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 My advice to everyone: treat the old APU rumours as being out of date.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

to me it's a ridiculous notion to release a 2 thread CPU for general purpose consumer hardware in 2020, even if it was zen3, but these were zen1 and single channel... smh

So true. My 12700K is unable to fully load the Win11 desktop with only two P-cores enabled. I gave it 10 minutes but the UI was still not functional and Task Manager showed 100% CPU usage the entire time.

 

[Shivansps]

Pollock were intended to compite in the same markets as the Intel Gemini Lake. And if you ever used one, Pollock is probably miles ahead.

Now Mendocino should take that place vs ADL-N, but it is very hard for AMD to enter this market for some reason. Rumor says they are going to try with a ARM-based soc next time.

 

[maddie]

So true. My 12700K is unable to fully load the Win11 desktop with only two P-cores enabled. I gave it 10 minutes but the UI was still not functional and Task Manager showed 100% CPU usage the entire time.

2 P cores should be at worst equal to a Ryzen 3 1300X or 3200G. Do these also become unusable with Win 11?

 

[Shivansps]

2 P cores should be at worst equal to a Ryzen 3 1300X or 3200G. Do these also become unusable with Win 11?

no, the 3000G works fine. Even the Celeron G5925 works ok. The only ones with problems are the Intel small cores older than ADL-N.

 

[maddie]

So true. My 12700K is unable to fully load the Win11 desktop with only two P-cores enabled. I gave it 10 minutes but the UI was still not functional and Task Manager showed 100% CPU usage the entire time.

2 P cores should be at worst equal to a Ryzen 3 1300X or 3200G. Do these also become unusable with Win 11?

no, the 3000G works fine. Even the Celeron G5925 works ok. The only ones with problems are the Intel small cores older than ADL-N.

Something else is happening then, not the 2 P cores performance.

 

[igor_kavinski]

Something else is happening then, not the 2 P cores performance.

Three likes and I start a new 2-core Win11 load-to-desktop performance benchmarking thread!

P.S. my 12700K is not being actively used and I'm too lazy to hook it up so the best I'll be able to test is an HP i7-10750H gaming laptop.

 

[jpiniero]

As i said it i think most of the problems with Phoenix 2 is that they did the wrong cutdowns, like they made it a 6 core when phoenix 1 already covers that.

The whole point of Little Phoenix was to replace Big Phoenix Ryzen 5 with something AMD could hit the price point needed with OEMs and still make money.

Will say it appears to have been successful on that point. Moving on from Cezanne isn't going to be that quick but (for instance) most of the new models at Lenovo are Phoenix based...

 

[igor_kavinski]

2 P cores should be at worst equal to a Ryzen 3 1300X or 3200G. Do these also become unusable with Win 11?

I think the problem is not the IPC but the onslaught of threads that Win11 throws during boot to desktop. Two cores is just not enough to make short work of them quickly enough.

 

[maddie]

I think the problem is not the IPC but the onslaught of threads that Win11 throws during boot to desktop. Two cores is just not enough to make short work of them quickly enough.

Read this

no, the 3000G works fine. Even the Celeron G5925 works ok. The only ones with problems are the Intel small cores older than ADL-N.

The G5925 is a 2 core/2 thread comet lake generation CPU. Something else is wrong with your system.

 

[igor_kavinski]

The G5925 is a 2 core/2 thread comet lake generation CPU.

Would love to know the boot time for that. To be clear, it should show the desktop and Start menu should be ready for launching apps.

 

[Shivansps]

Would love to know the boot time for that. To be clear, it should show the desktop and Start menu should be ready for launching apps.

It is fine, it dosent feel slow when using the OS or web browsing. Its not like a Gemini/Jasper Lake that you can feel them being slow. And even those boot fast.

But a 3000G is better in every possible way, there is not really a reason to use something like that.

 

[coercitiv]

Something else is wrong with your system.

This. I did a quick test on my own 12700K on Win 11, only 2c/2t enabled. From POST to rendering a page in a foreground Firefox tab, the system took about 25 seconds, and that included entering my OS account password. The load on the CPU lowered significantly and went close to idle after about 1m45s, but mind you I have 2 cloud clients doing sync checks after a fresh boot.

@igor_kavinski needs some spring cleaning on his machine.

 

[igor_kavinski]

@igor_kavinski needs some spring cleaning on his machine.

Great. Then it could be the slow RAM. I've an EXPO kit rated for 5600 that doesn't have an XMP profile and best I could get it to boot at on my Z790 ASROCK Sonic mobo is DDR5-4600 CL28. But it's been over a year since I turned my system on so it's possible that a BIOS update may fix the compatibility issue with that kit. And the reason for not turning that system on is that I'm an expat in UAE and don't have a permanent residence and can't afford an actual flat so I keep getting ousted when the person owning the flat leaves the country for good coz naturally, the new person leasing the whole flat wants me to pay more for the room. So my computing is currently just limited to my laptops.

