Question AMD Phoenix/Zen 4 APU Speculation and Discussion

[uzzi38]

I can finally make this thread.

https://twitter.com/x/status/1519669375283957760

Phoenix is indeed RDNA3. My advice to everyone: treat the old APU rumours as being out of date.

 

[igor_kavinski]

AMD Ryzen 8000G AM5 specs and release date leaks out, Ryzen 5000GT AM4 series allegedly in the works - VideoCardz.com

AMD Ryzen 8000G specs uncovered Iranian media outlet “sakhtafzarmag” claims to have exclusive specs for Ryzen 8000G series. The new desktop series powered by a duo of Phoenix1/Phoenix2 APUs is said to be launching in a few weeks. The site is claiming to have almost the full specs for the new APU...

videocardz.com

Performance improvement for the top part looks great.

 

[FlameTail]

Is not TSMC 6nm cheaper than 7nm?

 

[Joe NYC]

Does Cezanne cost less than Rembrandt?

Also DDR4 vs DDR5

 

[Shivansps]

Why is AMD still cranking out Cezanne in 2024?

1) Because there is demand for a cheap APU with decent IGP
2) They are not launching AM5 Athlons, they need to cover the low end with something.
3) The AM5 APUs are possibly very expensive.

 

[igor_kavinski]

Is not TSMC 6nm cheaper than 7nm?

Why should it be? If 7nm were expensive, AMD wouldn't be releasing additional SKUs on AM4. And all Zen 3 production would likely be halted.

 

[FlameTail]

IIRC according to Semianalysis, TSMC N4 is said to be cheaper than N5.

 

[Abwx]

1) Because there is demand for a cheap APU with decent IGP
2) They are not launching AM5 Athlons, they need to cover the low end with something.
3) The AM5 APUs are possibly very expensive.

Only the 8700G/8600G will be expensive, these should cost 300/200$ respectively, the 8500G/8300G should be at 150/110$.

8300G is the AM5 equivalent of an Athlon 3000G, at 2x the price though, but also at roughly 3x the CPU perfs and eventually 3x the GPU perfs.

 

[qmech]

Why should it be? If 7nm were expensive, AMD wouldn't be releasing additional SKUs on AM4. And all Zen 3 production would likely be halted.

There was a report at one point that TSMC was trying to move customers off N7 by providing very good prices on N6, a node that retained full IP compatibility with N7 (unlike N7+). The expectation was that the relatively inexpensive nature of the switch and the 18% higher density of N6 would make N6 worth it for most.

If Cezanne is a low-volume chip, even the *relatively* inexpensive switch might just make it not worthwhile. TSMC might also have N7 capacity to spare (N6 uses EUV), enabling AMD to get a better deal on that slightly older node. The performance and power characteristics are the same for N7 and N6, so the *only* issue is cost.

 

[jpiniero]

Only the 8700G/8600G will be expensive, these should cost 300/200$ respectively, the 8500G/8300G should be at 150/110$.

As I said before, since Raphael is cheaper to make there is no reason to think Big Phoenix will be cheaper than it.

 

[jpiniero]

If Cezanne is a low-volume chip

Cezanne is the exact opposite of low volume. Even now it's still the vast majority of AMD's laptop sales. That might only change several months from now.

 

[Abwx]

As I said before, since Raphael is cheaper to make there is no reason to think Big Phoenix will be cheaper than it.

By the time these APUs are released 7700/7700X prices will be down, they are currently overpriced in respect of a 7800X3D wich cost only 10-15% more at 360€ in Germany, 7600 is already at 196€ and 7500F at 180€, so the trend for lower pricings is already on motion.

 

[eek2121]

Why should it be? If 7nm were expensive, AMD wouldn't be releasing additional SKUs on AM4. And all Zen 3 production would likely be halted.

Machine time among other things.

Yes, TSMC is (or was? I don't know if things have changed or not) cheaper.

EDIT: Even with the price difference, AMD probably determined that Cezanne cost less, and Cezanne isn't a bad chip. I have 2 machines with it.

 

[moinmoin]

Cezanne is the exact opposite of low volume. Even now it's still the vast majority of AMD's laptop sales. That might only change several months from now.

Yes, Cezanne is currently still a rather big part of AMD's mobile pie. But that pie likely is still siginificantly dwarfed by both Intel's and especially Apple's and others' mobile pies.

 

[Joe NYC]

Yes, Cezanne is currently still a rather big part of AMD's mobile pie. But that pie likely is still siginificantly dwarfed by both Intel's and especially Apple's and others' mobile pies.

Isn't AMD and Apple similar in laptop market share?

 

[moinmoin]

Isn't AMD and Apple similar in laptop market share?

Mobile includes phones and tablets. That's why I wrote "especially Apple's".

 

[LightningZ71]

One thing that really cranks my gears:
AMD's Cezanne part uses a VEGA iGPU.
AMD has announced that they are discontinuing support for VEGA in ALL forms, save for critical security updates.
AMD, after making that announcement, has later announced NEW SKUs for Cezanne.

