It is better left disabled, no software (there may be some edge cases) rely on video interrupt and there's a lot of hardware that doesn't have it at all or enable it by default. You need that IRQ for stuff like intelligent MPU401 interface or something else that actually is going to have the IRQ used.

There was a topic here related to this 3 years ago. For practical purposes it is not necessary and you are not really missing anything, the PC platform hardly used this feature due to inconsistencies around the graphics hardware.

Interesting, so it's an old "vsync interrupt" that is presumably one of the things that lets 8088MPH do some of the CGA tricks, but irrelevant later

8088MPH doesn't use EGA or VGA graphics card. It uses a real CGA in ways that people believed it was not possible on a IBM PC.

Doesn't 32-bit windows require it?

AFAIK, no since I have two cards that doesn't use an IRQ and works on Windows without any problem.

At one point I couldn't get a Riva128 PCI to work in Windows 95 OSR2 properly until I turned on VGA Interrupt. The drivers just constantly insisted the device wasn't functioning properly.

This was on an MVP3 platform though, which is known to have some PCI *ahem* quirks though. I'm not sure, although I may find out soon, if installing the VIA PCI Latency patch would have also addressed my problem.

Accellerated cards are a different piece of cake. We might gather some games that use the vertical sync interrupts for color cycling effects (I guess my ten fingers are enough to count them), but otherwise, that interrupt didn't get much use on PCs. On Accellerators, especially 3D accellerators, the card is a coprocessor that gets batches of drawing commands issued from the processor, and often uses an interrupt to tell the processor that it is (nearly) done executing the commands, and the processor can now issue further commands to keep the card busy.

So my rule of thumb is: For framebuffer cards and DOS gaming, no IRQ is needed. For modern cards in 32-bit windows, you can try without, but shouldn't be surprised if the driver refuses to work.

Hi all,

I just read mkarcher‘s last post.
Accelerator is a noun that can be widely interpreted.
Do you think the „coprocessor“ comparison also applies on high End 2D GDI accelator (ISA) cards as the GD5434 with heavy drawing offloading as well? Meaning should you also active irq here to get best GDI acceleration?

The vsync interrupt (originally on IRQ2, and then later demoted to chained IRQ9 iirc) quickly became obsolete after it first appeared in EGA, and most VGA adapters do not implement it.

Many games use video effects that require synchronization to the vertical blank and/or the vertical sync (both are sometimes needed due to a way IBM botched up their VGA adapter design). Practically all DOS EGA/VGA/SVGA era games implement such synchronization via polling, and not via the interrupt. The only game that I know that used the interrupt to sync is the EGA implementation of Gauntlet on DOS.

In the PCI 3D accelerator era, when each vendor would have their proprietary 3D drivers the situation might be different, and it is not really described or observable (or particularly meaningful, unless you are a 3D driver developer) whether those drivers internally use a PCI hardware interrupt or not - the driver manages all of that, and programming is done via one of the 3D APIs.

Thanks for the answer. Strange that even a gd5434 from 95/96 had this implemented