Chess will never be solved, here's why

Bytes per second is a data transfer rate across an interface. Exactly which interface did you have in mind, or doesn't it matter?

All interfaces. All computer components and types of processing. All architectures.
What kind of computer processing or type of hardware or software is not subject to bytes per second and clock speed in cycles per second or number of cores or degree of parallelism among its cores?
Or doesn't it matter?
The word 'nodes' just magically makes all the physical limitations vanish?
I don't think so.
tygxc got shot down on that. And on taking the square root. And on trying to argue 'chess is a draw' because some currently top computers of similiar strength keep drawing each other.

I dont think anybody here is arguing one metric makes another one dissappear.... rather which metrics are practically useable for a solving process and which best measure the speed, efficiency, progress made

Subject to bytes per second where? Between main storage and HDD, between two high speed registers, across the internet? Where?

See my previous post.
If you don't see 'bytes per second' - what do you see?
You could tell us all about it Martin.

Somebody was.
For years.
Trying to pretend that 'nodes per second' knocks out 'bytes per second'.
At the time I was mentioning 'ops per second'.
Now 'nodes per second' is being revisited in the forum.
But another point being missed is that the various jargon terms could be discussed more.
But again whoever might say 'No! Go look it up! And don't use AI !'
Missing the point that however much whoever might play around with that word 'nodes' you're still not going to get around those basic limitations. I've mentioned four so far.
'Hey - we can make an aircraft out of lead! Entirely! Yes! It'll fly!'
No it won't ...

I see, "All interfaces. All computer components and types of processing. All architectures.".

It's a bit like you saying your car does 150, and when I ask what units, you say, "All units, the steering wheel, the spark plugs, all makes of car".

The data rates across all the different interfaces in a computer can differ by millions. Have you ever seen a claimed computer speed specified in bytes/sec.? Do you think the time it takes Stockfish on your desktop to solve a mate depends on the bytes/sec spec from your camera?

(Well, you asked me to tell you all about it.)

Only, I would suggest, to someone who also doesn't understand the difference.

Of course they vary - you're attacking your own ideas Martin.
I answered your question but you went off on a tangent.
But that's okay.
I think you've already conceded that the speed the computer can process information is limited by its central processors - by its clock speed - by its number of cores and the degree of parellelism it achieves among those cores.
You've conceded because you haven't challenged any of that.
Telling me that the speeds vary between the computer components is like saying water is wet.
That's already known.
What you could probably talk about is the different units of measurement (bytes - bytes per second - ops per second - IOPs per second - FLOPs per second ( easy to know something but not everything about those units by their definitions).
And of course the almighty nodes per second.
At least you haven't done a tygxc meaning you haven't tried to claim that 'nodes per second' knocks out the other units.
----------------------------
In theory - internet research could be done by any member regarding the term 'nodes per second' and its evolution and why its preferred in various situations.
Or people very knowledgable could choose to talk about that.
(and of course a particular someone (not I) who talks in very nasal tones will see it as an opportunity (won't name him this time in view of Wind's recent warning) to express his conceit.
Fortunately he has no authority here.
There are some who even take an extreme position that you shouldn't talk about something unless you have a degree in the subject.
Fortunately - such persons also have no power here.

It's worth underlying the last point. The reason top computers of similar strength keep drawing with each other is that they are of similar strength. The same top computers get healthy plus scores - nearer to 100% than 50% if you go far enough - against computers of a different era.

Of course, the fundamental match-up of interest to solving chess is white versus black. White gets a large percentage of the wins in top computer chess. In the original Alphazero versus Stockfish match of a few years back, 89% of the wins were with white with white scoring 75% and black scoring 56% for Alphazero, and white scoring 44% and black scoring 25% for black. Thus including both, it was 59.5% for white.

In CEGT, it's from 55%-56% for white across all the games, with what looks like a significant INCREASE at slower time controls.

If the latter is confusing to anyone, you generally get fewer wins as the time control gets longer, but that does not imply the difference between white and black's score goes down. The first is about the VARIANCE in results and the latter about the MEAN AVERAGE of the results.

It would certainly be reckless to extrapolate the upper trend in white's score with longer time controls to conclude that with a billion years or so per move white would win all the time, but it is not an impossible scenario.

As another easily misundestood nuance, it is often observed that in most games white's advantage dissipates as the game progresses (definitely NOT all of it on average, as white wins more games), BUT this is not inconsistent with the possibility that with greater strength, white retains the advantage longer into the game. We can't be sure where this leads to when strength is maximal.

I agree its worth revisiting that last point - but adding that tygxc got shot down when he tried to make conclusions from the fact of top computers drawing each other.
So did his supporter who tried to use circular reasoning.
(And who worries a lot about who's supporting who and 'cabals') (no names)
And tygxc was also 'dealt with' on 'taking the square root' (a ridiculous idea in the context) of John Tromp's number.
But tygxc also got shot down on 'nodes per second' too.
I don't recall the details of exactly how.
I took a position at the time that 'nodes per seconds' cannot knock out 'ops per second'.
After 19,000 posts that position continues to stand.
Maybe some will try to claim that tygxc wasn't claiming they could.
Often - people in a chatroom will try to decide the issue by popularity.
By who likes who. In the clubs - even moreso.

@Elroch
I edited and improved my reply to your post a bit.
So the quote you made underneath isn't quite the same.
No big deal though.

Yes, but I already said in the above that the high draw percentage only reveals low variance of results between players of very high and similar strength. It provides no basis for reasoning whether there exists any enormously stronger player, any more than a world championship match between Kasparov and Karpov (similar low win rate) refuted the possibility that Stockfish 15 would score over 95% against either of them.

And we seem to be in agreement. Then and now.
Or rather - total agreement on the issue of top computers drawing each other not proving 'that chess is a draw'.
We seem to also be in basic agreement it seems that 'taking the square root' to cut down John Tromp's number isn't valid too.
And I wasn't the only one back then contesting tygxc's use of 'nodes per second' to try to add to his claim of 'chess could be solved in five years if the money's there'

19K posts and still waiting for a breakthrough?

Try wearing jeans when you defend your Blitz title, and you get expelled from the tournament??

What's not to like with the Royal Game?  

You say that all the time, but then they also simultaneously say warmest year ever over and over again, so which is it? If it's really also causing deep freezes than those should cancel out the hotter summers, but it doesn't. Global warming is good if arctic air entirely could simply cease to exist. Nothing to "move anywhere" then.

Both I think. The two are not incompatible (as evidenced by the fact that that's what happens - they do measure these things).

I wonder if it's possible for chess computers to be programmed with factual endgame knowledge added on to alpha-beta pruning. By that I mean adding parameters for it to immediately "understand"/evaluate certain endgame piece combinations/pawn positioning as draws or wins. Alpha-Beta Pruning can't even find this easy drawing technique:

Forcing a drawn endgame out of a heavily disadvantaged position due to making the remaining pawns have the wrong color promotion square. 3 total moves. If it could recognize such bishop + pawns endgames as drawn, it would stumble upon that known evaluation 3 moves deep, and not give the initial position -3.5 as stockfish does.

That's what they do.

Usually by tweaking the alpha-beta static evaluations, but SF for example (at least pre NNUE, it may not still be true) allows the whole of specific endgames to be coded outside of alpha beta search.