571mhz AT, 879mhz 80287
f) Speedsys 45.04 CPU bench, 45.41 meg/sec memory speed

At 40mhz x 3, the SiS 496 chipset performs somewhat better, but the CPU performs slightly slower.

Compared to OP's system, the SiS chipset completely kills it for PCI performance.

System Specs:
* AMD 5x86-P75 (133mhz)
* Lucky Star LS486-E REV. C2 PCI motherboard with SiS 496 chipset and 256kb 15ms Cache
* 64meg EDO RAM
* S3 968 graphics card with 2meg RAM
* Onboard IDE controller
* Realtek 8029 with XT-IDE ROM
* Yamaha YMF719 sound card

OP, you might have much better success with a VLB card, if the FIC board presents a raw interface from the 486 CPU to the VLB slots.

Thanks to all, i changed the video card, now an ARK1000 vesa local bus 1024Kb , the L2 cache it's now 15ns for all the memory chips (tag15 ns+ram15 ns) and the performances are slighlty better.
--------------------------------
cachechk: same as before..no increase.
L1 cache 120 MB/s
L2 cache 47 MB/s
Main 20,5 mb/s
--------------------------------
Superscape Alternate Realities: 55.5 fps (was 47.6)
--------------------------------
Quake: 9.4 fps (was 9.2)
--------------------------------
PC Player Bench 13.2 (was 12.9)

In the bios the cache timing control setted as "TURBO" (was FAST)

i've noticed better performance setting in the bios "external cache scheme" as "WT" instead "WB".

Maybe i've reached the maximum performance for this configuration? Not very happy ... 🙁

This board scales back memory timings with different bus speeds. I intend to write an utility to override the memory timings set by BIOS in an effort to reach the maximum performance possible but that's isn't gonna happen in another month due to various life things in the way right now.

Tiido wrote:

This board scales back memory timings with different bus speeds..

higher the fbs speed=lower timings?

Yeah, BIOS compensates timings with FSB change. I have looked up registers in the chipset and there's no such automatic scaling, so eventually I'll make an utility to change timings to whatever you want and reach the theoretical maximums.
You can see how memory performance doesn't change much with increase of bus speed even though there should be a very significant change.

Tiido, thanks so much for this clarification !!! now i understand why my system performs the same way at different bus speeds..

Sorry for the necropost.
I'm messing around with this mb.
Apparently it doesn't like to work with 1024kb of cache, it makes up to the post screen but then it hangs, I was able to make it work with 512kb (128K x 8 ) with 128k as tag, chinese IS61C1024-10N.

I'm testing the stuff with a virge dx pci video card, an intel dx2 66, 16mb of memory, with latest bios version 4.27GN2A (02/14/96).

In the bios chipset features setup there are some interesting options, those are the options which I get the best result according to Cachechk:

Those are the weird results:

There is this weird warning about slow memory.

Strange enough with write back cache I don't get the warning but the results about the ram are noticeably worse but I get rid of the slow warning message:

Another weird thing if I go with cache timing to turbo, the results keep being the same, another thing to notice, I am unable to use the burst write option, otherwise the system reboot automatically as soon as dos is loaded

Sorry for necroposting but:
I found 16mhz fsb settings
JK1:2-3 JK2:1-2 JK3:1-2 JK4:2-3
im posting here if someone want to run 486dx-sx @16mhz
Not stable but working

You must use 3276gn1 (half rom).
Then AMD 5x86 is recognized properly, WB works. Full performance in benchmarks.

Just to add some remarks to this old and still prevailing issue:

1. Using the 3276gn1 BIOS kills PCI support and with this ther whole "VIP" feature, which is the unique feature of mainboards like this.

With the original 427gn1 BIOS
2. The main reason for decreased performance when using write-back CPUS like AM5x86 or Intel´s DX4-100 SK096 (and/or running higher frequencies 133, 150, 166) is that BIOS doesn´t know these CPUS and - according to my tests - not just increases timings, but turns off Dirty-TAG. The result is a sharp decline in L2 performance. Changing the timings for L2 in BIOS has no effect in this situation.

3. Best over-all performance seems to be AM5x86 @ 100MHz. Speedsys benchmarks are 25/40/87 in this case. As soon as you raise FSB or multiplyer to 33x4, memory and L2 performance will drop.
Note that RAM performance isn´t and doestn´t get much better than some average DX2 mainboard.
Later VIA based VIP platforms had the successor chip VT82C496G and the performance increased significantly.