Ever since their arrival in March of this year, AMD’s Ryzen desktop CPUs have proven themselves compelling alternatives to Intel’s chips at nearly every common price point. Ryzen CPUs generally bring more cores and threads to the table than equivalently-priced Intel CPUs, and recent sales have made it easier than ever to get gobs of multithreaded computing power for less.
Intel’s latest Coffee Lake CPUs have closed the multithreaded performance gap in much of our testing, but the company’s continued insistence on careful feature segmentation and the spotty availability of Coffee Lake parts in general has made the blue team’s next-gen broadside less damaging to the resurgent AMD than one might expect so far.
|Ryzen 7 2700U||4/8||2.2||3.8||10||1300 MHz||4MB||15W||Two channels
|Ryzen 5 2500U||2.0||3.6||8||1100 MHz|
AMD’s renewed competitiveness in the enthusiast desktop is a border skirmish compared to the war it’s getting ready to wage in the mobile-CPU marketplace. Most PCs sold these days are laptops of some kind, and AMD’s Ryzen Mobile chips are its latest in a long line of “accelerated processing units”: CPU cores with a powerful on-die Radeon graphics processor. For the first time since the advent of the APU, AMD has both competitive CPU cores and a cutting-edge graphics-processing unit that it can bring together. With four Zen CPU cores and eight threads, plus as many as 640 Radeon Vega compute units, the Ryzen 5 2500U and Ryzen 7 2700U stuff a ton of potential computing horsepower into their 15W TDPs.
In an unusual reversal of roles for this year of CPU upheaval, though, Intel beat AMD to the punch with its eighth-gen mobile CPUs, which have been available for some time in shipping systems since shortly after their August launch. Kaby Lake Refresh brings the same four cores and eight threads to 15W CPUs that AMD is targeting with the CPU side of the Ryzen Mobile family.
The fact that this first battle is being waged in a 15W theater may help AMD compete better with Intel’s chips in general. In the no-holds-barred arena afforded by 95W-and-higher TDPs, Intel can often clock its client cores to the moon and render discussions of IPC differences versus Zen moot in the wash of its chips’ clock-rocket exhaust. Shrink that thermal envelope to 15W, though, and those minor differences become a lot more relevant. The Ryzen 5 2500U boasts a single-core Turbo speed of 3.6 GHz, a clock-for-clock match for Intel’s Core i5-8350U, and its base clock speed is 300 MHz higher than that of the Intel chip’s.
The improved Precision Boost intelligence in Raven Ridge chips also affords the Ryzen 5 2500U much more granular control over its per-core Turbo speeds versus its desktop cousins, too, meaning that the chip could potentially run more of its cores at higher clocks for longer as it distributes its thermal budget over given workloads. That improved precision seems essential for a thermally-restrictive environment like a laptop chassis.
HP’s Envy x360 takes the stage
There’s more to Ryzen Mobile, but we’re here to test the thing, not dissect it further. For more information on the guts of Ryzen Mobile, be sure to check out my launch-day coverage. Let’s talk a bit about the first system with a Ryzen Mobile APU inside: HP’s Envy x360.
Given the importance of Ryzen Mobile’s performance to the competitive landscape of mobile computing, we went out and bought one at retail to see how it handles. Our $750 test system pairs a Ryzen 5 2500U APU with a one-terabyte, 7200-RPM mechanical hard drive, 8 GB of dual-channel DDR4-2400 RAM, a 15.6″, 1920×1080 screen with a convertible hinge. Most impressively, this rather beastly system is just 4.75 lb (2.15 kg) and slips into a body just 0.77″ (19.6 mm) thick. Svelte. The massive-for-a-laptop Bang and Olufsen speaker array above the keyboard on this thing is the best-sounding mobile audio setup I’ve heard, too.
We can never end our judgment at looks alone, though, and choosing a mechanical hard drive, no matter how good, as the primary storage device in any computer is an unfortunate decision in 2017. Any virtues of the Ryzen APU are eclipsed by the agony of tangling with the limits of mechanical storage. For the moment, HP is offering $100 off a custom Envy x360 from its website, making a 256GB SSD an affordable upgrade from the factory at $130. $760 for this same system with a 256GB SSD upgrade is an immeasurably better starting position for the Envy x360 and its Ryzen 5 2500U.
