I honestly feel that every Graphics developer in Hollywood, reasearch center bought and into Threadripper plenty of Enthusiasts around the world. When I bought my first TR4 CPU it was $300 for the 1900X and the MB was $400. I started with 2 sticks of RAM for $100. I was happy to get a 2920X for $700 and again decent boards were $400. Then they released 3rd Gen and again Hollywood was overjoyed but AMD realized that they did not need to use TR4 to market Ryzen to the enthusiast market. As selling 1 3960X for $3000 would mean 10-15 1900X, 4 2920X or even 3 2950X. Where AMD was good to us was CPU performance as my 5950X absolutely blows my 2920X completely away in everything. I had one of those 3300 CPUs and was definitely impressed enough that I really wanted a 3960X. I even thought about getting a TRx40 board and waiting for prices to drop. Unfortunately that has not occurred as the 3960X may be $300 less than when it launched. Having said that I do believe that AMD knows that they currently (entirely based on current CPUs) don't have anything in the budget space to compete with Intel's budget offerings. They might indeed be working on a Threadripper chip to respond to Intel's announcements around HEDT. I can actually see a dual 5800X3D chip being dropped if they need to respond to Intel's HEDT. However AMD would be genius if they released a chip on AM4 that would be the same thing.
You're overall not entirely wrong, but there are a ton of inaccuracies here. First off, you're misrepresenting (by omission) the price of 1st gen TR - the 1900X was $550 at launch, and due to its architectural quirks (NUMA issues in particular) generally performed worse than the 1800X outside of very specific applications. You might have bought yours for $300, but that's almost half off the original retail price, which says something about how that CPU did in the market. The 1920X and 1950X were much more impressive offerings, but came at a similar cost increase ($799 and $999).
As for pricing and profits, you currently get a 16-core Ryzen 5950X for ~$550, down from its $750 MSRP. That's
massively better value than even first gen TR. The market has moved on. So, sure, AMD makes mroe selling a 3960X than a 1950X, but they also make less selling a 5950X than any of those 1st gen TRs. As a consumer, even accounting for the massive price increases in the past couple of years (and excluding GPUs, talking about CPUs alone right now) we get more value for money overall, and more performance for less.
Saying Hollywood is a big TR adopter is also ... well, likely somewhat accurate, but they're an even bigger adopter of TR Pro and EPYC, and the film and TV industries have seen a massive move towards off-site render farms over the past decade. The problem TR represented a solution for in that space is a problem that has been solved better and at a much larger scale by the wide availability of either self-owned or for-rent render time on massively powerful server clusters. Even freelancers and small studios can now relatively easily get access to massively powerful render farms for much less than it would cost them to build up their own HEDT-based setups.
Also, as you yourself admit - most of that performance you mention, which typically isn't memory or PCIe bound, is now found in the now up to 16-core MSDT market - and at much higher clocks, lower power draws, and higher IPC. Which leads to the situation I outlined above: TR only makes sense if you see significant gains above 16 cores/32 threads, or if you really need the PCIe and/or RAM bandwidth (but also don't stand to benefit further from the extras TR Pro and EPYC bring to the table in those areas). Workers doing 3D/VFX in the entertainment industry are now most likely working on high core count MSDT systems with large render jobs offloaded to a render farm. The benefits to these workflows from a non-pro TR lineup would either be nonexistent or would be even larger with TR Pro.
As for releasing a theoretical 5900X3D or 5950X3D in response to Intel's HEDT efforts - that's unlikely for several reasons. It would still be on an MSDT platform with two memory channels and a relatively low number of PCIe lanes. It would also top out at 16 cores, which is already mostly comparable to Intel's 8+8 MSDT chips. If Intel launches a new HEDT platform, a new MSDT CPU wouldn't be much of a response to that beyond what already exists in the 5950X. That doesn't mean it might not launch - I don't think it will, but AMD seems set on keeping AM4 alive for a while yet, so I may be surprised - but any further X3D AM4 chips would be more of a bid towards keeping AM4 relevant than a "kinda-HEDT" response to Intel.
Best give up. Unless you make a point that is grounded in immutable logic or undeniable fact, you will fail.
Yet my arguments are rooted in both. You're just in denial. The industry and markets have moved on in ways that render traditional "enthusiast-grade" HEDT a
tiny niche, gobbled up from both sides by expanding MSDT performance and massively increased Pro HEDT/server performance. That's just plain facts.
And what you fail to understand is that you are wrong. This market sector is not small. The reason AMD and Intel both, at the exact same time, dropped out of that market was when? 2020 was it? Pandemic maybe? Things that make you go "Hmmm" indeed.
Except for the fact that AMD has been pretty clear that TR has never sold well, that it has been more of a hobby/show-off project than anything else, and they've been quite clear in their desire to move those efforts into more sustainable and profitable areas. Hence the launch of TR Pro. Heck, you won't even find TR 3000 non-pro on their site, which is frankly astounding - they exist, after all. But it shows clear intent: TR Pro is where TR is continuing. Traditional HEDT is no longer a sustainable market.
Oh please. Save it.
I'm going to show an example...
Here. Let's see what the community response is, shall we?
EDIT: 7 votes cast so far and with 28% in favor... But we'll see.
EDIT2: 14 votes cast and we're still at 28% in favor, but this is just day one...
Yeah, still early days, but I don't see four votes in favor in an enthusiast forum as much of an argument in your favor. Remember, such a thread is inherently biased towards those interested in the subject, so compared to the overall population you'll likely see an overrepresentation of positive answers - and the population of this forum is in turn
massively more likely to be interested in HEDT than the rest of the world. Yet it's currently about 3-to-1, and the discussion in the thread is quite in line with my beliefs. We'll see how this develops, but so far I'm not seeing anything that doesn't explicitly support my views.
I still remember the "Spectrum 48K+ MHz war days"; buying a then so-called "computing machine" and whispering the last tune from 80s pop & rock idols kept you proud and happy for a good couple of summers or so. Now you buy a zillion times more powerful "computing machine", costing a ton more money which is superseded about 3 days after being launched and succeded weeks before rolling on shop's shelves... And you have to hear to CRAP music too!!!
