Way, way back in the fall of 2006, I put together my first PC in my dorm room. I picked out a Core 2 Duo E6400 and a proper motherboard, guided by a friendly-sounding article from some PC hardware site I’d found while Googling around. My elation with that PC—Dual cores! 4GB of memory! A graphics card that can run Half-Life 2! Free Windows Vista!—probably wasn’t shared by the AMD boardroom at the time.
The Conroe cores in that E6400 and its friends helped touch off an Intel CPU performance lead that AMD hasn’t much challenged since. 2007’s Phenom family of chips suffered from a performance-robbing TLB erratum, and the Phenom II series only duked it out with Intel chips from prior generations in its time. Famously, 2011’s ambitious Bulldozer architecture trailed Intel’s seminal Sandy Bridge CPUs substantially when it launched aboard the FX-8150, and the Piledriver refresh of that architecture in 2014 didn’t help much. Our move to frame-time benchmarking between Bulldozer and Piledriver made the refreshed architecture’s shortcomings especially clear for gaming performance. Then-AMD CEO Rory Read eventually conceded that Bulldozer “was not the game-changing part [sic] when it was introduced three years ago,” but the ‘dozer’s derivatives have had to soldier on in various forms in AMD’s CPUs ever since.
It didn’t help that AMD’s bet on the fusion of Radeon graphics and traditional CPU cores over seven generations of APUs didn’t find many takers in the lower end of the market. We can’t forget the company’s long slide into data-center irrelevance, either, an attractive and high-margin business that Intel basically has to itself these days.
So, yeah. After 10 years and change, the Zen microarchitecture that’s launching this morning aboard AMD’s Ryzen CPUs has a lot riding on its shoulders. The entire company’s future, if I had to guess. No biggie.
Not to spoil things too much, but Zen is solid. Go ahead and breathe a sigh of relief now. We’ve had three Ryzen CPUs in the TR labs this past week: the Ryzen 7 1800X, the Ryzen 7 1700X, and the Ryzen 7 1700. We’ve spent nearly every waking hour of the past few days turning every knob and dial we can to make our Ryzen CPUs sweat. Before we see whether or how the first Zen chips live up to the deafening hype that AMD has drummed up over the past few months, it’s worth taking a peek under the hood to see just how the company fulfilled the promises it’s made about Ryzen’s performance.
From the ground up
The Zen microarchitecture is a complete re-imagining of what an AMD x86 processor should look like. The company’s engineers have tossed the tightly-coupled “module” concept of Bulldozer and friends on the scrap heap. Instead, Zen is a sleek, shiny new chassis that looks a bit like Sandy Bridge and its derivatives if you squint a bit. AMD has consistently touted a “40% IPC speedup” in its discussions of Zen from the beginning, and I’ll do the best I can to briefly explain how AMD got there with its latest and greatest.
At the highest level, I want to draw your attention to two clusters of rectangles in this high-level block diagram of the Zen CPU core. The first point of interest is that each core will have its own integer and floating-point units to work with. This coprocessor layout is quite a bit different from the dual-integer-unit and shared-floating-point-unit structure of the Bulldozer core. Another new AMD trick for Zen is simultaneous multithreading, or SMT—better known as Hyper-threading in Intel parlance—to take advantage of otherwise idle execution resources. Much of the Zen core can be competitively shared between multiple threads of execution, and only a few resources—a new structure for AMD chips called the op cache, the store queue, and the retire queue—are statically partitioned.
The op cache is one of the biggest improvements to Zen’s fetch-and-decode stage. This structure first made its appearance in Intel’s Sandy Bridge architecture, and it serves as a temporary home for the internal micro-ops generated as part of the decode stage. This bit of cache is important because it can let the core leave its power-hungry fetch-and-decode hardware spun down. Instead, recently-decoded micro-ops can be dispatched straight into the maw of the core’s execution units for processing if they’re needed again. That shortcut has benefits for both latency and power consumption. You can read more about the benefits of op-caching in David Kanter’s incredible Sandy Bridge deep-dive. (David’s deep-dives have been indispensable in laying the foundations for this article, and they’re required reading for anyone with even the slightest curiosity about modern CPU architectures. Do go check them out.)
Zen also features an improved hashed-perceptron branch predictor compared to its predecessors. AMD (accurately) calls this a “neural network” instead, because neural networks are cool right now. It’s also not a new concept for AMD chips: Bulldozer, Piledriver, and Jaguar have all used similar technology in their predictors. AMD didn’t share many details of what it changed in the Zen predictor relative to its prior architectures.