To make benchmarking the Envy bearable (and to put it on a level playing field with the Acer Swift 3s I just reviewed), I stuck a Samsung 960 EVO 500GB NVMe from our stockpile inside. This setup won’t be representative of the Envy x360 you might buy at retail, but it’s a much fairer starting point for the Ryzen 5 2500U.
HP also seems to have put a 6-bit IPS panel of some kind in the Envy x360 to show the output of the Vega 8 IGP to the world, another decision that puts the wrong foot forward for this system. Color banding in photos, on web sites, and apps with any kind of fine gradations between colors will be immediately obvious. I find this decision quite odd, given the Envy’s Wacom pen support and the potential content-creation chops of the Ryzen 5 2500U.
I could go on about the Envy’s feathery, feedback-free keyboard and aggressive fan, but this isn’t a review of the x360 in and of itself. HP’s engineers can certainly design a dashing and classy-looking notebook, but looks alone don’t make a solid PC. I’d recommend waiting and seeing how other Ryzen Mobile systems shake out unless you simply must have one of these APUs in a laptop today. The Envy x360 falls short in enough places that I’d have a hard time justifying its price tag.
For information about the Envy x360’s battery life, refer to our separate article published after this piece hit the wires.
Our testing methods
As always, we did our best to deliver clean benchmark numbers. We ran each benchmark test at least three times and took the median of the results.
Here are the specifications of our test systems:
|Acer Swift 3
|Acer Swift 3
|Acer Swift 3
|Alienware 13 R3||HP Envy x360|
|CPU||Intel Core i5-8250U||Intel Core i5-7200U||Intel Core i7-7700HQ||AMD Ryzen 5 2500U|
|Memory||8GB (2x4GB) DDR4-2400||8GB (2x4GB) LPDDR3-1866||8GB (2x4GB) DDR4-2133||16GB (2x8GB) DDR4-2666||8GB (2x4GB) DDR4-2400|
|GPU||Intel UHD Graphics 620||Nvidia GeForce MX150||Intel HD Graphics 620||Nvidia GeForce GTX 1060 6GB||AMD Radeon Vega 8|
|Graphics memory||N/A||2GB GDDR5, 6 GT/s effective||N/A||6GB GDDR5, 8 GT/s effective||N/A|
|Storage||Intel SSD 600p 256GB NVMe SSD||Samsung PM961 512GB NVMe SSD||Samsung 960 EVO 500GB SSD|
|Battery||3220 mAh Li-ion||76 Wh Li-ion||55.8 Wh Li-ion|
Our thanks to Intel for providing the three Acer Swift 3 systems for our testing. The Alienware 13 R3 playing host to our Core i7-7700HQ is my personal system and was not provided by a manufacturer for evaluation. The Tech Report independently acquired the HP Envy x360 and its Ryzen 5 2500U for testing.
Some additional notes regarding our testing methods:
- All systems were configured to use Windows’ default “Balanced” power plan over the course of our testing.
- Unless otherwise noted, all tests were conducted with display resolutions of 1920×1080 and refresh rates of 60 Hz. Vsync was disabled using the graphics-driver control panel, where possible.
- All drivers and system firmwares were updated to the most recent versions publicly available before testing.
Our testing methods are generally publicly available and reproducible. If you have questions regarding our methods, feel free to leave a comment on this article or join us in our forums.
Memory subsystem performance
Let’s kick off our testing with a quick look at the main memory performance from these systems using the built-in benchmarks from the AIDA64 utility.
Given that it has a hungry CPU and GPU in the same package, it’s good that the Ryzen 5 2500U nearly sweeps our synthetic memory bandwidth tests.
Despite the move to a single Zen core complex, however, the Ryzen 5 2500U doesn’t improve on its desktop counterparts’ tendency to post high memory access latency numbers. That higher latency could reduce the chip’s effective memory bandwidth compared to the competition.
Some quick synthetic math tests
AIDA64 also includes some useful micro-benchmarks that we can use to sketch out broad differences among CPUs on our bench. The PhotoWorxx test uses AVX2 instructions on all of these chips. The CPU Hash integer benchmark uses AVX, while the single-precision FPU Julia and double-precision Mandel tests use AVX2 with FMA.
