After more than a decade as an outsider and years of hyping his latest company, we have reached the moment of truth: AMD Ryzen processors are on our test bench and we can finally discuss our results.
Pricing and specs have been released and are known to most of you, but in a nutshell: AMD announced last week that its $ 500 Ryzen 7 1800X worth eight cores / 16 threads into a 95W Package with 3.6 GHz (4.0 GHz Boost) and performance on the level of Intel's $ 1,050 Broadwell-E Core i7-6900K with eight cores.
AMD also released the Ryzen 7 1700X for $ 400 with the same eight cores, 16 threads and 95 W TDP, but with a reduced clock speed of 3.4 GHz or 3.8 GHz base and boost. AMD is putting this chip against the Intel Core i7-6800K, a six-core processor that is currently retailing for $ 425.
The last processor on the market today is the Ryzen 7 1700 for $ 300. The same eight cores and 16 threads remain, but run at a base of 3.0 GHz and a boost of 3.7 GHz as well as a lower TDP of 65 W. The Ryzen 7 1700 is said to compete with the Core i7-7700K, which is currently is available for $ 350, although retailers appear to be implementing price cuts on most Intel processors.
The cheaper Ryzen 7 1700 CPU is arguably the most interesting chip in the series, as its multiplier remains unlocked for easy overclocking, which could result in a performance of 1700X or possibly even 1800X at a fraction of the price. Unfortunately, we don't have any on hand for today's review, but we'll have it for a proper review soon.
AMD said last week that Ryzen has exceeded the company's goal of achieving a 40% IPC improvement over its excavator architecture. The actual number is said to mean an impressive 52% improvement in IPC, enough to make their CPUs competitive and attractive to avid PC manufacturers for the first time in years.
The "Zen" architecture
According to AMD, the new x86 architecture called "Zen" is a rare event in the semiconductor industry because it is a "clean sheet" for the company. In terms of performance, the Zen microarchitecture aims not only to handle multithreaded tasks, but also to improve the performance of individual threads using advanced parallelism at command level (ILP).
The architecture has a much larger instruction scheduler window so that the CPU can schedule more and send more value to the execution units. A new micro-op cache means that the L2 and L3 caches can be bypassed if commonly used micro-operations are used.
The company also speaks of a "neural network-based" branch predictor that is used to minimize prediction errors by making Zen more intelligent in preparing optimal instructions and ways for future work.
The latest feature implemented by the latest AMD architecture is SMT (Simultaneous Multithreading), which allows two threads per core. This is a feature we've seen on some IBM, Intel, and Oracle processors, and is a departure from the CMT (Clustered Multi-Thread) design that was previously used in Bulldozer.
While Bulldozer used modules that worked similarly to a dual-core, dual-thread processor, Zen used SMT for what AMD calls the CPU Complex (CCX) design. CCX is a four-core / eight-thread module with its own L1, L2 and L3 cache.
The cache hierarchy is as follows: Each core receives a 64 KB L1 cache for instructions and data, a 512 KB L2 cache, and a shared 8 MB L3 cache that serves four cores. The L1 cache has been changed from write-through to write-back to enable lower latency and higher bandwidth. AMD advertises up to five times the cache bandwidth and we believe this claim applies to the L3 cache, while L1 and L2 are closer to twice the speed.
The Ryzen 7 series has two CCX units that enable eight cores and 16 threads. It is possible for AMD to disable individual cores with the CCX, and this is undoubtedly the way they create the six-core / 12-thread models.
For those who are wondering, the CCXs communicate via the high-speed model "Infinity Fabric", a successor to HyperTransport, which itself is the front-side bus in AMD's Opteron, Athlon 64, Athlon II, Sempron 64, Turion 64, Phenom Phenom had replaced II and FX families of microprocessors.
Infinity Fabric (essentially HyperTransport 2.0) is a flexible interface / bus that facilitates data exchange between CCXs, system memory as well as I / O and PCIe controllers. This provides Zen with powerful command and control functions for real-time estimates and adjustments to core voltage, temperature, socket power consumption, clock speed and more.
These command and control functions are critical to AMD's SenseMI technology. It is a package of five related "senses" that rely on sophisticated learning algorithms and the above-mentioned C&C functions to optimize the performance, performance characteristics of the cores and the management of speculative cache requests. and perform AI-based branch prediction.
Each Ryzen processor has a grid of interconnected sensors that are accurate to 1 mA, 1 mV, 1 mW and 1 degree C with a polling rate of 1000 / s. These sensors provide telemetry data that is fed into the Infinity Fabric control loop and allow the CPU to make adjustments based on current and expected operating conditions.
The five senses are described as follows:
- Pure strength A control system that optimizes clock speeds and frequencies to achieve the best performance with the lowest power consumption. According to AMD, Pure Power is used to enable "lower performance with the same performance".
- Precision Boost Works with Pure Power, but offers higher performance with the same power consumption thanks to incremental clock rate increases of 25 MHz.
- Extended frequency range (XFR) With XFR, perhaps the most interesting SenseMI function, the CPU can increase its clock rate depending on the temperature level beyond the nominal boost clock frequencies. This sounds similar to Nvidia's GPU Boost 3.0 technology, with which Pascal GPUs have often been pushed well beyond their nominal frequencies. According to AMD, XFR will reward enthusiastic coolers by scaling Ryzen's clock speeds with cooling solutions.
- Neural network prediction The use of the term "neural network" is somewhat loose here, but this technology preloads instructions by anticipating the actions that a user can take in advance.
- Smart prefetch Learns data access patterns to pre-fetch required data into the CPU cache so that it can be accessed immediately when needed.
The AMD AM4 platform
The Ryzen processors are complemented by the brand new AM4 platform from AMD, which comes on the market with three key chipsets: the entry-level A320, the mid-range or mainstream B350 and the X370 for enthusiasts.
|Chipset||USB (3.1 G2 + 3.1 G1 + 2.0)||SATA||SATAe||PCIe Gen 2||Multi-GPU||Overclocking|
|X370||2 + 6 + 6||4th||2nd||8 lanes||Yes||Unlocked|
|B350||2 + 2 + 6||2nd||2nd||6 lanes||No||Unlocked|
|A320||1 + 2 + 6||2nd||2nd||4 lanes||No||Blocked|
All chipsets support NVMe PCIe SSDs, SATA, SATA Express, dual-channel DDR4 memory, native USB 3.1 Gen 2 and more.
The X370 differs from the B350 model with more USB 3.1 Gen1 ports, more SATA ports, more PCIe 2.0 lanes and support for multi-GPU technologies. As you can see in the table above, the A320 sacrifices an additional USB 3.1 port and two additional PCIe 2.0 lanes in addition to locking (no overclocking).
The great thing about the AM4 platform is its flexibility. Socket 1331 optimizes AMD's socket infrastructure (AM3 and FM2 +) in a single interface that can support both seventh generation Ryzen and APUs. AMD also claims that future Raven Bridge and Zen processors will support the AM4 platform. The company plans to keep this socket for several years until 2020 and to integrate future technologies such as PCIe 4.0 and DDR5.