https://www.techpowerup.com/review/amd-zen-4-ryzen-7000-technical-details/ This interview answers a lot of questions that the AMD keynote address in Computex 2022 did not answer.
AMD opened Computex 2022 with a bang by announcing its next-generation Ryzen 7000 "Zen 4" desktop processors and dozens of design wins for notebooks across all price points. According to public roadmaps, the second half of 2022 promises to be very busy for AMD, with the company launching not just these desktop processors, but also its 4th Gen EPYC "Genoa" enterprise processors and Radeon 7000 series RDNA3 graphics cards. We have come to expect a double-digit percentage IPC gain with each generation of the "Zen" CPU architecture that breathed life back into the processor market, and in its fifth year in the market, "Zen 4" is expected to continue this trend.
During Computex, we caught up with Robert Hallock and asked him many questions we had, and also included things our community members brought up in numerous forum discussions in recent days. A big thank you goes out to the whole team at AMD who made this interview possible, and of course Robert who was kind enough to provide fascinating insights into how the Ryzen 7000 series is shaping up and what gamers and PC enthusiasts can expect from the next big AMD chip powering their battlestations.
[font="'helvetica neue', arial, helvetica, sans-serif; font-size: 22px"]What AMD Announced at Computex
Before we get into the interview itself, here's a quick recap of the Ryzen 7000 series—these are new-generation desktop processors by AMD built in the new Socket AM5 LGA package, which requires new motherboards. The switch to AM5 was required as the processors come with the latest I/O: DDR5 memory and PCI-Express Gen 5. The switch to a land-grid array makes the processors physically resilient, and as you'll soon know, there's good reason behind the odd polygonal shape of the integrated heatspreader (IHS) for these chips. At the heart of the Ryzen 7000 is the new "Zen 4" microarchitecture that comes with higher IPC and new capabilities compared to "Zen 3." The "Zen 4" CPU cores are built on the cutting-edge 5 nm EUV process, while the I/O die is made on the 6 nm process. Integrated RDNA2 graphics is now standard-issue, although there's more to this, as is a bigger PCI-Express 5.0 lane budget than what competing Intel processors offer.
In its Computex 2022 presentation, AMD showed off an unnamed prototype Ryzen 7000 series processor playing AAA games. This chip was shown to be 16-core, 32-thread, which made us wonder if 16 is the maximum core-count for the Ryzen 7000 series at launch, much like the Ryzen 5000 series. Since all these 16 cores are what Intel would describe as "performance" cores, going in with 16-core, 32-thread as the maximum multi-core offering would be an indication that AMD is unfazed by Intel's E-core strategy to shore up multi-threaded performance. We began by asking AMD if 16-core, 32-thread indeed is the top dog of the Ryzen 7000 stack.
The Interview16-core, 32-thread is the maximum core configuration for the Ryzen 7000 at launch?
That is correct.
At Computex, you showed a 15% single-thread performance gain over the Ryzen 9 5950X. Wouldn't that only put the gaming performance on par with 5800X3D?
I think it's too early to say actually. We were deliberately conservative with our number on single-thread performance. We do intend to publish the exact breakdown of IPC vs. frequency contribution later in the summer, also including performance, power, and area on the new process. As far as what stacks up against what, I think it's too early to say, we're still in silicon bring-up.
What are your thoughts on 3D Vertical Cache (3DV Cache) for Zen 4?
3DV Cache will absolutely be a continuing part of our roadmap. It is not a one-off technology. We are a big believer in packaging as a competitive advantage for AMD, something that could meaningfully enhance performance for people, but we have nothing specific to announce for Zen 4 yet.
Will Zen 4 be focused on the high-end only?
No.
Your presentation mentioned "AI acceleration." Is that AVX-512 or something more exotic, like Intel GNA?
Yes. Specifically, AVX 512 VNNI for neural networking and AVX 512 BLOAT16 for inferencing. Both are pretty nice speedups, we're not using a fixed-function acceleration, this could be something we could do with our Xilinx acquisition. We are starting to see more consumer applicability of AI workloads, like video upscaling, which has grown a lot in the last two years. I think there is a general trend for the average enthusiast to take on more AI-type workloads. The time felt right to bring these features into the chip, given we moved to a smaller process node with better performance, power, and area capabilities.
