Speculation: Ryzen 3000 series

[amd6502]

It is...

https://twitter.com/x/status/1199354469227487232

40mm x 40mm is package size.
Using pixels it comes around ~110 pixels x ~104 pixels of die within ~347 x ~347 pixels of package size, making the die around ~150 mm2. Which is definitely smaller from Raven0/Raven Ridge's ~210 mm2 die size. Which is the emboss on the IHS.

Should be fine for a lot of the budget notebooks, as long as the users are fine with an iGPU that's a class below Bristol Ridge A-series and the 8 ~ 10CU Picasso based Ryzens.

 

[Shivansps]

To me they are getting rid of a unsuccessful embedded product. A 14nm Raven 2C/4T Vega 3 will look really bad once Renoir hits the market...

2C/4T Vega 3 at $50 to (probably) 4/8 Zen 2 Vega 8 at $100... it may be good for office use but it leaves a huge gap in performance.

And on notebooks... Vega 3 has a chance of getting outperformed by the next gen Atom.

 

[lobz]

Video is interesting!

OK I won't accuse you of trolling just because it's your first post here, but please read at least a bit of the thread before posting ancient and 140x argued-through stuff.

 

[amd6502]

2C/4T Vega 3 at $50 to (probably) 4/8 Zen 2 Vega 8 at $100... it may be good for office use but it leaves a huge gap in performance.

And on notebooks... Vega 3 has a chance of getting outperformed by the next gen Atom.

As long as it plays multimedia smoothly at 60fps+ at the native laptop resolution (or better yet 1080p external monitor) it wouldn't be the least bit relevant.

Low end to midrange notebook gamers will get a Vega 8 model or better anyway.

 

[Thunder 57]

As long as it plays multimedia smoothly at 60fps+ at the native laptop resolution (or better yet 1080p external monitor) it wouldn't be the least bit relevant.

Low end to midrange notebook gamers will get a Vega 8 model or better anyway.

I dunno, even I'll admit Vega 3 is weaksauce. Especially when it's by far the most common option found for Ryzen 3 in laptops. Vega 6 seems more appropriate but it's only in the 3300U I believe. Then you would have Vega 8 for Ryzen 5, and Vega 10 for Ryzen 7. That seems to make more sense IMO. If you want to use a Vega 3, make it an Athlon model.

 

[naukkis]

This 2c version is meant to replace Stoney-Bridge based products on chromebooks etc. And will be a massive upgrade. As it's still competitive against low-end i3 they also sell it as Ryzen3 but that probably wasn't their main plan.

 

[Shivansps]

This 2c version is meant to replace Stoney-Bridge based products on chromebooks etc. And will be a massive upgrade. As it's still competitive against low-end i3 they also sell it as Ryzen3 but that probably wasn't their main plan.

For that is OK, but they are placing it on all kinds of notebooks and on desktop as well...

 

[NostaSeronx]

HP 87B7____ - Geekbench

Benchmark results for a HP 87B7____ with an AMD 3020e with Radeon Graphics processor.

browser.geekbench.com

Dual-core 2c/2t w/ Dali... unknown title...
Athlon Gold 3150u
Athlon Silver 3050u

With that I assume it will fall into...
"Athlon Gold" 3120e 2c/4t 6w
"Athlon Silver" 3020e 2c/2t 6w

14nm at 6w vs 28nm at 15w(earliest)/6w(newest) =>

a6-8500p - Geekbench Search - Geekbench

a4-9120e - Geekbench Search - Geekbench

 

[amd6502]

I wonder if we'll see an 1600 AF version of the 3400g Picasso part (2400G AF). It would be quite nice to see a discounted 3400g priced a little upwards of the 3200g.



Regarding Stoney:

a4-9120e - Geekbench Search - Geekbench

current 6w high bin versus medium bin are both useable:

A4-9120C

2 / 2

1.6 GHz

2.4 GHz

A9-9420e

2 / 2

1.8 GHz

2.7 GHz

And low 15w low bin and 10w medium bin is also useable:

A4-9120

2 / 2

2.2 GHz

2.5 GHz

A6-9200

2 / 2

2 GHz

2.8 GHz

A way to dispose of more low bins would be cTDP 15W-25W bga ITX boards (AM1 successors) at 2.7/3.3GHz


The 10w high bin clocks 0.4GHz higher and is quite nice.