In any case, better branch prediction is critical for allowing the chip to speculatively execute instructions without choosing the wrong path in the instruction stream. Get it wrong, and you have to flush the pipeline, an extraordinarily wasteful and performance-degrading operation in most cases. You can read more about the hashed-perceptron predictor in Daniel Jiménez’s introductory paper on the subject.
Intel has trumpeted better branch predictor accuracy in virtually every one of its recent microarchitectures, and it’s been quite reluctant to share any details of what it’s changed to get there. That guardedness suggests the company’s branch-prediction secret sauce is a major competitive advantage. Given Haswell CPUs’ uncanny accuracy in branch prediction, for example, there’s a reason for that.
Zen features a ton of other architectural improvements that contribute to its impressive performance gains over prior generations of AMD CPUs. We’d love to cover them all in depth for you, but we’ve been running tests on Ryzen right up until the NDA lift this morning and beyond. If you’d like to know more, be sure to check out David’s Zen write-up at the Microprocessor Report for more detail than we can possibly offer.
The Ryzen lineup
Zen is riding in this morning on three new CPUs: the Ryzen 7 1700, the Ryzen 7 1700X, and the Ryzen 7 1800X. You’ll already be familiar with these eight-core, 16-thread chips from AMD’s launch event last week, but a couple details have changed since we last checked in. Most notably, all three CPUs now feature Extended Frequency Range, or XFR, support.
|Model||Cores||Threads||Base clock||Boost clock||XFR||TDP||Price|
|Ryzen 7 1800X||8||16||3.6 GHz||4.0 GHz||Yes||95W||$499|
|Ryzen 7 1700X||3.4 GHz||3.8 GHz||Yes||95W||$399|
|Ryzen 7 1700||3.0 GHz||3.7 GHz||Yes||65W||$329|
Contrary to what we’ve heard about XFR until now, however, the technology basically applies a 100-MHz clock bump if a Zen chip’s internal sensors detect thermal headroom to work with. There’s no way to configure or turn off XFR, either (short of overclocking); it’s just a thing that will happen with adequate cooling. We found that even AMD’s Wraith cooler is enough of a heatsink to let XFR kick in on the Ryzen 7 1700 and Ryzen 7 1700X, so it seems as though many folks will be able to enjoy some additional out-of-the-box clock speed headroom without overclocking.
Next quarter, two Ryzen 5 CPUs will launch. The Ryzen 5 1600X will offer six cores running at 3.6 GHz base and 4.0 GHz boost clock speeds. It’ll be accompanied by the Ryzen 5 1500X, a four-core, eight-thread CPU with 3.5 GHz base and 3.7 GHz boost clock speeds. AMD says these chips will be priced below $300, but it didn’t offer further details. Those chips will be followed in the second half of this year by Ryzen 3 CPUs, although we don’t know anything about those presumably budget-priced chips yet. We can say that all of these Ryzen parts will be unlocked for those who want to try their hands at overclocking on the appropriate platform.
A quick tour of AM4 platforms
Socket AM4 will launch on a dizzying array of motherboards this year. Most PC builders will be interested in AMD’s high-end X370 chipset and the more entry-level B350 platform. Those two chipsets are the only way to get access to Ryzen CPUs’ unlocked multipliers. You can see how the spec breakdown shakes out in the complicated table below. Most notably, the A320, B350, and X370 chipsets will enjoy native USB 3.1 support, a feature that Intel has yet to integrate into its chipsets.
X370 is also the only AMD chipset that will offer builders the opportunity to employ dual-GPU setups in SLI or Crossfire through splitting a Summit Ridge CPU’s 16 lanes of dedicated PCIe 3.0 connectivity. We have a wide range of X370 motherboards in the TR labs now, and we’ll try and offer our thoughts on them when we can.
What we’re not testing today
Thanks to shipping delays, a constant stream of BIOS and software updates, and other headaches, we simply ran out of time to complete our testing before this morning’s NDA lift. Because of those circumstances, we elected to paint as complete a picture of the Ryzen 7 CPUs’ performance as possible now while leaving some other details only informally explored. We’ve completed all our usual productivity and gaming tests, and we think we can offer a solid idea of Ryzen’s value for system builders of all stripes.