Is integrated graphics standard on most SKUs?
IGP is standard. It's included on all 6 nanometer IO dies, which has a small number of compute units built in, specifically to enable video encode & decode and multiple display outputs. Integrated graphics is very relevant for the commercial market, where most of our CPUs have not had graphics, that's a big customer appetite, they don't buy discrete GPUs. Now we have a much richer portfolio of processors that can play into the commercial space. For enthusiasts, it will help diagnose a bad graphics card to get the system up and running when you're still waiting on the GPU to show up. The iGPU configuration [specifications] are consistent, and all of the CPUs will have it.
Does this mean the end of APUs on the AM5 desktop platform?
Not at all. We actually don't think of the Ryzen 7000 Series as an APU. It's a processor that has graphics, and I know that's a subtle difference. To us, when we say "APU," it really means the product has powerful graphics, is capable of playing a game, has all the bells and whistles for video encode, display, drivers, etc. The IGP in Ryzen 7000 is designed to light up a monitor, handle video encode/decode, run a home theater PC, do productivity, but it's not gaming-grade graphics. APUs with big graphics are absolutely a continued part of our roadmap, and you'll see more.
Does the IGP support AV1 decode?
Yes, it is similar in capabilities to the Ryzen 6000 Series. It's the same RDNA2 compute units, same VCN [Video] and DCN [Display] IP.
Why the new heatspreader design? Why the holes on the sides?
It's actually how we achieve cooler compatibility. If you flip over one of the AM4 processors, you'll find a blank spot in the middle without pins, which has space for capacitors. That blank space is not available on Socket AM5, it has LGA pads across the entire bottom surface of the chip. We had to move those capacitors somewhere else. They don't go under the heatspreader due to thermal challenges, so we had to put them on top of the package, which required us to make cutouts on the IHS to make room. Because of those changes we're able to keep the same package size, length and width, same z-height, same socket keep-out pattern, and that's what enables cooler compatibility with AM4.
Will AM4 coolers provide the optimum experience or just be "compatible"? i.e., will there be new cooler releases for AM5?
I think it's a little bit of both. We still intend to offer 65 W and 105 W CPUs even though the socket power limit has gone up to 170 W. Not every single processor will use that power envelope. Coolers designed for the 65 W and 105 W parts will be equally appropriate for AM5. For the 170 W socket power CPUs, I expect the existing higher-end air and liquid-cooling units to be perfectly fine, but there will probably be a wave of new solutions positioned as "designed for" these CPUs. Personally, I have a 5950X in my system on a Noctua NH-D15, and I fully intend to reuse that cooler on Socket AM5.
We saw several exciting new power-management capabilities on the Ryzen 6000 mobile processors. Are some of those included on the new Ryzen 7000 CPUs?
Yes, but not in the way you would expect. In fact, these technologies actually went into the IO die, which is a new design for this platform. We'll talk more about the technologies this summer. You should expect the lion's share of power savings and power improvements to come from the new IOD, and the 6000 Series technologies we integrated there.
In the photos, the compute dies look gold-plated?
They are not gold-plated, it's a process called "back-side-metallization" that we use to solder the dies to the heatspreader. Depending on how that's manufactured, it can refract light in different colors [like the surface of a DVD], and in this case it refracts in gold.
Is CPU overclocking supported on the lower-end chipsets? What about memory overclocking? Or will memory just run at some JEDEC baseline on these chipsets
Yes, CPU overclocking is supported on the lower-end chipsets. There is no change in the overclocking strategy on our chipsets. Memory, multipliers, voltage, all will be available on B650 and X670.
What can we expect from the processors in terms of CPU overclocking?
I'm not gonna make a commitment yet on frequency, but what I will say is that 5.5 GHz was very easy for us. The Ghostwire demo was one of many games that achieved that frequency on an early-silicon prototype 16-core part with an off-the-shelf liquid cooler. We're very excited about the frequency capabilities of Zen 4 on 5 nanometer; it's looking really good, more to come.