A9-9400

2 / 2

2.4 GHz

3.2 GHz

and is similar to the mass produced popular 15w part:

A6-9225

2 / 2

2.6 GHz

3 GHz

and finally the golden Stoney

A9-9425

2 / 2

3.1 GHz

3.7 GHz

 

[Thunder 57]

I wonder if we'll see an 1600 AF version of the 3400g Picasso part (2400G AF). It would be quite nice to see a discounted 3400g priced a little upwards of the 3200g.



Regarding Stoney:



current 6w high bin versus medium bin are both useable:

A4-9120C

2 / 2

1.6 GHz

2.4 GHz

A9-9420e

2 / 2

1.8 GHz

2.7 GHz

And low 15w low bin and 10w medium bin is also useable:

A4-9120

2 / 2

2.2 GHz

2.5 GHz

A6-9200

2 / 2

2 GHz

2.8 GHz

A way to dispose of more low bins would be cTDP 15W-25W bga ITX boards (AM1 successors) at 2.7/3.3GHz


The 10w high bin clocks 0.4GHz higher and is quite nice.

A9-9400

2 / 2

2.4 GHz

3.2 GHz

and is similar to the mass produced popular 15w part:

A6-9225

2 / 2

2.6 GHz

3 GHz

and finally the golden Stoney

A9-9425

2 / 2

3.1 GHz

3.7 GHz

What is this crap about Stoney? It's dead, done gone, kaput. Why do you and Nostra keep going on about this nonsense? All AMD is pushing is Zen. Get over your fantasies about cheapo construction cores. You can already get an Athlon 3000G for $55. You think AMD is going to waste fab orders on $20 garbage? AMD said themselves they are no longer the "budget brand". Get over your fan fiction, please.

 

[amrnuke]

What is this crap about Stoney? It's dead, done gone, kaput. Why do you and Nostra keep going on about this nonsense? All AMD is pushing is Zen. Get over your fantasies about cheapo construction cores. You can already get an Athlon 3000G for $55. You think AMD is going to waste fab orders on $20 garbage? AMD said themselves they are no longer the "budget brand". Get over your fan fiction, please.

Agree. At Microcenter you can snag a 3000G for $43.

BUT I do think there is some advantage to getting AMD into low end laptops and HP all-in-one and microSFF units for instance, and that would be to grab mindshare both of companies/developers and end-users. AMD might be looking at it as a way to get people more familiar with the brand. And if vendors want to put the chips in well-built solutions, that's a great place to start. Getting HP/Dell and other big vendors to integrate these chips makes sense because those companies have a vested interest in their stuff "just working" in the business world.

 

[NostaSeronx]

What is this crap about Stoney? It's dead, done gone, kaput. Why do you and Nostra keep going on about this nonsense? All AMD is pushing is Zen. Get over your fantasies about cheapo construction cores.

Stoney is on a 28nm node, it is technically better than Raven2/Dali.

Stoney = 125 mm2 / less masks / older, slower technology / cheaper at Dresden
Raven2/Dali = 150 mm2 / more masks / newer, faster technology / expensive at Malta

22fdx port within the same 14LPP(Fin1-FEOL) -> 12LP(Fin2-FEOL)/12LP(Fin2-FEOL+) metrics. The added benefit of that is increased performance, decreased power, and higher yields.

Add a 22fdx port similar to GF28SHP(Kaveri(6x 90nm)) to GF28A(Carrizo(8x 90m)), then Stoney could be dropped straight onto Renoir's platform, since LPDDR4(X) support could be added.

However, 12FDX seems to be the focus right now. With a new module architecture:
- Split-cache L2(definitely)/L3(maybe) hierarchy (Instruction and Data caches)
- FPU moved out of central sub-module and into a vector sub-module; central sub-module, vector sub-module, L2 hub sub-module all have their own clocks and IVR/power.

 

[amrnuke]

Stoney is a 28nm node, it is technically better than Raven2/Dali.

Stoney = 125 mm2 / less masks / older, slower technology / cheaper at Dresden
Raven2/Dali = 150 mm2 / more masks / newer, faster technology / expensive at Malta

You'll have to explain that further.

 

[NostaSeronx]

You'll have to explain that further.

Raven2/Dali has faster intellectual property.

Stoney => 35.2 GB/s(max) northbridge bandwidth.
Carrizo => 41.6 GB/s(max) northbridge bandwidth.
Raven/Picasso/Raven2/Dali => 51.2 GB/s(max) northbridge bandwidth, potentially at lower latency.

Raven2/Dali has 128-bit DDR4 support, while Stoney only has 64-bit DDR4 support.

The distance between Raven2/Dali and Stoney is easily fixable by making a new SoC on 22FDX. With 12FDX the benefits are larger since density/power/ams favor FDSOI over FinFET.

