AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

It’s been just over a year since AMD launched its new Ryzen microprocessor based on the Zen architecture. Zen’s impact can scarcely be overstated; the CPU market has changed more in the past 16 months than in the previous six years combined. Today, AMD is launching its second generation of Ryzen microprocessors, and angling to regain the overall performance crown from Intel.

Before we dive into the review, let’s take a quick look at where things stand. The Ryzen 7 and Ryzen 5 families that AMD launched last year struck the heart of Intel’s desktop lineup. For years, Intel kept its product segmentation the same. Its Core i7 enthusiast desktop CPUs were quad-cores with Hyper-Threading (4C/8T), the Core i5 packed four cores and no Hyper-Threading (4C/4T), the entry-level enthusiast Core i3 was a dual-core with HT (2C/4T), and the low-end CPUs in the Pentium and Celeron families were dual cores without Hyper-Threading.

This status quo held for six years until Ryzen hit it like a bomb. Even before the family launched, Intel added Hyper-Threading support to its low-end processors. Once Ryzen was in-market, Intel slashed its 10-core CPU price in half. It introduced a new, short-lived set of Kaby Lake-X CPUs in an attempt to spur adoption of its High End DeskTop (HEDT) product lines. All of these efforts culminated with the launch of last year’s Core i7-8700K, an aggressively clocked six-core / 12-thread CPU on a new spin of Intel’s 14nm (14nm++). We put the Core i7-8700K up against the Ryzen 7 1800X last year and concluded that Intel’s potent mixture of clocks and IPC made the Core i7-8700K a superior choice to the Ryzen 7 1800X overall.

Since then, the top of the market has been quiet. Today, that changes. AMD intends to take back the overall performance crown, and it thinks the Ryzen 7 2700X is the CPU to do it.

Second Generation Ryzen, Not Ryzen 2

These new CPUs will almost inevitably be called Ryzen 2, but that’s not the nomenclature AMD uses. Ryzen 2 is reserved for an architectural refresh expected next year on GlobalFoundries’ 7nm process node. The CPU we’re reviewing today still uses the same fundamental architecture as the Ryzen CPUs we reviewed last year — with a few improvements. First, there’s a small amount of additional IPC gain, thanks to the cache and memory latency improvements detailed below. All slides can be enlarged by clicking on them.

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

The “up to” in the specs leave some wiggle room as to where the improvements are, but we’ll take a look at how the Ryzen 7 2700X compares with the 1800X in these metrics. Overall, a 3 percent IPC boost and better cache latencies are in line with what we expected to see 12 months after Ryzen’s initial debut.

The Ryzen 7 and Ryzen 5 2xxx CPUs pick up most of their improved performance from clock speed increases, not raw IPC improvements. At first glance, these clock speed gains are underwhelming; the Ryzen 7 2700X’s 3.7GHz base and 4.3GHz boost clock aren’t much different from the Ryzen 7 1800X’s 3.6GHz base and 4.1GHz boost. But AMD didn’t just nudge base and boost clocks upwards — it made several changes to the underlying algorithm and pushed TDP a nudge higher to give the CPUs more breathing room.

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

In the past, AMD’s Precision Boost had a strict per-core demarcation. There was some light allowance for workloads, but frequencies only varied within a very narrow range based on XFR. The new Precision Boost 2 algorithm is designed to shift in a linear fashion, balancing core load, frequency, temperature, and voltage.

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

In this diagram, the red line corresponds to the Ryzen 7 1800X’s actual Turbo frequencies, the yellow line are the clock speeds the Ryzen 7 2700X is expected to hold at a given thread count in OCCT, and the dotted white line represents an idealized linear curve. Unlike the Ryzen 7 1800X, which drops to 3.7GHz once the number of cores rises above two, the Ryzen 7 2700X maintains a higher turbo frequency, peaking just above 4GHz.

These higher sustained clocks are key to AMD’s claim that the Ryzen 7 2700X can deliver meaningful performance improvements compared with its predecessor. A 4GHz all-core boost frequency is a significant improvement on the 1800X’s 3.7GHz limitation. And speaking of frequency boosts, AMD’s Extended Frequency Range system has been tweaked as well.

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

Extended Frequency Range (XFR) debuted with Ryzen last year and is designed to offer a bit more overclocking headroom. The CPU is capable of adjusting its clock speed upward by as much as 100MHz if conditions are right and can adjust its clock in 25MHz increments. None of that capability has changed. What is new this time around is the way the CPU responds to its thermal environment and the way AMD is communicating XFR functionality going forward. Instead of advertising the CPU as having a “standard” maximum Turbo clock with XFR as the proverbial cherry on top, AMD is advertising maximum clock rate, period. This is a messaging change, not an engineering shift; XFR2 still uses the same fine-grained 25MHz clock shifts to maximize performance.

