Games behaving differently on the Next

[Pegaz]

@1024Mak

Now we have reached the stage, where hdmi is totally irrelevant, as expected...
The point is, that hdmi was important to many potential bakers and therefore promised in a very explicit way, as a very important feature.
Without it, the kickstarter would be much less successful, Henrique knew that back in 2016 and he asked Ben Versteeg to help him with that, not by coincidence.
You also know all this, but for some reason you relativize things again, with this rose tintend approach so I'm not surprised at all.
Following your logic, we don't even need a VGA output for Next, Sinclair certainly didn't predict that Spectrum would attach to the monitor, nor do I know anyone who had it, back in the day.

[1024MAK]

@1024Mak

Now we have reached the stage, where hdmi is totally irrelevant, as expected...
The point is, that hdmi was important to many potential bakers and therefore promised in a very explicit way, as a very important feature.
Without it, the kickstarter would be much less successful, Henrique knew that back in 2016 and he asked Ben Versteeg to help him with that, not by coincidence.
You also know all this, but for some reason you relativize things again, with this rose tintend approach so I'm not surprised at all.
Following your logic, we don't even need a VGA output for Next, Sinclair certainly didn't predict that Spectrum would attach to the monitor, nor do I know anyone who had it, back in the day.

Of course HDMI is important. The intention was that the ZX Spectrum Next would be able to work with modern TVs.

And I fully understand that the current situation has lead to disappointment.

However, in real life, sometimes there are bumps in the road.

Until I looked into it, I did not fully realise the limitations of the HDMI system. In the past HDMI has been promoted as the best thing ever (as far as connecting up video systems to your TV is concerned). And TV manufacturers are busy dropping the existing analogue inputs faster than you can blink.

You would expect that a HDMI Video Source should just connect to a HDMI video display. But it turns out it is not always as simple as that

The reasons why have already been discussed.

To be honest, the TV / video display already has a frame buffer (all LCD TVs and monitors have to have one, due to the way they work). It’s just a shame that the consumer equipment manufacturers and movie companies did not implement a better data format. They have been far too busy promoting 4K. And ignoring the way the human eye works. A faster display update would be far better (say 100Hz) at improving the viewers experience of ‘moving pictures’ (as opposed to a static picture).

Of course if all the people here had known about all these problems back before the Kickstarter was launched, it’s likely the HDMI limitations would have been listed. I honestly cannot remember exactly what was said in relation to the HDMI, as the HDMI was not important to me personally (I use SCART or composite for most of my old computer video connections).

Although my Next is connected via HDMI.

I don’t think Sinclair really had a vision for the ZX Spectrum once it was launched. He was more interested in other things. Never mind that the humble ZX Spectrum was the most successful thing that he was ever involved with.

It’s possible that the hardware designers of the ZX Spectrum were thinking slightly ahead, in terms of video outputs, as there was an option for a composite video or component video output, there are links on some versions of the ZX Spectrum board for these. And the same signals can be routed to the edge connector.

At the time, of course VGA was in the future...

VGA itself is not exactly a good standard either...

But the main point is that HDMI is a lot harder than was expected.

And the ZX Spectrum Next is very compatible with the earlier ZX Spectrum models. Especially if using analogue video connections (RGB SCART or VGA). It is not as good when using HDMI, due to technical limitations.

The possible work arounds have already been discussed. Are any of them ideal, no. But some will give users an additional choice.

I don’t know how many backers will be happy or unhappy with this. But given that the Next does very nearly get to the goals that were set, I think that’s reasonable.

Mark

[Lethargeek]

About Z80 CPUs. There are at least four NMOS versions: Zilog’s original version and the official licensed versions by other manufacturers (such as Synertek, Mostek [MK3880] and SGS [Z8400]). It was reverse engineered and ‘copies’ produced by several Japanese, East European [U880] and Soviet manufacturers [T34BM1, КР1858ВМ1]. Indeed, the NEC version [µPD780] found in most Sinclair made ZX Spectrum computers is an “illegal” (as in, not officially licensed) reverse engineered version.

Later a CMOS version [Z84C00] was developed, and a version of this is still manufactured. This is what is likely to be used if your machine is repaired.