Photoworxx proves a strong start for the Ryzen 5 2500U. The AMD chip pulls about even with the Core i5-8250U and its freshly-doubled core count in this test.
CPU Hash would seem to deliver a resounding win for the Ryzen 5 2500U, but remember that the Zen architecture supports Intel’s SHA Extensions. The Ryzen 5 2500U can accelerate SHA-1 hashing, while Intel’s chips have to do it the hard way with good old CPU muscle. Since AIDA64 uses SHA-1 for this benchmark, the AMD chip can blast past the Intel CPUs in this test.
In these AVX-intensive floating-point tests, the Ryzen 5 2500U pulls ahead of the Core i5-7200U and its dual cores, but it can’t quite catch the Core i5-8250U. That’s because the 2500U’s SIMD units are half as wide as those of the i5-8250U’s, so even with a core-count and potential clock-speed advantage, we should only expect the AMD chip to end up slightly ahead of the i5-7200U.
Zen CPU performance is kind of a known quantity at this point, so while it’s exciting to have four of those cores in 15W, it’s even more exciting to see how the Vega 8 integrated graphics processor performs. Let’s do that now.
Dota 2 (Fastest)
Let’s kick off our gaming tests with one of the most popular games on the planet today. Dota 2 has a reputation for running on everything, even Intel integrated graphics, thanks to a wide and forgiving range of graphics options. More importantly, Dota 2 has a powerful built-in replay system that allows us to precisely replicate the most intense moments in a match from a given player’s point of view, letting us reliably repeat the game’s chaotic battles for benchmarking.
We’ll begin our Dota 2 trek with the Fastest preset at 1920×1080, which cranks down the resolution scaling setting to 52% or so and disables virtually all of the game’s eye candy. It ain’t much to look at with these settings, but budding e-sports pros can at least get in on the fun with Fastest.
Well, these numbers aren’t quite what I expected. It would seem that Dota 2 is pretty CPU-bound with its Fastest preset, even on integrated graphics, and that’s bad news for IGPs sharing a common memory interface with a CPU. The Vega 8 IGP is definitely somewhat faster than the Intel parts, but its 99th-percentile frame time isn’t much better. The GeForce MX150 and its dedicated GDDR5 RAM score a major victory over the IGPs here in both performance potential and delivered smoothness.
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 our graphics card spends beyond certain frame-time thresholds, each with an important implication for gaming smoothness. Recall that our graphics-card tests all consist of one-minute test runs and that 1000 ms equals one second to fully appreciate this data.
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 30-Hz refresh rate. Go lower than 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, while a constant stream of frames at 8.3-ms intervals would correspond to 120 FPS.
As the spiky frame-time plots above suggested would happen, all of the IGPs post enough time past 50 ms that noticeable hitches and stutters will likely be the order of the day. The Vega 8 IGP fares the worst of this bunch by that metric. Click over to the 33.3-ms graph, and AMD’s IGP spends twice as much time as Intel’s UHD Graphics 620 working on those tough frames.
We have to shift our attention to the 16.7-ms graph to find the crossover point where the Vega 8 delivers superior results to the Intel IGP duo. Although the Vega 8 spends two fewer seconds past 16.7 ms compared to the UHD Graphics 620, its higher time-spent-beyond figures at 33.3 ms and 50 ms means that the Radoen’s potentially more fluid experience will be punctuated by more hitches and stutters than the Intel IGPs deliver.
For all that, I wouldn’t read too much into these chips’ performance with Dota 2‘s Fastest preset. If the game truly is CPU-bound with these settings, our results would suggest that it might be beneficial to find a recipe of settings that shift more of the work to the IGP. We may need to retest with the “Fastest+” preset and see what kind of effects it has on these results.
Dota 2 (Best Looking)
You can’t have fast without slow, and Dota 2 offers a preset called Best Looking that turns up all of the game’s eye candy and returns resolution scaling to 100%. I turned on Best Looking while keeping resolution at 1920×1080.
Before I dive into these numbers, I should be clear that we’re not trying to slag any of these chips by going with this setup. We’re not expecting playable frame rates, for the most part. This is a stress test designed to really throw the differences in processing power among these IGPs into stark relief.
