PCem

leilei wrote: ↑Wed 02 Dec, 2020 3:16 pm
That is not a bug. That's the back porch/border color area (which is missing emulation on EGA/VGA cards), and if you ask me, that border could be made even larger considering what monitors those cards were for.....

Ok, then is it possible to remove this invisible area for CGA video cards? This invisible area gets in the way. Since, for example, in 86Box or DOSBox-X, this invisible area is no in CGA mode. And rightly so, as it allows you to fully immerse yourself in the gameplay.

hey guys,
been having some fun with the new build like always thanks!
I've seen a bug or somethings been pledging this game in all my builds since like pcem 11 or 12 I can't remember. Centipede the game runs fine in 3dfx normal mode but when you run adventure mode it would just crash after a minute or so with no error msg under win 95, 98, and me, You're just back at the desktop with a straight face. With all kinds of builds and I've done a lot trust me Anyways I felt like doing a win2000 pro build for pcem 17 idk wild hair. Plus I didn't want to try to rebuild one of my old installs right now. So wouldn't you know I'd have to try my old problem game Centipede adventure mode. It crashed after a minute or so like always but 2000 gave me a error msg! =p So I thought I'd share it here .
The instruction at "0x2000e3b0" referenced memory at "0x00bbf1d8". The memory could not be "read" idk what it means exactly but I do know its crashed about the same way in every pcem v- I've ever built up
I have tried playing with the virtual memory plus other memory things to no avail

A fabulous release indeed with lots of pleasant surprises. A number of games have been fixed, sound stuttering which affected certain DOS games running under Windows 9x or in pure DOS mode is gone and there also is a slight speed improvement.

Nevertheless, there some relatively minor issues related to v17.

Gigabyte GA-686BX
When using my old BIOS file dated October 13 1998 PCem always crashes during one of the last sections of the Windows 95 or 98SE installation procedure, in particularly at the point when the installer is about to install the Windows help file (this comes after the P'N'P devices configuration section is complete).
Out of curiosity, I tried to use the Windows image I had from FIC VA-503+ installation and PCem crashed while loading Windows (normal mode), however, I was able to boot normally into safe mode. I fixed the issue I had by going to Gigabyte's website that had (and still has) 3 bios files available for this particular board (the GA-686BX). The 2.9 version from February 1999 also caused the crash and inability to boot into Windows, however, both 2001 versions work fine.
Another issue that I experienced with Gigabyte board was the system freeze when using ViRGE/DX drivers from August 1997 (the same drivers which are also bundled with DX6 due to being the most stable ones). Whenever I attempted to run any DOS application under Windows 95, the system would hang. Obvious examples: Doom II and Return to Zork. The only way to fix the issue was to switch the driver version. I had no such issues with FIC VA-503+ on v17.

Voodoo 3/Banshee
Vast majority of games I tried (including some obscure old shareware DOS games) worked, however, I would like to point two examples out that didn't:
- Gex 3D: Enter the Gecko (crashes PCem as soon as it's run - works fine with Voodoo 1 and 2, though)
- Resident Evil 1 (hangs as soon as the first FMV is about to load. RE2 works fine).

That's pretty much all for now. Definitely worth getting Rocket Lake, although I'll be interested to know how far is it possible to go with v17 in terms of CPU speed on Ryzen 5000 series too.

After testing v17, I discovered that SMI/SMM instructions were added, and it appears that suspend/sleep mode works on the Intel VS440FX motherboard under Windows 95 OSR2, (and wakes up upon pressing a key or moving the mouse) however, upon waking up from sleep mode the emulated machine hangs, and makes a beeping sound every few seconds. On the ASUS P/I P55T2P4, the sleep mode does not appear to work, with 3 short beeps before it enters sleep mode, but after a few seconds, without pressing a key or moving the mouse, it wakes up from sleep mode without hanging. If you shutdown Windows 95 after this, before the screen "It's now safe to turn off your computer", one beep is heard before this screen appears (which according to the motherboard manual for the ATX version of this motherboard, the XP55T2P4, which shares the same BIOS ROMs as the P55T2P4, are supposed to be heard upon initiation of sleep mode or Windows 95 shutdown, after which the system PSU turns off).

