Speculation: Ryzen 3000 series

[PotatoWithEarsOnSide]

You don't get market penetration by offering a similar product at the same price.
AMD badly needs market penetration, so 8c at last year's 6c price (or lower) is where it'll go.

 

[jpiniero]

You don't get market penetration by offering a similar product at the same price.

What would help more is faster IPC and/or higher clocks, not extra cores.

 

[Shivansps]

Seems like it'd be far too much engineering effort to do anything other than the 1/4 IO die.

Not sure about that, if they are really going for non-monolothic, non-I/O die on desktop Ryzens, it just the a navi chiplet who acts as the I/O chip. Its either Navi 20 with extra 128bit DDR4 and 2 satas, or there is a smaller navi that we dont know.

 

[DrMrLordX]

https://wccftech.com/amd-launching-ryzen-3000-series-cpus-apus-radeon-gpu-ces/amp/

Guess we'll find out soon enough if his souces are reliable.

All eyes are on CES.

 

[chrisjames61]

You don't get market penetration by offering a similar product at the same price.
AMD badly needs market penetration, so 8c at last year's 6c price (or lower) is where it'll go.

Nah, that isn't going to happen. At this point they aren't in a position where they need to do that with Ryzen. If Ryzen 3 is good as we think it may be then prices will stay the same or rise. Why on earth would they lower prices?

 

[maddie]

Nah, that isn't going to happen. At this point they aren't in a position where they need to do that with Ryzen. If Ryzen 3 is good as we think it may be then prices will stay the same or rise. Why on earth would they lower prices?

Is it possible to sell a product based on production cost + some consciously chosen markup?

A lot here seem to think that the only worthwhile strategy is always follow your competitor's pricing. Do we have any idea of AMD's short term (3-5 yrs) goal and their strategy to get there? AMD has what I believe to be an extremely low cost of production (all costs including design) and they will use that advantage as much as possible. their one die strategy for Ryzen 1 was brilliant.

Hector Ruiz, when he was at AMD did exactly what you're proposing (maximized short term gain) and we all know the result of that.

 

[AtenRa]

Don't you want to see AMD make money? They aren't a charity. Why would they sell an next generation eight core for what they were selling a last generation six core? A Ryzen 3 eight core should go for what a 2700X currently goes for at a minimum and probably more.

Its called progression, something both Intel and AMD were doing in the past but people have forgotten what PCs used to be the last 5-6 years. Core 2 Duo to Core 2 Quad to Nehalem to SandyBridge is what we want as consumers and finally we may get it after all those stagnated years.
This is what PC ecosystem used to get every 2 years, its time to return to the old days

 

[TempAcc99]

If the top end 3xxxX is a 16c 4.0GHz base, I can see it cannibalising future sales.
Where do you even go from there?
The consumer would probably buy the 6c model, keep it for a few years, and then swap it for the 16c model, and that'd keep them going for a very long time.
Short of IO changes, what would a consumer ever really need to swap to another CPU for?
Long term, I think that core proliferation is going to kill the market.

Fully agree. On the other hand it will be similar to 2500k up to 7700k. There wasn't that much performance increase to be had especially if your 2500k had a nice OC. Upgrading was mostly helpful due to platform improvements.

If they offer a 16c with ryzen3000-series what can they offer for 4000-series? small clock bumps? How would they price it? either very high or the 6-core would essential be in the sub $100 bracket.

Ryzen3 and navi not having pcie4 would kind of be a disappointment. Having higher bandwidth + IF + their memory tech (don't remember it's name) must be worth something? Like only 4gb hbm on card to save cost on card.

 

[Joe Braga]

Vega20 is already PCIe 4.0.

I Guess that it wiil be retrocompatible e mixed PCIe3.0/4.0

 

[ub4ty]

You don't get market penetration by offering a similar product at the same price.
AMD badly needs market penetration, so 8c at last year's 6c price (or lower) is where it'll go.

I paid $170 for my 1700 (8 core0. So, i'm kind of spoiled by that.
I did however pay a grand total of $1,000 when the 1950x (16 core) first came out and had my CC pm it to a sale price of $850. I didn't do that lightly and it was a hard pill to swallow. I'd love for the price of a 16 core to come down so I can consolidate. 1950x went for $450 at its lowest sale price. That seems fair to me.

