VMware works great, but VPC/VBox are faster. However I've never had a problem getting any USB peripheral working in it.

The only thing it doesn't do is DirectX which is only available when XP is installed. DirectX8 would be great if they added a layer for it... but alas it's not to be. VMware is more aimed at testing OS's and not playing games 😒

RoyBatty wrote:

VMware works great, but VPC/VBox are faster. However I've never had a problem getting any USB peripheral working in it.

Does USB joystick work with VPC?

Not that I know of, I was speaking of vmware. =]

I've checked out virtualbox 3.12. I still have to edit the machine config with an editor instead of their slow gui to turn on acceleration, but 2d now works as well vs 3.11 which I could only get 3d enabled.

Personally I use a non-recommended at all solution - DOSBox, if you're brave 😀

Only advantages of that:
- OPL3!
- Input responsiveness
- s3 driver
- Video capture

Disadvantages:
- completely unsupported by the team as it is not intended to run Win9x
- DOES NOT RUN DOS GAMES (sorry, no extreme paintbrawl and other win32/dos hybrid games)
- you'll pull your own hair out trying and erroring to get it to work.
- some games will have key mapping issues (Killing Time)
- no disk or cd mounting

It's not the best, but it WorksForMe(TM)

Well I eventually managed to get past the USB joystick problem --or rather, the general USB problem. I just disabled USB 2.0 from VirtualBox VM setting, and then I'm able to attach my USB devices to the guest O/S.

BUT.

I'm very disappointed with its performance. Yes, I know, being Intel Atom and such. But still, what would one say when Star Rangers is significantly slower in VirtualBox than in DOSBOX? And I thought virtualization is faster than emulation.

Still no hope of playing Novalogic's F-22 Lightning II though. The game crashed on the Windows 98 guest O/S, with "the application just performed illegal operations" error message.

Kreshna Aryaguna Nurzaman wrote:

And I thought virtualization is faster than emulation.

It is - virtualization works best if your CPU supports virtualization extensions (i.e. AMD-V), and if your CPU is the correct architecture (x86).

VirtualBox is unbearably slow without virtualizaiton extensoins. By that I mean, Win95 taking 12 minutes to boot on a 2ghz system.

Perhaps you can try Bochs (link) http://bochs.sourceforge.net/

I also have a new netbook (Asus Eee PC) which I bought 2 days ago with Win XP, and I've also d/l Bochs a week or so before getting the netbook, and I have yet to try Bochs with Win 98SE guest on my netbook.

And you're absolutely sure that bochs is fast?

leileilol wrote:

Kreshna Aryaguna Nurzaman wrote:

And I thought virtualization is faster than emulation.

It is - virtualization works best if your CPU supports virtualization extensions (i.e. AMD-V), and if your CPU is the correct architecture (x86).

VirtualBox is unbearably slow without virtualizaiton extensoins. By that I mean, Win95 taking 12 minutes to boot on a 2ghz system.

IIRC, currently VT doesn't work with 16bit code. So any 16bit code run under current VM's is using dynarec same as DOSBox. AFAIK, DOSBox is fastest at dynarec for 16bit code.

I remember 9x booting pretty fast under VPC so not sure why VBox would be so slow guess it's simply the matter of no optimization being done for their dynarec OS's for older than NT.

Obviously Windows 95 is mostly 32bit with 16bit boot code/WOW but with VT turned off it's still using dynarec.

Actually vmware was faster during my tries, and it depends on the techniques
of the codes (like qemu sucks hell at selfmodifying code). Don't remember if
virtualbox inherited this type of slowness from qemu.

>DOSBox is fastest at dynarec for 16bit code.

DOSBox is faster in overall emulation, but qemu is quite a bit faster in dynrec. The problem is it doesn't work in a lot of instances, but for the cases where it misses its failing points, it will fly past DOSBox.

DosFreak wrote:

leileilol wrote:

Kreshna Aryaguna Nurzaman wrote:

And I thought virtualization is faster than emulation.

It is - virtualization works best if your CPU supports virtualization extensions (i.e. AMD-V), and if your CPU is the correct architecture (x86).

VirtualBox is unbearably slow without virtualizaiton extensoins. By that I mean, Win95 taking 12 minutes to boot on a 2ghz system.

...............

I remember 9x booting pretty fast under VPC so not sure why VBox would be so slow

Oh, the Windows 98 guest boots pretty fast on my Intel Atom host. Sadly, the fastness ends there; once the Win98 guest finishes booting, everything become slow. Installing the hi-res video driver helps a bit, but not much.

I'm not sure if Intel Atom has virtualization extensions, but the virtualization extension option in my VM is both checked and greyed out, which strongly indicates that it is enabled and cannot be disabled. 😖

DOSBox is faster in overall emulation, but qemu is quite a bit faster in dynrec.

This is plain wrong. Try some real games before posting such information,
like duke3d crawls under qemu, or quake gets nice 100 "claimed" fps whereas
you clearly see that the screen is only drawn like 2x per second.

You need to reread, and screen updates aren't dynamic recompilation.

How do you determine that the recompilation is faster?

Running something CPU based, CPU benchmarks for example. As I said, qemu fails a lot, and buggy, it's only in instances when that isn't happening does it show.

Like i said, when skipping all the nasty things (screen refresh is just one thing),
you can't judge the two even with a seemingly raw cpu benchmark.

You can when you're only looking to determine which is faster in dynamic recompilation, which was what I asserted the first time. DOSBox is faster in overall emulation, but when comparing dynamic recompilation where QEMU works, it's faster. If it weren't so buggy and incomplete, porting the QEMU CPU emulation to DOSBox would be very attractive since it's faster.