Dota 2‘s Best Looking settings certainly put some room between the integrated graphics processors on the bench today. Even if it’s not turning in playable frame rates, on average, the Vega 8 IGP is at least flirting with them. The Radeon also shaves nearly 20 ms off the 99th-percentile frame time of Intel’s latest and greatest entry-level IGP, even if it’s still far higher than what we’d like for a smooth experience.
For all its spikiness at these settings, the Vega 8 spends far, far less time than the Intel IGPs past the 50-ms mark. The Radeon also struts its stuff at 33.3 ms, where it spends less than a third of the time on those tough frames compared to the UHD Graphics 620’s results. Even if one wouldn’t want to play Dota 2 on the Radeon at these settings, these results prove that AMD’s Vega IGP is in its own class for integrated graphics performance. We’d still have to relax some eye candy to get Dota 2 above the 30-FPS bar that’s widely accepted as “playable” for this class of product, but the AMD IGP is a lot closer to clearing that bar than Intel’s latest.
Rocket League (1280×720)
Rocket League is another wildly popular title with a runs-anywhere reputation. We dialed in some fairly high settings at a low resolution to see how it played on this trio. Like Dota 2, Rocket League has a wonderful replay interface that lets us precisely play back a given match in-engine and from a player’s viewpoint.
The Vega 8 IGP gets off to a strong start with Rocket League. Even at this low resolution, we avoid the apparent CPU-bound coffin corner that Dota 2 hits with its lowest preset. In both performance potential (as measured by its average FPS) and in delivered smoothness (as measured by its 99th-percentile frame time), Vega 8 stomps all over the Intel IGPs.
Thanks to one big spike toward the middle of our test run, the Vega 8 does post some time past 50 ms and 33 ms in our time-spent-beyond filtering of the data, but that isn’t typical of the experience it delivers. The most important threshold to consider is at 16.7 ms, where the Vega 8 spends almost a quarter of the time that the UHD Graphics 620 does on frames that would drop the delivered frame rate below 60 FPS. Entry-level gamers can generally be assured of a smooth and fluid Rocket League experience at these settings.
Rocket League (1920×1080)
Our second round of e-sports stress testing cranks up the resolution to 1920×1080 while leaving all of our previous graphics settings in place.
Where Intel’s UHD Graphics 620 can’t quite clear the 30-FPS average that defines “playable” for many folks at these settings, the Vega 8 can. Its 99th-percentile frame time is just a hair above the 33.3-ms mark, too, so the Radeon is clearing that bar for the vast majority of our test run. If you prefer high resolution to high frame rates in Rocket League, the Vega 8 IGP can do it.
In fact, the Vega 8 spends just half a second beyond 33.3 ms in this test—a figure you’d barely notice in practice over the course of our one-minute test run. The UHD Graphics 620 spends an order of magnitude more time past this threshold.
For the e-sports portion of our gaming tests, the Vega 8 IGP proves itself far more capable than the GT2 configurations of the Gen9 IGP shipping in Intel’s Kaby Lake and Kaby Lake Refresh CPUs. One still might need to tweak the graphics settings in these titles for the best experience on AMD’s Vega IGPs, but the red team offers considerably more breathing room than Intel’s IGPs do.
The Witcher 3 (1280×720)
The Witcher 3 kicks off our primarily AAA-focused phase of testing. I wanted to test Tomb Raider‘s 2013 reboot on the Ryzen 5 2500U to get a sense of how it compared to the Intel IGPs in that title, but the Radeon’s drivers forced vsync on with Tomb Raider regardless of the in-game settings I chose. Although Tomb Raider certainly runs well on the Ryzen 5 2500U, it’s hard to get a sense of its performance at a locked 30 FPS.
That left the significantly-more-demanding The Witcher 3 to serve as my older triple-A test mule. I selected a blend of medium and low settings to really give these IGPs a workout with this demanding title.
Despite the hair-raising nature of that match-up on paper, the Vega 8 runs The Witcher 3 just 15% slower than the discrete GeForce does at these settings. The Radeon’s significantly spikier graph and higher 99th-percentile frame time don’t paint as favorable a picture for its performance, but hey—this is integrated graphics we’re talking about. Even with that 99th-percentile result, it’s still kind of a wonder that the Vega 8 IGP can turn in a playable experience with The Witcher 3 when the blue team’s IGPs are running the game so slowly that Geralt literally hangs in midair while he runs.
