Ryzen 5000 Mobile: SoC Upgrades

While the introduction page focuses mainly on the change to Zen 3 cores, AMD has explained to AnandTech that there are plenty of other changes in this update which enable both performance and efficiency, as well as battery life enhancements, for users.

From this point on I will start using the silicon codenames, such as

  • Cezanne (Ryzen 5000 Mobile with Zen 3),
  • Lucienne (Ryzen 5000 Mobile with Zen 2),
  • Renoir (Ryzen 4000 Mobile, all Zen 2),
  • Vermeer (Ryzen 5000 Desktop, all Zen 3),
  • Matisse (Ryzen 3000 Desktop, all Zen 2)

Double Cache and Unified Cache for Cezanne

To reiterate the primary SoC change for Cezanne compared to Renoir, the eight cores now have a unified cache rather than two cache segments. On top of this, the cache size has also doubled.

This is similar to what we saw on the desktop, when AMD introduced Vermeer – Vermeer with Zen 3 had a unified cache over Matisse with Zen 2. At that time, AMD was pointing to the unified cache enabling better gaming performance as it lowered the ‘effective’ latency for CPU memory requests in that combined cache region. The same thing is expected to hold true for the new Cezanne silicon in Ryzen 5000 Mobile, and will play a key part in enabling that +19% IPC increase from generation to generation.

Improved Memory Controller for Cezanne and Lucienne

One of the key metrics in mobile processors is the ability to eliminate excess power overhead, especially when transitioning from an active state to an idle state. All major silicon vendors that build laptop processors work towards enabling super-low power states for when users are idle, because it increases battery life.

A lot of users will be used to features that keep the processor cores in low power states, or the graphics, but also part of this is the interconnect fabric and the memory controller. One of the new developments for Ryzen 5000, and in both Cezanne on Zen 3 and Lucienne on Zen 2, is that AMD has enabled deeper low-power states for the memory physical layer (PHY) interface. This enables the system to save power when the memory subsystem is either not needed or in a period of low activity. This means putting the fabric and memory on its own voltage plane, but also enabling the required logic to drive it to a lower power when idle. AMD states that the low-dropout regulators (LDOs) are configured to enable this transition, and in certain circumstances, allow the PHY to be bypassed to further lower power consumption.

The tradeoff with having a part of the processor in such a low power state is the time it takes to recover from idle, which is also a metric to keep track of. AMD is stating that the design in Ryzen 5000 also enables a fast exit to full activity, meaning that the high performance modes can be entered quickly.

Also on the memory front, it would appear that AMD is doubling capacity support for both LPDDR4X and DDR4. For this generation, Cezanne systems can be enabled with up to 32 GB of LPDDR4X-4267 (68.2 GB/s), or up to 64 GB of DDR4-3200 (51.2 GB/s). The benefits of LPDDR4X are lower power and higher bandwidth, while DDR4 enables higher capacity and a potentially upgradable design.

Per-Core Voltage Control for Cezanne and Lucienne

In line with the same theme of saving power, not only should the periphery of the core be managed for idle use, but the cores should as well. In Ryzen 4000 Mobile, AMD had a system whereby each core could have a separate frequency, which saved some power, but the drawback was that all the cores were on a single voltage plane and so even if a core was idle when another one was heavily loaded, all cores were running at that top voltage. This changes with all members of the Ryzen 5000 Mobile family, as both Cezanne and Lucienne will both feature voltage control on a per-core level.

The slide from AMD shows it best – the cores running at higher frequencies get higher voltage, and the cores that are idling can reduce their voltage to save power. One of the main limits to enabling this sort of profile, aside from actually having the control to do it in the first place, is to do it fast enough for it both to count towards power consumption but also such that it is transparent to the user – the cores should still be able to come to a high voltage/high frequency state within a suitable time. AMD’s design works with operating system triggers and quality of service hooks to apply high-frequency modes in a task-based format.