Unfortunately, we had to make a few cuts from our schedule to achieve that goal. Overclocking performance and power efficiency measurements will have to wait for a separate article, as will platform performance measurements for X370 like USB 3.1 transfer speed and NVMe storage performance. We apologize in advance for the omissions, but we think you’ll enjoy the rest of our review. Let’s get to it.
Our testing methods
For each of our benchmarks, we ran each test at least three times, and we’ve reported the median result. Our test systems were configured like so:
|Processor||AMD Ryzen 7 1700||
|AMD Ryzen 7 1800X|
|Chipset||AMD X370 (Promontory)|
|Memory size||16 GB (2 DIMMs)|
|Memory type||G.Skill Trident Z DDR4-3866 (rated) SDRAM|
|Memory speed||2933 MT/s|
|Memory timings||13-13-13-33 1T|
|Processor||AMD FX-8370||Intel Core i7-2600K||Intel Core i7-3770K|
|Motherboard||Gigabyte GA-990FX-Gaming||Asus P8Z77-V Pro|
|Chipset||990FX + SB950||Z77 Express|
|Memory size||16 GB (2 DIMMs)|
|Memory type||Corsair Vengeance Pro Series DDR3 SDRAM|
|Memory speed||1866 MT/s|
|Memory timings||9-10-9-27 1T|
|Processor||Intel Core i7-4790K||Intel Core i7-6700K||Intel Core i7-7700K||Intel Core i7-6950X||Intel Core i7-5960X|
|Motherboard||Asus Z97-A/USB 3.1||Asus ROG Strix Z270E Gaming||Gigabyte GA-X99-Designare EX|
|Memory size||16 GB (2 DIMMs)||16 GB (2 DIMMs)||64GB (4 DIMMs)|
|Memory type||Corsair Vengeance Pro Series
|G.Skill Trident Z
|G.Skill Trident Z
|Memory speed||1866 MT/s||3866 MT/s||3200 MT/s||2400 MT/s|
|Memory timings||9-10-9-27 1T||18-19-19-39 1T||16-18-18-38 1T||15-15-15-35 1T|
They all shared the following common elements:
|Storage||2x Kingston HyperX 480GB SSDs|
|Discrete graphics||Gigabyte GeForce GTX 1080 Xtreme Gaming|
|Graphics driver version||GeForce 376.33|
|OS||Windows 10 Pro|
|Power supply||Corsair RM850x|
Thanks to Corsair, Kingston, Asus, Gigabyte, Cooler Master, Intel, G.Skill, and AMD for helping us to outfit our test rigs with some of the finest hardware available.
Some further notes on our testing methods:
The test systems’ Windows desktops were set at a resolution of 1920×1080 in 32-bit color. Vertical refresh sync (vsync) was disabled in the graphics driver control panel.
- After consulting with our readers, we’ve decided to enable Windows’ “Balanced” power profile for the bulk of our desktop processor tests, which means power-saving features like SpeedStep and Cool’n’Quiet are operating. (In the past, we only enabled these features for power consumption testing.) Our spot checks demonstrated to us that, typically, there’s no performance penalty for enabling these features on today’s CPUs. If there is a real-world penalty to enabling these features, well, we think that’s worthy of inclusion in our measurements, since the vast majority of desktop processors these days will spend their lives with these features enabled.
The tests and methods we employ are usually publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Memory subsystem performance
To get a sense of where Ryzen’s dual-channel memory architecture slots into the pantheon of bandwidth, we employed AIDA64’s directed memory read, write, and copy tests.
For eight hungry cores and 16 threads in a single socket, Ryzen CPUs fall decidedly mid-pack in these synthetic benchmarks. Intel sees fit to give its high-end desktop processors quad-channel memory controllers for plenty of breathing room, and the Core i7-5960X and Core i7-6950X generally enjoy much faster memory transfers as a result in these tests. Considering that the memory multiplier settings on our Gigabyte motherboard are locked out above 3200 MT/s right now, one can’t just shove DDR4-3866 into an X370 mobo and get around this issue, as one can with Z270 motherboards.
We’d usually test memory latency with AIDA64 and cache latencies with SiSoft Sandra at this point, but AMD warned us that neither utility performs correctly with Zen’s caches or memory controller. As a result, we’re holding off on reporting those numbers through independent testing.
AMD did provide reviewers with its own internal measurements of cache bandwidth and latency data for Zen. We won’t be diving deep into these numbers, but it is interesting to see how Ryzen chips’ cache hierarchies stack up against their Broadwell-E nemesis.
