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 My advice to everyone: treat the old APU rumours as being out of date.

 

[TESKATLIPOKA]

Personally I hope all designs using dGPUs will build on Dragon, with Phoenix mostly left to its iGPU. I also expect both to be announced at CES, though I'd be surprised Phoenix to see any relevant availability in the market before Dragon.

I don't think Phoenix will be limited to only IGP. The reason being that Dragon Ridge will be most likely only 12C and 16C and for gaming even 6-8 cores are enough and that is Phoenix. 12CU RdNA3 will be pretty strong, but nowhere near comparable to a dGPU solution. Even a cutdown 24-28CU N33 should be at least 2x faster.
8C16T Phoenix + N33 will be a pretty decent gaming solution.

 

[moinmoin]

I don't think Phoenix will be limited to only IGP. The reason being that Dragon Ridge will be most likely only 12C and 16C and for gaming even 6-8 cores are enough and that is Phoenix.

Phoenix won't be "limited" to iGPU, but I sure hope the selection of products not including a dGPU will increase significantly. Also I don't think Dragon Range will be limited to above 8 cores, Dragon Range + dGPU are a perfect fit since in that configuration TDP and battery life are usually not considered limiting factors anyway.

 

[LightningZ71]

If current speculation holds true, Drago Range being based on Raphael will give a larger L3, which is important for many games. That leads me to believe that there may be at least 8 core Dragon Range CPUs as they would still have a notable performance benefit over a half-L3 phoenix part. As for when they will actually hit the market? Well, going by AMD's recent track record, they will introduce the two products at CES, then, about 6 months later, a handful of OEM designs for the major business markets will start to appear as well as a small number of super-high end parts from one or two laptop makers that will be in extremely limited volume at astronomical prices. Then, about 9 months later, you'll start to see mainstream notebook parts that were announced in the spring and summer to become available in the normal channels. It'll be about a year from they are announced for products with reasonable pricing for their market segments to become widely available on the market.

As for RDNA3 on Phoenix, please note that, from what we're seeing from driver commits on linux, RDNA3 as it exists on Phoenix will be notably different from the dGPU version, and seems to be more of RDNA2 with expanded and restructured caches and a couple of functional unit tweaks. This makes some sense as it will exist in a highly bandwidth constrained environment, won't have sufficient transistor budget to make some features worth implementing, and will likely be more focused on providing better performance for FSR/XeSS so that it can provide good performance at 1080p/1440p.

 

[scineram]

I don't think Phoenix will be limited to only IGP. The reason being that Dragon Ridge will be most likely only 12C and 16C and for gaming even 6-8 cores are enough and that is Phoenix. 12CU RdNA3 will be pretty strong, but nowhere near comparable to a dGPU solution. Even a cutdown 24-28CU N33 should be at least 2x faster.
8C16T Phoenix + N33 will be a pretty decent gaming solution.

It won't be enough in gaming or competitive at all if it still has pathetic 16MB L3.

 

[TESKATLIPOKA]

Phoenix won't be "limited" to iGPU, but I sure hope the selection of products not including a dGPU will increase significantly. Also I don't think Dragon Range will be limited to above 8 cores, Dragon Range + dGPU are a perfect fit since in that configuration TDP and battery life are usually not considered limiting factors anyway.

It's true that Rembradt without dGPU was mostly the lower TDP one, but Rembrandt laptops weren't that many to begin with. Hopefully It will be much better in this gen.
Is It better to use Phoenix or Dragon Ridge with one missing CCD for an 8C16T gaming laptop? Probably 8C16T Dragon Ridge could be cheaper to produce, but power draw won't be much better than Raptor from 45-88W based on that table from ComputerBase, but that needs to be verified, If It applies or not to the mobile chips.

As for RDNA3 on Phoenix, please note that, from what we're seeing from driver commits on linux, RDNA3 as it exists on Phoenix will be notably different from the dGPU version, and seems to be more of RDNA2 with expanded and restructured caches and a couple of functional unit tweaks. This makes some sense as it will exist in a highly bandwidth constrained environment, won't have sufficient transistor budget to make some features worth implementing, and will likely be more focused on providing better performance for FSR/XeSS so that it can provide good performance at 1080p/1440p.

I though Phoenix IGP is(will be) the same as N33.

It won't be enough in gaming or competitive at all if it still has pathetic 16MB L3.

I think they will finally move to 32 MB. At least I hope so.

 

[moinmoin]

@TESKATLIPOKA The name is Dragon Range.

I'd think Phoenix Point will use the usual denser mobile Zen cores. I personally expect it to be the same as the cores in Zen 4c Bergamo, which is already confirmed to have 16MB L3$.

I'd expect Dragon Range to have a lower idle/base power draw than Raphael, but beyond that and binning the table from ComputerBase should apply.

Regarding the scarcity of Rembrandt laptops, that's quite odd thinking about it. I argued before that like in previous years corporate demand hooked all supply, but then AMD reported mobile units and shares crashing last quarter. So Rembrandt laptops apparently weren't being made in the same quantity as previous gens at all (and AMD officially didn't see that coming)? In that case one can only hope companies are more daring with upcoming laptops using Dragon Range and Phoenix Point.

