Modern computer complexity has heavy impact on keyboard latency

[Millennium]

The PC, as such, is rapidly becoming a niche product for enthusiasts, and I question how much longer that niche will exist? For how much longer will we be able to buy components, build a machine to our spec and choose our OS? Or will we end up with a choice between offerings from a few a mega-corps (Apple, MS, Google, whatever)?

Hmm I did NOT expect such a deep level of discussions on a Wednesday night. I salute you. This is correct, though hopefully there is enough invest in the Personal C Market that it will persevere.

[Saracen]

It's a shame he can't run the same test using the same methodology with different hardware and the same OS. That would probably give a much better indication of hardware differences.

The OS is the real variable here, even with multiple runs.

Up to a point, we can. If you (as I assume was done) you film keypress to screen timings while the PC is at a pre-OS load point, like in a BIOS screen, and do it repeatedly to get a large sample size, and then aversge it out, you could get a pretty decent metric for hardware latency.And it'd bde OS-independent.

But at least to my way of thinking, pointless too.

After all, a computer, and OS, and software are really all merely parts of a system designed to do a job, be it run your accounts, edit photos, surf the web or shoot/blow up virtual game enemies, or whatever.

Personally, the ONLY use for PC performance metrics is to guide me in assessing how good, and at what cost, components are in doing the jobs I want to do.

For instance, if I'm spec'ing a photo-editing, I don't give a hoot about killing virtual enemies, and vice-versa. If I'm after an all-rounder i have go try to balance different needs.

But I struggle to see how key-to-screen hardware latency tells me anything useful, hence my remark about it being academically interesting but ultimately, unimportant.

[DDY]

Not seeing the point is very different to missing the point.

I'm always open to seeing a point if someone cares to make one.

Has anybody in the history of anything actually found this to be a problem?

Actually, don't answer that, I really don't care.

It comes across to me that you were expecting this work to address a problem that doesn't exist.

The work was not meant to address a problem but rather to disseminate research findings.

That's where I think you've missed the point.

[Iota]

Personally, the ONLY use for PC performance metrics is to guide me in assessing how good, and at what cost, components are in doing the jobs I want to do.

Well that sums it up in a nutshell for me when I'm building a new PC, how much does it cost for what I would like it to do?

Which is where I'm currently sat at, because my PC constantly crashes every single time I want to shoot / blow up virtual game enemies. It's done well, but it's time to build a new one.

[Saracen]

Seems to be sound upgrade-rationale to me, Iota. It's how I do it, too.

[aidanjt]

You mean adding giant piles of code between keyboard switch activation and pixel updates on the display adds a lot of latency? Surely not!

[DanceswithUnix]

Oh, and I still have a fully operational Apple IIe, with a variety of extras and peripherals, bought in the late '70s. I wonder who else does, too?

Apple II joysticks measured how long it took to charge up a capacitor, so latency on those sucked really badly. The keyboard was polled, so read really fast but only if you were looking for it.

So if they had measured analogue joystick response, the picture would have been different. Even the Dragon used a software A/D converter that took quite a few of it's 0.9MHz cycles.

[Saracen]

Funnily enough, Dances, a joystick is one Apple peripheral I never had. Upgraded RGB video, check. Z80 board for CP/M, check. 16K upgrade, check. Even another HUUUUGE 128k upgrade, check. And so on.

But joystick? Don't think so.