This really gets my goat so I am going to go on a little, I hope you don't mind There is nothing personal here, just an airing. Full disclosure -- I took over the fpga design of the Next a year and half ago when Victor wanted some free time back.Nienn Heskil wrote: ↑Tue Aug 04, 2020 2:32 pm But then of course they had to increasingly start 'overflowing the banks', with all sorts of unrealistic and unnecessary bs, and just overall catering to fb randoms that clearly have, at best, only seen a Spectrum in a picture book before.
....
Anyway, if you're going as far as to drum up a pretense of designing some sort of semi-official (tm) successor to the Spectrum, the reasonable thing to do is listen to the people that actually understand a thing or two about it (because, as painful it is to admit in an age of idiots, they do exist ), first and foremost. And maybe less to the kind of folks that requests wasting an unused Z80 opcode just to have an instruction to rotate A 4 times.
I realize braindead fanboys aren't exactly conversation material, but this is simply not the context where you get to use the word 'elitism' or similar, not even close.
The Next team is composed of experts from the spectrum community. I am a former vlsi design engineer (I designed ASICs way back when), owned a ts2068 and ran many spectrum games, read the technical magazines, self-taught basic and assembler as a child. I know all about the timex machines, all spectrum models and original pentagons. I know all about the z80 line: the z180, z280, z380, ez80, rabbit semiconductors and others. I have a lot of experience in embedded systems and am an expert in z80 assembly language. I have been part of the spectrum community since 1993ish when the online existence of the spectrum was in its beginning. I have been an active contributor to the spectrum scene for the past 20+ years, most notably through the z88dk c compiler and various things like the sp1 sprite engine which has been used in around 70 spectrum games of the modern era.
Amongst other team members there are professional game developers that started on the zx81/spectrum and have seen their careers pass through many machines to the modern day. There are spectrum luminaries, particularly Garry Lancaster who has been extending +3 dos for heading toward twenty years and that has culminated in the NextZXOS we have today -- a true descendant of +3 dos.
When you talk about "consulting the experts", many of the biggest experts were in the team, and that team comes from the spectrum community. A lot of the time the people calling themselves experts and demeaning others as "fanboys" were not experts at all and on top of that only had a cursory grasp on what they were criticizing. When I read those insulting comments, I cringed and shrugged. They've never been ignored though -- the end result today comes from direction within the team and suggestions from the community. Many times those suggestions were conflicting; don't do this, do do this, etc.
However the main point:
This was taken very seriously.if you're going as far as to drum up a pretense of designing some sort of semi-official (tm) successor to the Spectrum
Because the community in general only has a weak understanding of hw, the result we have today looks misplaced to them. But it is not. If this were 1990ish and the task was to make a new compatible spectrum competitive with the 16-bits, and we were involved, we would end up close to the machine we have today. Naturally there are some concessions made to the modern day -- in 1990 you wouldn't have sd cards, you wouldn't have a pi / wifi, you wouldn't try to be compatible with all spectrums (I would have aimed for the 128K), you wouldn't get 28MHz, maybe you'd only have one palette, etc. The machine has also benefitted from 30 years of hindsight in the set of peripherals it got, many of which weren't made until the 2000s.
But you would get a layer 2, hw sprites, scrolling ula, tilemap, possibly timex modes (they are only there because of people in the team familiar with timex and the fact they'd been done already on the uno). This would have happened because any spectrum aimed for 1990ish would have had to have been at least competitive with the 16-bits and tasked with that, along with acknowledgement that the spectrum was a games machine, we ended up there. Budgets and technical know-how at Sinclair / Amstrad be damned Although this would have been an expensive Sinclair, it would have been cheaper than the 16-bits it would have had to compete against.
How hard are these things to implement? Not very hard. The extensions to the ula (timex modes, hw scrolling) are trivial. Of course you would do them. Layer 2 is even more trivial -- it's a few lines of vhdl and much smaller than the ula. This is a good way to increase colour resolution while at the same time maintaining backwards compatibility with the original system. Sprites are more difficult and would require a separate 16K pattern memory but this is well within the realms of 8-bit machines made in the mid 80s. The later TMS99xx chips often had 16K to 128K of its own memory for things like sprites. You can look at the msx or msx2 for examples. The tilemap is also smaller than the ula and is made to use the same memory as the ula. On the original 128K spectrums, the ula owns 32K of ram (bank 5 and bank 7); on the Next it's only 24K (16K bank 5, 8K for the first half of bank 7) -- more limited than the spectrum! -- and this is where the ula and tilemap coexist. The design depends on faster memory which would have been easily achievable over a distance of 10 years when ram performance would have been doubling every few years. If not, the features can degrade. 50 sprites per line instead of 100. No high resolution mode for the tilemap, etc. It also depends on integrating everything into an asic. This would have been bigger than Amstrad's +3 ula, the Sam's asic but much smaller than the asics used in the Amiga, eg. If required, a single asic could be split into two with one likely devoted to sprites.
So although you may not agree where the Next landed, I hope it's clear it is still remaining honest with a 1990ish what-if machine in the important aspects. However, whether you agree with the choices made, a new spectrum would have had to attack colour clash and limited colours. Those limitations were a function of 1982 costs and technology and would not reflect anything new or competitive made in 1990+.