Undocumented or illegal opcodes, and effects on registers have various effects, some of which are useful. However, as they are not part of the formal definition of the instruction set, different implementations of the Z80 are not guaranteed to work the same way for every undocumented opcode.

This is irrelevant if the "total compatibility" with the original spectrum models is claimed. This means compatibility not with just the old software most likely not using obscure effects, but with any new software for the old hardware that might use them. Original models didn't use random Z80s. Those repaired machines with odd processors might be not totally compatible anymore as well. Also regarding Patrik's tests, there are only small differences between NMOS and CMOS variants and just in one test afair. Not passing at least these tests completely at least as one of the official Z80 variants is inexcusable.

[1024MAK]

About Z80 CPUs. There are at least four NMOS versions: Zilog’s original version and the official licensed versions by other manufacturers (such as Synertek, Mostek [MK3880] and SGS [Z8400]). It was reverse engineered and ‘copies’ produced by several Japanese, East European [U880] and Soviet manufacturers [T34BM1, КР1858ВМ1]. Indeed, the NEC version [µPD780] found in most Sinclair made ZX Spectrum computers is an “illegal” (as in, not officially licensed) reverse engineered version.

Later a CMOS version [Z84C00] was developed, and a version of this is still manufactured. This is what is likely to be used if your machine is repaired.

Undocumented or illegal opcodes, and effects on registers have various effects, some of which are useful. However, as they are not part of the formal definition of the instruction set, different implementations of the Z80 are not guaranteed to work the same way for every undocumented opcode.

This is irrelevant if the "total compatibility" with the original spectrum models is claimed. This means compatibility not with just the old software most likely not using obscure effects, but with any new software for the old hardware that might use them. Original modes didn't use random Z80s. Those repaired machines with odd processors might be not totally compatible anymore as well. Also regarding Patrik's tests, there are only small differences between NMOS and CMOS variants and just in one test afair.

How do you know which Z80 chip is fitted in which machines? Sinclair used multiple different suppliers who bought components from various sources. Look at photos of original ZX Spectrum machines and there are at least four different makes of Z80 CPUs used (not including CMOS types). Further, over time, the manufacturing of the silicon chip that the Z80 is made of changed as the process evolved.

It’s extremely unlikely that you will ever get total 100% compatibility, as it is always possible that at some point in the future, someone will find another edge case.

Maybe the Kickstarter should not have used such terms, but instead should have said something slightly less exact.

But the ZX Spectrum Next gets really really close.

When run with an analogue video connection, do you have any examples of the Z80 core failing to run any ZX Spectrum software?

Does this software run on all existing ZX Spectrum models? And on ZX Spectrum compatibles like the Harlequin?

Mark

[DouglasReynholm]

I had some problems running Manic Pietro over VGA. Tried every video mode at 50Hz, but the multi colour was badly corrupted. The only way to get it to work was to use the Pentagon .tap and start the game in Pentagon timing mode. Anyone got the 'vanilla' version going? It's not a big deal, but thought I'd mention.

[Lethargeek]

How do you know which Z80 chip is fitted in which machines? Sinclair used multiple different suppliers who bought components from various sources. Look at photos of original ZX Spectrum machines and there are at least four different makes of Z80 CPUs used (not including CMOS types). Further, over time, the manufacturing of the silicon chip that the Z80 is made of changed as the process evolved.

Right now i vaguely recall at least one example of fast border effects code using an undocumented (but working perfectly on any Z80 used in original models) instruction "out (c),0" for maximum speed. This instruction behaves differently on CMOS and NMOS chips, and the prior code actually does a CPU detection using other undocumented (but working perfectly on any Z80 used in original models) features. Don't remember the prod title and specific detection method (was it NMOS interrupt bug, or undocumented flags effects, or both), will search for it.

But heck, now i can imagine some future software written by some guys specifically to NOT work on the Next.

It’s extremely unlikely that you will ever get total 100% compatibility, as it is always possible that at some point in the future, someone will find another edge case.

This is not a passable excuse for not having 100% compatibility with all the already known cases.

[1024MAK]

It’s extremely unlikely that you will ever get total 100% compatibility, as it is always possible that at some point in the future, someone will find another edge case.

