Emulators of PC/XT: accurate speed?

[carlos12]

Which emulator do you think gives the closest speed to the real hardware's on PC/XT machines? PCem, 86Box or Varcem?

When I try some processor intensive things, such as CGA scrolling, PCem's speed is, I'm afraid, too good to be true. I'm trying the same programs with the same machines emulated both on 86Box and PCem. For example, I use a custom AMI XT clone at 10mhz, Compaq Portable or an Amstrad PC1512, among a few others. If I start the scroll program at once on both emulators, with the same machine, the one running on PCem always finishes first for a few seconds. Also the difference of speed is visually noticeable.

I just cannot check it with real hardware as I no longer have it. I'd love to read your opinions about this. Thank you very much!

 

[ropersonline]

To latch on to this: Is there, by the way, any emulator that can optionally throttle disk access to closely match the original speed of a 5150's real floppy drives?

 

[Chuck(G)]

I don't see why not--floppies are simple devices.

 

[VileR]

86box uses a more accurate 808x CPU core (reenigne's XTCE) than PCem. It also does a surprisingly good job of running almost the entire 8088 MPH demo, so the proof is in the pudding.

For most practical things however, they should be pretty close to each other, and close enough to PC/XT hardware. I had to test that on a few occasions, with things like 8086 sorting algorithms as well as CGA video timing, and the performance difference of PCem/86box vs. an actual 5160 was only barely distinguishable from the 'noise' caused by interrupts etc. With interrupts disabled, there were apparent differences, but tiny ones indeed.

That's not to say that even 86box is 100% accurate (it isn't), but the difference isn't enough to matter in most cases.

I'm pretty sure that both of them (and MAME as well) emulate disk access speeds fairly accurately, but that's something I haven't actually tested.