VOGONS

3 (+3 more) retro battle stations

Reply 260 of 376, by maxtherabbit

Rank Oldbie
Rank
Oldbie
pshipkov wrote on 2021-06-18, 03:07:

With C2 or D revisions i see around 18fps in Quake 1 at 4x40MHz.
Q1 is hard cpu bound, so given some decent vga, processor speed is the dominant factor.
dont think i ever checked 3x50 on these late isa/pci boards. so maybe the perf difference we are seeing comes from there.

My processor is stable at 160 so I will try 4x40. You were using EDO iirc, I'm using FPM because I have the 'NV' version of the north bridge

Reply 262 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie
maxtherabbit wrote on 2021-06-18, 03:38:

My processor is stable at 160 so I will try 4x40. You were using EDO iirc, I'm using FPM because I have the 'NV' version of the north bridge

I see.

Please post your findings if you get to run some 4x40 tests. Also bios settings and video card.
I was thinking to share some info about bunch of late 4860isa\pci motherboards at some point soon, so extra bits will be welcome.

retro bits and bytes

Reply 263 of 376, by feipoa

Rank l33t++
Rank
l33t++

I spent more time trying to get the UM8886BF PCI IDE driver working in DOS and Windows 95c.

In DOS, v3.1, I used the Install4.exe file. Here you can setup your IDE transfer speed. On my system, only PIO-3 seemed stable from previous tests and I determined that this translates to the 180 ns option in the installer driver. The installer places device=c:\um8673.sys into your config.sys file.

Upon reboot, you are faced with this:

1UM8886BF PCI IDO DOS DRIVER VTEST2COPYRIGHT 1994-1995 UMC. All Rights Reserved.34Press F5 to skip driver installation,5or other keys to continue ...

I could never get passed that screen. So I went on to installing v3.2.

I used install4.exe again and went through the setup process. This time, I get a more useful screen in DOS

1Drive 2 Status:2Using the table settings speed3This is the optimal and reliable speed for this HDD.45Drive 0, 1, and 3 show "not installed"6Drive 2 reports:7Cylinder Number: 156368Head number: 169Sector Number: 6310Buffer Size: 1 KB11Sector/Interrupt: 112Drive Speed: 14  (this is setup during installation)13FIFO Status: Enable14PnP option status: Disable15IDE Controller: UM8886BF1617Drive 2 Model: CF Card18UM8886BF Disk Controller has 25 logical drivers from D: to Z:1920You have too many block devices specified in your CONFIG.SYS file. Remove some disk drivers from your CONFIG.SYS file, and then try again.

Huh? This is what my config.sys file looks like:

1DOS=HIGH,UMB,NOAUTO2LASTDRIVEHIGH=F3Country=004,850,C:\WINDOWS\COMMAND\country.sys4DEVICE=C:\WINDOWS\HIMEM.SYS /TESTMEM:Off /V5REM DEVICE=C:\WINDOWS\EMM386.EXE RAM /MIN=0 I=B000-B7FF /V6DEVICEHIGH /L:2 =C:\LS120\ATAPIMGR.SYS /T:0 /F:IOS.INI 7DEVICEHIGH /L:2 =C:\WINDOWS\COMMAND\DISPLAY.SYS CON=(EGA,,1)8DEVICEHIGH /L:2 =GSCDROM.SYS /D:MSCD000 /V9DEVICEHIGH /L:1 =C:\WINDOWS\IFSHLP.SYS10SHELL=C:\WINDOWS\COMMAND.COM C:\WINDOWS\ /E:1024 /P11REM add the following line for LS120 drive access in DOS. Remark if booting to Windows 95. DOS line doesn't work if two devices on same IDE port12REM DEVICEHIGH /L:2 =C:\LS120\MKELS120.SYS /Q /F:IOS.INI1314device=c:\um8673.sys

I tried to put a remark in front of the LS120\ATAPIMGR.SYS and the GSCDROM.SYS, but it didn't help.

At this point, I remarked out the um8673.sys device in config.sys and proceeded to boot Windows. I am presented with:

1UMC UM8886BF PCI IDE Windows driver V3.12Copyright 1994 UMC. All Rights Reserved.34The Microsoft Windows 32-bit disk driver (WDCTRL) cannot be loaded on this computer because of interrupt conflicts.56To continue starting Windows without using the 32-bit disk driver, press any key.

huh? I'm guessing this driver is only for Win3.x.