My generally favorable impression of The Witcher 3 on the Vega 8 is borne out by our time-spent-beyond-33.3-ms graph. While the Radeon IGP does spend enough time past 50 ms and 33.3 ms to be noticeable in The Witcher 3, it’s not enough to be ruinous over the course of our one-minute test run. I’d happily play The Witcher 3 on either of these systems, even if I would prefer the GeForce.
The Witcher 3 (1600×900)
Time for another stress test. I bumped The Witcher 3 to 1600×900 while keeping my graphics settings the same as I did at 1280×720.
The extra resolution harms the performance potential of both of these graphics processors, but surprisingly, it doesn’t increase the Vega 8 IGP’s 99th-percentile frame time at all (not that the Radeon had a stellar result to begin with, but hey). The Ryzen 5 2500U can’t quite clear the 30-FPS bar that we want to see.
The GeForce’s average-FPS result with these settings is still flirting with a playable experience, but its 40.5-ms 99th-percentile frame time suggests some bumps in the road. Let’s look at our time-spent-beyond-X graphs to see just how bumpy things got.
If you’re willing to rough it in exchange for every possible pixel of resolution, the GeForce MX150 spends just over a second and a half of our one-minute test run on tough frames that would drop the frame rate below 30 FPS. The Radeon is thoroughly outclassed at these settings, so gamers will want to stick with lower settings, less eye candy, or both approaches to keep frame rates playable and frame delivery smooth-ish.
To close out our triple-A testing, it only seemed fitting to give Doom‘s 2016 reboot a try. I cued up the game’s Vulkan renderer at 1280×720 with the low graphics preset, compute shaders, and a few of the game’s minimally-intensive post-processing settings enabled.
This is the kind of result that frame-time analysis was built to capture. Even though the Radeon isn’t churning out the frames nearly as fast as the MX150 can, both cards turn in roughly the same 99th-percentile measure of delivered smoothness. You’d expect as much, given the furriness of the frame-time graph for the Nvidia chip.
Although the graph for the MX150 may look ugly, that kind of result is repeatable. After my first glimpse of those results, I re-ran the test from scratch to see if I could get better performance, to no avail. I tried monitoring the MX150 with the GPU-Z utility to see if I could get to the bottom of the issue, and the card frequently reported running into a thermal limit during the course of my tests. That thermal limit resulted in swings in clock speeds of a couple hundred MHz, so it could be that Doom just doesn’t react well to that kind of variance.
For its part, the Vega 8 comes quite close to clearing the “playable” bar in Doom at these settings, although I have to note that a 30- to 40-FPS range in this title really does not do its lightning-quick gameplay any justice. You can play Doom on the Ryzen 5 2500U, but it’s not the most fulfilling experience. Still, it could be worse.
Starting at the 33.3-ms mark, the Vega 8 turns in an admirable performance. The chip spends barely any time on tough frames that take longer than 33.3 ms to produce, so one can be assured of a consistently smooth experience at Doom‘s tougher moments, even if its gameplay isn’t as fluid as I’d like.
For all its spikiness, the MX150 still turns in less time past 33.3 ms than the Vega 8 does by a small margin, although we’d really hope to see no time at all in the green team’s bucket given the card’s high average frame rate. The 16.7-ms threshold really isn’t that concerning for the GeForce, either. The net result of all that furriness in the graph is that the MX150 spends just about six seconds past 16.7 ms working on tough frames. We’d still like to see less variance, but it’s clear that the MX150 is still delivering a different class of experience than the Vega 8 in this title.
At this early stage, the Vega 8 graphics processor is a bright spot for the Ryzen 5 2500U. Although Dota 2 players might need to carefully tweak their graphics settings for the best experience on integrated graphics, that’s not a complaint that’s unique to AMD’s APUs. In a range of other titles, the Vega 8 IGP provides much better gameplay experiences than Intel’s entry-level IGPs, presuming that a given title can be made playable on the blue team’s IGPs at all.
