AMD Ryzen Threadripper 7000 Series Processors Tipped for September 2023 Release, including HEDT and Workstation Models

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
12,634
Points
113
AMD isn't planning on launching its Ryzen Threadripper 7000 Series processors until Autumn, according to a tweet shared by hardware leaker Chi11eddog today that suggests the upcoming lineup of chips, which have been rumored to feature up to 96 cores, will be released in September 2023 and comprise both workstation and HEDT models.

See full article...
 
Why bother when the 9000 series will be right around the corner?
 
Is there ever a time you can't say this about a computer upgrade though?
Sure, but there are degrees. I could legitimately understand not buying a 7700X now when the V3d models are a few weeks away for example.

I don’t know when the 9000 series will be out, but let’s say it’s Nov. a 9950x will cost a heck of a lot less than a threadripper and will likely cover performance differential of probably 4-6 extra cores
 
I don’t know when the 9000 series will be out, but let’s say it’s Nov. a 9950x will cost a heck of a lot less than a threadripper and will likely cover performance differential of probably 4-6 extra cores
Ok, true. I don’t own a HEDT platform, but my understanding was it was less about raw performance and more about PCI lanes, max RAM, etc, which probably isn’t going to change for the next Zen release - or even the next few, until they decide to start looking away from AM5
 
In the end, I think it's a moot point unless you've got some really deep pockets for this. Weren't there rumors a few months back that the next TR platforms were going to be insanely more expensive?
 
As we single out, and rationalize each occurance, I think in the end, the only thing happening is insane dollar devaluation, its just not hitting everything at the same time, but it is hitting many things at a fast pace.
 
Why bother when the 9000 series will be right around the corner?

I kind of wish they would launch the high end parts first in each generation.

HEDT first, then general consumer.

When they launch them last, and so far behind, you essentially have to make the sacrifice of practically getting last gen architectures for the privilege of going HEDT, which throws a lot of water on the whole experience.

If anyone should have the premium first dibs on the new arch, you'd think it would be your HEDT customers who are paying a premium.

That, and I'm still pretty bitter about never getting a drop in Zen3 TRX40 chip. I spent more than I've ever spent on a motherboard givne the promise of long term support, and then it never arrived.
 
In the end, I think it's a moot point unless you've got some really deep pockets for this. Weren't there rumors a few months back that the next TR platforms were going to be insanely more expensive?

That was for the Pro models I believe.

The Pro models are essentially EPYC chips with twice the PCIe lanes and RAM channels, and priced much higher.

The only reason this looked like a massive price increase was because they neglected to release a non-pro version this gen.

If these rumors are accurate, it seems like they are reversing that decision.

That said, this doesn't mean the new ones won't be more expensive. That seems to be the way everything is going these days.
 
I kind of wish they would launch the high end parts first in each generation.

HEDT first, then general consumer.

When they launch them last, and so far behind, you essentially have to make the sacrifice of practically getting last gen architectures for the privilege of going HEDT, which throws a lot of water on the whole experience.

If anyone should have the premium first dibs on the new arch, you'd think it would be your HEDT customers who are paying a premium.
So, consider that AMD does launch the high-end parts first - they launch Epyc. The difference here being that the same eight-core CCD that goes into a 7700X can also go into an Epyc CPU, whereas Intel produces entirely different parts for Xeons.

AMD is probably going after revenue sources first. Epyc sells, and the margins have to be tremendous. Consumer desktop (AM5) SKUs likely sell in volumes that eclipse 'workstation' SKUs, but sit lower on the margin scales.

While Threadripper is its own socket.

I think for AMD, the decision to go prosumer for a specific generation is going to come down to an alignment of stars; not only does there have to be demand for the parts, but there has to be enough revenue and enough margin in producing them on top of having the parts to spare vs. Epyc and their Ryzen AMx parts. Caveating TR Pro here since that is just a rebranded Epyc.

I'll also point out the stupendous increases in core count and thread count in the consumer space, as well as some general progress in terms of connectivity. As much as I'd love to build a Zen4 TR machine, I find my own reasoning to be based on the idea of combining functions from different systems into one, something that I personally would prefer to move away from so long as I'm using 'prosumer' parts that lack the reliability of real server room hardware, to include consumer power delivery.

Things like the VM server should be separate from the file server, and anything having to do with internet access should be on its own device, and all of them need separate UPSs, and so on.

As for my personal machine - can't say that I need more than one CPU socket, two RAM slots, one GPU, and more than two or three NICs, all of which can be supplied by a consumer platform that's likely to stay more up to date.
 
As long as you have pcie channels and slots for mapped controllers doing your file server and other not network device items and non gaming devices on your hypervisor should be a non issue. You could even repurpose your server doing file server as a hypervisor as well to let you use the compute on it more favorably as it's real role is probably pcie lanes. If licensing cost isn't an issue.

The real downer to running most hypervisors at home is the cost of licensing those sockets.
 
As long as you have pcie channels and slots for mapped controllers doing your file server and other not network device items and non gaming devices on your hypervisor should be a non issue. You could even repurpose your server doing file server as a hypervisor as well to let you use the compute on it more favorably as it's real role is probably pcie lanes. If licensing cost isn't an issue.
It's not about taking advantage of compute, really - it's about separating of duties in order to separate failure domains.

With consumer gear, the likelihood of something going down is much higher, and if an 'eggs all in one basket' approach is used, then certain critical services are likely to go down more often too.

Really just a personal choice on my part.
 
Become a Patron!
Back
Top