Question AMD Phoenix/Zen 4 APU Speculation and Discussion

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

ahimsa42

Senior member
Jul 16, 2016
225
194
116
i was planning on getting a 6800U in 2022 but with this announcement will likely instead wait until early 2023 to see how phoenix laptops shake out. i think that perhaps waiting for zen4 & RDNA3 may be the best course at this point. of course, as always it all comes down to price & availability. hopefully there will be more phoenix details forthcoming from AMD at Computex later this month.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Given typical AMD mobile APU rollout availabilities, you wouldn’t actually find more than a token few x800u laptops before the end of the year, and they’ll be priced higher than better performing systems that have dGPUs until next year anyway.
 

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
hopefully there will be more phoenix details forthcoming from AMD at Computex later this month.
I doubt it, it's surprising that they acknowledged a future APU at all with Rembrandt so soon out of the door and lacking wide availability across market segments.

Unless Phoenix fills out the very lowest end of Rembrandt's market like it's planned exclusively for ULV laptops and embedded.

I'd expect no more than general Zen4 and RDNA3 details at Computex lacking even SKU info to give Intel and nVidia to plan against - IPC, perf/watt efficiency increases and all that.
 

ahimsa42

Senior member
Jul 16, 2016
225
194
116

wonder how long it will be until phoenix actually hit the marketplace-perhaps june 2023 at the earliest?
 

ahimsa42

Senior member
Jul 16, 2016
225
194
116
Raven Ridge: Feb 2018
Picasso: Jan 2019
Renoir: Jan 2020
Cezanne: Jan 2021
Rembrandt: Jan 2022
Phoenix: ???

Probably Jan 2023, but hey you never know. Actual availability is another matter.
exactly-actual availabilty after the paper lauch seems to be around 6 months later
 

Timorous

Golden Member
Oct 27, 2008
1,532
2,535
136
Since AMD sell APUs with the GPU turned off / defective is it possible for them to go the other way and sell parts with broken CPUs as a dGPU? If this is really around 3060m 60W tier performance that could make for a good 7400 tier dGPU, especially if they could make it such that when the CPU is fused off that L3 cache becomes Infinity Cache. It would save AMD designing, and building an N34 so might work out as a cheaper way to service that tier.
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
They may be skipping RMB on AM5, it may not be a bad idea at this point, with dgpus stocks recovering, AM4 sticking around for at least another year due to DDR5 prices and AM5 cpus having a basic IGP. Going directly to Phoenix for AM5 is not a bad idea.

We are 3 months away from the date the Cezanne launched to the public and have had a good availity ever since. There should be an announcement soon, otherwise it may not be coming.
 

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
I don't think they are skipping RMB on AM5, rather budget builds using AM5 are far off for obvious reasons (new platform, DDR5 priceup) so it makes little sense to launch budget parts that can be sold with higher margins elsewhere (like in laptops). Same may still apply for Phoenix as well. But salvaged parts of both with the iGPU fused off are likely to appear down the road anyway as those have no use in mobile parts.

Since AMD sell APUs with the GPU turned off / defective is it possible for them to go the other way and sell parts with broken CPUs as a dGPU? If this is really around 3060m 60W tier performance that could make for a good 7400 tier dGPU, especially if they could make it such that when the CPU is fused off that L3 cache becomes Infinity Cache. It would save AMD designing, and building an N34 so might work out as a cheaper way to service that tier.
Very very unlikely. For that the monolithic APUs would need contain an external interface to connect to the iGPU part through PCIe, such I/O interfaces take up plenty space and just aren't necessary for internally connected iGPUs. This may change once APUs go MCM as well.
 
  • Like
Reactions: Tlh97 and yuri69

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
Oh thank god RGT finally got it right

(2) Zen 4 Phoenix Is RIDICULOUS - Specs & Performance Leaks | Zen 5 Uses 5nm - YouTube

Most of the video is about Phoenix, which is cool, and if someone can post the summary that'd be great because I can't, but yes, Zen 5 uses a variant of N4.
I got you. Let me summarize it here.

Phoenix:
- Reiterated that Phoenix is designed to put pressure on the low-end GPU market
- 6 WGP for Phoenix iGPU
- >3 GHz for iGPU
- 2023 CES debut
- 8 Zen 4 cores w/ 16 MB L3 cache
- Doubled L2 cache, presumably no Infinity Cache
- How will the iGPU achieve its performance targets without Infinity Cache? Apparently, it will be due to high clocked memory (8500 MT/s) using a really good memory controller

1652119948561.png
1652119961247.png
1652120075939.png

Zen 5:
- Zen 5 and RDNA4 uses TSMC N4 (again, highly customized by AMD)

TSMC Updates:
- AMD intends to pay 6.5B to TSMC/GloFo for the rest of the year's chip capacity
 

desrever

Member
Nov 6, 2021
108
262
106
Theres no way the OEMs will put 8500MT memory on their laptops just cause AMD designs for it. If AMD is designing it for that, I hope they have some partnered designs that actually come out with that. APUs are so memory limited at some point they will need infinity cache.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
I doubt 16C is likely for a gaming APU while game consoles remain 8C max for the PS5/XSX generation.

IMO if Dragon Range differs significantly from the Phoenix hardware spec beyond LPDDR5 -> DDR5 support I imagine that it would be in adding stacked Infinity cache and/or greater GPU WGP count.

The former should be especially helpful for APUs sharing system memory and alleviating the bandwidth constraints this causes.

The LPDDR5 limitation of Phoenix noted on the slide makes me think it is possibly destined for a standalone VR HMD at some point, or at the very least being marketed to attract interest to this end.