AMD is about to produce, as a new product, a part that they have ZERO intentions of providing ANY real support for going forward at day one, save for critical security updates. I realize that VEGA is a (now) old gpu architecture, but if you are going to produce new models of parts with it, you should be providing "game ready" drivers for new games for it.

This is absolutely anti-consumer. I get ending support two or more years after you release a part. I am absolutely against not providing anything but the absolutely minimal amount of support on day-one of a new product. I have criticized Intel endlessly for how they handled their licensed GPUs in older atoms, and the Kady Lake G parts. AMD deserves the same level of derision for this decision.

 

[Schmide]

One thing that really cranks my gears:
AMD's Cezanne part uses a VEGA iGPU.
AMD has announced that they are discontinuing support for VEGA in ALL forms, save for critical security updates.
AMD, after making that announcement, has later announced NEW SKUs for Cezanne.

They are not discontinuing support in all forms and it isn't legacy status.

https://www.anandtech.com/show/21126/amd-reduces-ongoing-driver-support-for-polaris-and-vega-gpus

At some point the old hardware is unable to support, or at the very least lend itself to, new functionality.

It's not great but not that bad.

 

[Joe NYC]

Mobile includes phones and tablets. That's why I wrote "especially Apple's".

I see. I was thinking just laptops.

 

[moinmoin]

This is absolutely anti-consumer.

More like anti-Windows actually. Under Linux graphics driver support for older Radeon chips is actually excellent, often improving far beyond support EOL under Windows.

 

[Joe NYC]

Raphael is probably not cheaper to make. Binning/Validating/packaging 2 chiplets costs more than ddoing it with one. It's [71mm N5 + 122mm N6] vs [178mm N4], so probably the same cost.

I think it is the opposite. Validating (or just getting it done on time) is easier with smaller chips with more limited functionality. And additionally, there is some reuse or the chiplets.

For example, reusing Zen 4 IO die in Zen 5 is improving time to market, and validating only Zen 5 CCD chiplet alone is going to be less complex.

N6 is less expensive than N4 and the cost difference will only grow with more as the leading node advances. N6 is going to be it for I/O and caches for foreseeable future.

 

[Shivansps]

One thing that really cranks my gears:
AMD's Cezanne part uses a VEGA iGPU.
AMD has announced that they are discontinuing support for VEGA in ALL forms, save for critical security updates.
AMD, after making that announcement, has later announced NEW SKUs for Cezanne.

AMD is about to produce, as a new product, a part that they have ZERO intentions of providing ANY real support for going forward at day one, save for critical security updates. I realize that VEGA is a (now) old gpu architecture, but if you are going to produce new models of parts with it, you should be providing "game ready" drivers for new games for it.

This is absolutely anti-consumer. I get ending support two or more years after you release a part. I am absolutely against not providing anything but the absolutely minimal amount of support on day-one of a new product. I have criticized Intel endlessly for how they handled their licensed GPUs in older atoms, and the Kady Lake G parts. AMD deserves the same level of derision for this decision.

They did the exact same thing when they launched the FM2 APUs A6-7480 and A8-7680 they launched with legacy gpu drivers. Unless im miss remembering, they definately did it on FM2, they were selling APU with igps that they had already said they were not longer supporting.

 

[Shivansps]

Only the 8700G/8600G will be expensive, these should cost 300/200$ respectively, the 8500G/8300G should be at 150/110$.

8300G is the AM5 equivalent of an Athlon 3000G, at 2x the price though, but also at roughly 3x the CPU perfs and eventually 3x the GPU perfs.

Those prices arent that bad, everything is about $50 more expensive than it should be, a little less for p2. The dissapointing part is how only the 8700G will provide a real jump in igp performance after 6 years of Vega and a memory upgrade.

The 8300G will be a really good APU IF it is $100-$110 AND it has the full 4CU. That would be the only case were AMD gives you more perf for less money. Not to mention R3 APU are non-existant after the 3200G, they never really launched the 4300G or 5300G to the open market. I hope this is the case.

 

[Panino Manino]

It's so frustrating that consumers never have what they want.
These APUs with small GPU are a waste. AMD could design one with only C cores giving space on the die to more CUs, but they'll never do this.

The APU of dreams will always remain a dream.

 

[hemedans]

Those prices arent that bad, everything is about $50 more expensive than it should be, a little less for p2. The dissapointing part is how only the 8700G will provide a real jump in igp performance after 6 years of Vega and a memory upgrade.

The 8300G will be a really good APU IF it is $100-$110 AND it has the full 4CU. That would be the only case were AMD gives you more perf for less money. Not to mention R3 APU are non-existant after the 3200G, they never really launched the 4300G or 5300G to the open market. I hope this is the case.

Radeon 760M in 8600G isn't bad , it has 8CU and should provide significant boost over Vega igp.

 

[Shivansps]

Radeon 760M in 8600G isn't bad , it has 8CU and should provide significant boost over Vega igp.

oh its 8?I trought it was 6. ok yeah 8 could work.