Also, when shutting down Windows 95 on the Intel VS440FX, it goes to the "It's now safe to turn off your computer" screen, (and power management is enabled in the BIOS) instead of shutting down and closing the emulator session window, the PIIX3 used on these motherboards supports APM, which should cause this to happen.

Are the differences between AT and ATX PSUs emulated now in PCem?

What causes this, because the sleep mode appears to work on the Intel VS440FX apart from it hanging upon the system being waken and beeping afterwards?

Soft power is only currently implemented on machines implementing ACPI. To date that's the GA686BX (using PIIX4) and VA503+ (using the VIA equivalent).

Thank you for your efforts and dedication, Sarah. I have a lot of fun with the newest version.

• One thing I noticed is that as soon as I start MS-DOS on WIndows 98 SE, the emulator freezes.

Gigabyte GA-686BX
Intel Pentium 233
128 MB
S3 Virge/DX
Sound Blaster PCI 128

• Also, a post on reddit that could be helpful to improve the emulation:

I wonder if I can finally get LEGO Chess to work at a playable speed. In PCem16 I always slow down to 50% on the cutscenes when using hardware rendering

Edit: Wow the cutscenes play at 100% now!

But a new issue popped up for me that wasn't in PCem16: Rendering issues --> https://imgur.com/a/3FD4YUO

• And lastly, when I toggle fullscreen in a lucas arts game (Ctrl+ALT+Page down) and go back to window mode again, I can't activate the menu (like sam and max) (F1) anymore.

I did not find it mentioned anywhere, so let me just report that v17 runs FreeDOS (www.freedos.org) with both "minimum" and extended "jemm" memory manager. Running games works without problems (it did not test extensively, but encountered no problems).

In addition, booting from CD-ROM and installing FreeDOS from CD-ROM image (iso) works now. You seem to have to use the "legacy" version, but it runs. I tried it on the [Socket 7] Shuttle HOT-557 with both Intel Pentium and AMD K6 CPUs.

V16 could boot FreeDOS from floppy image, but the Installer did not work, and the memory managers were broken. V17 is a significant improvement.

This opens up significant possibilities. FreeDOS is not only free, but it can run pratically anything MS-DOS or DR-DOS can run. It also has good CD-ROM drivers, network support, energy saving (its "shutdown" command works now on PCem, by the way), etc.

Maybe we could include a bootable harddisk image with FreeDOS installed with PCem? This way, novices would have a much easier time of setting PCem up for the first time. Now all that is missing is a free PC BIOS (Pentium class, ideally), and PCem could run out of the box, just like DosBox.

Oh, and of course, congratulations for V17 to Sarah and everybody!

Congratulations on releasing another update to PCEm. Looking back at PCEM releases, the last 8 years have brought massive improvements that are highly appreciated. Although many functions are implemented right, I will focus on two that are consistently (for many releases) unsolved.

1. Typematic rate settings or BIOS support for control of keyboard port seems to be completely missing. Although it may seem minor, the inability to reduce the delay between keypresses in DOS (which relies directly on BIOS) and thus increase the repeat rate I am pretty sure affects more than just every text-mode application that is running. In comparison with other less accurate emulators, this still seems like the most important feature miss that I wholeheartedly recommend to be studied.

2. Erroneous modifier key operation (could be related to buffer). In PCEM, if you press then depress ALT it seems to register as if the ALT key always bounced first. This can be easily reproduced by comparing behaviour of ALT+a specific keypress in MSDOS Edit or any other Windows App (the same happens in Windows 3.x or 9x). In MSDOS Edit if you press ALT+F the menu opens but as soon as you depress any key it closes instead of remaining open. If you open Windows Explorer and press ALT+F the File menu should remain open after you press it but, instead, it quickly appears and then disappears. To achieve a normal behaviour you have to keep pressing ALT after you depressed the other key.

