The end of THREADRIPPER?

[Drumroll]

Just wondering if the news around that the new Ryzen launch would be the end of the Threadripper?
As the core count will increase, allied to new resources and technologies, there's only few scenarios where the TR4 platform would make sense?
Too bad the quad channel won't be available.
Would make sense to AMD to keep with the TR4 platform?

 

[opethdisciple]

Doubt it. The cores/threads available on the Threadripper platform will just increase surely.

 

[anthony99]

I'm sure once 16c/32 Become mainstream threadripper will go more cores.There's always people who need more cores in their work

 

[DragonQ]

Ryzen 1 was 4-8 cores, Threadripper 1 was 8-16 cores. Ryzen 3 will be 4-16 cores, so Threadripper 3 will probably be 16+ cores. Maybe they'll top out at 32 cores again, maybe it'll be 48 or 64.

 

[LePhuronn]

I've said this many times before, AM4 and TR4 are completely different platforms for completely different purposes. There is no comparison to be made.

If you only need core count then, of course, a 12 or 16 core Ryzen 3000 will do you quite nicely and spares you the extra expense of Threadripper. We're talking hobbyist and prosumer work here: simple video editing and streaming, Blender work, photo manipulation, etc.

But Threadripper is for professional work: broadcast-level video production, animation and 3D, extensive VMs, low to mid level computational sciences, that sort of thing. This workload will eat up all the cores you can give it, all the RAM you can feed, multiple sources and arrays of fast storage for insanely large raw footage files, multiple compute GPUs, etc.

Ryzen 3000 may give us 16c/32t CPUs and a pair of M.2 slots at full speed, but AM4 doesn't give us 128GB+ RAM, ECC RAM support as standard, multiple 16x PCIe slots, multiple full-speed M.2 slots and still have PCIe lanes left over for extra features like 10Gb ethernet and the like. Ryzen may top out at 16 cores, but Threadripper will start at 16 cores and go all the way up to 48 or even 64.

Threadripper was a skunkworks side project that I don't even think AMD expected to be as popular as it's become. As a result, Zen 2 isn't the end of Threadripper, Zen 2 is Threadripper reaching maturity.

 

[Tee Hee Johnson]

They are for different markets, core count will increase in TR.

Maybe pull the speeds up a bit.

Will remain 2 different entities.

 

[Scougar]

TR is more than just cores. It is lanes, memory and other functionality that is pro/semi-pro level.

 

[smilingcrow]

If the AM4 platform evolves enough in terms of PCIe lanes, core counts, memory bandwidth and maximum size then it will eat away at the demand for the lower end of the TR platform.
Workstation users will continue to benefit from more cores at the very least, but in some areas the extra supply may outstrip demand.
For example, how many current TR2 have maxed out the supported RAM size and are using all the PCIe lanes?
So increasing these will become much less relevant at some point although there may be new areas where TRx pulls ahead of AMx to compensate.
Only AMD know what the current mix of sales for the TR platform is in terms of low, mid and high end and how the road-maps for both platforms might lead to cannibalisation.
But then even they don't know exactly what people are using TR systems for in terms of RAM and PCIe usage.
Until consumers definitively know what the new AM4 chipset offers as well as the overall platform performance and features then it's not possible for them to say whether Ryzen 3 makes TR2 redundant for their usage patterns.
Ideally one would want to have info on TR3 before purchasing Ryzen 3 for those in the cross-over zone, although if Ryzen 3 looks good enough for you that should be clear once the reviews are in.

 

[LePhuronn]

Only AMD know what the current mix of sales for the TR platform is in terms of low, mid and high end and how the road-maps for both platforms might lead to cannibalisation.

As pointed out to me elsewhere, the 8-core 1900X was never repeated with Threadripper 2 so it's likely that AMD are aware that Threadripper customers want more cores than the top-end Ryzen.

But to say again, there isn't cannibalisation because these products are like chalk and cheese. It's like suggesting the Focus RS might cannibalise Ferrari 488 sales.

 

[smilingcrow]

As pointed out to me elsewhere, the 8-core 1900X was never repeated with Threadripper 2 so it's likely that AMD are aware that Threadripper customers want more cores than the top-end Ryzen.

I pointed that out for one. I don't think they needed to update that as the 8C chip was for people who needed the platform and not the extra cores. In that scenario improving the CPU performance by 10 or 15% was an irrelevance really. If people needed the platform and more grunt then they could jump to 12C or higher.

But to say again, there isn't cannibalisation because these products are like chalk and cheese. It's like suggesting the Focus RS might cannibalise Ferrari 488 sales.

Of course's there's potential for cannibalisation as some people primarily just want more cores and have no use for most of the other platform features although they might well use quad channel memory as it comes for free with the platform. Not sure why you are so blind to this!
The question is to what degree will cannibalisation occur with Zen 2?
I haven't said that it will be a significant issue as you'd have to run a poll to get hard data to find that out.
But don't tell me that there weren't people looking at the blow out deals on TRx 12/16 core who changed their mind once AMD hinted at dual chiplet AM4 chips.

 

[LePhuronn]

I pointed that out for one

Was it you? I have the memory of a goldfish these days. Well then there's no need for me to tell you

Of course's there's potential for cannibalisation as some people primarily just want more cores and have no use for most of the other platform features although they might well use quad channel memory as it comes for free with the platform. Not sure why you are so blind to this!

