AMD Ryzen 4000 Mobile APUs: 7nm, 8-core on both 15W and 45W, Coming Q1

[harlan4096]

At last year’s CES, AMD showcased its then Ryzen 3000 mobile processors as part of the announcements. In what is becoming a trend, at this year’s CES, the company is doing the same in announcing its next generation Ryzen 4000 mobile processors. This year is a little different, with AMD showing off its manufacturing strategy at TSMC 7nm for the first time in the mobile space. There’s a ton of options on the table, both at 15W and 45W, offering some really impressive core counts, frequencies, and most importantly, design wins. Here are all the details.

On AMD’s desktop processor portfolio, the company released a combination package with 7nm and 14nm chips on it – for the new Ryzen 4000 series, codenamed Renoir, it’s a fully integrated monolithic piece of 7nm silicon, containing 8 Zen 2 cores and up to 8 compute units of enhanced Vega graphics. That’s 2x the core count of the previous generation Ryzen Mobile, but 3 fewer compute units, however AMD has reasons for that decrease and emphasized that due to the process node they can get 8 CUs to perform better than 11 CUs.

The new 8 core / 8 CU chip on 7nm is also very small. AMD wouldn’t let us measure it directly, or take pictures before the keynote, but by putting the chip next to a Zen 2 desktop chiplet, I was able to make out that the new APU is pretty much double the size of a single Zen 2 chiplet, and double 74mm2 is 148mm2 or 150mm2 for a round number. Given TSMC’s defect rate of 0.09 per cm2, this would give a yield around 87%.

AMD is going to be launching two variants of Ryzen Mobile 4000, based around two power points. These will both be based on the same 8-core silicon, and be separated into 15 W hardware, traditionally called the U series, and 45 W hardware, traditionally called the H series. There is a special ASUS-only exclusive of the H series based in 35 W, and we’ll cover that as well. The main reason why AMD can offer the same silicon at two different power envelopes is due to transistor scalability, binning, disabling some of the graphics, and base frequency adjustments. It also means that in 15 W mode, there are different power delivery requirements from AMD’s partners.

AMD stated that the Zen 2 design in this chips follows the same CCX layout as the desktop hardware, which means the 8 cores are split into two CCX units which communicate over the internal infinity fabric. AMD has also adjusted the L3 amount, to 4 MB per CCX, which is half that of the consumer desktop line.

With the graphics, there’s a bit more of a granular compute unit deficit as we go down the stack, and that’s an important way to differentiate the hardware similar to the CPU side. If you notice, the frequencies are super high – AMD said that this enhanced version of Vega, architected for 7nm, actually responded really well in terms of frequency and power, and so they were able to boost the clocks up a lot. So despite the drop from 11 CUs in the silicon design down to 8 CUs, the 1750 MHz frequency really piles on the performance. AMD is promoting +28% GPU performance in 3D Mark Time Spy against Intel’s Core i7-1065G7, as well as a single threaded performance lead over both Ice Lake and Comet Lake.

These CPUs all support LPDDR4X memory, up to 64 GB, and AMD says that the infinity fabric is not tied to this memory clock. This helps the chip reach even lower power in its idle states, and the company said that they have rearchitected a good portion of the power delivery in the APU in order to be able to power down and power gate more elements of the SoC than was previously possible. AMD said that this decoupling of the infinity fabric and memory support, especially with both CPU and GPU accessing it, was made substantially easier due to the APU being a monolithic solution (with that in mind, it’s likely that AMD might not be going down the chiplet APU route any time soon). Also worthy to note is that AMD is saying that they have reduced the latency for parts of the chip to enter/exit idle states by 80%, and it’s this that helps enable the power gating in such a way to remain responsive. In previous products, certain elements of the design had to remain powered in order to be as responsive as the user required.

In terms of power, AMD is touting a full 2x performance per watt on the new 15 W CPUs, made possible by doubling the cores in the same power envelope and keeping the frequency high. AMD stated that this was possible due to +30% efficiency from the core and the SoC design, and +70% efficiency in the process compared to the previous products. Overall SoC power for the same frequency of the APU is down 20% as well, allowing AMD to push more out of the hardware.

One question that AMD should be answering about the U series chips this year is if they have succeeded in meeting the 25x20 Goal they set themselves in 2014 - provide 25x more performance over their older Kaveri platform by 2020. As it stands, AMD's main issue with meeting this goal seemed to be idle power, so given that the company is saying it has improved, we should hopefully hear some progress on this. It should be noted that for both U-series and H-series, the chips only support PCIe 3.0 - I suspect that PCIe 4.0 might be too power hungry for these form factors, and for modern graphics PCIe 3.0 is sufficient. As for other features IO features on the chips, the exact SoC SATA/USB/PCIe support will be announced closer to when the products actually ship.
...

Continue Reading