On AMD’s desktop processors, we saw that the introduction of a feature called CPPC2 helped enable this, and the same is true on the mobile processors, however it took another generation to do the required design and firmware changes.

Power and Response Optimization (CPPC2) for Cezanne and Lucienne

As we accelerate into the future of computing, making the most out of each individual bit of silicon is going to matter more. This means more control, more optimization, and more specialization. For Cezanne and Lucienne, AMD is implementing several CPPC2 features first exhibited on desktop silicon to try and get the most out of the silicon design.

‘Preferred Core’ is a term used mostly on the desktop space to indicate which CPU core in the design can turbo to the highest frequency at the best power, and through a series of operating system hooks, the system will selectively run all single-threaded workloads on that core assuming no other workload is present. Previously, threads could bounce around to enable a more equal thermal distribution – AMD will now selectively keep the workload on the single core until thermal limits kick in, enabling peak performance and no extra delays from thread switching. For overclockable systems, this typically also represents the best core for boosting the frequency, which becomes relevant for Ryzen 5000 Mobile and the new HX series processors.

Another part of CPPC2 is frequency selection, which reduces the time for the transition from low-frequency to high-frequency from 30 milliseconds down to under 2 milliseconds. This equates to a 2-frame adjustment in frequency being reduced down to sub-frame adjustments. The consequences of this enables workloads that occur for shorter than 30 milliseconds can take advantage of a momentarily higher frequency and get completed quicker – it also enables the system to be more responsive to the user, not only in idle-to-immediate environments, but also in situations where power is being distributed across the SoC and those ratios are adjusting for the best performance, such as when the user is gaming. Also enabling load-to-idle transitions on the order of 2 milliseconds improves battery life by putting the processor in a lower power state both quicker and more often, such as between key presses on the keyboard.

The third part of CPPC2 is the migration away from discrete legacy power states within the operating system. With an OS that has a suitable driver (modern Windows 10 and Linux), frequency control of the processor is returned back from the OS to the processor, allowing for finer grained transitions of when performance or power saving is needed. This means that rather than deal with the several power states we used to, the processor has the full continuous spectrum of frequencies and voltages to enable, and will analyze the workflow to decide how that power is distributed (the operating system can give hints to the processor to aid in those algorithms).

GPU Improvements on Cezanne and Lucienne: Vega 8 to Vega 8+

As mentioned on the previous page, one of the criticisms leveled at this new generation of processors is that we again get Vega 8 integrated graphics, rather than something RDNA based. The main reason for this is AMD’s re-use of design in order to enable a faster time-to-market with Zen 3. The previous generation Renoir design with Zen 2 and Vega 8 was built in conjunction with Cezanne to the point that the first samples of Cezanne were back from the fab only two months after Renoir was launched.

If we look at the change in integrated graphics from the start of Ryzen Mobile. The first generation Raven Ridge was built on 14nm, had Vega11 graphics, and had a maximum frequency around 1200 MHz. The graphics in that Renoir design were built on 7nm, and despite the jump down from Vega11 to Vega8, efficiency was greatly increased and frequency had a heathy already a jump up to 1750 MHz. Another generation on to Cezanne and Lucienne, and the graphics gets another efficiency boost, enabling +350 MHz for added performance.

Part of this update is down to tweaks and minor process updates. AMD is able to control the voltage regulation better to allow for new minimums, reducing power, and has enabled a new frequency sensitive prediction model for performance. With the greater power controls on the CPU and SoC side, this means that power budget can be more readily accessible by the integrated graphics, allowing for higher peak power consumption, which also helps boost frequency.

Note that these features apply to both Cezanne and Lucienne, meaning that the Zen 2 products in the Ryzen 5000 Mobile do get a sizeable boost in graphics performance over Renoir here. Ultimately it is that 15 W market for which this update is aimed, given that the H-series (including HS and HX) are likely to be paired with discrete graphics cards.

