6502s are faster than all modern computers!

[Justin]

OK, not really. However, I found this input latency comparison of old vs new quite interesting. What is especially interesting is that the winning system is a 6502 that (if I remember correctly) does not use interrupts for keyboard input. This reminds me of past comments from Garth about 6502 interrupt latency being superior to 68K and newer chips.

http://danluu.com/input-lag/

[whartung]

When I try to imagine the byzantine code path that a character on a keyboard must travel to get to the screen on a Symbolics computer, I think 300ms is down right speedy.

[GARTHWILSON]

Ed brought this up and we discussed it a couple of years ago, but I'm having trouble finding it. He posted the same link.

[hoglet]

GARTHWILSON wrote:

Ed brought this up and we discussed it a couple of years ago, but I'm having trouble finding it. He posted the same link.

Try this thread: OT: on the responsiveness of old computers vs new ones

[BigEd]

Thanks for the trip down memory lane! That was quite the OT thread.

[BillO]

Click bait title. I don't mind, but it almost seems like venerable 6502 is surfing into the future. Very validating.

(Can anyone else see this connection?)

[Chromatix]

Well, there are many definitions of "fast". For computing, two of the most important are throughput and responsiveness.

The 6502 doesn't do all that well at throughput, at least until you compare it to some much older machines (PDP-8/E anyone?). But it *can* do very well at responsiveness, largely because both the software architecture and the hardware are under the designer's control.

I define "responsiveness" as the reciprocal of latency; if it can reliably detect a keypress and update the screen to match within 1ms, you can claim a keyboard-to-screen responsiveness of 1000Hz. (This is much faster than most actual display devices, so it'll be perceived as instantaneous. But some applications, particularly those involving haptic feedback, have been known to *require* that level of responsiveness in order to feel natural.) In my line of active research, computer networks can also be described in similar terms.

[cb3rob]

considering your use of the word 'modern'. what's so 'modern' about deliberately selling crap that breaks all by itself within a year or 2 after leaving the shop and refusing to sell new units of the same type. i thought scams pretty much where the oldest commercial concept besides the prostitution industry. (that means you: intel and arm that pretty much leaves the 6502 and z80 without any competition whatsoever.

also, it's 2019, and lightswitches are still 1 bit computers. nobody ever complains that a lightswitch would not be 'modern' it does the job. just like 6502s do their job.

[BigEd]

Umm, let's be sure to keep the discussion on track, with interesting information. Too many emotive words and sideswipes and we end up with an argument, which is quite a different thing.

[drogon]

cb3rob wrote:

also, it's 2019, and lightswitches are still 1 bit computers. nobody ever complains that a lightswitch would not be 'modern' it does the job. just like 6502s do their job.

Alexa, dim the livingroom lights

Is possible now...

The real thing this thread is about is latency though. Typing here, my keystrokes are being detected by a microcontroller, then serialised over USB into my computer, de-serialised, then through the kernel and into the application (Firefox in my case) which passes them to a large bit of javascript (itself an interpreted language) which then sends them to the display via the font renderer and a network layer (for this is running under X windows after all) then finally the pixels are poked to the display... Phew! I wonder just how many million lines of code all that took...

Sometimes I think we've just gotten a bit too clever for our own good. Or maybe I'm just having a "Nutrimatic" moment...

The colours are nice though.

Cheers,

-Gordon

[sark02]

drogon wrote:

Alexa, dim the livingroom lights

Is possible now...

This right here.

There's a somewhat romantic notion that the 8-bit era was pure, the computers were fast, and the world was shiny and bright. The thing is, they weren't (and the world wasn't). Working on little Atari 800XL games back then, I remember always wanting more speed that wasn't there. That's why I jumped onto the 68000 as soon as those machines appears, and for a while, it satiated me. Then it didn't.

Our ideas about what we can even accomplish with computers is often shackled by the limitations of what's on hand. Sure, any specific computer can solve a set of problems. You could spend a lifetime writing programs for a 1KB ROM + 1KB RAM system...

You may not appreciate what the "bloated" systems of today provide, and you may be appalled by the long, twisty journey a key-press takes before being received by the application you're using, and then by the almost inconceivable chain of processing that results in a character being echoed on the screen... but all this is the result of people standing on the shoulders of those that came before them, and pushing ideas forward.