Synthetic math performance with Y-Cruncher
Normally, this spot is where we’d share the performance results from the synthetic benchmarks built into the AIDA64 utility. Unfortunately, those benchmarks haven’t been updated for Ryzen as we go to press, either. Instead, we turned to Y-Cruncher, a program that calculates pi out to arbitrary billions of digits. Not only does this program require a powerful CPU to run well, it also benefits from fast memory since it’s working with a large pool of RAM to store its calculations.
Y-Cruncher comes with a handful of executables that are tuned for various x86 extensions. Y-Cruncher also offers a Bulldozer-optimized binary that can use that chip family’s unique SIMD instructions. We used the newest version of the executable that ran on each chip without throwing warnings about ISA compatibility. We ran the program in its multithreaded mode and chose a 2,500,000,000-digit test size.
One thing is clear from our Y-Cruncher results right away: AVX2 SIMD support seems to help a lot. Ivy Bridge, Sandy Bridge, and Bulldozer don’t have it, and they suffer accordingly. The Ryzen CPUs have AVX2 support, but their 256-bit AVX throughput is half that of the Haswell and newer chips because of the 128-bit width of their FP units. Despite their high core and thread counts, the Ryzen chips land smack between Haswell and Skylake here. The Intel Extreme Edition chips put their copious memory bandwidth and execution hardware to good use by leading the pack in number crunching.
Doom likes to run fast, and especially so with a GTX 1080 pushing pixels. The game’s OpenGL mode is an ideal test of each CPU’s ability to keep that beast of a graphics card fed. We cranked up all of its eye candy at 1920×1080 and went to work with our usual test run in the beginning of the Foundry level.
Doom‘s OpenGL renderer demands plenty of single-core throughput to keep frame rates high and 99th-percentile frame rates low. While Intel’s menagerie achieves higher frame rates and lower 99th-percentile frame times than the Ryzen chips here, it’s worth noting that all of the CPUs but the FX-8370 here are rarely dipping below about 83 FPS. That’s still a plenty smooth and playable Doom experience. Still, this first gaming test puts a point on the fact that even with all of their generational improvements, Ryzen CPUs don’t have quite the same IPC oomph as Intel’s latest architectures.
These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid—or at least less than perfect. The formulas behind these graphs add up the amount of time the GTX 1080 spends beyond certain frame-time thresholds, each with an important implication for gaming smoothness. The 50-ms threshold is the most notable one, since it corresponds to a 20-FPS average. We figure if you’re not rendering any faster than 20 FPS, even for a moment, then the user is likely to perceive a slowdown. 33 ms correlates to 30 FPS or a 30Hz refresh rate. Go beyond that with vsync on, and you’re into the bad voodoo of quantization slowdowns. 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame. And 8.3 ms corresponds to 120 FPS, an even more demanding standard that Doom can easily meet or surpass on hardware that’s up to the task.
None of the CPUs we tested have more than a trace of frames that would drop frame rates below 60 FPS, so it’s worth clicking over to the more demanding 8.3-ms plot to see what’s happening. There, we can see that the Ryzen CPUs spend about as much time churning on tough frames that would drop animation below 120 FPS as Intel’s Sandy Bridge and Ivy Bridge CPUs do.
The switch to Vulkan levels the playing field a bit for the Ryzen chips. With this API, only the FX-8370 has any problems keeping a 99th-percentile frame time of under 10 ms. That performance translates to a gaming experience of over 111 FPS 99% of the time. Still, the Ryzen CPUs (save for the top-end 1800X) can’t quite match the high average frame rates of even the Core i7-3770K. Hm.
With the help of our “time-spent-beyond-X” graph, we can confirm that all of the CPUs at hand spend only fleeting moments past 16.7 ms working on tough frames. The picture at 8.3 ms is quite encouraging, as well. Only the FX-8370 turns in a figure past this threshold that we’d call noticeable. Moving on.
Although Crysis 3 is nearly four years old now, its lavishly detailed environments and demanding physics engine can still stress every part of a system. To put each of our CPUs to the test, we took a one-minute run through the grassy area at the beginning of the “Welcome to the Jungle” level with settings cranked at 1920×1080.