All that performance comes from the same total CPU power budget on paper as that of Intel’s latest chips, too, so AMD has undeniably crammed an impressive amount of graphics oomph into one low-power package.
The Vega 8 is no match for a GeForce MX150, of course (though it comes surprisingly close in The Witcher 3). The integrated Radeon could be a better performer than older versions of Nvidia’s entry-level discrete chips, though. Without testing, it’s hard to say, and we don’t have a GeForce 940MX-equipped laptop handy to find out.
I’m certainly curious to see what the more-capable Vega 10 IGP in the Ryzen 7 2700U is capable of now, and the prospect of a higher-TDP desktop chip with this CPU-and-graphics combo inside is especially exciting. Hopefully we’ll be able to find out with time.
Now that we’ve got an idea of the Vega 8 IGP’s gaming performance, it’s time to find out how the CPU side of the die handles.
We don’t usually include Futuremark’s wide-ranging PCMark benchmark in our test suite, but I’m making an exception for these general-purpose laptops. PCMark 10 tests some things that we don’t have tools for, like video conferencing, app launch times, and general productivity performance with common office apps like LibreOffice Writer and Calc. Considering that PCMark is a one-click test, it was hardly a burden at all to rotate it in for this review. Thanks to Futuremark for providing us with access to a PCMark 10 Professional license for these tests.
Let’s start off with these systems’ overall scores, including gaming results. Even with the Ryzen 5 2500U’s powerful Radeon IGP in the picture, it doesn’t open that much of a lead over the Acer Swift 3 with a Core i5-8250U alone inside. Hm.
The essentials benchmark tests web-browsing performance, video conferencing, and a range of app start-up times. This test would seem to favor single-threaded performance and responsiveness most of all, so it’s perhaps not a surprise that Intel’s chips still hold a lead. For folks who use their PCs for the most basic of tasks, Intel’s single-threaded performance edge and technologies like Speed Shift seem to give it quite an edge.
The productivity test checks word-processing acuity and spreadsheet performance with both lightweight and heavy-duty number-crunching. Some portions of the test are even OpenCL-enabled. Still, the nature of this test would seem to favor per-core performance above all, and the Ryzen 5 2500U still can’t eke out an edge.
The digital-content-creation portion of this test covers three phases: photo editing tasks with large raw files from a range of DSLRs (some parts of which are OpenCL-enabled), video editing across both CPU and OpenCL code paths, and rendering and visualization of 3D models. Given those workloads, it’s no shock that the Alienware 13 R3 comes out on top, but the Ryzen 5 2500U finally has an opportunity to demonstrate superiority over even the MX150-powered Acer Swift 3.
Although we already know the story the PCMark gaming test tells, it’s nice to see it confirmed in yet another benchmark.
PCMark may not be the most demanding benchmark, but its workloads seem typical of how the middle of the bell curve of users will experience PC performance. For common web-browsing and office tasks, Intel CPUs would still seem to have a substantial edge in responsiveness and snappiness. Let’s see if we can tease out some of those differences further with some directed benchmarks.
Compiling code with GCC
Our resident code monkey, Bruno Ferreira, helped us put together this code-compiling test. Qtbench records the time needed 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.
Our compiling benchmark has always had a scent of Amdahl’s Law about it. Even with four cores and eight threads across the board, the Ryzen 5 2500U can’t catch the Core i5-8250U.
File compression with 7-zip
Although the Ryzen 5 2500U is a little behind the i5-8250U’s file-compression performance with 7-zip, it’s just about as fast in the more common decompression task. Not bad.
Disk encryption with Veracrypt
Veracrypt, a continuation of the TrueCrypt project, offers a built-in benchmark that tests both the AES algorithm (which many of today’s CPUs can accelerate) and a variety of other algorithms that require good old CPU elbow grease to crunch. In the accelerated AES portion of the benchmark, the Ryzen 5 2500U can’t quite catch the i5-8250U, core-for-core and thread-for-thread.
In the unaccelerated Twofish portion of the benchmark, the Ryzen 5 still trails the Core i5-8250U.
The evergreen Cinebench benchmark is powered by Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. The test runs with a single thread and then with as many threads as possible.