Most of everything in RV2/Dali(ignoring CPU IP(Zen arch)) can be ported up(22FDX) or across(12FDX) with minimal cost. Leading to a Stoney that achieves better performance/power than Rv2/Dali at lower cost. With a larger fab source: Fab1 Dreden(~90,000 wpm(planned 22FDX cap)) vs Fab8 Malta(~40,000 wpm(current FinFET cap)). More wafers per month means that GloFo sells at a instrinically lower cost.

Fitting Zen/Ryzen into the low budget market is bad, but continuing the budget Excavator through lower budget is good.

Cost to build and return from customer purchase makes the Ryzen 3 3250U(Dali) an awful thing to sell compared to any Ryzen x 4xxxu(Renoir).

 

[moinmoin]

Raven2/Dali has faster intellectual property.

Stoney => 35.2 GB/s(max) northbridge bandwidth.
Carrizo => 41.6 GB/s(max) northbridge bandwidth.
Raven/Picasso/Raven2/Dali => 51.2 GB/s(max) northbridge bandwidth, potentially at lower latency.

Raven2/Dali has 128-bit DDR4 support, while Stoney only has 64-bit DDR4 support.

The distance between Raven2/Dali and Stoney is easily fixable by making a new SoC on 22FDX. With 12FDX the benefits are larger since density/power/ams favor FDSOI over FinFET.

Most of everything in RV2/Dali(ignoring CPU IP(Zen arch)) can be ported up(22FDX) or across(12FDX) with minimal cost. Leading to a Stoney that achieves better performance/power than Rv2/Dali at lower cost. With a larger fab source: Fab1 Dreden(~90,000 wpm(planned 22FDX cap)) vs Fab8 Malta(~40,000 wpm(current FinFET cap)). More wafers per month means that GloFo sells at a instrinically lower cost.

Fitting Zen/Ryzen into the low budget market is bad, but continuing the budget Excavator through lower budget is good.

This reminds me of my personal semiconductor fairy tale fantasy: Imagine if you could put the SNES' Ricoh 5A22 CPU (or pick the CPU of your favorite console/home computer from the 80s/90s) on the latest process node! Oh so tiny, that makes it extra cheap to manufacture! And the power usage, it likely could run on a tiny button cell, oh so efficient!

But the era is gone. The market is gone, it moved onto different products. So the fantasy won't ever happen. Same with your fantasy Stoney.

 

[NostaSeronx]

Same with your fantasy Stoney.

Do realize the new product = new name. Also, new name comes with newer revolutions in CPU/GPU/NB, etc. While expecting 100% Excavator going to 22FDX/12FDX is fantasy. Expecting TAGE, new Int cores, new FPU, overhauled L2, etc isn't.

No Raven2/Dali part satisfies the low cost/low power market like Stoney does. Which means Stoney is HVM/lower price, Dali is LVM/higher price in the same market. Anything else from AMD is an enormous loss especially with that Ryzen 3 4300u salvaged die.

Raven -> Renoir
Bristol -> Raven2/Dali
Renoir and Raven2/Dali compete being of same die size 150mm2. Higher productivity at TSMC compared to GloFo also means 7nm is globally cheaper than 14nm.

Stoney -> nothing yet
Anything being a true successor(primarily smaller die than 125mm2) from Fam15h/Stoney is better than any Fam17h/Raven2/Dali part.

 

[moinmoin]

No Raven2/Dali part satisfies the low cost/low power market like Stoney does.

AMD is no longer interested in the low cost market. So all the low cost market is going to get is scraps that didn't bin into higher margin products and would otherwise be thrown away altogether, so no loss there for AMD.

 

[NostaSeronx]

AMD is no longer interested in the low cost market.

Missing an asterisk saying for Zen products. Ryzen isn't allowed to be the budget alternative. Any other family can be budget, but not Zen. Athlon Zen and Ryzen Raven2/Dali are budget parts that have lower return than even Stoney.

 

[moinmoin]

Missing an asterisk saying for Zen products. Ryzen isn't allowed to be the budget alternative. Any other family can be budget, but not Zen. Athlon Zen and Ryzen Raven2/Dali are budget parts that have lower return than even Stoney.

On desktop, maybe. The higher margin markets for it are embedded and mobile.

 

[amrnuke]

Raven2/Dali has faster intellectual property.

Stoney => 35.2 GB/s(max) northbridge bandwidth.
Carrizo => 41.6 GB/s(max) northbridge bandwidth.
Raven/Picasso/Raven2/Dali => 51.2 GB/s(max) northbridge bandwidth, potentially at lower latency.

