AMD Ryzen in 2018
Spoilt for choice
AMD has had an eventful 2018. Hot on the heels of the launch of its second-generation Ryzen processors and Ryzen APUs with Vega graphics, the chipmaker launched its refreshed Threadripper line-up, helmed by the 32-core Threadripper 2990WX.
The company now has something for pretty much any segment of the market, regardless of whether you’re looking to build a low-cost productivity machine, game, or work with character modeling.
That said, the new Ryzen chips and Raven Ridge APUs comprise a total of six processors, and that’s what we’re going to be focusing on today. While it’s pretty obvious who Threadripper is for, namely, hardcore gamers and serious enthusiasts, the mainstream Ryzen chips cater to a much broader swath of consumers who may have a harder time settling on the right chip.
For starters, the biggest divide sits between the Pinnacle Ridge Ryzen processors and Raven Ridge APUs. In the following paragraphs, we’ll talk about the differences between the two, their respective chip designs, and key new features before moving on to the performance numbers.
Zen meets Vega
AMD has made available two Raven Ridge APUs, the S$145 Ryzen 3 2200G (3.5GHz, 4MB L3 cache) and S$237 Ryzen 5 2400G (3.6GHz, 4MB L3 cache). Both of these are quad-core chips, but the Ryzen 5 2400G supports simultaneous multi-threading for a total of 8 threads.
Compared to the Ryzen 3 1200 (3.1GHz, 8MB L3 cache) and Ryzen 5 1400 (3.2GHz, 8MB L3 cache), the Raven Ridge chips sport higher base and boost clocks, which will help in applications that are sensitive to clock speeds, such as games.
Here's an overview of their specifications:
|Cores/Threads||Base/Boost Frequency||Graphics Model||Graphics Core Count||Max. GPU Clock||L2/L3 cache||TDP||Price (SGD)|
|Ryzen 5 2400G||4/8||3.6GHz/3.9GHz||Radeon Vega 11 Processor Graphics||11||1,250MHz||6MB||65W||$234|
|Ryzen 3 2200G||4/4||3.5GHz/3.7GHz||Radeon Vega 8 Processor Graphics||8||1,100MHz||6MB||65W||$144|
We've already covered the basics in this article looking at their graphics performance, but here's a recap anyway.
The basic building block of the Zen microarchitecture is a quad-core CPU complex (CCX), where each CCX houses 8MB of L3 cache. This comes together with another quad-core CCX to form an 8-core Zeppelin die.
The Raven Ridge APUs use the same underlying approach, but with key differences. Instead of a 2+2 design where two cores are symmetrically disabled across two CCXs, both chips use a single CCX configuration and a 4+0 design.
In addition, there's less L3 per core – 1MB instead of 2MB – so you get a total of 4MB of L3 cache instead of 8MB.
One benefit of having all four active cores in the same CCX is the possibility of reduced latency because any application that uses multiple cores will not have to use the Infinity Fabric interconnect to talk to cores and cache on another CCX.
The Vega graphics engine sits where the second CCX would have been, effectively combining Zen cores with Vega graphics on the same piece of silicon.
The Zen core complex and Vega graphics are linked together via AMD's Infinity Fabric, which in turn also services clients like the multimedia engines, display engine, DDR4 memory controllers, and the I/O and system hub.
Finally, the Raven Ridge chips are based on a 14nm+ FinFET process that is a density-optimized version of the 14nm node it was using before. This is different from the 12nm LP process that AMD's Pinnacle Ridge chips are using.
Zen+ is a new and better Zen
GlobalFoundries' 12nm LP process offers the second-generation Ryzen processors 10 to 15 per cent better transistor performance than preceding nodes, which extends the clock speed range of the processors and reduces the required current at all frequencies.
That's reflected in the higher base and boost clocks, which you can see below:
|Cores/Threads||Base/Boost clock||L3 cache||TDP||Cooler||Price (SGD)|
|Ryzen 7 2700X||8/16||3.7GHz/4.3GHz||16MB||105W||Wraith Prism (LED)||$485|
|Ryzen 7 2700||8/16||3.2GHz/4.1GHz||16MB||65W||Wraith Spire (LED)||$448|
|Ryzen 5 2600X||6/12||3.6GHz/4.2GHz||16MB||95W||Wraith Spire||$335|
|Ryzen 5 2600||6/12||3.4GHz/3.9GHz||16MB||65W||Wraith Stealth||$292|
One of the key weaknesses of the first generation of Ryzen processors was its relatively weak performance in games at 1080p, so the higher overall clock speeds should do something to alleviate that.
In addition, AMD is claiming a 50mV reduction in core voltage across the entire operating range of the chips.
Zen+ also features design optimizations to speed up access to cache and memory, which can help boost performance in latency-sensitive tasks. According to AMD, you can expect the following improvements:
- 13% reduction in L1 cache latency
- 34% reduction in L2 cache latency
- 16% reduction in L3 cache latency
- 11% reduction in DRAM latency
The new processors also officially support DDR4-2933 memory now, up from DDR4-2667 from before.
What new features are there?
Both the Pinnacle Ridge and Raven Ridge processors employ Precision Boost 2, an improved version of the Precision Boost frequency boosting algorithm that AMD used on its Ryzen 1000-series desktop processors.
Precision Boost distinguished between dual-core and all-core boost frequencies. This means that in cases where a certain workload utilized multiple threads but did not utilize each core fully, the frequency would still drop to the lower all-core frequency, even though there was technically no thermal or other utilization limitations that preventing the frequencies from boosting higher.
It's not difficult to see why this wouldn't be the most effective implementation for the best possible performance, and AMD's looking to address this issue with Precision Boost 2.
Precision Boost 2 does away with this dualistic model and replaces it with a new and more opportunistic algorithm. It utilizes Dynamic Voltage Frequency Scaling (DVFS) technology, and relies on data on CPU temperature, current, and load.
Instead of a steep drop-off when moving from an a dual-core boost to an all-core boost, Precision Boost 2 seeks to enforce a smoother curve, while at the same time seeking to implement the highest possible frequency (while sticking to built-in thermal, electrical, and frequency limits).
It retains the same 25MHz granularity as the original Precision Boost, but it allows select cores to boost higher in certain scenarios such as games. In other words, you should enjoy a performance boost because the processor is freed up to run at higher clock speeds than before.
This process is a continuous adjustment loop, relying on information gathered by AMD's SenseMI Technology about things like voltage, current, and temperature.
Then there's a new version of AMD's Extended Frequency Range (XFR) technology dubbed XFR2. AMD allows the processor to maintain higher frequencies for longer if users have a high-end cooling solution installed that exceeds the recommended thermal specifications, but this was restricted to just a small number of cores on the Ryzen 1000-series chips.
In comparison, XFR2 now operates across all cores and threads, just like Precision Boost 2. This should result in an overall higher average clock speed, and give users a modest uplift in performance.