This is not a passable excuse for not having 100% compatibility with all the already known cases.

So if you find problems, please do post up examples and report to the Next team by the requested route. Then the Next team can investigate. And maybe fix the issue.

Mark

[Lethargeek]

So if you find problems, please do post up examples and report to the Next team by the requested route. Then the Next team can investigate. And maybe fix the issue.

i don't know what route you're talking about (and don't plan to join FB if you mean it) but looking how this was already done in the very 1st reply in this same thread and then the excusestorm ensued i doubt it will be worth the hassle in any case

[Swainy]

I had some problems running Manic Pietro over VGA. Tried every video mode at 50Hz, but the multi colour was badly corrupted. The only way to get it to work was to use the Pentagon .tap and start the game in Pentagon timing mode. Anyone got the 'vanilla' version going? It's not a big deal, but thought I'd mention.

It works perfectly on mine via RGB on a CRT.

[1024MAK]

So if you find problems, please do post up examples and report to the Next team by the requested route. Then the Next team can investigate. And maybe fix the issue.

i don't know what route you're talking about (and don't plan to join FB if you mean it) but looking how this was already done in the very 1st reply in this same thread and then the excusestorm ensued i doubt it will be worth the hassle in any case

As I understand it, they want issues reported via their forum https://www.specnext.com/forum/index.php and via https://gitlab.com/thesmog358/tbblue/-/issues

Mark

[Alcoholics Anonymous]

I had some problems running Manic Pietro over VGA. Tried every video mode at 50Hz, but the multi colour was badly corrupted. The only way to get it to work was to use the Pentagon .tap and start the game in Pentagon timing mode. Anyone got the 'vanilla' version going? It's not a big deal, but thought I'd mention.

Works fine here. Make sure you are keeping up to date as many things have been fixed about how the machine is configured when loading legacy software.

[Alcoholics Anonymous]

This is irrelevant if the "total compatibility" with the original spectrum models is claimed. This means compatibility not with just the old software most likely not using obscure effects, but with any new software for the old hardware that might use them.

Compatibility with all these spectrum models: 48k, 128k, +3, pentagon in both hardware and software is very high. The Next is more compatible with the spectrum than any other original spectrum model. You can plug in original peripherals and they will work too. It does not create some effects like snow (does someone want that?) and an analog effect that was shown many years ago with an animated willy iirc.

All problems communicated in this thread have to do with HDMI except one which was found to be with how a snapshot was loaded.

Original models didn't use random Z80s. Those repaired machines with odd processors might be not totally compatible anymore as well. Also regarding Patrik's tests, there are only small differences between NMOS and CMOS variants and just in one test afair. Not passing at least these tests completely at least as one of the official Z80 variants is inexcusable.

This is wrong too. They used whatever source of z80 was cheapest. Have a look on the internet -- you will see photos of original pcbs using, eg, Zilog and SGS sources. The 48k model specifically is perhaps predominantly nec.

New spectrum builds like the harlequin, evo etc use cmos z80s. These are different from nmos z80s in a few respects :- they fix the "ld a,i/r" bug, the out(c),0 thing, some older versions were missing the sll undocumented instructions and I would be shocked if some of the more obscure undocumented instructions aren't missing. Stuff like "rlc a,(ix+n)" -- and I have seen this used in a spectrum demo so if this is missing, the cmos z80 will fail to run such demos properly too. When moving to nmos from cmos you are laying out the entire chip from scratch and there will be many opportunities to be different.

Some of those things being tested are meaningless. A zilog z80 cannot pass those tests because the reverse-engineered nec z80 has different undocumented flag behaviour. This comes down to a different implementation in the hardware.

It's almost like a cult that has grown that thinks you have to reproduce exactly what an nec z80 does. Not even a real z80 ffs. And it's easy for non-technical people to get swallowed up in it because they don't really understand what is being tested.

[Alcoholics Anonymous]

You also know all this, but for some reason you relativize things again, with this rose tintend approach so I'm not surprised at all.
Following your logic, we don't even need a VGA output for Next, Sinclair certainly didn't predict that Spectrum would attach to the monitor, nor do I know anyone who had it, back in the day.