In the single-threaded portion of the benchmark, the Ryzen 5 2500U beats out the Speed Shift-less i5-7200U, and it only narrowly trails the Core i5-8250U and i7-7700HQ.
Cinebench is really about its multithreaded portion, though. The Core i7-7700HQ still reigns supreme here, but the Ryzen 5 2500U handily outperforms the i5-8250U.
Cinebench seems to like AMD CPUs, and AMD certainly likes talking up Cinebench in its CPU marketing, so let’s see whether this win carries over to other CPU rendering tasks.
Blender is a widely-used, open-source 3D modeling and rendering application. The app can take advantage of AVX2 instructions on compatible CPUs. We chose the “bmw27” test file from Blender’s selection of benchmark scenes to put our CPUs through their paces.
Cinebench isn’t an anomaly among the systems we have at hand. The Ryzen 5 2500U shaves more than a minute off the Core i5-8250U’s performance here.
So why the major gap? My instinct is that the form factors of the systems on our test bench play a major part. The 15.6″ HP Envy x360 that hosts the Ryzen 5 2500U has lots of metal in its expansive chassis, and it has a powerful-sounding and aggressive fan inside. The Core i5-8250U is in a 14″ chassis with a quieter fan. Acer plans to release a 15.6″ Swift 3 with Ryzen Mobile APUs inside, so it’d be interesting to get the i5-8250U version of that same Swift 3 and see just how large the performance gap between those two systems remains.
Here’s a new benchmark for our test suite. Corona, as its developers put it, is a “high-performance (un)biased photorealistic renderer, available for Autodesk 3ds Max and as a standalone CLI application, and in development for Maxon Cinema 4D.”
The company has made a standalone benchmark with its rendering engine inside, so it was a no-brainer to give it a spin on these CPUs. The benchmark reports both a rays-per-second and time-to-completion figure, and we’re reporting the time-to-completion result.
Here, the Ryzen 5 2500U and the Core i5-8250U emerge dead-even.
Handbrake is a popular video-transcoding app that recently hit version 1.0.7. To see how it performs on these chips, we’re switching things up from some of our past reviews. Here, we converted a roughly two-minute 4K source file from an iPhone 6S into a 1920×1080, 30 FPS MKV using the HEVC algorithm implemented in the x265 open-source encoder. We otherwise left the preset at its default settings.
Handbrake seems to favor the wider SIMD units and higher per-core performance of the Core i5-8250U in this workload, but the Ryzen 5 2500U isn’t far behind.
CFD 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. We configured Euler3D to use every thread available from each of our CPUs.
It should be noted that the publicly-available Euler3D benchmark is compiled using Intel’s Fortran tools, a decision that its originators discuss in depth on the project page. Code produced this way may not perform at its best on Ryzen CPUs as a result, but this binary is apparently representative of the software that would be available in the field. A more neutral compiler might make for a better benchmark, but it may also not be representative of real-world results with real-world software, and we are generally concerned with real-world performance.
With that in mind, the Ryzen 5 2500U beats out the dual-core i5-7200U, but it can’t close the gap in Euler3D performance that typifies AMD CPUs versus their Intel competitors.
Digital audio workstation performance
One of the neatest additions to our test suite of late is the duo of DAWBench project files: DSP 2017 and VI 2017. The DSP benchmark tests the raw number of VST plugins a system can handle, while the complex VI project simulates a virtual instrument and sampling workload.
We used the latest version of the Reaper DAW for Windows as the platform for our tests. To simulate a demanding workload, we tested each CPU with a 24-bit depth and 96-KHz sampling rate, and at two ASIO buffer depths: a punishing 64 and a slightly-less-punishing 128. In response to popular demand, we’re also testing the same buffer depths at a sampling rate of 48 KHz. We added VSTs or notes of polyphony to each session until we started hearing popping or other audio artifacts. We used Focusrite’s Scarlett 2i2 audio interface and the latest version of the company’s own ASIO driver for monitoring purposes.
A very special thanks is in order here for Native Instruments, who kindly provided us with the Kontakt licenses necessary to run the DAWBench VI project file. We greatly appreciate NI’s support—this benchmark would not have been possible without the help of the folks there. Be sure to check out their many fine digital audio products.