Here I am typing on this laptop with a color LCD screen... I perceive no delay between key-press and character appearing.

Yesterday I was using Oculus Rift VR glasses, flying an F-14 and landing on an aircraft carrier... looking around at a virtual world, with no perceivable delay as I rotated my head. This isn't 1990s VR with 500ms latency between turning your head and the image changing, and a crude flat-shaded low-polygon, low-resolution world. This is 2019 VR, with < 12.5ms latency and a hundreds of thousands of polygons, texture mapped, virtex and pixel-shaded, multi-pass rendered at high resolution. The latency is below the perceptual threshold. 12.5ms to respond to an IMU motion report, compute a new world transform, render a new world view, post-process to apply the VR image warp, and send to the display. Twice (once for each eye.. with a slightly different viewpoint). This is on a Windows 10 PC. And, to boot, I'm flying with 20 other people across an Internet of connected computers thousands of miles apart, with 60-80ms latency - allowing for busy flight decks, formation flight, and dogfights.

Tell me more about how today's latency is so bad...

[GARTHWILSON]

sark02 wrote:

Here I am typing on this laptop with a color LCD screen... I perceive no delay between key-press and character appearing.

<snip>

Tell me more about how today's latency is so bad...

I often have to look away from the screen when I type on my 3.2GHz, dual-core, 64-bit PC, because the delay makes me make mistakes. The delay problem is much reduced on my laptop, partly because, I suppose, the keyboard doesn't have to go through USB or even PS/2. My typing mistakes on the laptop are due to lack of cupped keys (so my fingers don't stay centered on the keys as well by feel), and the short travel of the keys.

[commodorejohn]

sark02 wrote:

Tell me more about how today's latency is so bad...

Sure: today's latency is so bad. And that fact is not altered by the fact that you can do your fancy VR fun stuff or whatever. It may not bother you personally or impact your own particular workflow, but it's still a fact (as attested by the research conducted by the author of that article.) Other things being nice does not make a crummy thing not crummy.

[sark02]

GARTHWILSON wrote:

I often have to look away from the screen when I type on my 3.2GHz, dual-core, 64-bit PC, because the delay makes me make mistakes.

You might be running crappy applications. Hardly the fault of the OS, but if that's the narrative you're suggesting then I won't debate it.

Quote:

The delay problem is much reduced on my laptop, partly because, I suppose, the keyboard doesn't have to go through USB or even PS/2.

At least on my laptop, the keyboard is attached to the Intel chipset PS/2 interface. Yours might be different, but there's no direct keyboard matrix on Intel chipsets.

[sark02]

commodorejohn wrote:

sark02 wrote:

Tell me more about how today's latency is so bad...

Sure: today's latency is so bad. And that fact is not altered by the fact that you can do your fancy VR fun stuff or whatever. It may not bother you personally or impact your own particular workflow, but it's still a fact (as attested by the research conducted by the author of that article.) Other things being nice does not make a crummy thing not crummy.

Uh, I think the fact that I can do my fancy VR thing disputes the supposed fact quite nicely.

This link: https://pavelfatin.com/typing-with-pleasure/ (linked from the subject webpage) illustrates how editor latency is a function of the editor application more than any intrinsic system problem. Editors that do on-the-fly spell-checking, autocompletion, or code suggestion are, necessarily, burdening every keypress with algorithms, pattern-matching, database lookups, etc. It's no wonder that they can feel sluggish, or behind the curve. But in the end it's an application complaint, not a platform constraint.

As soon as a keyboard matrix is removed from the memory bus and put on the end of a serial link (be it PS/2 or USB 3.0), the absolute keyboard to photon latency is increased, but from a human interface perspective that increase is not significant. What is significant is how application programmers write their code, and as the link here shows, there's a whole range or low and high latency apps.

I think it's absurd to make any real claim that modern computers are slower or higher-latency than 70s and 80s examples. I mean: you can if it makes you feel happy, but I don't think it's useful.

Just because you were young and vital in the 70s doesn't make computers of the 70s better in any practical and useful measure than those of today.