Good grief. Pegaz, your problem is you don't want to understand. That's fine but don't raise BS by saying attempts to explain are a relativization. Those of us who have to *do* things try to explain why things are the way they are because some people want to know.

The HDMI issue has been known for a long time and for a long time we have said we will be working on it. What I can't do for you is break any limitations we have -- the size of the fpga, the laws of physics. So let's see where things are in a few months, yeah?

The zx hd is inadequate as a solution. As was tkpie, which was Victor's solution for the KS, and is basically the same thing. I have mentioned what is a possible solution along these same lines in this thread already -- and that is the Next sending its digital RGB stream to its own pi zero to generate the HDMI. That's one possible path but it's not the best one because it severely hampers the pi as a useful co-processor. There are several other possible paths too, and I can think this one and another would almost certainly work, as unattractive as they are.

Rather than zx hd, which is fine for the original models, I would select one of the RGB to HDMI converters out there which usually sell for about the same price. The advantage is it would work for everything the zx next generates -- 48k timing, 128k timing, +3 timing, Pentagon timing, and all the graphics modes -- and it would work on other systems you may have that output RGB.

[Lethargeek]

Compatibility with all these spectrum models: 48k, 128k, +3, pentagon in both hardware and software is very, very high. The Next is more compatible with the spectrum than any other original spectrum model. This includes the obscure effects. You can plug in original peripherals and they will work too.

Now this is just CRAZY. Thanks to Patrik and other people, now we have a well documented knowledge about how to write code working perfectly on the original machines (whatever Z80 they use) but NOT working on the Next. This is freaking DEFINITION of incompatibility and denying it really smells of cultist mindset.

This is wrong too. They used whatever source of z80 was cheapest. Have a look on the internet -- you will see photos of original pcbs using, eg, Zilog and SGS sources. The 48k model specifically is perhaps predominantly nec.

And all of these were officially licensed and passed Patrik's tests all the same except NMOS/CMOS difference. And the Next is not similar to either in these tests, and is even worse than reverse-engineered eastern clones having some other discrepancies but not that much.

New spectrum builds like the harlequin, evo etc use cmos z80s. These are different from nmos z80s in a few respects :- they fix the "ld a,i/r" bug, the out(c),0 thing, some older versions were missing the sll undocumented instructions and I would be shocked if some of the more obscure undocumented instructions aren't missing. Stuff like "rlc a,(ix+n)" -- and I have seen this used in a spectrum demo so if this is missing, the cmos z80 will fail to run such demos properly too. When moving to nmos from cmos you are laying out the entire chip from scratch and there will be many opportunities to be different.

What's your source? Never heard of ANY Z80 missing actual instructions overall (and used in any of the original Spectrums in particular).

Some of those things being tested are meaningless. A zilog z80 cannot pass those tests because the reverse-engineered nec z80 has different undocumented flag behaviour. This comes down to a different implementation in the hardware.

Repeat, according to numerous tests performed by many different people i'm aware of, any NMOS Z80 gives similar results to any other NMOS Z80, any CMOS Z80 gives similar results to any other CMOS Z80, with eastern clones having a few distinctions. That's all.

It's almost like a cult that has grown that thinks you have to reproduce exactly what an nec z80 does. Not even a real z80 ffs. And it's easy for non-technical people to get swallowed up in it because they don't really understand what is being tested.

Nobody said it ought to be the NEC Z80. Well, not me surely. Problem is, the Next fails to reproduce ANY real Z80 variant.

[Pegaz]

[mention]Alcoholics Anonymous[/mention]

This is just great.
You pulled one paragraph (in fact, just one word) from my message to make your point, just as a way to change the subject.
My whole message was a reaction to Mark's post, I reply to him, because I felt that it somewhat relativized the importance of the hdmi.
He answered me reasonably and I had no further objection.
Mark and I had discussions before about different things, sometimes we could agree, sometimes not, but I know for sure that he doesn't need anyone to represent him in the forum debate.
You allready went too far when you started to divide the forum members into honest and dishonest and now you go even further, ripping off my message, putting words in my mouth and presenting your own projections as my claims.
btw, I was already quite satisfied with the answers I get today regarding zx-hd and video converters, but thanks for the tip anyway, now I know that there is a decent solution to this problem, even if it will cost extra.