The DAWBench VI test at 96 KHz and 64 samples is the most punishing of this bunch, and none of our mobile CPUs can handle it. Even the i7-7700HQ delivered crackling and popping with no voices of polyphony in play. Relax the buffer size to 128 samples, and the Ryzen 5 2500U stays more competitive with the Core i5-8250U than most any Ryzen CPU has with an equivalent Intel chip in our testing so far. Perhaps the single CCX of the Ryzen 5 2500U is helping here.
The DAWBench DSP test is less about agility and more about pure multithreaded grunt. Even so, the Ryzen 5 2500U can’t keep up at 96 KHz and 64 samples. Relax the buffer size again, and the 2500U hangs right with the i5-8250U.
With our sampling rate reduced, we return to the VI portion of DAWBench. Once again, the Ryzen 5 2500U hangs right with the i5-8250U.
Returning to the more straightforward DSP test at a 48 KHz sampling rate, the 2500U is slightly behind the i5-8250U, but its performance remains competitive at both buffer depths.
Aside from its one unusual stumble in the DAWBench DSP test at 96 KHz, the Ryzen 5 2500U is—strangely enough—the most competitive AMD CPU for DAW performance that we’ve tested of late. The single-CCX design of the 2500U seems to afford it better performance in the latency-sensitive DAWBench VI test than any of its dual- or quad-CCX counterparts can manage, and it hangs right with the i5-8250U in DSP testing.
The Ryzen 5 2500U is the ultimate expression of AMD’s APU concept that we’ve tested so far. Its four Zen CPU cores finally offer competitive performance for Intel’s latest chips in a wide range of tasks, and its Vega 8 integrated graphics processor generally offers smooth and responsive entry-level gaming experiences that the blue team’s comparably-positioned CPUs can’t match.
The Vega 8 IGP is the brightest spot for the Ryzen 5 2500U. Where Intel’s UHD Graphics 620 IGP might run out of gas even at the lowest possible resolutions and graphics settings in some titles, Vega 8 usually provides playable experiences even with more demanding games and sometimes even higher resolutions. It’s no replacement for a GeForce MX150, but Vega 8 certainly sets a high new bar for integrated graphics performance in this power envelope.
On the CPU side of the die, Zen still has a modest single-threaded performance deficit versus a similarly-clocked Skylake core in our testing, and that seems to hurt the Ryzen 5 2500U in productivity tasks like web browsing and office work where responsiveness is paramount. Outside of its wins in Blender and Cinebench, the Ryzen 5 2500U hits multithreaded parity with the i5-8250U more often than not, though whether that parity will hold outside of a 15-inch notebook with plenty of metal casing and an aggressive fan remains to be seen.
Overall, AMD finally has the competitive CPU cores it’s so desperately needed to go with the powerful graphics processors in its APUs, but for the middle of the bell curve, single-threaded performance and snappiness still matters most. Considering how precious that kind of performance is for a responsive user experience, AMD and its partners will have to make the case that the all-around competence of the Ryzen Mobile APU is worth trading away a bit of the single-threaded oomph of Kaby Lake Refresh for a broad swath of users.
The strongest way for AMD’s partners to make that case will be with price tags. As I outlined in my i5-8250U review, Intel’s partners can put together Core i5-powered ultrabooks with GeForce MX150 graphics, 8 GB of RAM, and 256 GB SSDs inside for around $700 to $800. If one of AMD’s partners can get a Ryzen 5 2500U into a thin-and-light laptop with a similar SSD and RAM configuration for around $600, we’ll really be talking.
With only one Ryzen Mobile system on the market right now, though, it’s too early to make broad proclamations about the success of AMD’s return to competition in mobile computing. Without a doubt, the Ryzen 5 2500U and Ryzen 7 2700U are AMD’s best and most competitive mobile APUs ever. Whether that achievement can persuade notebook buyers to choose a Ryzen system over one with Intel inside remains to be seen.
Like this review? Help us continue to perform the dozens upon dozens of hours of hands-on testing that go into articles like this one by becoming a TR subscriber today. Subscribers get exclusive benefits as part of their membership, and you can contribute any amount you’d like to support us. Best of all, we were doing crowdfunding before it was cool, so your entire contribution (less transaction fees) helps us continue our work. Thanks in advance for your support.