Crysis 3 lets the Ryzen trio shine a bit thanks to its affinity for lots of cores and threads. Even so, all three chips finish just midpack in average-FPS terms, although the race is a tight one. Each Ryzen CPU is just a couple milliseconds off the 99th-percentile frame time pace set by the Core i7-6950X and the Core i7-7700K, too.
Our “time-spent-beyond-X” graphs paint a pretty picture for Zen’s performance in Crysis 3. None of the Ryzen chips cause our GTX 1080 to spend more than an instant past the critical 16.7-ms mark, and they only spend a couple seconds of our test run on frames that cause the GTX 1080’s output to drop below 120 FPS. We’ll chalk that up as a success.
Watch Dogs 2
Here’s a new addition to our CPU-testing suite. We heard through the grapevine that Watch Dogs 2 can occupy every thread one can throw at it, so we turned up the eye candy and walked through the forested paths around the game’s Coit Tower landmark to get our CPUs sweating.
Unfortunately, because of the DRM baked into this title, we were only able to complete testing on six of our test CPUs before the NDA lift. We’ve updated all of our graphs with complete data now, but what a headache. Note to publishers: if you’d like to make your game useful for hardware reviewers, don’t lock us out just because we switch machines a bunch.
We thought we were doing the Ryzen chips a favor by running them with a title that does indeed take advantage of all of their cores and threads. Instead, the Ryzen 7 1700 had an embarrassing moment when Watch Dogs 2 advised us that the chip didn’t meet the game’s minimum specs. Intel’s Extreme Edition CPUs and the Core i7-7700K didn’t take Watch Dogs 2 lying down, either. With this almost entirely CPU-bound setup, the Ryzen chips and the Intel competition all fall into a neat line that’s purely reflective of their performance.
At the 16.7-ms threshold, the Ryzen 7 1700 spends by far the most time on tough frames: three seconds of our test run on scenes that make our GTX 1080 drop below 60 FPS. The other two Ryzens perform better, but they still can’t quite match the Intel chips here. A flip over to the 8-ms graph shows an Intel lead at that demanding threshold, as well.
Deus Ex: Mankind Divided
After our Core i7-7700K review, where Deus Ex: Mankind Divided proved GPU-bound, we tweaked the game’s settings to see if that remained the case across its entire range of eye candy. Happily, we discovered that turning off MSAA and lightening a few other loads on the graphics card turns DXMD‘s polygon-rich environments into a real torture test for any CPU.
Unshackle the GTX 1080 with some strategic settings changes, and Deus Ex can actually run quite swiftly. Its CPU scaling seems to top out at about eight threads, however, so the Core i7-7700K and company rule the roost by a decent margin. Just like Watch Dogs 2, the performance progression in the graphs above is more or less what we expect from a completely CPU-bound test.
While there’s little churning of note for these CPUs at the 16.7-ms threshold in DXMD, a click over to the 8.3-ms mark shows that the Intel CPUs spend substantially less time holding up the GTX 1080 than the Ryzen chips do. Even with an apparent eight-thread workload to fully occupy them, AMD’s latest just can’t keep up with the GTX 1080’s thirst for work at higher frame rates.
Grand Theft Auto V
Grand Theft Auto V can still put the hurt on CPUs as well as graphics cards, so we ran through our usual test run with the game’s settings turned all the way up at 1920×1080. Unlike most of the games we’ve tested so far, GTA V favors a single thread or two heavily, and there’s no way around it with Vulkan or DirectX 12. In that way, it’s a perfect test of whether a CPU can keep the graphics card fed.
Noticing a pattern yet? While the Ryzen CPUs deliver a fine 99th-percentile frame time, they just can’t match the higher average frame rates that the Core i7-6700K and Core i7-7700K can produce. In fact, considering the weird wall that the Ryzen 7 1800X hits with its 99th-percentile frame time, we have to wonder if there’s not a memory bandwidth issue in play.
As expected, the FX-8370 is the only CPU that can’t keep out of the past-16.7-ms doldrums. We’re more interested in what happens past the 8.3-ms mark with these systems, and the Core i7-7700K is good for keeping the GPU waiting for less than half the time at that threshold than the Ryzen 7 1800X is. The Ryzen 7 1700X and the Ryzen 7 1700 both keep the Core i7-3770K company toward the back of the pack.
So what are we to make of Ryzen as a gaming chip? Give one enough threads, as a couple of our benchmark titles do, and a Ryzen CPU can be a fine, if not exceptional performer. Most games still don’t take advantage of n threads, however, and in situations like GTA V, lower-end Ryzens can’t keep our GTX 1080 fed any better than 2012’s Core i7-3770K. Let’s get into some non-gaming tests and see why that might be.