FCLK vs. Mem Clock is 1:2 for optimum performance (= ~1500 MHz FCLK), or will 1:1 be possible (3000 MHz FCLK)?
We will get into more of that over the summer. We are encouraged by both the fabric and memory overclocking capabilities.
How do you feel about the transition from DDR4 to DDR5?
AMD is betting on DDR5, there's no DDR4 support in Zen 4. In the last months, we talked to many component vendors, module makers, etc., to confirm their supply roadmap to confirm timings and avoid shortages. Everybody is coming back with very optimistic answers. DDR5 will be abundant in the lifetime of Socket AM5. The abundance and new demand from Socket AM5 will help bring pricing down.
At this time, the demand for DDR5 is limited because our competitor lets people skip DDR5 for DDR4. We think there will be cost parity, or very near parity, coming as a result of Socket AM5. We're so excited about DDR5 because the frequencies are fantastic, which is exactly what Ryzen loves. Just like Zen 3, Zen 4 is fabric-based, and we are seeing better overclocking characteristics as we're moving to Zen 4 on both the memory and fabric side. Memory validation is one of the last things we undertake because we need a nearly finished platform to do that kind of work. Even at this early phase, we're able to hit DDR5-6400, there's more to come, it's very encouraging.
There's a bit of confusion about the CPU's PCIe 5.0 configuration. We've seen 24 and 28 lanes mentioned. Could you clarify?
There are 28 total lanes from the CPU, all Gen 5, of which 4 are peeled off for downlink to the chipset and the remaining 24 are available to the user. On X670 Extreme that means graphics operates at x16 Gen 5 or x8/x8 Gen 5, and there's one M.2 NVMe x4 Gen 5. On the X670 (non-Extreme) only the M.2 NVMe slot is required to be Gen 5, the top slot for graphics will optionally be Gen 5. On B650, only M.2 storage will be Gen 5. Of course, other components, like companion controllers or additional NVMe devices, can connect to Gen 5 on the CPU.
All X670E boards use PCIe 5.0 PEG? Or are motherboard makers free to downgrade to Gen 4?
Yes, that is a requirement. It must be the top two slots as Gen 5; if you have one GPU, the top slot is x16, if you have two cards installed, both slots are x8.
With 28 Gen 5 lanes from the CPU, 16 PEG lanes, and 4 chipset lanes, 8 lanes are left. Is it possible for motherboards to have two M.2 slots wired to the processor?
Yes, that is a possibility.
Why the introduction of chipset E SKUs? Such as "X670E" vs. "X670"
It's actually a direct response to consumer feedback when we launched X570. We had some users who were still on Gen 3 graphics cards or storage devices. Additional PCI-Express generations are a cost-adder to motherboards, there are retimer and redriver components that must go on the board to extend the signal from the CPU. Customer feedback was that they liked the features and design, but they don't really need the Gen 4 stuff and would rather save the cost. As we were thinking about how to make motherboards more accessible at a wider range of price points, offering the premium motherboards with and without Gen 5 requirements made a lot of sense. It will offer more affordable options for those users who don't need Gen 5 graphics capability.
Is the X670E chipset fanless?
It's fanless.
You announced USB4 on Rembrandt, but we haven't seen anything about it on the Zen 4 slides.
We'll get into USB configurations later in the summer, but you should have already seen some motherboards announced with USB4 support, I think from Gigabyte this week.
AMD and MediaTek recently announced the development of new Wi-Fi 6E modules, are those a requirement for all AM5 motherboards?
Motherboard vendors have the freedom to choose modules, they will pick the module that makes most sense from a cost and feature point of view. We don't place any requirements on what they should use, but we do have an AVL, of course. [Approved Vendor List, a recommendation which components to use in a design]
[end of interview]
We profoundly thank Robert Hallock for taking the time to speak with us and are really excited to learn more about the other new capabilities of "Zen 4" Ryzen 7000. Stay tuned!