I boot into Windows 95 and decide to disable the Standard IDE controller provided by Microsoft, then reboot. Same DOS screen error. The DOS installer places these lines in your C:Windows\System\system.ini file, namely under [386Enh]

132BitDiskAccess=on2device=c:\windows\system\int13.3863device=c:\windows\system\um8673.3864;device=*int13

I then install the UMC Windows 95 driver from within Windows 95. The files provided in the Win95 folder are:
Install.exe
Itewin95.dll
Umc86b.inf
Umc86b.mpd

I reboot. In Device Manager, the Standard Dual IDE controller is absent and in its place I see UMC86B DUAL PCI IDE Controller, however it has a yellow exclamation. Recall I had disabled the standard dual ide controller in a previous step. I tried to re-enable it, but the check boxes to disable it were not present. Shoot. So I uninstall the UMC86B via the install.exe program, reboot, re-enable the standard dual IDE controller, then reinstall the UMC86B driver. Reboot, and system hangs, just as it did during yesterday's Windows driver tests. I'm not sure how you got it working, but I'd like to know!

Reply 265 of 376, by feipoa

Rank l33t++
Rank
l33t++

I checked the CF card throughput on another PCI 486 I have setup, which is based on the SiS 496/497 chipset. However, the IDE controller implemented is a "Symphony PCI IDE controller". That poor thing only gets 1700 KB/s read/write, making the 4200 KB/s on the Biostar seem like a real winner. Unfortunately, there is no option to set the PIO mode in the BIOS, not even an "auto" feature, so maybe it is running on PIO-1.

I then checked the HDD throughput on another 486 I have setup, however it is using a VLB SCSI card and gets 9885 KB/s.

Reply 266 of 376, by maxtherabbit

Rank Oldbie
Rank
Oldbie

what's even more vexing to me is that I was unable to get any variant of the Adaptec 2940 to work in my LS-486E

tried a plain 2940, 2940U, and 2940UW - all cards known working in other machines - and all of them would just hang in a death spiral of rescanning the SCSI bus

Reply 267 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie

Well, while this is certainly annoying at least your board is alive, while one of the Biostar 486 board here seems to went silent.
I turned it on early today to check on Feipoa's questions. The cpu got burning hot in few seconds. Turned-off power immediately.
Tried again - same - cpu got super hot.
Puzzling. I left the board a while ago fully setup and working for 3x60mhz and 4V.
Lowered voltage to 3.45 - suddenly CPUs got back to normal temperature, but no picture anymore.
No idea what was that really. 4v never produced such extreme temperatures. In fact 4v is just fine with air cooling only.

This reminds me of these Biostar boards acting up once in a while.
Until today it was mostly around interpreting bios timings differently, but today was voltages it seems.

Anyone else hitting issues like that and finding a cure ?

retro bits and bytes

Reply 268 of 376, by feipoa

Rank l33t++
Rank
l33t++
maxtherabbit wrote on 2021-06-19, 15:51:

what's even more vexing to me is that I was unable to get any variant of the Adaptec 2940 to work in my LS-486E

tried a plain 2940, 2940U, and 2940UW - all cards known working in other machines - and all of them would just hang in a death spiral of rescanning the SCSI bus

Did you try the 2940U2W? I remember when I was playing with another SiS 496-based PCI board that it seemed to favour the 2940U2W, but I had to reduce the speed in the card's BIOS to 40 MB/s rather than 80 MB/s.

pshipkov wrote on 2021-06-20, 08:27:
Well, while this is certainly annoying at least your board is alive, while one of the Biostar 486 board here seems to went silen […]

Well, while this is certainly annoying at least your board is alive, while one of the Biostar 486 board here seems to went silent.
I turned it on early today to check on Feipoa's questions. The cpu got burning hot in few seconds. Turned-off power immediately.
Tried again - same - cpu got super hot.
Puzzling. I left the board a while ago fully setup and working for 3x60mhz and 4V.
Lowered voltage to 3.45 - suddenly CPUs got back to normal temperature, but no picture anymore.
No idea what was that really. 4v never produced such extreme temperatures. In fact 4v is just fine with air cooling only.

This reminds me of these Biostar boards acting up once in a while.
Until today it was mostly around interpreting bios timings differently, but today was voltages it seems.

Anyone else hitting issues like that and finding a cure ?

I've never run into that problem on any 486 motherboard unless the CPU was installed in the wrong orientation. It is also possible that you had the jumper set on 5 V. The VRM jumpers on this board aren't labelled sequentially which can make for easy mistakes if you aren't extra alert, e.g. JP36, JP39, JP37. What crack pot set that up?

Do you have another one of these MB-8433UUD boards? If so, perhaps you can use that.

Reply 269 of 376, by maxtherabbit

Rank Oldbie
Rank
Oldbie
feipoa wrote on 2021-06-20, 08:38:

Did you try the 2940U2W? I remember when I was playing with another SiS 496-based PCI board that it seemed to favour the 2940U2W, but I had to reduce the speed in the card's BIOS to 40 MB/s rather than 80 MB/s.