The thermal responsiveness is a new feature AMD is introducing with Ryzen 2. Enthusiasts tend to deploy better CPU coolers and to pay more attention to thermal management, but the performance impact of a higher-end cooler is typically nil, assuming your chip wasn’t overheating to start with. AMD’s XFR2 monitors the CPU’s temperature and overall operating state and allows the chip to clock slightly more aggressively if it maintains an appropriate thermal envelope. According to AMD, the gap between a basic CPU cooler and an ultra high-end model is about 7 percent, with its own Wraith Prism accounting for about a 4 percent clock boost over base.

Since AMD ships a Wraith Prism with every retail-packaged Ryzen 7 2700X, this means you should get a small clock boost out of the gate, with a little bit of performance headroom left over if you’re the kind of person who splurges for a high-end Noctua (or equivalent) cooler.

AMD’s Wraith Prism

We normally don’t spend much time evaluating stock CPU coolers, but AMD made a major point of announcing that Ryzen 7 2700X would ship with a new version of the Wraith cooler they’ve used before. The Wraith Prism includes multiple direct contact heat pipes, a copper base, integrated LEDs (hence the “Prism” moniker), switchable fan profile support, and a “Dark mode for RGB Heathens.”

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

During the Ryzen 7 2xxx launch event AMD told us it believes this cooler should stand up well against entry-level enthusiast coolers. While a full CPU cooler comparison is beyond the scope of this review, we deliberately tested the Ryzen 7 2700X with the AMD Wraith Prism and AMD’s default thermal interface material (TIM).

AMD’s New X470 Chipset, StoreMI

Once upon a time (the late 1990s), chipset launches were a genuinely momentous event. Granted, much of the buzz was driven by the fact that chipsets of the era often resembled micro-scale trainwrecks, but finding out if your favorite chipset vendor had managed to fix any of the bugs they’d been shipping for the past year was practically a national pastime. Today, chipset updates slip by nearly unnoticed.

X470 boards are supposed to include improved power infrastructure to allow for higher clocks on Ryzen 2xxx chips, but the degree to which this matters will depend on how solid your motherboard’s power infrastructure was in the first place. If you’re buying into Ryzen for the first time, there’s no reason not to use X470, but if you’ve got an X370 board you’re happy with, it should work perfectly well with a 2000-series Ryzen CPU. The only major new feature, and one we sadly didn’t have time to test for this review, is StoreMI — a new software caching solution that appears to operate in much the same fashion as an SSD cache drive.

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

StoreMI is a software solution AMD purchased and is including with all X470 motherboards. It’s designed to fuse two different storage solutions like an SSD and hard drive into a single unified pool of storage. AMD claims StoreMI can even support Intel Optane, which is kind of hilarious if true, but we weren’t able to put that claim to the test. (AMD also took pains to note that it’s StoreMI that can support Optane, rather than AMD itself).

Everything AMD has shared about StoreMI to-date suggests that it offers the same performance benefits as a standard SSD cache drive, even if it differs in the way it pools multiple storage devices together instead of creating an invisible partition where data from an HDD is cached to an SSD for faster performance. Conceptually, it’s similar to a RAID 0, and it carries the same risks. Just like a RAID 0, the loss of one drive means the loss of data on both drives.

What We’re Looking For

Before we hit the results, let’s talk a bit about the shape of the fight. AMD’s Ryzen typically isn’t as efficient as Intel’s Coffee Lake in single-threaded code. Its strength is in well-threaded applications, particularly if the application scales well with SMT (Intel calls its application of this technology Hyper-Threading). The Core i7-8700K won our overall nod for best desktop processor because it combined best-in-class single-threaded performance with very strong multi-threaded scaling. While the Ryzen 7 1800X can still beat the Core i7-8700K in a number of multi-threaded tests, we felt the Intel CPU presented a better overall value.

We don’t expect to see AMD winning single-threaded competitions against the Core i7-8700K outright; the clock gains AMD has eked out often aren’t large enough. What we’re watching for is to see if AMD’s new Ryzen 7 2700X can open a large enough gap between itself and the Core i7-8700K to take the overall pole position again.

Test Setup

We tested the Ryzen 7 2700X and Ryzen 7 1800X using an MSI X470 M7 AC and 32GB of DDR4-3200 RAM clocked at that frequency. All of our testbeds used an M.2 Samsung 960 Pro for primary storage, and the latest version of Windows 10 with all updates installed. Our test results are displayed in the slideshow below:

Gaming Benchmarks

The game benchmark side of the equation is lighter than I’d like — I had to leave Hitman out due to oddly low performance I’m still trying to track down. The figures we’ve included, however, are enough to make the point. With a GPU like the GTX 1080 Ti, there’s not much point to testing 1080p, and above 1080p, there’s virtually no difference between AMD and Intel. In the few tests where this isn’t true, the Ryzen 7 2700X turns in faster results than the 1800X, implying that the chip also helps close the 1080p gap on sheer clock speed.

We may revisit the 1080p question in days to come, but nothing we’ve seen over the past year has led us to think there’s any meaningful difference between how AMD and Intel drive top-end GPUs at 1440p or 4K across the general game industry. Individual titles can obviously show variance, but on the whole the two platforms are neck and neck.