I know this was posted a while ago, but I haven't had time to respond. There are console refreshes scheduled (according to rumors from a solid source) for next year. Unsure as to the extent of the refresh. I do not expect them to be a simple shrink. It is much more likely we will see updated hardware. Regardless, they likely won't have more than 8 cores, but I did want to bring this up.

Realistically AMD is going to have to go beyond 8 cores on mobile in order to maintain performance leadership on mobile.

As for anyone asking about "U" series chips. I'm 100% expecting 15W chips to disappear. You can keep a 30-35W chip cool on a thin/light laptop. Look at the Macbook Air. 15W just isn't enough of a budget for 8 wide cores, insanely fast IO, and a decent GPU block.

EDIT:

For the record, my uninformed speculation is that the console refreshes will use a custom spin of Rembrandt. It would allow the consoles to run faster and cooler while providing more compute power.
 
Last edited:

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
Theres no way the OEMs will put 8500MT memory on their laptops just cause AMD designs for it. If AMD is designing it for that, I hope they have some partnered designs that actually come out with that. APUs are so memory limited at some point they will need infinity cache.

I'm still waiting on the rumored DDR5 12000+ modules that are in development.
 

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
Theres no way the OEMs will put 8500MT memory on their laptops just cause AMD designs for it. If AMD is designing it for that, I hope they have some partnered designs that actually come out with that. APUs are so memory limited at some point they will need infinity cache.
Considering how much of the eventual performance depends on the actual build (not only selection of memory but also allowed and achievable TDP available to CPUs and GPUs) it's overdue that Intel, AMD and Nvidia enforce some transparency to consumers by the OEMs regarding performance and power usage for any given laptop built and sold. Stuff like Nvidia's Max-Q went the opposite extreme in opaqueness.
 

jamescox

Senior member
Nov 11, 2009
637
1,103
136
Theres no way the OEMs will put 8500MT memory on their laptops just cause AMD designs for it. If AMD is designing it for that, I hope they have some partnered designs that actually come out with that. APUs are so memory limited at some point they will need infinity cache.
Is it possible that they mounted the DDR5 chips on the cpu package like Apple? That would easily be able to run at very high clock.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Notably, in certain cases, the 680m iGPU on Rembrandt can already exceed the oerformance of the 6500xt, 3050, and essentially any of the 4GB dGPUs due to being able to allocate and use more than 4GB of RAM for VRAM. So, saying that it will exceed those isn't a huge stretch. The 3060m maxQ is a curious beast, as the 3050ti can usually outperform it in low ram use situations already. So, the claims are believable from a VRAM point of view. From a GPU throughput view, a 3ghz+ 6WGP iGPU made on RDNA3 ip should be capable of keeping up with those parts well enough.

The question remains, will the raw VRAM bandwidth of such an iGPU be enough for consistent frame rates at those averages?
 

MadRat

Lifer
Oct 14, 1999
11,909
229
106
Consoles are all purpose-built soldered memory configurations, so you never know what they may do. Would probably be something you wouldn't want to fail, so more likely to be conservatively robust by design. The best part of purpose-built is that you can build tighter specifications for the layout, keeping trace lengths tight. And because of economy of scale, you can build something more sophisticated at a cheaper final cost. I don't think with $1,000 phones being the norm, that a $1,000 console with 8K capability is out of realistic expectations.

Sony should steal Apple's playbook and offer $1,000 PS6 baseline and start with 'deluxe' models launched between $1,200-1,500 in various limited editions. Offer a supercharged version at $2,000 within six (6) months. Tease the supercharged model from the beginning to break sticker shock. If its really special, especially when it comes to 8k performance and load times, you goad your competitors to compete.
 
Last edited:

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
So this means that Dragon Range will have more than 16 Cores right?


View attachment 61294


More than 16 Zen 4 cores? Unlikely. Heck, I'm not even sure how they'll do 16 cores but if I had to guess, it's a larger monolithic die with 16 cores.

- Highest core count? Yes, if you only include big cores. Alderlake-HX still only has 8 P cores.
- Highest thread count? Yes, if it's 16C/32T.
- Highest cache? 2x16 MB ought to do it.
 

uzzi38

Platinum Member
Oct 16, 2019
2,565
5,575
146
More than 16 Zen 4 cores? Unlikely. Heck, I'm not even sure how they'll do 16 cores but if I had to guess, it's a larger monolithic die with 16 cores.

- Highest core count? Yes, if you only include big cores. Alderlake-HX still only has 8 P cores.
- Highest thread count? Yes, if it's 16C/32T.
- Highest cache? 2x16 MB ought to do it.
Hm? No, it's probably just the formerly rumoured Raphael-H.
 
  • Like
Reactions: Tlh97 and moinmoin

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
More than 16 Zen 4 cores? Unlikely. Heck, I'm not even sure how they'll do 16 cores but if I had to guess, it's a larger monolithic die with 16 cores.

- Highest core count? Yes, if you only include big cores. Alderlake-HX still only has 8 P cores.
- Highest thread count? Yes, if it's 16C/32T.
- Highest cache? 2x16 MB ought to do it.

Who said it was monolithic? Also, that does imply > 16c, since Intel has a '16 core' gaming CPU.

People keep saying chiplets on mobile can't be done. Adaptive IF and other power optimizations would allow chiplets to happen easily.

EDIT: That slide surely can't be from AMD. It has a misspelling.
 

SteinFG

Senior member
Dec 29, 2021
400
454
106
"Highest core count for a mobile cpu" does not imply >16c, it just implies 16c. AMD made that phrase very carefully, read it again ;)