I am really excited (again) for what AMD has in store. Intel hasn't done stirred me in this manner in a while.
It's really because I saw AMD's roadmap and platform years ago when Ryzen was launched and its exciting to see it unfold with all the new tech and value it brings. What people also fail to realize sometimes is that lower prices is in itself innovative. It brings amazing technology, performance, and potential to people who otherwise wouldn't have been able to afford it. This creates larger and more sustainable markets in the future. This allows new people to forge new use cases. Value performance is quite innovative at this stage in computing. When I sell my 1700/mobo, i'm fine with 50% off what I paid... An 8 core w/ heatsink/fan for $85. A mobo for $50. That's unheard of. Some person whose being dying for such performance who was locked out due to affordability on some dated intel machine is probably waiting for this kind of opportunity upgrade. That's where the action is at and it makes me excited. A regular person now has a pretty beefy processor that was unheard of in terms of core count outside of data centers some years ago. He/she could go on to make tomorrow's Google.

So, indeed... Sometimes it is much wiser to focus on market penetration and proliferation instead of gouging the snot out of your customers because you can. There's always a backlash in doing so. Intel likely lost me for a decade after their quad core gouging. I don't even think about that with AMD. I will if they go Intel pricing in this new series but I doubt they will which I think is commendable and innovative and builds longer term and more profitable relationships and potential.

Merry Christmas everyone !!!
I just got wholesale fiber interconnects as a gift.. Mucho value (70% off retail)

 

[Kedas]

Ryzen 3000 CPU and GPU are 7nm
Will Ryzen 3000 APU be 12nm?
Why are they so much out of sync?

 

[chrisjames61]

Its called progression, something both Intel and AMD were doing in the past but people have forgotten what PCs used to be the last 5-6 years. Core 2 Duo to Core 2 Quad to Nehalem to SandyBridge is what we want as consumers and finally we may get it after all those stagnated years.
This is what PC ecosystem used to get every 2 years, its time to return to the old days

Look what the old days got AMD. Nearly bankrupt. Just because you want them to basically give away a much improved product doesn't mean they should or will.

 

[maddie]

Look what the old days got AMD. Nearly bankrupt. Just because you want them to basically give away a much improved product doesn't mean they should or will.

AMD didn't nearly go bankrupt by pricing low deliberately. This is a false rememberence of the history. They had to price close to production costs due to a performance disadvantage coupled with a much larger, relative to Intel, costly die. Do you remember the prices they initially wanted for Bulldozer?

Things have radically changed. There is a lot more leeway presently in their pricing choices.

 

[Centauri]

AMD almost went bankrupt by blowing their piggy bank on the Dresden fab and overpaying for ATi. If they hadn't made either of those moves they wouldn't have even sniffed at liquidity (or R&D funding...) problems.

 

[Tuna-Fish]

Ryzen 3000 CPU and GPU are 7nm
Will Ryzen 3000 APU be 12nm?

Yes. AMD has published several roadmaps that clearly say that 3000-series APUs are 12nm products.

Why are they so much out of sync?

Historically, the AMD response has been "APU needs both GPU and CPU parts to be ready, and then to be integrated, so this will take longer to release than just a CPU." To me, this does have a bit of the feel of a cop-out, however. They get to make plans years in advance, they should be able to figure it out.

I think the real reason is simply that the company only has a limited amount of people who are good at doing the final physical design part of fitting a design to a process, and those people only do that in the company. So, as designs get rolled out on a new process, they should be spaced out so there is always the same amount of work for them. AMD seems to have gone with the higher-margin products in the lead. This is a sensible decision for them, but it's of course disappointing if what you personally really want is an APU.

If modular APUs (in the sense of Rome) become a thing, this might change in the future.

 

[chrisjames61]

AMD didn't nearly go bankrupt by pricing low deliberately. This is a false rememberence of the history. They had to price close to production costs due to a performance disadvantage coupled with a much larger, relative to Intel, costly die. Do you remember the prices they initially wanted for Bulldozer?

Things have radically changed. There is a lot more leeway presently in their pricing choices.

I never mentioned the reason they sold dirt cheap cpu's. The reason is obvious. They weren't very good and it almost destroyed their business. My point is they aren't going to give away the next generation of Ryzen cpu's like some people here have posted. Why on earth would they sell a 3000 series Ryzen 8 core at the price of a Ryzen 2000 series six core? The answer is thay aren't going to do that.

 

[chrisjames61]

AMD almost went bankrupt by blowing their piggy bank on the Dresden fab and overpaying for ATi. If they hadn't made either of those moves they wouldn't have even sniffed at liquidity (or R&D funding...) problems.