 

[eek2121]

Looks like AMD finally got around to publishing more details about clock speeds for their APUs.

8500G Zen4: 4.1 base/5.0 boost, Zen4c: 3.2 base/3.7 boost

8300g Zen4: 4.0 base/4.9 boost, Zen4c 3.2 base/3.6 boost.

I do wish AMD had given us an 8+16 variant dual CCD chip. It would run circles around Intel’s top offering. Zen4c is significantly faster than *mont cores even with a much lower clock.

 

[SteinFG]

I thought it hits the boost wall at 3.2 and that's why all Siena SKUs don't go above that number. 3.7 boost is actually higher than I expected.

 

[Shivansps]

Looks like AMD finally got around to publishing more details about clock speeds for their APUs.

8500G Zen4: 4.1 base/5.0 boost, Zen4c: 3.2 base/3.7 boost

8300g Zen4: 4.0 base/4.9 boost, Zen4c 3.2 base/3.6 boost.

I do wish AMD had given us an 8+16 variant dual CCD chip. It would run circles around Intel’s top offering. Zen4c is significantly faster than *mont cores even with a much lower clock.

Yes but Zen4C is petty much Zen4 with capped vcore and clocks... it is not small AT ALL, it even has AVX512. It is a big core, you cant really put 8+16.

 

[coercitiv]

It is a big core, you cant really put 8+16.

You can, the 16c Zen 4c CCD is just slightly bigger.

However, the truly stunning thing here is the die size. 16 Zen 4c cores are barely larger than 8 Zen 4 cores. At ISSCC 2023, AMD disclosed Zen 4’s CCD to be 66.3mm². This is the design area without die seal and scribe lines at the edges. Zen 4c’s CCD design area is just 72.7mm², not even 10% bigger! Keep in mind that there are double the cores, double the L2 cache, and the same amount of L3 cache on each die. The cores must have shrunk greatly to fit even more cache per die with only a small area increase.

Source.

 

[Abwx]

I thought it hits the boost wall at 3.2 and that's why all Siena SKUs don't go above that number. 3.7 boost is actually higher than I expected.

It was possible to estimate the boost clock with the CB R23 score/power they once published for the 6C part, this pointed to roughly 3.55GHz, so nothing new here.

AnandTech Forums: Technology, Hardware, Software, and Deals

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

www.anandtech.com

 

[eek2121]

You can, the 16c Zen 4c CCD is just slightly bigger.

Source.

You went further (farther?) than I. I just assumed he couldn't do math. (to be fair, I can't english! )

 

[MadRat]

You'll never see the best options of a design until EOL.

 

[Shivansps]

You can, the 16c Zen 4c CCD is just slightly bigger.

Source.

The point stands, Zen4C is smaller but not that small for that(i think it is 30% less), im not exactly sure what did they cut off but you cant make a 8+16 because it will need more than double the space requiered to fit all those cores.

What they could do is make a 16C Zen4C.

 

[moinmoin]

it will need more than double the space requiered to fit all those cores.

Can't you read? The quoted text states that the 16c Zen 4c CCD is only 10% larger than the 8c Zen 4 CCD.

 

[coercitiv]

The point stands, Zen4C is smaller but not that small for that(i think it is 30% less), im not exactly sure what did they cut off but you cant make a 8+16 because it will need more than double the space requiered to fit all those cores.

Zen 4 chiplet is 66.3mm² design area.
Zen 4C chiplet is 72.7mm² design area.

The Zen 4C CCD is just ~10% larger than Zen 4. To achieve this, everything is tightly packed to improve density. Some features like the TSVs for 3D cache are gone. If we ignore the L2, the core shrank by about 44%. Doubling the core count meant double the L2 as well, but that was compensated by less L3 per core.

 

[Mopetar]

to me it's a ridiculous notion to release a 2 thread CPU for general purpose consumer hardware in 2020, even if it was zen3, but these were zen1 and single channel... smh

It's a salvage part that they'd otherwise throw out. There's probably some overseas market where they'll be used just due to affordability or some company that wants to push out a craptop that Walmart can stick on their shelves to market a $200 PC for the next Black Friday.

Honestly it'd probably be fine to use to run a minimalist Linux install for Grandma to use for web browsing and email. There are a few people who really don't need a four core CPU.

 

[Shivansps]

Can't you read? The quoted text states that the 16c Zen 4c CCD is only 10% larger than the 8c Zen 4 CCD.

Yes but that dosent means you can fit a 8C Zen4 AND 16C Zen4c in there. It would be easier to make it 16C Zen4 and call it a day.