Overall, on the first impression, the new emulator version being released seems to be smoother and a step further in the right direction. Unfortunately emulation of powerful systems (a pentium II with a 3dfx Voodoo 3 card), at 100% speed, with maximum consistency in maintaining the performance, cannot be done with the current CPU performance limits even in modern systems and the problem of PC emulation is hard to be split more effectively on multiple cores. I know that game console emulator developers had some success in coordinating cores split on CPU, GPU, Sound. Although it is difficult to ensure a more effective task split in emulation, I am confident that more powerful emulation capabilities using general purpose means will be available so that at least part of the I/O can be directed to another core and, thus, split the emulating task on 4 cores. It seems that the only reasonable choice for emulation, without dedicated or custom hardware, may be the current envisioned path. It is not easy.

Getting back, a major appreciation goes to Sarah and all developers that were involved in this project.

j1forPandE wrote: ↑Sun 06 Dec, 2020 3:40 pmMaybe we could include a bootable harddisk image with FreeDOS installed with PCem? This way, novices would have a much easier time of setting PCem up for the first time. Now all that is missing is a free PC BIOS (Pentium class, ideally), and PCem could run out of the box, just like DosBox.

Don't know if PCem emulates any chipsets that'd satisfy SeaBIOS. The information on SeaBIOS seems rather vague and feature-list focused. Ditto for any of the Cirruses involved with Bochs/Qemu.

You can already kind-of Free it up with Xi8088 that'd need a much older freedos though.

Seems the text display of the ATI EGA Wonder 800+ (ATI-18800) is off by a few pixels:



My emulated machine:
[286] AMI 286 clone
CPU: 286/25
FPU: 287
MEM: 1664KB
Video: ATI EGA Wonder 800+ (ATI-18800) (Speed: Default)

It was working correctly in V16

te_lanus wrote: ↑Mon 07 Dec, 2020 6:02 am Seems the text display of the ATI EGA Wonder 800+ (ATI-18800) is off by a few pixels:



My emulated machine:
[286] AMI 286 clone
CPU: 286/25
FPU: 287
MEM: 1664KB
Video: ATI EGA Wonder 800+ (ATI-18800) (Speed: Default)

It was working correctly in V16

The cause of this problem is that ATI EGA Wonder 800+ card's text mode display timing is different from other cards. This problem would have occurred after VGA byte panning emulation was implemented. Attached patch should fix this problem.

Don't know if PCem emulates any chipsets that'd satisfy SeaBIOS. The information on SeaBIOS seems rather vague and feature-list focused. Ditto for any of the Cirruses involved with Bochs/Qemu.

Interesting, I didn't know about SeaBIOS. Its here, in any case: https://www.seabios.org/SeaBIOS

Apparently it is used in qemu and kvm. Seems worth a try.

Regarding FreeDOS, are you sure it doesnt run on PC/XT? According to its wiki, it should run (in real mode) "on any PC". The original PC/TX is mentioned specifically. It seems they didnt try, though

So far v17 of PCem is a complete success. I have some more testing to do but so far performance of the Pentium 2 and the Voodoo 3 is very promising.

For some reason the PII 350 reads as a PII 333 on the boot screen. I do not know if this is easy to fix. Performance of the Pentium 2 350 is pretty good. Star Trek Armada works flawlessly at 233 Mhz and holds at 73% at 350 Mhz.

The Voodoo 3 is performing ok. There are occasional glitches in some games, but this is nowhere as serious as in NHL 2000. That game's in-game text corruption is consistent across any Win98 emulated machines.

Besides the already mentioned bugs in keyboard BIOS implementations, I noticed a new one.