The bigger problem was being non competitive as far as cpu offerings for at least five years.

 

[chrisjames61]

Yes. AMD has published several roadmaps that clearly say that 3000-series APUs are 12nm products.


Historically, the AMD response has been "APU needs both GPU and CPU parts to be ready, and then to be integrated, so this will take longer to release than just a CPU." To me, this does have a bit of the feel of a cop-out, however. They get to make plans years in advance, they should be able to figure it out.

Just because you design something years in advance and "should have it all figured out" doesn't mean the fabrication gets done by magic. In any business whether you are building widgets or cpu's the higher margin products will get higher priority than niche products.

 

[DrMrLordX]

Ryzen 3000 CPU and GPU are 7nm
Will Ryzen 3000 APU be 12nm?
Why are they so much out of sync?

AMD has switched up their CPU development roadmap. From 2013 to 2016, they became an APU company, selling products like Kaveri, Carrizo, and Bristol Ridge as their main product. Now those products are on the back-burner.

AMD's current pattern is to emphasize product in the following order of prorioty:

EPYC (Server) first - Rome is first out of the gate on 7nm
Ryzen (desktop) second - Matisse is second out of the gate on 7nm
Threadripper (HEDT) third - Whatever is their internal codename for the 7nm HEDT will probably follow somewhere in Q3 2019
Ryzen mobile (low-end APU) last - 2019 will see 12nm Zen+ based APUs, under the broad name of Picasso. They will aim at mobile and low-end desktop, with very little emphasis on cheap AM4 solutions.

We probably won't see 7nm APUs until 2020.

This is the reverse of how they used to be pre-Zen, where they were first to market with APUs and launched pretty much nothing elsewhere.

The confusion is that 3000-series APUs will be 12nm Picasso while 3000-series CPUs will be 7nm Matisse. If WCCFTech is to be believed, AMD may put Matisse under the 300-series monicker, so instead of having things like the 3800x or 3850x, we'll have the 380x and 385x. That will leave the 4-digit 3000-series numbers to Picasso.

 

[Kedas]

They can put their Zen2 R3300 50W (6 core 3.2Ghz base, based on non official info) with some GPU in there.
It will have an obvious higher clock than intels 6core laptop with i7-8750H CPU (2.2Ghz base)
Making the need for a 6core APU even lower now.

 

[Panino Manino]

Ryzen 3000 CPU and GPU are 7nm
Will Ryzen 3000 APU be 12nm?
Why are they so much out of sync?

With the chiplets, does this mean that AMD can release a chip with a bunch of different lithographies? 14nm IO + 7nm CPU + 12nm GPU?
"Could" this make sense?

 

[Tuna-Fish]

With the chiplets, does this mean that AMD can release a chip with a bunch of different lithographies? 14nm IO + 7nm CPU + 12nm GPU?
"Could" this make sense?

In principle, yes, but that's not what's happening right now. The next AMD APU line is codenamed Picasso, and some support for it has already landed in Linux, and that makes it very clear that it's just a 12nm respin of Raven Ridge, the same way Pinnacle ridge (12nm CPUs in the 2x00 series) was a 12nm respin of Summit Ridge.

 

[Topweasel]

Ryzen 3000 CPU and GPU are 7nm
Will Ryzen 3000 APU be 12nm?
Why are they so much out of sync?

Because the other two come first. For AMD to do a Zen 2 + Navi APU. They first have to have a foundation to begin developing on. So they can't start developing that CPU till they have the other ones going which will always put it about 6 months to a year behind. I am guessing this maybe the last time that happens though. There is a good chance the Zen2+Navi APU will be chiplet based and if it's not the next one will be. Which will mean in the future the to dies will be agnostic to each other and AMD won't have to wait for after either one is ready to launch a new CPU for the other.

 

[Shivansps]

if they are going for chiplets it may be a capacity problem, they are still using the old 12nm process because they cant get enoght 7nm to serve APU as well.
Specially if they are bringing out the Vega 20 al CES, kinda strange because they just launched the RX590.

But the most likely reason, speacially after seeing the Athlons 220/240 is that AMD just dosent care about APU because Intel has nothing to come closer to the current APUs.

 

[jpiniero]

Specially if they are bringing out the Vega 20 al CES, kinda strange because they just launched the RX590.

If they do launch any Vega 20 (or really any 7 nm dGPUs at CES), it's going to be Frontier Edition expensive; not anything in the price range of the 590.