Power Consumption and the Wraith Prism

This isn’t going to be a full power consumption comparison; the power supply I used to use for testing has developed a short and I’m stuck on a much older, less-efficient 1200W unit. As such, none of the data gathered from it is comparable with previous tests, and we’re keeping this Ryzen-centric.

We’ve tested the Ryzen 7 1800X and Ryzen 7 2700X in a 20-minute Prime95 torture session using in-place FFTs for maximum power consumption. Power consumption figures are shown below:

AMD Ryzen 7 2700X Review: Can AMD Cream Intel’s Coffee Lake?

I decided to test the difference in power consumption at DDR4-1866 and stock voltages compared with 1.35v and DDR4-3200. The gap is much larger than I expected. Increasing RAM clocks to DDR4-3200 and using 1.35v for DRAM voltage increased power consumption by a massive 56W. I’ve noted these kind of jumps when testing Intel platforms as well (I’ve never tried comparing relative efficiency between the two and can’t comment on it at this point). Even for 32GB of RAM, a 56W jump in power consumption seems hefty — but that’s what we measured.

As far as CPU power consumption is concerned, we saw power consumption figures that align with what AMD told us to expect to see. Power consumption has ticked up from the 2700X to the 1800X, but so has overall performance. This is a worst-case power consumption figure — Prime 95 hammers a CPU’s AVX units — so don’t expect to see these kinds of figures if you’re doing other kinds of work.

As for AMD’s Wraith Prism cooler, it kept the CPU at 83.75 degrees after a 20-minute stress test in Prime 95, with our RAM clocked at DDR4-3200 and 67C when we clocked our RAM at DDR4-1866. No, that’s not a typo. The CPU clock stayed pegged to 3.9GHz throughout the torture test, with occasional excursions to 3925MHz when we tested DDR4-3200 and held a 4GHz constant clock when we tested DDR4-1866.

Now, 83.75 degrees isn’t great, but remember we’re using AMD’s stock cooler and a stock thermal pad alongside settings that put a great deal of thermal stress on the CPU. The 67C temperature after 20 minutes of Prime95 is quite good for a cooler in this budget category.

If there’s a caveat to the Wraith Prism, it’s that the cooler isn’t really all that quiet. This is the sort of thing you can tune on a modern motherboard, and there’s enough headroom in the Wraith Prism to give you space to reduce the fan noise to keep the system quieter, but be aware going in that this isn’t a whisper-silent cooler. Those of you who opt for Noctua’s giant coolers to minimize system noise won’t find yourselves switching. But the Wraith Prism does a solid job.

AMD Reclaims Ryzen’s Previous Pole Position

The Ryzen 7 2700X is not the show-stopper the Ryzen 7 1800X was. But it’s something every bit as important: A measured, genuine improvement executed on-time. As important as first-generation Ryzen was to AMD, the company needed to do more than just launch an architectural revision. It needed to show it could keep executing its roadmap and deliver a steady cadence of improvement over time.

Fans of Intel’s Core i7-8700K will still have a lot to like after this review, unlike poor Core i7-7700K owners, who cheerfully picked up Intel’s latest 7th Generation CPU in January 2017 only to be run over by Ryzen a few months later. High clocks, six cores, and strong IPC make the 8700K a formidable foe and it still punches above its weight class. But the additional 10 percent performance the Ryzen 7 2700X delivers gives AMD’s latest core room to maneuver. Everywhere the 1800X was outperformed, the 2700X is outperformed by less. Everywhere the 1800X won, the 2700X wins by more.

If you know you need single-threaded performance above everything else, the Core i7-8700K is still the top CPU on the market. But in well-threaded code, especially rendering applications, the Ryzen 7 2700X is superior. The fact that it’s $20 cheaper and ships with a decent stock cooler hurts nothing, either. Intel may already be planning its next counter-stroke, but until those rumors become reality, AMD has our nod for top overall CPU.

Continue reading

Intel’s Raja Koduri to Present at Samsung Foundry’s Upcoming Conference
Intel’s Raja Koduri to Present at Samsung Foundry’s Upcoming Conference

Intel's Raja Koduri will speak at a Samsung foundry event this week — and that's not something that would happen if Intel didn't have something to say.

Ryzen 9 5950X and 5900X Review: AMD Unleashes Zen 3 Against Intel’s Last Performance Bastions
Ryzen 9 5950X and 5900X Review: AMD Unleashes Zen 3 Against Intel’s Last Performance Bastions

AMD continues its onslaught on what was once Intel's undisputed turf.

Leaked Benchmarks Paint Conflicting Picture of Intel’s Rocket Lake
Leaked Benchmarks Paint Conflicting Picture of Intel’s Rocket Lake

Rumors about Rocket Lake have pointed in two opposite directions recently, but the more competitive figures are more likely to be true.

Intel’s Iris Xe Max Discrete GPU Is Slower Than the Integrated Version
Intel’s Iris Xe Max Discrete GPU Is Slower Than the Integrated Version

Intel's Iris Xe Max has debuted, but the discrete GPUs performance is a bit odd, as new benchmarks show.