AMD's Octa-Core 4nm Phoenix Point Zen 4 APU Breaks Cover

Ryzen CPU
Ryzen CPU (Image credit: AMD)

AMD recently shared a roadmap of its desktop and mobile client processors at Financial Analyst Day 2022. However, it didn't take long for hardware leaker Tum_Apisak to dig up one of the chipmaker's upcoming processors, presumably Phoenix Point.

Phoenix Point is the successor to AMD's Ryzen 6000 (Rembrandt) mobile APUs. Although AMD didn't reveal the exact time frame for Phoenix Point's release, the next-generation chips should hit the market next year. Phoenix Point will see AMD's jump from TSMC's 6nm manufacturing process to the cutting-edge 4nm manufacturing process. The latter offers a 6% smaller die area and will empower chipmakers to optimize their chips' performance and power consumption.

Besides the transition to a smaller process node, Phoenix Point will also wield AMD's mighty Zen 4 cores, which reportedly bring a 10% IPC uplift and more than 35% higher overall performance. However, the more compelling trait will be the APU's integrated graphics. AMD will incorporate its RDNA 3 graphics into Phoenix Point, potentially allowing the APUs to compete with entry-to mid-range discrete mobile graphics. That's big news, considering Phoenix Point competes in the 35W to 45W category, powering thin and light gaming laptops.

As expected, Phoenix Point arrives with support for PCIe 5.0 and DDR5 memory, specifically LPDDR5. Nonetheless, AMD may extend memory support to include other formats in the future. In addition, with Phoenix Point, we'll see the first inclusion of AMD's Artificial Intelligence Engine (AIE), an IP the chipmaker obtained after absorbing Xilinx for $54 billion.

AMD Eng Sample: 100-000000709-23_N (Image credit: Astroinformatics Group)

The AMD engineering sample (ES) sports the 100-000000709-23_N codename and belongs to AMD's Family 25, which houses both Zen 3 and Zen 4 chips. Hardware detective Benchleaks found the same chip in the MilkyWay@Home project on the BOINC platform. It's weird, but AMD's unreleased Ryzen processors have started to see a cozy home at the MilkyWay@Home project. Back in January, two Ryzen 7000 (Raphael) showed up in the same project, indeed a weird place to be since it isn't even a benchmark.

According to the entry, which carries today's date, the Phoenix Point chip in question wields 16 threads, meaning it's likely an eight-core part with simultaneous multithreading (SMT). Other than that, the report doesn't tell us anything useful.

We'll likely hear more about Phoenix Point in the upcoming months since the Zen 4 and RDNA 3 APUs will, in all likelihood, debut in 2023 to challenge other chips on our list of best CPUs for gaming.

Zhiye Liu
News Editor and Memory Reviewer

Zhiye Liu is a news editor and memory reviewer at Tom’s Hardware. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

  • Quenepas
    This is the APU that will get handheld PC gaming (Steam Deck) interesting. Zen 4, RDNA 3, 4nm, PCIe 5.0, DDR5 and a new AI engine that for sure will have gaming enhancing elements. Pretty sure we'll see a slimmer and more powerful with better battery life Steam Deck 2 with this APU.
    Reply
  • StuWiFi6DDR5
    I hope, it gets to the mini computers, with ddr5 and pcie5, I’d be happy with a 4 core, as long as I can use fast short and long term memory. I don’t want the hassle of integrating the motherboard and case, but I do want to be able to buy a barebones and make sure it has ddr and pcie 5. The last 6 core APU I bought, didn’t have pcie 4, the motherboard I bought did, the ssd I bought did, but the 4000 series, 7nm, 6 core APU didn’t support pcie4, imagine my annoyance.
    Reply
  • P1nky
    Quenepas said:
    This is the APU that will get handheld PC gaming (Steam Deck) interesting. Zen 4, RDNA 3, 4nm, PCIe 5.0, DDR5 and a new AI engine that for sure will have gaming enhancing elements. Pretty sure we'll see a slimmer and more powerful with better battery life Steam Deck 2 with this APU.

    Not going to happen. This APU will cost almost just as much as the entry level Steam Deck.
    Reply
  • -Fran-
    Intel better get their "I4" node ready by then, otherwise they'll most likely get floored by this on mobile.

    Regards.
    Reply
  • digitalgriffin
    P1nky said:
    Not going to happen. This APU will cost almost just as much as the entry level Steam Deck.

    Maybe, maybe not. What generally happens is when this becomes an older more mature node, AMD will have excess and will be willing to cater to bulk purchases like Valve at a severely discounted rate. But this will take years. Valve is notoriously slow, and this is still a new node. I wouldn't expect before 3 years to get a steam deck 2. Look at the typical console life cycle before updates.
    Reply
  • StuWiFi6DDR5
    P1nky said:
    Not going to happen. This APU will cost almost just as much as the entry level Steam Deck.
    I’m with you, you could drive a new Xbox, with a chip like that.
    Reply
  • cfbcfb
    AMD's problem now is the ancient Osbourne paradox.

    I see the internal struggles. CEO says 15%. Marketing guy the next day says 40% or more. And then there's a "revised statement" that puts it back to the lower number.

    About 40 years ago a company named after its founder, a guy named Osbourne, released the Osbourne 1. And it was pretty nice. Until he mentioned the soon to be arriving Osbourne 2, which was much better.

    Everyone agreed with him, and then nobody bought the 1, since the 2 was going to be so much better, and Osbourne went out of business before getting the model 2 to market.