[Firefox]

Edit: https://medium.com/@uto_dev/the-awakeni ... 32c7377788

• Sinclair weren't going to do a 128k Spectrum

• The original toast rack PCB was designed by Flare Technology in Cambridge (them wot did the Atari Jaguar)

• Investronica improved the ROM

• Then Sinclair de-improved it a bit for their version

[1024MAK]

If you think that the NEC µPD780 is a official licensed version, you are mistaken. Read what it says in this Wiki. Then read this transcript (you want page 13 onwards). Yes, in their reverse engineering, they reproduced most of the undocumented instructions and behaviour, but not all of it was exactly the same as a Zilog Z80 CPU (or of a second source official licensed version).

At least some differences exist, see here

Mark

[TheSMoG]

code working perfectly on the original machines (whatever Z80 they use) but NOT working on the Next. This is freaking DEFINITION of incompatibility and denying it really smells of cultist mindset.

That statement actually seems (to me) like the definition of a cultist mindset but I digress; you want to write software that doesn't work on the Next?
That's fine if you want to attempt it and it's absolutely your choice but eventually (as written in the discussion that preceded) one could just port a core from say the ZX DOS that will implement what you're trying to block. So one way or the other the Next WILL run the software you're trying to block it from. And you'll have spent a lot of your time and effort to do exactly what?
All of this makes no sense really as a discussion.
It's been said above and I shall reiterate: The machine is a work in progress with a very high degree of completeness and it IS a work in progress precisely because the team just doesn't want to stand still and say: "this is it; we're not doing any more". All original Spectrum models were incomplete in some way, shape of form, due to Sinclair's (and then Amstrad's) mindset of rushing things to market. Yet they succeeded despite all that. You can accept that the effort is being made or you can believe that there's no effort to improve anything. Your choice entirely.

Additionally nobody's forcing you to buy -or keep if you have one already- one.

My personal preference is that people would support efforts to enhance the Spectrum as a platform that is being made from wherever that effort comes; the Evo, the Omni, the Uno/Dos and everything else in between and of course the Next. But that doesn't mean that everybody needs to like everything and of course nobody's forcing you to like the Next or the technical explanations any team member uses when asked a question.
Obviously some people like to nitpick, others to look at the bigger picture; it's fundamental differences and nothing can be done about that.

Finally: eventually ALL old software will run on the Next perfectly; that's why we invite reports of incompatibilities and we fix them. For future software it is ALWAYS a good author practice to contact platform maintainers (of ANY platform not just the Next) in order to make them as compatible as possible.

Have a great evening

[Lethargeek]

If you think that the NEC µPD780 is a official licensed version, you are mistaken. Read what it says in this Wiki. Then read this transcript (you want page 13 onwards).

it just says the NEC version was copied, nothing about it becoming or not becoming licensed later

Yes, in their reverse engineering, they reproduced most of the undocumented instructions and behaviour, but not all of it was exactly the same as a Zilog Z80 CPU (or of a second source official licensed version).

At least some differences exist, see here

Ok, maybe i didn't recall this particular bit correctly. But this is just two instructions having effects so obscure that these weren't discovered until 2012. Whereas the Next fails with too many and too well known other instructions, so my main point still stands true.

Even making it fully compatible with just that one NEC variant is still infinitely better than keeping it incompatible with any one.

[DouglasReynholm]

If you think that the NEC µPD780 is a official licensed version, you are mistaken. Read what it says in this Wiki. Then read this transcript (you want page 13 onwards).

it just says the NEC version was copied, nothing about it becoming or not becoming licensed later

Yes, in their reverse engineering, they reproduced most of the undocumented instructions and behaviour, but not all of it was exactly the same as a Zilog Z80 CPU (or of a second source official licensed version).

At least some differences exist, see here

Ok, maybe i didn't recall this particular bit correctly. But this is just two instructions having effects so obscure that these weren't discovered until 2012. Whereas the Next fails with too many and too well known other instructions, so my main point still stands true.