Raven2/Dali has 128-bit DDR4 support, while Stoney only has 64-bit DDR4 support.

The distance between Raven2/Dali and Stoney is easily fixable by making a new SoC on 22FDX. With 12FDX the benefits are larger since density/power/ams favor FDSOI over FinFET.

Most of everything in RV2/Dali(ignoring CPU IP(Zen arch)) can be ported up(22FDX) or across(12FDX) with minimal cost. Leading to a Stoney that achieves better performance/power than Rv2/Dali at lower cost. With a larger fab source: Fab1 Dreden(~90,000 wpm(planned 22FDX cap)) vs Fab8 Malta(~40,000 wpm(current FinFET cap)). More wafers per month means that GloFo sells at a instrinically lower cost.

Fitting Zen/Ryzen into the low budget market is bad, but continuing the budget Excavator through lower budget is good.

Cost to build and return from customer purchase makes the Ryzen 3 3250U(Dali) an awful thing to sell compared to any Ryzen x 4xxxu(Renoir).

Why would AMD want to invest any money into porting Dali features over to Stoney? Is there some massive high-margin untapped market there?

 

[NostaSeronx]

Why would AMD want to invest any money into porting Dali features over to Stoney? Is there some massive high-margin untapped market there?

There is a big market, yes.

Raven2/Dali on 14nm = 150 mm2
Stoney on 28nm = 125 mm2

Getting rid of negatives of the 28nm node with a transition is ideal.

Stoney's successor would thus be on 22FDX/12FDX have a sub-cm2 die w/ high production quantity compared to Raven2/Dali.

Features gained but aren't included with Raven2/Dali:
- Higher CPU GHz(Excavator is on 28nm, all nodes in 14nm generation are faster, so clock rate intrinsically goes up)
- Access to 64-bit LPDDR4X, and do to timing of launch possible upgradability to LPDDR5 later. (Super-Combo Interface: DDR54/LPDDR54X)
- Smaller die size, higher production capacity(intrinsically lower costs/maturer production line), higher yields, etc.
- 125*C operation support, w/ 150*C 12FDX-HiTemp later on.
- Space/HighTude operability
- Support of a true FIVR do to planar node, reducing mobo costs.
- Longer supported node increasing 10 year support to the absurd 25 year support.

IoT/SBC -> Embedded -> Mission-critical -> Laptop -> AiO -> Credit-card NUCs

However, Stoney on 14LPP-derivatives/12FDX-derivatives => not the goal
Something that does low-cost better and doesn't reduce the latest Zen's cost is preferable.
It also needs to be unrushed, allowing for extended optimization over time. For 22FDX, 22FDX+, 22FDX++, 22FDX#, 22FDX#+, 22FDX#++, 22FDX##, etc. Which might included upgraded substrates(Al2O3), gates(Full metal gate), BEOL(carbon doped copper, CNTs, Nanosheet copper, etc), process(new machines), etc.

 

[amd6502]

If Raven2 does 6W well there may not be need. Pay higher production cost but no development cost. How well does it do 6W?

Get over your fantasies about cheapo construction cores. You can already get an Athlon 3000G for $55. You think AMD is going to waste fab orders on $20 garbage? AMD said themselves they are no longer the "budget brand". Get over your fan fiction, please.

It's a ULP is big market, even the lower end. You have 2-in-1's, sub $150 emmc, $200 chromebook competitors, and low end laptops ($230-300). Stoney along with Raven2 addresses this price range pretty well, but I don't know about battery life and 6W performance, compared to similarly priced products.

 

[Shivansps]

Is something wrong with Picasso production? im just been told that there is not more 3400Gs for maybe months here.

 

[moinmoin]

Is something wrong with Picasso production? im just been told that there is not more 3400Gs for maybe months here.

Still seems to be rather readily available here in Germany (though there are some shops waiting for supply, 15 out of 33 on geizhals.de).

 

[Ajay]

Is something wrong with Picasso production? im just been told that there is not more 3400Gs for maybe months here.

Apparenlty, there are serious disruptions facing many technology manufacturing business due to the novel Corona Virus. Chinese manufacturers have extended their lunar new year holiday shutdowns while the government assesses risk levels. I've seen hints that a major rollout for a top level semicon producer has been delayed because of it (that would be Intel, AMD or Nvidia).

All that to say, I wouldn't be surprised if the 3400G has been impacted (probably relies on China for part of the supply chain).