AMD has been heavily criticized when it announces that its new Ryzen 5000 Mobile U-series processors will not all use the latest core design. At the product announcement, we were told that some of the U-Series processors would be based on the previous Zen 2 generation, and this was primarily for partners to take advantage of the new naming scheme, but also to reuse designs with the same ballpark performance. A number of tech enthusiasts (including myself, I must say) mocked it because it made the whole system complicated. It’s still complex, but we’ve realized that these latest Zen 2 based mobile processors also include a whole range of updates that make them a better version of what they are.
To simplify things, I call these products their AMD code. The older Zen 2 processors are called Renoir, and the newer Zen 2 processors are called Lucienne. Here is a list of the new Ryzen 5000 U series, with Lucienne in yellow.
Renoir has been a very successful product for AMD for all intents and purposes. Placed in the Ryzen 4000 Mobile range, it has become the basis of AMD’s mobile portfolio and has been installed in approximately 100 design gains since market marketing. Lucienne, on the other hand, is a minor player in the latest Ryzen 5000 Mobile series. It does not have the updates that the new Zen 3 cores have, but we have since learned that it’s on the power side of things, rather than being a copy of Renoir, but it’s almost Renoir Plus.
What Lucienne brings to the table about Renoir appears in discrete categories.
Memory Manager
The memory controller in Lucienne is now able to disconnect its voltage from the core and enter a lower power state when not in use or for low bandwidth reasons. This ultimately saves power, and AMD has made it possible to bypass certain voltage indicators to help stay in the low voltage state. Apart from the cores and graphics, the other two consumers in power within a mobile processor are the internal communication and the external communication, the memory controller of which falls under the latter. AMD has also introduced a system whereby the memory controller can wake up faster than before to a full bandwidth state, enabling better response of the deep sleep states.
In addition, the memory controller can now support twice the memory capacity of Renoir: up to 64 GB DDR4-3200, or up to 32 GB LPDDR4X-4267. Using DDR4, the system may have more memory but also be user-adjustable, but LPDDR4X swaps it for faster bandwidth (68.4 GB / s versus 51.2 GB / s).
Voltage control per core
In similar conditions as the memory controller, the voltage of each core in a mobile processor is one angle to maximize performance when needed and reduce power loss when idle. In Renoir, all the cores can adjust their frequency, but they all had to run at the same voltage. Lucienne changes it so that each core can adjust its voltage independently, enabling finer grain current management and a more optimal power-saving system. There are also additional brackets that operating systems can use if he knows that core performance is needed in advance.
Preferred core
When we talk about turbo, it is historically accepted that any core can reach the highest single core turbo frequency, and that the workload is sometimes shifted between the cores to help with thermal management. However, when a system uses a preferred kernel, it means that a system can be optimized for that particular kernel, and can extract more performance. AMD introduced its Preferred Core technology to the desktop two generations ago and now it’s coming to mobile processors. One core out of eight Lucienne silicone is named the best core, and through an operating system driver (standard in Windows) all workloads are preferably placed on that core.
Frequency ramp
One of the characteristics that sums it all up is how fast a core can move from lazy to peak performance and back again. If a system takes too long to speed up, or move down again, response and power are lost. A typical modern system is expected to increase from standstill to peak frequency within two frames at 60 Hz or 32 milliseconds, but the latest systems from AMD and Intel have done so much faster, often within 16 ms. AMD’s improved clock gate technology enables Lucienne to reduce it to 1-2 milliseconds. This means that a system can easily go up and down between each keystroke on a keyboard, thus providing an immediate response to a user, while keeping total power consumption low. Typing on a keyboard may mean that a kernel is active almost continuously during the 16-32 milliseconds, but making this change faster offers many power savings through these transitions.
Continuous performance levels
The legacy of an operating system to command performance is through performance conditions, or P-states. In this case, the operating system will request a specific level of power and performance from the processor based on its workload, and the processor will respond. It was originally implemented at a time when turbo first came to modern processors, and that workload analysis could be done better by the operating system. Now we can perform this level of monitoring on the processor directly, and through an operating system driver (which is already part of Windows), with system support, the level of frequency control can be transferred to the processor. The processor also gets an effective continuous distribution of performance, rather than discrete P conditions.
While Renoir has had P-states, Lucienne gets the benefit of performance requests at CPU level.
Faster integrated graphics
With the additional power control elsewhere in the core, the operation of the power supply to the integrated graphics has also been adjusted to allow better regulation and ultimately a lower minimum voltage. Using firmware, AMD has enabled a frequency-sensitive prediction model that enables the GPU to adjust its voltage and frequency based on its dynamic power management. Along with the better regulation and the balance between power budget done between CPU, interconnection, DRAM and the GPU, more power budget is available for the GPU. For Lucienne, this means +150 MHz at the highest IGP speed compared to Renoir.
Slide shows Cezanne numbers, but also applies to Lucienne
But I thought Lucienne Silicon is the same as Renoir Silicon?
That’s the big question. We asked AMD if Lucienne was taking the same step as Renoir, and the answer was not exactly connected in one direction or another. The simple answer is yes, but AMD would like to make clear that significant changes have been made to firmware and manufacturing, which means that, despite the fact that the transistor layout is identical, Lucienne’s features would never have worked in Renoir without the changes. which has been affixed. .
Although it is the same silicone layout and floor plan, some of these features were not possible in Renoir. AMD built these features in, perhaps knowing that it could not be enabled in Renoir, but sufficient changes and improvements were made in the manufacturing phase and the firmware phase so that these features were enabled in Lucienne. These ideas often have very strict time windows to implement, and even if they are designed in the hardware, there is a strict cut-off point against time if it does not work as intended, it is not activated. The best result, of course, is to make everything work on time, but building CPUs is harder than we realize.
Sometimes I wonder how we ever let these rocks powered by lightning ever work in the first place.