    AMD now has an Osbourne problem. Why would anyone buy the 5000/6000 series cpus or the current 6000 series run of graphics cards, when waiting a little bit could get you the 7000 series of both, with far greater performance?

    I know I was looking to build a 5950x system with a 6900 or 6950. But since there have been hints of fantastic performance and more cores with the higher end 7000 series, I'm going to wait for the next gen, and have a look at Intels 13th gen and 14th gen products.

    Yeah, I know...this mindset can keep you waiting forever. But I can wait less than a year to see quantum leaps in cores and compute power.
    Reply
  • Vorador2
    digitalgriffin said:
    Maybe, maybe not. What generally happens is when this becomes an older more mature node, AMD will have excess and will be willing to cater to bulk purchases like Valve at a severely discounted rate. But this will take years. Valve is notoriously slow, and this is still a new node. I wouldn't expect before 3 years to get a steam deck 2. Look at the typical console life cycle before updates.

    Errh, you do know Steam Deck uses a custom APU whose retail version won't be out until fall? Valve IS wiling to get custom silicon made from AMD.

    In any case, this is a 45W part so no way they're going to stuff it into a Deck. Deck runs pretty hot with a 25W part and battery is less than ideal.
    Reply
  • digitalgriffin
    Vorador2 said:
    Errh, you do know Steam Deck uses a custom APU whose retail version won't be out until fall? Valve IS wiling to get custom silicon made from AMD.

    In any case, this is a 45W part so no way they're going to stuff it into a Deck. Deck runs pretty hot with a 25W part and battery is less than ideal.

    I'm sorry I'm missing your point.

    The steamdeck custom APU is using a combination hybrid of old tech. It's not leading edge. Look at all the chopped PS5 APU units being sold.

    Until this point, there were no APU's available that met Valve's requirements. Hence a custom one was made.

    A 3400g is also a 65W part, but I can throttle it to 25W. AMD will just cherry pick the parts that run at the lowest voltages with stability.

    It all depends on price, capability, power use, and availability.
    Reply
  • StuWiFi6DDR5
    cfbcfb said:
    AMD's problem now is the ancient Osbourne paradox.

    I see the internal struggles. CEO says 15%. Marketing guy the next day says 40% or more. And then there's a "revised statement" that puts it back to the lower number.

    About 40 years ago a company named after its founder, a guy named Osbourne, released the Osbourne 1. And it was pretty nice. Until he mentioned the soon to be arriving Osbourne 2, which was much better.

    Everyone agreed with him, and then nobody bought the 1, since the 2 was going to be so much better, and Osbourne went out of business before getting the model 2 to market.

    AMD now has an Osbourne problem. Why would anyone buy the 5000/6000 series cpus or the current 6000 series run of graphics cards, when waiting a little bit could get you the 7000 series of both, with far greater performance?

    I know I was looking to build a 5950x system with a 6900 or 6950. But since there have been hints of fantastic performance and more cores with the higher end 7000 series, I'm going to wait for the next gen, and have a look at Intels 13th gen and 14th gen products.

    Yeah, I know...this mindset can keep you waiting forever. But I can wait less than a year to see quantum leaps in cores and compute power.
    I’m not sure that’s true, the industry, goes into long stagnations as well, look at how long Intel, was on 14nm. Recently things changed a lot, from 10nm, to 7nm, then 5nm, from DDR4, to DDR5, from PCIe3, to PCIe4, PCIe5, this is significant, memory bottlenecks and transistor numbers make a lot of difference. With the usual incremental changes, the law of diminishing returns applies, that makes little difference, just architectures, it’s been a long time since we changed from 32bit to 64bit.

    When we get to 4nm, 3nm, the physics problems get very difficult, to go farther, DDR4 has been the standard, for a long time, so DDR5 will be the standard for a long time, PCIe3 was the standard for a long time. Things will slow down, in 9 months I went through a 300 series motherboard, to a 400 and then 500 series motherboard in 9 months as they became price available. Now we have 600 series motherboards, but with the new chip mount standard, DDR5, PCIe5, 700 series motherboards, will remain the standard for some time. You might remember when we had the Pentium chip, Windows 98, Explorer 4, once we’d solved the 2000 bug, nobody wanted to upgrade.

    Hence the tech wreck, the law of diminishing returns, a huge tech boom, 97 to 2000, followed by much slower development, first falling off a cliff, then difficulty, finding the money, to advance things. After Windows XP, the software, became more confused, slower, so it wasn’t until the advent of the smartphone and then tablet, before things started to move again, the iPhone, Android, iPhone, Nexus 7. Because of the memory, flash changing the game, lower power consumption chips and RAM, the number of pixels soared, for a time, until that plateaued. Pixel numbers remained the same 2K, RAM and flash, advanced slowly, after M1, Qualcomm, expect things to slow, with Apple and windows, once the customer has something that’s not advancing rapidly, they don’t buy new stuff, my sister to my horror is using a 2011 Mac.

    Sure I’m running a Mac mini M1, iPad mini 6, Apple TV, Xbox One X, 5G Android 12 phone, Chromebook tablet, 16GB RAM, PCIe3, i7, USB C mini PC, running Windows 11, but that’s not the average punter, they run the same old bomb, for half a decade or more, until something significant happens. A lot of them, don’t even update their software, look forward to things slowing down again, chip shortages ending, demand decreasing, as the pandemic weakens, other industries taking the high ground.

    Real life changes, energy, transportation and food, leaving computers, in the background, until something fundamentally different comes along, VR/AR, optical computing, this is the peak, the rest of the world isn’t like us geeks, they don’t like upgrading, especially when there isn’t something really compelling.
    Reply