I'm not blind the anything. Up until now, if AMD purchasers wanted more than 8 cores then they'd look at Threadripper. Ryzen 3000 offers (let's say) 16 cores so those purchasers who just want more cores are now spared the expense of the HEDT platform and kit they won't use. But like I said, Threadripper is not about having more cores, it's everything else that comes with it. And if Threadripper 3 starts at 16 cores, then where's the cannibalism? The crossover is a tiny, tiny portion of the Venn diagram for those looking solely at core counts.

I haven't said that it will be a significant issue as you'd have to run a poll to get hard data to find that out.
But don't tell me that there weren't people looking at the blow out deals on TRx 12/16 core who changed their mind once AMD hinted at dual chiplet AM4 chips.

So we need to run a poll to get crossover data, but you're happy to make an assumption to prove a secondary point? Don't go down that route.

 

[Stephanie Peterson]

Imho TR will go to 48 cores later this year.

 

[Deleted member 66701]

Threadripper + on package HBM should keep it relevant.

 

[Skynet5]

Not just about cores. But that the thing that is marketed more than other features, so most people that dont need the other stuff don't know about it.

 

[Rroff]

TR is more than just cores. It is lanes, memory and other functionality that is pro/semi-pro level.

Kind of like my 4820K v 4770K - with the extra PCI-e lanes, extra RAM channels, etc. stuff that won't make much odds for a lot of people but can be very beneficial for certain usages not typical of mainstream consumer use.

 

[Journey]

As long as they don't discontinue the Zen+ range of TR CPU's then it is all OK. So many of the designs that I've implemented have moved from X299 to the TR4 platform, mainly for the 64-PCI-E lanes on offer. Obviously the core count is handy too, but now that Intel compete in that range it is not the end of the world, but no way could I envisage losing 20 PCI-E lanes.

 

[Vince]

Not just about cores. But that the thing that is marketed more than other features, so most people that dont need the other stuff don't know about it.

Anybody buying TR knows about all of it. Trust me, I bought it on release. People who bought am4 may not know but people who paid 500 quid for a board and 1k for a cpu knew exactly what they were buying.

 

[Skynet5]

Anybody buying TR knows about all of it. Trust me, I bought it on release. People who bought am4 may not know but people who paid 500 quid for a board and 1k for a cpu knew exactly what they were buying.

That was kinda my point. Those that don't know, typically don't need to. And then some of them ask questions on whether threadripper is dead.

I re-read what I wrote and clearly it was poorly written.

 

[smilingcrow]

I'm not blind the anything. Up until now, if AMD purchasers wanted more than 8 cores then they'd look at Threadripper. Ryzen 3000 offers (let's say) 16 cores so those purchasers who just want more cores are now spared the expense of the HEDT platform and kit they won't use. But like I said, Threadripper is not about having more cores, it's everything else that comes with it. And if Threadripper 3 starts at 16 cores, then where's the cannibalism? The crossover is a tiny, tiny portion of the Venn diagram for those looking solely at core counts.

I think I see the basis of the misunderstanding.
This isn't about people who actually need the features of HEDT downgrading to a mainstream platform so all the talk of the benefits of HEDT is really irrelevant.
This is about mainstream users who purely want more cores moving to HEDT just for that.
I saw it with Intel HEDT when DAW users and others jumped to HEDT purely for the extra cores.
Sure, some also wanted some of the extras and some gamers even convinced themselves that they actually needed more PCIe lanes when if I recall correctly the difference between 8x and 16x in dual GPU setups was insignificant.

Maybe this will make it clearer.
If Ryzen 3 topped out at 8C and TR3 16C is £500 do you think there is a market for 16C at that price even with the X series boards costing an extra ~£160 or so for TR?
Threadripper 1950X is just slightly over £500 today at certain stores and at 7nm there should be a decent profit on TR3 16C in that price range.
There is clearly interest in the supposed 12C and 16C AM4 chips based on this thread so clearly there's a market for them.
Do you really think that if 16C is TR only that hardly anybody who would otherwise would have bought it for AM4 at a similar price would buy it?
I think AMD would noticeably expand sales of TR if they did that.
So in not doing that they are definitely losing extra sales for TR.
Not losing TR sales to AM4 but losing out on people choosing AM4 over TR.
This is what I think you have missed.

I don't think this is a bad thing for consumers at all as most have no real interest in HEDT as a platform.
So in sacrificing more potential sales for TR they can really push ahead and totally annihilate Intel for the mainstream desktop platform which is great for consumers.
Surely TR is strong enough that it doesn’t need propping up anyway by limiting AM4?
Maybe you saw the word cannibalisation as a negative thing whereas I see it as a positive.

 

[Tom B]

I think they'll keep TR but have a slower release cycle. Releasing a new TR chip every year is probably not very profitable, the market for HEDT chips is tiny in comparison to desktop chips so developing/marketing a TR chip probably isn't great if they have to do it every year. However, a 2 year release cycle would mean that the development/marketing budget then gets two years worth of sales. If you want a 32 core CPU you can still get the 2990wx and that can happily remain the highest core count CPU until the 4xxx ryzen series.