 

[scineram]

Wtf would they use 4c cores when they want these to be clocked above 5 gigahurtz?

 

[moinmoin]

Wtf would they use 4c cores when they want these to be clocked above 5 gigahurtz?

Phoenix Point won't clock that high, mobile first chips never clock as high as desktop chips. That's what Dragon Range will be for.

 

[TESKATLIPOKA]

@TESKATLIPOKA The name is Dragon Range.

I used the wrong name for some time now. I still like It better than Dragon Range.

 

[scineram]

My dude they will not put 4-GHz cores in mobile. They will go beyond 5.

 

[LightningZ71]

The current 6000 series mobile chips boost to 5+ghz on N6. I don't see them backtracking on that in the slightest. I do see them sharing the same CCX layout between bergamo and phoenix. Its a hallmark ofAMD to reuse as much IP as possible wherever it makes sense. This would be no exception. For lower mid to midrange mobile, 16MB of L3 is both plenty and competitive. For those that want to wring every last FPS that they can from their CPU, there is Dragon Range. For those that just need their laptop to turn on, sip power, and put a picture on the screen while they surf the internet and do office type tasks on web-apps, mendocino covers that market well enough.

 

[Bigos]

No way AMD will use Zen4c, which sacrifices single thread performance for higher density, in a client product. At least not without a non-c core in the same package (and I doubt we will see hybrid cores from AMD anytime soon).

 

[DrMrLordX]

I used the wrong name for some time now. I still like It better than Dragon Range.

They could have called it Dragon Rage. All it would have taken was one typo!

 

[moinmoin]

sacrifices single thread performance for higher density, in a client product

All of AMDs mobile APUs so far did just that. I don't recall people complaining about the frequency difference before, and this time there's Dragon Range as alternate choice.

All it would have taken was one typo!

Maybe it actually was a typo.

 

[Bigos]

All of AMDs mobile APUs so far did just that. I don't recall people complaining about the frequency difference before, and this time there's Dragon Range as alternate choice.

Not really. For example Rembrandt boosts up to 5GHz, which I believe won't be possible with Zen4c due to the use of high density libraries.

 

[uzzi38]

It won't be enough in gaming or competitive at all if it still has pathetic 16MB L3.

Unfortunately that's not changing any time soon... for the most part anyway.

 

[uzzi38]

@TESKATLIPOKA The name is Dragon Range.

I'd think Phoenix Point will use the usual denser mobile Zen cores. I personally expect it to be the same as the cores in Zen 4c Bergamo, which is already confirmed to have 16MB L3$.

I'd expect Dragon Range to have a lower idle/base power draw than Raphael, but beyond that and binning the table from ComputerBase should apply.

Regarding the scarcity of Rembrandt laptops, that's quite odd thinking about it. I argued before that like in previous years corporate demand hooked all supply, but then AMD reported mobile units and shares crashing last quarter. So Rembrandt laptops apparently weren't being made in the same quantity as previous gens at all (and AMD officially didn't see that coming)? In that case one can only hope companies are more daring with upcoming laptops using Dragon Range and Phoenix Point.

Zen 4c vs Zen 4 is a bit of a complicated affair. It has it's own v/f characteristics - it doesn't behave like Zen 4 always does. Not going to give too much context but the efficiency cross over point is a bit too low.

As for Dragon Range, if AMD can get Genoa idling around 15W, I'm gonna freak if they can't get Dragon Range idling lower than Raphael does.

 

[naad]

if AMD can get Genoa idling around 15W,

Definitely interested where you've seen that.

 

[BorisTheBlade82]

Me 2. Apart from Binning I would not put too much hope in Dragon Range idle consumption.

 

[DrMrLordX]

Maybe it actually was a typo.

AMD/Gamefreak crossover confirmed!?!?!?

 

[Kaluan]

Do customers for DTR laptops even care about idle that much tho?

What they care about is speed and laptops not burning a whole through their desk while they complete that 3h render in Cinema4D.

With the new media engines in RDNA3, what are the odds we see AMD claim some wild 30 hour video playback battery time for Phoenix designs? 😅

 

[Glo.]

It won't be enough in gaming or competitive at all if it still has pathetic 16MB L3.

The CPU performance should be between Zen 3 desktop and Zen 4 desktop. Closer to Zen 3, but still a little above it.

It'll be fine, even by today standards.

 

[scineram]

It sure is fine if AMD is comfortable with Raptor Lake and M2(3) running circles around them.

 

[DrMrLordX]

It sure is fine if AMD is comfortable with Raptor Lake and M2(3) running circles around them.

You mean the 13900HX? Because nothing else Intel has in mobile is actually Raptor Lake.

 

[Glo.]

It sure is fine if AMD is comfortable with Raptor Lake and M2(3) running circles around them.

Right, and what Intel has in mobile that even has comparable iGPU performance, or in the case of Phoenix - will have?