Tested on a Shuttle HOT 557 with a Phoenix S3 Trio 64, Pentium 75 CPU, standard IDE drive emulation, it seems that some disk emulation routines generate errors when a lot of files are transferred.

The procedure is as follows. Run Windows 3.11 (WFW), try to transfer or delete as many files to/from a diskette. With or without 32bit disk mode 32bit file mode, the transfer fails with a disk not ready, disk full error, depending on application (DOS or Windows). The same problem does not occur in DOS mode. The problem may be linked with protected mode operation in PCEM 17.

Note: no smartdrive is active. 32 bit disk or file support in Windows 3.1 has no influence if enabled or disabled (and it shouldn't, since floppy support means CPU interrupts and BIOS calls). The above problem can occur also with IDE disk transfers but it appears in less predictable circumstances (for instance, on a long app setup run, it fails when searching a certain file to decompress, always on the same file, after a couple of them were decompressed successfully).

dodleh wrote: ↑Tue 08 Dec, 2020 11:55 pm Tested on a Shuttle HOT 557 with a Phoenix S3 Trio 64, Pentium 75 CPU, standard IDE drive emulation, it seems that some disk emulation routines generate errors when a lot of files are transferred.

Transferred from and to what?

SarahWalker wrote: ↑Wed 09 Dec, 2020 5:04 pm

dodleh wrote: ↑Tue 08 Dec, 2020 11:55 pm Tested on a Shuttle HOT 557 with a Phoenix S3 Trio 64, Pentium 75 CPU, standard IDE drive emulation, it seems that some disk emulation routines generate errors when a lot of files are transferred.

Transferred from and to what?

The transfer problem can be easily reproduced between hdd and floppy. However, it can also occur on long file operation routines with hundreds of files and directory to copy from one directory to another on the same hdd partition.

The floppy to hdd transfer bug (which is also reversible) is well presented also here: https://pcem-emulator.co.uk/phpBB3/view ... 6c9a73105f

Noticed TVGA8900's cursor blinking being overly fast

seems the Diamond Stealth 32 (Tseng ET4000/w32p) isn't displaying correctly in textmode

Tested with [386SX] Amstrad MegaPC & [286] GW-286CT GEAR

te_lanus wrote: ↑Fri 11 Dec, 2020 10:36 am Screenshot from 2020-12-11 12-29-36.png

seems the Diamond Stealth 32 (Tseng ET4000/w32p) isn't displaying correctly in textmode

Tested with [386SX] Amstrad MegaPC & [286] GW-286CT GEAR

My Diamond Stealth 32(Tseng ET4000/w32p) ROM doesn't have this problem. However, in order to use your ROM, it need to do a patch similar to the one I posted earlier.

Hi,
here some news about working games.

Config:
GA-686BX, Intel Pentium II 233 with Creative 3D Blaster Banshee

Working:
-Frogger
-Frogger 2 Swampys Revenge

Broken/Crash
- Centipede (Hasbro, 1999) --> Crash by passing the bridge to level 2 and back to windows desktop

best regards

As of v17 PCEM now has some weird conflicts with MSI afterburner that weren't present in previous versions like screenshots sometimes not working or locking up if you try to resize the window during the boot up with Afterburner running

It's very strange

Marc Behrens wrote: ↑Fri 25 Dec, 2020 9:25 pm Hi,
here some news about working games.

Config:
GA-686BX, Intel Pentium II 233 with Creative 3D Blaster Banshee

Working:
-Frogger
-Frogger 2 Swampys Revenge

Broken/Crash
- Centipede (Hasbro, 1999) --> Crash by passing the bridge to level 2 and back to windows desktop

best regards

i tested the game on other machines, previous version (PCEM 16), 86box, OS Win 98 and 95, result? same exact problem, happens even before crossing the bridge, some problem in the game I suppose.

aitus95 wrote: ↑Sun 27 Dec, 2020 5:44 pm

Marc Behrens wrote: ↑Fri 25 Dec, 2020 9:25 pm Hi,
here some news about working games.

