Great video man! Advices like these would encourage PC Enthusiasts on not to spend unnecessarily on tech.

Can confirm the memory constraints with 16 cores. I can push past 200gb in a serious virtualization workload with older cores, imagine how much moar workload modern Ryzen can handle. But even that is a seriously niche category.

Absolut understandable

Like you with warranty?

And I am Anthony! lol

I thought consumer threadripper users were mostly doing YouTube production?
That'll eat all the cores it can though, right? It'll just scale with cores linearly?

I’m glad I found this. This is the context I’ve been looking for. I’m looking to build for the very first time. I’m a developer that wants to run a bunch of VMs, occasionally gaming will also be done. I am very conflicted what amd cpu I want. I want a bunch of cores but I wasn’t sure how many is “enough”.

2:10 So they're enforcing by law the need for profit over ANY other positive things the CEO would like to practise. So that limits the way you can can run your company. Also limits MOST of environmental practices beyond the bare minimum legally enforced.

I need a thread ripper with 69 cores and a 6900xt for reasons…..

Well, your whining is a bit exaggerated, Linus: the non-pro Threadrippers were throttled by their memory bandwidth, which is why they were replace by Threadrippers Pro; although I agree that it is a pity that the sTRX4 socket got abandoned after a single generation.

So does this mean we will be seeing a 12980xe or 13980xe anytime soon?

I disagree with your conclusion. I feel that the chip shortage is the reason AMD have gone with just Pro chips.

They have limited chips to sell, so are selling those with the greatest profit margin.
If we didn't have the chip shortage, I think they would have had a full line-up; ready to take as much business as possible away from Intel.

Well Zen 4 cannot come soon enough. Intel has killed off HEDT despite competition from AMD where they actually had a good HEDT platform when they were always on top and AMD was a distant memory almost bankrupt before Ryzen. Now Intel despite their being competition is just stacking more e-waste cores on us while maxing out at 8 p core and forcing hybrid arch down our throats which is not for everyone with no option for more than 8 cores with great IPC. Intel was actually better and more well rounded even when there was no competition. Their HEDT platform was great. You could get up to 8 cores that gamed just as well as 4 core quads on mainstream and could do so much more. Only $1000 for 8 core Haswell-E or Broadwell-E and $300 to $400 for motherboard in an era where no game could think of more than 4 cores. Now games often require at least 6 or even 8 and having a little more could provide overkill headroom. Intel has no such option now on modern arch with more than 8 good cores. They did back then where they had modern option with more than 4 good cores in 2014-2015 despite no competition from AMD. Those prices were a bit high, but quite reasonable for HEDT platform in contrast to nothing existing now except Super high end server class stuff with EPYC and Threadripper Pro CPUs that may be sop well binned and can clock so high, but cost $5K or more nevermind $2K motherboards. The HEDT scene is worse with competition than it was when Intel had total control.

If only Intel would release a 10 P core Alder Lake CPU on current platform since they have no HEDT. AMD just please release Zen4 anytime now.

Sample use case of a developer who would actually benefit from TR being around: someone dealing with extreme-sized codebases. If you have to recompile constantly, and even small changes can cause huge compiles – let alone just doing test runs on new branches – performance matters. Even on something like the Linux kernel, running allmodconfig – the gold standard for testing kernels work – on my main system, let alone VM, is already painful enough with an R7 3700X, and there I use all my cores. Dispatching out to a CI server doesn't even make that much sense because I need the binaries local for testing and I'm not doing anything while I wait for the load. Not to mention this has to be repeated for each patchset as each patchset typically consumes a local git branch – meaning recompiling. Don't forget bisecting on any project; something with 1M commits (like the Linux kernel) takes 19-20 full compiles just to find a bug's source, let alone all the compiles in the attempt to patch it.

We the DiYer and enthusiasts are being wedged out. They’re going to make our culture dumber than ever before because one day we won’t have the choice to build a pc and learn the things involved in it. Way to stunt the human race government corruption and corporate greed!

"Ahh linus is at It again ranting about evil corporations"; as one should – God.

I have a feeling that zen4 might push the x670 platform with 24-32 cores to compensate for thread ripper.
If they can push 16 cores to 5.5ghz, they can put 32 cores at -/+ 4ghz easily.

This is depressing but Linus and Anthony are right 💯. It made sense for my grad school to pay for my HEDT in the early 2010s because I was performing numerical simulations that were either trivially parallelizable or were concurrent enough to warrant use of CUDA and a dedicated Nvidia Tesla card. If someone came to me today with a similar use case I'd set them up with an EC2 cluster on AWS.

And now I'm looking at 7:44 and noticing that I accomplish all that and more at home through three separate machines (one of which is a Pi). Not only is that way cheaper than having a single HEDT, but if the VM server goes down, my wife's 3D print isn't interrupted, and my kid is still able to access his project files on the family file share.

All of this is nothing new–HPC (High Performance Computing) has been synonymous with cluster compute since I was in grad school, so honestly the only thing shocking about HEDT being on its deathbed is honestly that it stuck around so long.