As and when AMD decides to move from Vega to RDNA, we’re likely going to see some of the Cezanne be re-used such that we might see Zen3 + RDNA in the future, or the combined Zen 4 + GPU chip might be a full upgrade across the board. This is all speculation, but AMD’s CEO Lisa Su has stated that being able to re-use silicon designs like this is a key part of the company’s mobile processor philosophy going forward.

Security Updates in Cezanne

One of the features of Zen 3 is that it enables AMD’s latest generation of security updates. The big update in Zen 3 was the additional of Control Flow Enforcement Technology, known as CET. This is where the processor will create shadow stacks for return calls to ensure that the correct return addresses are called at the end of functions; similarly indirect branch jumps and calls are monitored and protected against should an attacker attempt to modify where an indirect branch is headed.

Both AMD and Intel have spoken about including Microsoft Pluton security in their processors, and we can confirm that neither Cezanne nor Lucienne have Pluton as part of the design. Both AMD and Intel have stated that it will be integrated ‘in the future’, which seems to suggest we may still be another generation or two away.

Process Node Updates on Cezanne and Lucienne

Perhaps one of the smaller updates this time around, but AMD has stated that both Cezanne and Lucienne use the latest intra-process node updates on N7 for these products. While both previous generation Renoir and these two use TSMC’s N7 process, over the lifecycle of the manufacturing node minor changes are made, sometimes to reduce defect density/increase yield, while others might be voltage/frequency updates enabling better efficiency or a skew towards better binning at a different frequency. Usually these additions are minor to the point of not being that noticeable, and AMD hasn’t said much beyond ‘latest enhancements’.

AMD Ryzen 9 5980HS Cezanne Review CPU Tests: Core-to-Core and Cache Latency
POST A COMMENT

218 Comments

View All Comments

  • Meteor2 - Thursday, February 4, 2021 - link

    Great point. Reply
  • ikjadoon - Tuesday, January 26, 2021 - link

    It's great to see AMD kicking Intel's butt in a much larger market (i.e., laptops vastly outsell desktops): AMD really should be alongside, or simply replacing, Intel in most premium notebooks. Gaming notebooks are not my cup of tea, but glad to see for upcoming 15W Zen3 parts.

    Will we see actual, high-end Zen3 notebooks? Lenovo, HP, ASUS, Dell: for shame if you keep ramming toasty Tiger Lake down customers' throats. Lenovo's done some great offerings with both AMD & Intel; that means some compromises with notebook design (just go all AMD, man; if/when Intel is on top, switch back!), but beefier cooling for Intel will also help AMD.

    Still, overall, I don't see anything convincing me that x86 is really right for notebooks, either. So much waste heat...for what? The M1 has rightly rejiggered expectations: 20 hours on 150 nits should be ordinary, not miraculous. Limited to no fan spin-up and max CPU load should yield a chassis maximum of 40C (slightly warmer than body temperature). And, all the while with class-leading 1T performance.

    As this is a gaming laptop, it's not too relevant to compare web benchmarks (what most laptops do), but this is peak Zen3 mobile and it still falls quite short:

    Speedometer 2.0
    35W Ryzen 5980HS: 102 points (-57%)
    125W i9-10900K: 119 points (-49%)
    35W i7-1185G7: 128 points (-46%)
    105W Ryzen 5950X: 140 points (-40%)
    30W Apple M1: 234 points

    You can double / triple x86 wattage and still be miles behind M1. I almost feel silly buying an x86 laptop again: just kilowatts of waste heat over time. Why? Electrons that never get used, just exhausted and thrown out as soon as possible because it'll throttle even worse otherwise.
    Reply
  • undervolted_dc - Tuesday, January 26, 2021 - link

    because you here are benchmarking javascript engine in the browser
    but not being enough you are comparing those in single thread so here you are comparing 1/16 of the 5950hs vs 1/4 of the m1
    a 128core epyc or a 64core threadripper probably will be even worse in this single threaded benchmark ( because those are levaring threads and are less efficient in single threaded app )
    if you like wrong calculations then 1 core of the 15w version use less tha 1w for what result ? ~ 100 points ? so who is wasting electrons here ?
    ( btw 1 core doesn't use 1/16 because there are boosts , but it's even less wrong than your comparison )
    Reply
  • ZoZo - Tuesday, January 26, 2021 - link

