AMD Ryzen 7 2700X Review

👤by Matthew Hodgson Comments 📅19-04-18
Zen+ Revisions
Zen, early in 2017, set an astounding starting block for AMD to build from. Today we get to see the fruits of their labour, with a full year of adjustments, tweaks and experience to improve their CPU range with Zen+.

The 12LP Process from GLOBALFOUNDRIES

The 12nm lithography of Zen+ improves on last year’s 14nm FinFET process, seeing improvements of 10-15% over preceding nodes. This helps to extend clock speeds and reduces the current required to perform a specific task. This delivers improved performance in all the ways that enthusiast customers requested: latency, overclocking, memory speed and clock speeds.

We also see a drop in core voltage across the operating range, vs. 14nm FinFET, as well as improved all-core overclocks.

L1, L2 and L3 cache all see improvements in latency, around 13%, 34% and 16% respectively. This should contribute to a system-wide performance improvement, ensuring the CPU can execute tasks more effectively.

AMD SenseMI Technology

AMDs Ryzen CPUs have a grid of interconnected sensors, monitoring every aspect of performance; they’re accurate to 1mA, 1mV, 1mW and 1°C and are monitored 1000 times every second, to ensure the optimal performance is attained at all times. These sensors feed information to the Infinity Fabric control loop, the “brain” of the CPU, making decisions based on information available to it right now, and predicting future workloads and operating conditions.

This smart grid of sensors allows Zen+ CPUs to fine-tune performance and power characteristics of the cores, manage cache pre-fetches and perform AI-based branch prediction.

Understanding these sensors is crucial to understanding how the CPU behaves.

Precision Boost 2

With AMDs first Gen Ryzen processors, we saw Precision Boost introduced for the first time, continually monitoring all aspects of the CPU, in real time, to ensure the optimal clock speeds and boost frequencies are applied at all times. However, this technology only applied when two of the 16 cores were “taxed” in lightly threaded applications.

Precision Boost 2 improves on this idea. AMD noticed that scenarios existed where 3 or more cores were in use, yet the overall workload was relatively small. This previously resulted in a huge drop in core frequency, when no imminent issues were actually present.

Thermal, electrical and utilisation headroom has been analysed in-depth, arriving at Precision Boost 2.

When a limit is encountered, in terms of power draw or heat generation, Precision Boost 2 will dither within the frequency range, in 25MHz increments, to attain a core frequency that is manageable at that exact moment in time. The process is continuous, with the CPU adjusting parameters 1000 times per second to ensure the Ryzen 2nd Gen CPU runs at the optimal core speed for any given task.

Precision Boost 2, then, could be considered as being greedy or opportunistic – if the clock speeds are there and they won’t cause any imminent problems to the thermals or power draw, the CPU will boost its frequency to as close to the 4.3GHz Boost Clock as possible.

Extended Frequency Range 2 (XFR2)

Now that you have a grasp of Precision Boost 2, XFR2 is aimed at the enthusiast who earns improved processing power by allowing the CPU a nicer place to live, for lack of a better term. Precision Boost 2 operates within a specific window, but that specification is conservative to allow for warmer or more humid climates, which would hamper CPU performance somewhat.

The thermal environment within a PC case is something the user can influence easily, usually by increasing airflow with the addition of fans, or purchasing a high-performance heatsink for the CPU; even running the PC within an air-conditioned room. These improvements will widen the thermal envelope and allow higher frequencies for longer periods of time, pushing the onus onto power delivery instead.

Essentially, if you can largely remove one of the two major contributing factors towards clock speed degradation, you will likely end up with a faster CPU.

These architecture improvements allows greater performance in a huge range of scenarios by allowing the CPU to run at the highest clock speed possible, providing it’s inside the thermal/electrical/load envelope, regardless of the number of cores and threads active. The improvements in moderately-threaded applications, such as games, could be as big as 500MHz when compared to the 1800X.

Improved Memory Support

Perhaps one of the largest issues with the Ryzen launch was memory compatibility issues, with expensive, high-performance memory sticks sometimes not working at all, or if they did, they had to be run at a much lower frequency than advertised on the packet.
With Ryzen 2000 series CPUs, memory support has been increased from DDR4-2667MHz to DDR4-2933MHz. This is something everyone expected from the 2nd Gen Ryzen launch due to the Infinity Fabric’s reliance on memory speed.

Better support for high-speed memory kits should provide a boost to cross-core and cross-CCX communication, improving CPU performance more than you would usually see on an Intel system, for example.

20 pages « 2 3 4 5 > »

Comments