It's possibly a good time for you to take some time to relax a while about this. I (personally) see that this is a problem for you, but please try and see the bigger picture? The Next team have done well despite adversity to bring us a great, but not perfect product. However, we all need to take a deep breath right now and remind ourselves of what is really important, and the fact the ZX Next is not to your high standards, is really quite grating right now.

As I said not so long ago on this thread, video standards are HARD. If you are not happy with your Next, that's okay, sell it and find another solution.

In the meantime, of course you have the right to ask questions and complain. But please respect those of us who don't want to hear it too.

EDIT: alternatively show us your own HDL chops, and how you might go about it better.

AKA, Put up, fix it, or shut up.

[Lethargeek]

That statement actually seems (to me) like the definition of a cultist mindset but I digress;

And this your statement seems to me like you don't understand the meaning of "compatibility" as well.

you want to write software that doesn't work on the Next?
That's fine if you want to attempt it and it's absolutely your choice but eventually (as written in the discussion that preceded) one could just port a core from say the ZX DOS that will implement what you're trying to block. So one way or the other the Next WILL run the software you're trying to block it from. And you'll have spent a lot of your time and effort to do exactly what?
All of this makes no sense really as a discussion.

Doesn't matter if it's me or someone else. Some people just don't want their software running in emulators. And right now the Next has worse CPU emulation than many software emulators. This is a fact.

Additionally nobody's forcing you to buy -or keep if you have one already- one.

This borders with ad hominem arguments and has absolutely nothing to do with the question does the Next live up to initial promises or not.

Finally: eventually ALL old software will run on the Next perfectly;

and as i said, this is not enough for "total compatibility"

that's why we invite reports of incompatibilities and we fix them.

I don't see it. These test results are almost the same as two years before. And what a nice reaction i'm getting here for bringing them.

For future software it is ALWAYS a good author practice to contact platform maintainers (of ANY platform not just the Next) in order to make them as compatible as possible.

definitely not the ones not understanding the real meaning of "compatible"
especially if the author wants his software to run properly on a real Spectrum

[Lethargeek]

As I said not so long ago on this thread, video standards are HARD. If you are not happy with your Next, that's okay, sell it and find another solution.

and i wasn't even talking about the video

In the meantime, of course you have the right to ask questions and complain. But please respect those of us who don't want to hear it too.

well, nobody's forcing you to listen

EDIT: alternatively show us your own HDL chops, and how you might go about it better.

Why me? It wasn't me making promises about it.

AKA, Put up, fix it, or shut up.

don't tell me what to do and i won't tell you where you may go

[1024MAK]

If you think that the NEC µPD780 is a official licensed version, you are mistaken. Read what it says in this Wiki. Then read this transcript (you want page 13 onwards).

it just says the NEC version was copied, nothing about it becoming or not becoming licensed later

Yes, in their reverse engineering, they reproduced most of the undocumented instructions and behaviour, but not all of it was exactly the same as a Zilog Z80 CPU (or of a second source official licensed version).

At least some differences exist, see here

Ok, maybe i didn't recall this particular bit correctly. But this is just two instructions having effects so obscure that these weren't discovered until 2012. Whereas the Next fails with too many and too well known other instructions, so my main point still stands true.

Even making it fully compatible with just that one NEC variant is still infinitely better than keeping it incompatible with any one.

My point is that a NEC µPD780 is a different silicon design to a Zilog Z80. Hence it is not fully 100% compatible. The official licensed versions of the Z80 were made by companies that had access to everything they needed, so they could manufacture a chip to the exact same silicon masks that were used to make the Zilog chips. Hence they were 100% compatible.

Yet here you are arguing that the Next should be 100% compatible. But 100% compatible to what? A NEC µPD780? Or a Zilog NMOS Z80? Or a CMOS Z80?

If the Next core runs 99.9% of software that is available for the ZX Spectrum (which is a huge amount mind) does that not demonstrate how much effort the team has put in? And they have said loud and clear that if you find any ZX Spectrum software that does not run (using analogue video modes) that they will investigate.

I honestly don’t see a problem here.

Even the different rubber key ZX Spectrum with different issue boards and different versions of the ULA has problems with some games.

As to software emulators, all of them have limitations. The vast majority are unable to run accurately at T state timings. And how many can communicate to any real ZX Spectrum expansion/interface edge-connector hardware (assuming the host computer had suitable I/O ports)?