Compiling code in GCC
Our resident developer, Bruno Ferreira, helped put together this code compiling test. Qtbench tests the time required to compile the QT SDK using the GCC compiler. The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.
Here’s a compelling start for Ryzen in our non-gaming tests. The R7 1700 goes blow-for-blow with the i7-7700K, and the R7 1700X only slightly tails the Core i7-5960X. The R7 1800X almost catches the Core i7-6950X. If Zen’s floating-point performance leaves a bit to be desired, the integer side of the core can be a beast when it’s churning away at full tilt.
These three benchmarks are about as single-threaded as it gets, so they’re an excellent indication of how Ryzen CPUs perform in lightly-threaded workloads. Pay close attention to these numbers if you’re curious about the IPC increases that AMD achieved with the Zen architecture.
There’s a bit of variance in how these tests shake out, but they all paint largely the same picture. At 4 GHz or so, the Zen architecture lands somewhere between Broadwell and Haswell in single-threaded throughput. As clock speeds start to decrease, however, the picture grows less rosy. The Ryzen 7 1700X isn’t much faster in this lightly-threaded workload than the Core i7-3770K at times, while the Ryzen 7 1700 can fall behind even the Core i7-2600K. Those measures have a direct correlation with how “snappy” a machine feels in common tasks, and the lower-end Ryzen 7 chips arguably won’t feel much faster than Sandy Bridge or Ivy Bridge desktops running around the same clock speed. Trust us, though: if you’re upgrading from an FX-series processor to Ryzen, you’ll immediately notice that common tasks like web browsing are much snappier.
In this common desktop workload, Zen exhibits a rather large performance delta between its compression performance and decompression performance. Considering that I probably unzip 50 zip archives for every one I compress, that’s probably not a bad tradeoff to make. Zen is only bested by Intel’s high-end desktop chips in compression, and it puts the Core i7-5960X to shame when unpacking archives.
TrueCrypt disk encryption
Although the TrueCrypt project has fallen on hard times, its built-in benchmarking utility remains handy for a quick test of these chips’ accelerated and non-accelerated performance when we ask them to encrypt data. The AES test should take advantage of hardware acceleration on the chips that support Intel’s AES-NI instructions, while the Twofish test relies on good old unaccelerated number-crunching prowess.
All of these chips support AES acceleration in hardware, so their performance scales roughly with the number of cores, threads, and Hertz on offer. The story is much the same in Twofish rates. This is another test where Ryzen excels.
Scientific computing with STARS Euler3D
Euler3D tackles the difficult problem of simulating fluid dynamics. It tends to be very memory-bandwidth intensive. You can read more about it right here.
For this set of chips, Euler3D seems to tell two different stories. For chips with lots of memory bandwidth but few threads (like the Core i7-7700K), the execution resources the chip has to offer are the bottleneck. For big, wide machines like Ryzen, memory bandwidth seems to be the bottleneck. Both the Core i7-5960X and the Core i7-6950X deliver tremendous performance in Euler3D thanks to their potent combination of many execution resources and bountiful memory bandwidth. Makes one wonder what Ryzen could do with an extra two memory channels.
3D rendering and video processing
The Cinebench benchmark is based on Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. This test runs with just a single thread and then with as many threads as CPU cores (or threads, in CPUs with multiple hardware threads per core) are available.
AMD favors Cinebench for its demonstrations of Zen’s single-threaded performance, and it’s not hard to see why. The Ryzen 7 1800X nearly matches the Core i7-6950X here, but it can’t catch the higher-clocked Intel mainstream desktop parts.
Surprise! More cores, higher scores. All of the Ryzen CPUs best the Core i7-5960X and its relatively slow all-core Turbo speed, but they can’t quite catch the Core i7-6950X with its unfair advantage of two extra cores and four extra threads. Still, this is another solid win for Ryzen.
Until recently, Blender was another common sight at Ryzen demo events. Its recent absence may be because of the version 2.78b update, which includes a number of optimizations for SSE and AVX2-compatible CPUs that improve performance. Our guess is that those updates might favor Haswell and friends more than they do Zen, as we’ve seen throughout this test.
The Blender project offers several standard scenes to render with Cycles for benchmarking purposes, and we chose the CPU-targeted version of the “bmw27” test file to put Cycles through its paces.