I don't have that one. The only LVD SCSI card I have is an LSI logic one with a 64-bit PCI card edge which physically won't fit in the slots on the LS-486E

Yes I know most of the time you can just plug 64-bit PCI cards into 32-bit slots and they work fine, but these 32-bit slots obstruct it physically

Reply 270 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie

i have couple of PCI SCSI adapters somewhere here.
Not sure if 2940 models are among them - if so can give them a spin.

As for the Biostar board not lighting up.
Well, that was me.
I was soo 101% that the clock generator jumpers are configured properly. Kept looking at everything else but them.
Checking on Feipoa's questions now ...

retro bits and bytes

Reply 271 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie

@Feipoa
Here is what i found.

Vanilla Windows 95 OSR2,
2x Biostar MB-843UUD-A rev.2 motherboards
Amd 5x86 @160MHz
1x 64Mb EDO, 50ns
4x 16Mb FPM (64Mb), 60ns
CompactFlash IDE adapter, as shown in the attached image (Jp2 state didn't matter - checked both options).
Sandisk ultra 50Mb/s (silver top), 1 partition FAT16
All BIOS settings on max, except the one shown in the screenshot below.
PIO mode set to Auto.

The DMA option in the standard Windows95 drivers for the UMC IDE controller does not persist between restarts.
It can be set from the driver's property dialog, but upon restart is back to OFF.

The UMC driver from archive UMC_8886BF_IDE-DMA-v3.2 worked just fine as before.
However its performance is disappointing.

Since i have been playing with FPM/EDO memory recently, i used the occasion to see if there will be measurable difference.

So in short - here are the numbers from WinTune2 disk test:

1_______________________________weighted average    uncached speed     cached speed2default Win95 driver, FPM RAM:       14501             16595             124073default Win95 driver, EDO RAM:       13512             17019              92864   UMC custom driver, FPM RAM:        2380              2553              2207

Notice that when using the UMC driver, even the model of the CF card is recognized.
So things are kind of working, just veeeery slow.

Attachments

Last edited by pshipkov on 2021-06-21, 00:21. Edited 2 times in total.

retro bits and bytes

Reply 272 of 376, by feipoa

Rank l33t++
Rank
l33t++

I'm puzzled by your benchmark result. I wonder if there is any substantial difference with OSR2 vs. 2.5 in this regard.

Some questions: What is your PIO mode set to? What CPU are you using? What is the FSB set to? Which CF card are you using? Could you try running ATTO Disk Benchmark and RoadKil DiskSpeed?

Reply 273 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie

edited my previous post with the requested details.

sure, can run atto and roadkill tests later today.
Preferred versions ?

in general - things work well out of the box with the windows driver and the custom umc one didnt crash, just not very fast.
Thats why i dont remember details from previous runs. There was no drama.

retro bits and bytes

Rank Oldbie
Rank
Oldbie

3x50:

Rank Oldbie
Rank
Oldbie

4x40:

Reply 276 of 376, by pshipkov

Rank Oldbie
Rank
Oldbie

great. thank you.

8mb/s disk i/o is normal for this chipset with CF.
the best i have seen is with Asus PVI 9.5mb/s, but most other boards gravitate around 7-8, give or take one.

what was the video card ?
write through is giving you better numbers ?

retro bits and bytes

Reply 278 of 376, by feipoa

Rank l33t++
Rank
l33t++

pshipkov, I'll have to check the versions of ATTO and Roadkil later after the wife and kids are in bed. What result did you get in Speedsys? I don't understand what is going on here. If you set your PIO mode to PIO-3, what result do you get in Speedsys for the Linear read speed and buffered read speed? What about for PIO-4? DOS results should be independent of the Windows OSR version. I wonder if the AUTO setting is doing some magic for you. I am not using AUTO. Which version of the BIOS are you using? What is the size in GB of your CF card?

EDIT: Are you also able to check using a Cyrix/IBM 5x86, preferably at 133 MHz at 66 MHz FSB, if not, then at 3x33, or any other configuration with the PCI BUS at 33 MHz.

Last edited by feipoa on 2021-06-21, 05:02. Edited 1 time in total.

Reply 279 of 376, by maxtherabbit

Rank Oldbie
Rank
Oldbie
feipoa wrote on 2021-06-18, 03:58:
maxtherabbit wrote on 2021-06-18, 01:04:

BTW in Speedsys I get about 8-9MB/sec

'eh? So you get about 4.5 MB/s in Win9x and 8.5 MB/s in DOS ? What's going on here?

I retested in Win98SE. On roadkill I'm seeing a peak of 6.77MB/s at 50MHz FSB

So not as bad as I was remembering, but indeed less than speedsys.

I also found there is in fact a SiS 496 IDE driver for Win3.x. It works on 98, but is actually a little slower than the standard IDE driver