It’s my impression that you just want to argue regardless.

Mark

[toot_toot]

I think the circles this thread is running in really highlights the problems the Next had from the start. On reflection the Next was trying to do too much for too many user types.

future proofed with HDMI output
fully compatible with the Spectrum library
enhanced Next functionality like ULA+, Turbo modes et
fully compatible with original Spectrum interface expansions
ability to simulate multiple Spectrum variations, including non-official ones like the Pentagon
a PCB that's fully compatible with the original Spectrum case/keyboard


Again on reflection it's clear to see that the Kickstarter pitch tried to claim far too much that was feasible, but was probably made to try and get a wide enough user base in order to get enough backers for the Campaign. If the Next was just focused on the first three points - which is really why I backed it - then they may have not got enough backers.

I understand that compromises have been made since the original Kickstarter campaign was launched, but I'm frustrated at the defensive response for some of the points not working as originally claimed is one of the other points above, e.g. "but.... it supports original interface expansions" or "but... it supports other Spectrum variations", "but....it supports the original Spectrum case" and somehow that makes it OK to not admit there's an issue with the incompatibility points that are being raised, like in this case the fact the Next isn't fully compatible with the Spectrum Library, using HDMI or not. A point is raised that the Next has many errors in the Z80 test and similar arguments comes back from the official team. Maybe the original posting wasn't carried out in the least emotional way possible, but it's still feedback on the Next. It's maybe why we seem to end up in a technical rabbit hole that's often very argumentative.

I suppose the reality is that the Next can't do all of the points above, because it's physically impossible or there aren't the resources to implement the fixes. Maybe the Next team should take a look at what they really want to support out of that list and prioritise it based on what users are looking for, but based on actual user feedback.

Back to the incompatible games. There have been several games listed on here as not working on the Next, I haven't seen a specific list anywhere of games that don't work on the Next, so having a list is good as it hopefully gives a feedback loop to the Next team and it actually saves users from wasting time trying to make games work on the Next that won't, or we might be able to find alternative files that do work. I've even posted a specific game, Football Manager Revisited, that absolutely does not work on the Next but works perfectly on real and emulated hardware, which isn't down to HDMI as it's giving a "Nonsense in basic" error, I'd imagine it's because it's a snapshot of a 128k machine but for some reason the Next isn't picking up it's a 128k machine. There isn't an alternative TAP or TZX file as it's a new game. Maybe it's down to the Next not picking up it's a 128k machine in the snapshot, but it would be nice to know that there's a fix being worked on or there's an alternative way to make the game work. But I don't see any official acknowledgement of the game not working or of a solution being offered. Likewise doing a quick search on this thread, there's no official response on the original poster regarding Dingo not working. The replies from the official Next people don't even acknowledge Dingo not working or offering a reason why it might not work and come across as technical and argumentative! The two links posted earlier - the Spectrum Next forum and the GitLab repository do not have a "not quite working" game list, which is probably why this thread was created.

[Pegaz]

As to software emulators, all of them have limitations. The vast majority are unable to run accurately at T state timings. And how many can communicate to any real ZX Spectrum expansion/interface edge-connector hardware (assuming the host computer had suitable I/O ports)?

Mark

In terms of compatibility, SpecEmu and Baremulator always come first to my mind, with an impressive level of accuracy.
Communication with the real hardware is rare among software emulators, but I know that good old Ramsoft's "RealSpectrum" works great with real floppy disks in various formats (Beta, +3, Disciple/+D) as well as with real tape attached to the line-in connector.
You can see here, how it works with the physical flopy drive on my old Toshiba Portege laptop.
https://www.youtube.com/watch?v=NnqOa35kEk4&t=185s

It is also capable to attach up to four real joysticks on a PC serial port.
btw, "RealSpectrum" is the first emulator I know that supported floating bus, 18 years ago.
http://ramsoft.bbk.org.omegahg.com/floatingbus.html

Russian UnrealSpeccy, even supports the use of real HDD drives, CD Roms and SD cards.
Spectaculator also supports physical flopy drives but I don't know exactly to what extent.