Whatever the Blender devs did to Cycles under the hood, every chip with AVX2 support enjoys huge gains compared to our last round of tests in our Core i7-7700K review. AMD needn’t have been bashful about Ryzen’s performance in these tests, either. Only the Core i7-6950X runs better.
Handbrake is a popular video-transcoding app that recently hit version 1.0. To see how it performs on these chips, we converted a roughly two-minute 4K source file from an iPhone 6S into the legacy “iPhone and iPod touch” preset using the x264 encoder’s otherwise-default settings.
x264 doesn’t seem to be scaling across all of the Core i7-6950X’s cores and threads, so the Ryzen chips all bunch up roughly under it. The gap between run times for even the more modest chips in this suite aren’t that far apart, however, so perhaps the program isn’t scaling beyond eight threads. The only CPUs that really suffer under Handbrake are those without AVX2 support, as we’ve come to expect. Perhaps that’s a good reason to consider moving up to a newer chip.
Digital audio workstation performance with DAWBench DSP
Here’s perhaps the most interesting addition to our benchmarking suite. DAWBench DSP is a freely-available project file for a number of digital audio workstation applications that lets us turn on a large number of instances of a standard VST (or effects plugin) while monitoring a looping audio track. The moment one starts hearing pops or crackles from the loop, it means the chip has reached its limit.
We chose the Reaper version of the project file and used the included ReaXcomp compressor plugin in its 64-bit form. To monitor the audio track, we plugged in a Focusrite Scarlett 2i2 USB audio interface using the USB 3.1 port (where available) on each of our test motherboards. We then installed Focusrite’s ASIO driver and selected the Scarlett as our playback device.
After some toying around, we decided that an ASIO buffer depth of 32 struck a good balance of low latency and CPU demand with our test setup. For the sake of time, we elected not to test at higher buffer depths, which decrease CPU load and increase performance. This is a CPU review, after all. Our graph describes the number of compressor instances we were able to turn on before overloading the CPU.
Given how similarly DAWBench scales compared to the Y-Cruncher results on our opening pages, it’s probably safe to say that Reaper and our plugin of choice both lean hard on AVX instructions (and AVX2, where available) to do their thing. We think that’s evidenced by the big leap in performance enjoyed by newer chips with AVX2 support like Zen.
Intel’s cores still have an undeniable advantage in SIMD throughput, though. The four-core, eight-thread Core i7-7700K does about as well as the eight-core, 16-thread Ryzen 7 1700 in this test, highlighting the fact that Zen’s floating-point unit has to halve its throughput in order to execute 256-bit AVX instructions. In contrast, the Haswell-E Core i7-5960X enjoys almost perfect performance scaling compared to the Core i7-4790K. The Zen CPUs trail it despite having the same number of cores and threads on tap (plus relatively higher base clocks, to boot). We’d be curious to see what Ryzen could do with similar SIMD throughput as Haswell and company.
Right now, though, music pros may still be elated by the R7 1800X’s value proposition. The hottest Ryzen 7 is just 18% behind the Haswell-E chip in the number of VST instances it can handle, but it’s a whopping 54% less expensive. Assuming Intel doesn’t cut the prices of its Broadwell-E chips to compensate, we think the Ryzen 7 lineup could be a great friend to audio producers on a budget.
AMD’s Ryzen 7 CPUs are arguably its best ever. Our tests show that the Zen microarchitecture can deliver single-threaded performance that’s about on par with Intel’s Broadwell core. In fact, AMD exceeded its ambitious 40% instructions-per-clock improvement target. Some of our directed tests actually showed as much as a 50% single-core boost from Piledriver to Zen. AMD is deservedly proud of this accomplishment.
Zen’s impressive single-thread potential is tempered by the delivered clock speeds of the eight-core, sixteen-thread parts that AMD is debuting today, however. The Ryzen 7 1800X trades blows with Intel’s Broadwell-E parts by dint of its high base and boost speeds, but that parity is a ceiling for the Ryzen lineup, not a floor. Lower-clocked Ryzen 7s seem to perform somewhere between Sandy Bridge and Ivy Bridge Core i7s in lightly-threaded tasks. As a result, they won’t feel like a substantial upgrade from an older Sandy or Ivy system while browsing Facebook and Twitter—out of the box, at least.