    128-core EPYC? Where?
    His comparison is indeed misleading in terms of energy efficiency, but it's sad that no x86 is able to come even close to that single-threaded performance.
    Reply
  • WaltC - Tuesday, January 26, 2021 - link

    Doubly sad for the M1 that we are living in the multicore/multithread era...;) Reply
  • ikjadoon - Tuesday, January 26, 2021 - link

    The energy efficient comparisons are pretty clear: the best x86 (Zen3) has stunningly lower IPC than M1, which barely cracks 3 GHz. The only way to make up for such a gulf in IPC is faster clocks. Faster clocks require the 100+W TDPs so common in high-performance desktop CPUs. It's why Zen3 mobile clocks so much lower than Zen3 desktop (3-4 GHz instead of 4-5 GHz)

    A CPU that needs 3x power to do the same work (and do it slower in most cases) must exhaust an enormous amount of heat, when considering nT or 1T benchmarks (Zen3 requires ~20W for 5 GHz boost on a *single* core). Look at those boost power consumption measurements.

    Specifically in desktops (noted in my comparison about tripling TDP...), the CPU *alone* eats up an extra 60 to 90 watts during peak usage. Call it +20W average continuously, so we can do the math.

    20W x 8 hours x 7 days a week = +1.1 kWh excess exhaust heat per week. x86 had two corporate giants to do better. It's been severely litigated, but that's Intel's comeuppance. If Intel can't put out high-perf, high-efficiency x86 architectures, then people will start to feel less attached to x86 as an ISA. x86 had billions and billions and billions of R&D.

    I see no reason for consumers to religiously follow x86 Wintel or Wintel-clones in laptops especially, but desktops, too: where is the efficiency going to be coming from? Even if Apple *had flat 1T* for the next three years, I'd still feel more optimistic about M1-based CPUs in the long-term than x86.
    Reply
  • Dug - Tuesday, January 26, 2021 - link

    "I see no reason for consumers to religiously follow x86 Wintel or Wintel-clones in laptops especially, but desktops, too: where is the efficiency going to be coming from?"

    Software, and getting work done. M1 is great and all, but just need to convince the boss that Apple or 3rd party has software available for our company....... Nope, oh well.
    Other negatives-
    For personal use, people aren't going to spend thousands of dollars to get new software on new platform.
    They can't play games (or should I say they can't play a majority), which is probably the largest market.
    They can't change anything about their software
    They can't customize anything.
    They can't upgrade any piece of their hardware.
    They don't have options for same accessories.

    So I'll go ahead and spend the extra $15 a year on energy to keep Windows.
    Reply
  • Spunjji - Thursday, January 28, 2021 - link

    "A CPU that needs 3x power to do the same work"
    It doesn't. It's been demonstrated a few times now that if you scale back Zen 3 cores to similar performance levels to M1, M1's perf/watt advantage drops to about 30%. It's still better than the node advantage alone, but it's not crippling, and M1 is simply not capable of scaling up to the clock speeds required to match x86 on desktop / HPC workloads.

    They're different core designs matched to different purposes (ultra-mobile first vs. server first) and show different strengths as a result.

    M1 is a significant achievement - no doubt about it - but you're *massively* overstating the case in its favour.
    Reply
  • GeoffreyA - Friday, January 29, 2021 - link

    Thank you for this. Reply
  • Meteor2 - Thursday, February 4, 2021 - link

    "M1 is simply not capable of scaling up to the clock speeds required to match x86 on desktop / HPC workloads" ...Yet. In a couple of years x86 will be behind ARM across the board.

    Fastest HPC in the world is ARM *right now*. Only the fifth fastest is x86.
    Reply

Log in

Don't have an account? Sign up now