Config:
GA-686BX, Intel Pentium II 233 with Creative 3D Blaster Banshee

Working:
-Frogger
-Frogger 2 Swampys Revenge

Broken/Crash
- Centipede (Hasbro, 1999) --> Crash by passing the bridge to level 2 and back to windows desktop

best regards

i tested the game on other machines, previous version (PCEM 16), 86box, OS Win 98 and 95, result? same exact problem, happens even before crossing the bridge, some problem in the game I suppose.

If you run it in a compatibility modes on a Windows XP VM and it does the same thing I'd say would be the final nail in the coffin the version of the game you're using is just broken

CryptidWorks wrote: ↑Sun 27 Dec, 2020 10:34 pm

aitus95 wrote: ↑Sun 27 Dec, 2020 5:44 pm

Marc Behrens wrote: ↑Fri 25 Dec, 2020 9:25 pm Hi,
here some news about working games.

Config:
GA-686BX, Intel Pentium II 233 with Creative 3D Blaster Banshee

Working:
-Frogger
-Frogger 2 Swampys Revenge

Broken/Crash
- Centipede (Hasbro, 1999) --> Crash by passing the bridge to level 2 and back to windows desktop

best regards

i tested the game on other machines, previous version (PCEM 16), 86box, OS Win 98 and 95, result? same exact problem, happens even before crossing the bridge, some problem in the game I suppose.

If you run it in a compatibility modes on a Windows XP VM and it does the same thing I'd say would be the final nail in the coffin the version of the game you're using is just broken

I could run this game in software mode on my 32bit vista rig with no problems years back ago

CryptidWorks wrote: ↑Sun 27 Dec, 2020 10:34 pm

aitus95 wrote: ↑Sun 27 Dec, 2020 5:44 pm

Marc Behrens wrote: ↑Fri 25 Dec, 2020 9:25 pm Hi,
here some news about working games.

Config:
GA-686BX, Intel Pentium II 233 with Creative 3D Blaster Banshee

Working:
-Frogger
-Frogger 2 Swampys Revenge

Broken/Crash
- Centipede (Hasbro, 1999) --> Crash by passing the bridge to level 2 and back to windows desktop

best regards

i tested the game on other machines, previous version (PCEM 16), 86box, OS Win 98 and 95, result? same exact problem, happens even before crossing the bridge, some problem in the game I suppose.

If you run it in a compatibility modes on a Windows XP VM and it does the same thing I'd say would be the final nail in the coffin the version of the game you're using is just broken

it seems strange, but on virtualbox with windows xp it doesn't die

An interesting way to reproduce the Centipede access violation consistently is to use software 320x240 and head straight for the water...

if it helps, this is what i get on windows 2000:

Re: SeaBIOS: Back when I checked, it did support i440BX, so now that there's the GA686BX, maybe it could be added. Also given that it's open source, it should be easier to get good support than for closed-source legacy bioses. Finally, don't forget there's also an VGAbios. (don't know what does it emulate)

Re: FreeDOS: Obviously it requires the "legacy" one, since PCem emulates legacy hardware.

Posted this in my thread as well but there seems to be a bug in the Socket 7 CPUs where they take a lot more host CPU than they should and lag when decoding Quicktime video of any kind.

Pentium II/233 works flawlessly in Quicktime heavy games

Pentium MMX 233 has noticeable slowdown and I have to drop it down to a Pentium MMX 200 for those titles


Especially odd because in literally everything else it's the REVERSE and it's the Pentium II that experiences slowdown

Borland Turbo Pascal 6 and 7 (for DOS) causes a full lock up a few seconds after loading in Windows 3.11 when the dynamic recompiler is enabled.

On a 486 machine with it disabled it runs absolutely fine, with it enabled it locks up. On Pentium machines it seems you can't disable it, and it locks up exactly the same way.