In an exciting hat tip to PC enthusiasts, all Ryzen CPUs will have unlocked multipliers for easy overclocking, so it might be simple enough to claw back some clock speed with a relatively affordable CPU. Remember those good old days? On early firmware, we got our $330 Ryzen 7 1700 up to a 3.9 GHz all-core overclock using just the modest AMD Wraith cooler. With those settings, the mildest Ryzen turns into a rather brisk single-threaded performer and a real fire-breather on the cheap for multithreaded workloads. We expect buyers willing to tweak a bit will be happy with the performance they can extract from a Ryzen 7 1700 and an affordable tower heatsink like the Cooler Master Hyper 212 Evo. We didn’t enjoy as much overclocking success with the already-speedy 1700X and 1800X parts, though. We’ll be exploring Ryzen overclocking in-depth in a separate article at some point.
Although AMD would like builders to think that a Ryzen 7 1800X is like getting a Broadwell-E Core i7-6900K for half the price and so on down the line, the reality is more complicated. We were surprised by how many programs in our test suite now appear to take advantage of AVX instructions, and Ryzen can only achieve half the 256-bit SIMD throughput as its Intel competitors when running those operations. Memory-bandwidth-constrained applications like Euler3D can also run far better when paired with Haswell-E and Broadwell-E’s quad-channel memory controller, where Ryzen seems to run into a bottleneck. That’s not comforting for a chip with eight data-hungry cores to feed.
If an application can take advantage of AVX, as our DAWBench DSP test seems able to, Intel’s high-end desktop parts can open a large lead on their Ryzen competitiors thanks to their beefier and higher-throughput SIMD hardware. Core-for-core, however, Ryzen still manages to hang pretty close with its Haswell-and-newer competitors like the Core i7-5960X despite this disadvantage. It helps that Ryzen CPUs are “discounted” far more than the percentage by which they trail the Intel competition, too.
Many of our other productivity tests didn’t run into memory or SIMD bottlenecks, and in those cases, the Ryzen 7 lineup truly does bring a new class of computing performance to the $500-and-under price point. Going by our index, the $400 Ryzen 7 1700X is basically a Core i7-5960X for about a third of the coin, and the Ryzen 7 1800X provides even slightly higher performance overall for just $500. That’s an incredible value in high-performance desktop computing, and we imagine that Intel won’t be able to avoid dropping prices on some of its Broadwell-E CPUs in response. For non-gaming applications, we think these Ryzen 7 chips will be difficult to ignore for those with a need for sheer computing power.
Even though Ryzen redefines the performance available at a given price point for highly multithreaded applications, gamers looking for a similar revolution from Ryzen are likely out of luck. We’ve been trying to find more multithreaded games to test with of late, and Watch Dogs 2 certainly qualified. It seems that game favors high IPC, memory bandwidth, and clock speeds, though, and the Intel chips we tested offer more of some or all of those things. Heck, in the suite of six games plotted in our value chart above, the Core i7-6950X just barely ekes out the top spot. That may be a first for an Intel high-end desktop CPU. For play, Intel’s Core i7-7700K remains the chip to beat, but the future seems to hold multithreaded promise, and that’s good for AMD.
AMD countered our questions about Ryzen’s performance at the modest resolutions we test CPU gaming prowess with by asserting that higher-resolution displays are becoming ever more popular. By extension, AMD seems to think the gaming market is moving toward being more graphics-card bound than CPU-bound. The Steam Hardware Survey doesn’t support this argument, but it is true that gaming at higher resolutions will lessen the differences in performance between Ryzen chips and Intel’s seventh-generation Core CPUs if a gamer chooses to play that way.
The company also suggested that the Core i7-7700K and its ilk will appeal more to “pure gamers” who just, well, play games. AMD sees Ryzen as a one-socket shop for those who want to game and stream to Twitch in the highest possible quality all at once. That may be, but we think gamers would rather not make a tradeoff between wide-shouldered grunt and the smoothness-enhancing goodness of high clock speeds and instructions-per-clock muscle. We probably owe it to ourselves to see how Ryzen and the Intel competition perform under those circumstances at some point to see just what the deal is, regardless.
Small wrinkles and our differences in performance priorities aside, it bears repeating that AMD is well and truly back in the high-performance x86 CPU game. If the company can further refine the Zen architecture over time, take advantage of future process improvements, and push clocks higher, we expect that future Ryzen chips will be competitive for many years to come. For now, we say bravo, AMD, and welcome back.