Today we're kicking off what is set to be many weeks of laptop component testing, as both AMD and Nvidia have lifted their embargos on next generation mobile hardware. On the AMD side, we have new Ryzen Mobile 5000 parts in the differentiated H and U-series, while Nvidia has launched the new GeForce RTX 30 mobile GPUs.
There's quite a number of new CPUs and GPUs launching today, so we're not going to be able to cover them all at once, but the goal will be to go over most of the key players in the coming weeks as they filter in over time. Of course, we've already made the introductions and we covered the news a few weeks back, but this marks the day when actual shipping products will become available and we are able to publish hands-on reviews and testing.
We're kicking off the journey with a look at Ryzen Mobile 5000, specifically the Ryzen 9 5980HS. The main improvement AMD has made to the new series is the upgrade to Zen 3 CPU cores, at least for the entirety of the high performance H-series. This brings all the advantages we saw from the 5000 series desktop line: higher IPC thanks to a redesigned core architecture, and the unification of the core, meaning we have just one 8-core CCX to deal with, rather than two 4-core CCXes. Same maximum eight core design, different layout that improves performance and latency.
On the mobile line specifically, AMD have also increased the total amount of L3 cache. We're going from 8MB split into 4MB per CCX, to 16MB of unified L3, a much needed improvement. There are also slightly higher frequencies available, as well as significant improvements to battery life with updates to power management across the APU.
There are also other areas that haven't received a substantial upgrade. The GPU design is virtually untouched, it's still 8 Vega compute units, just clocked higher. The memory controller has been updated to be more efficient, but supports the same memory tech. PCIe, display engines, multimedia capabilities, also the same. Still built on TSMC's 7nm process node.
When looking at AMD's Ryzen 5000 H-series, there are eight offerings: four "H" and "HX" parts, which have a default 45W power configuration, with the X denoting overclocking capabilities; and four "HS" series parts, at a lower 35W for power optimized designs.
The Ryzen 9 5980HS is the top of the line HS series chip, which brings an 8-core, 16-thread design, 16MB of L3 cache, a base clock of 3.0 GHz and a boost clock of 4.8 GHz. There's also 8 Vega compute units clocked up to 2.1 GHz.
Compared to last year's Ryzen 9 4900HS flagship, we're getting a bump from the Zen 2 to Zen 3 architecture, double the L3 cache, and a boost clock increase from 4.3 to 4.8 GHz with the same 8 core design. We're also seeing the same 8 Vega CUs, but with a 20% higher maximum frequency to work with.
Our test system for benchmarking the Ryzen 9 5980HS is a laptop that we're honestly a bit shocked can exist: the Asus ROG Flow X13. After testing the impressive Zephyrus G14 last year, we thought "yup, that's as small as an H-series device can realistically get while also providing discrete GPU acceleration". We were wrong. The X13 takes it one step further to provide an even more compact laptop body without sacrificing much in terms of performance - in fact, this Ryzen CPU inside will probably be faster in many areas.
The Asus ROG Flow X13 is seriously insane: it's got a 13.4-inch 3840 x 2400 display, which can give you an idea of the overall footprint. It weighs just 2.8 pounds and is 15.8mm thick, despite packing inside the Ryzen 9 5980HS, 32GB of LPDDR4X memory, and a GeForce GTX 1650 Max-Q GPU running up to 35W.
Granted, this GPU isn't as powerful as the RTX 2060 Max-Q in the Zephyrus - and that laptop has also been updated with Ryzen 5000 - but the Flow X13 is no larger than your typical 13-inch ultraportable, and no heavier either, packing an H-series CPU.
It does get better when you realize the Flow X13 can run the CPU at 35W and GPU at 35W simultaneously, but it can also power up the CPU to 45W when the discrete GPU is not being utilized in some power profiles. There's also other form factor benefits like the screen that flips around into a tablet mode, 62Wh battery, and a great keyboard for this sort of laptop.
Asus includes a neat port on the left side that provides direct PCIe lane access for their new external GPU solution, the XG Mobile, which means you can hook up the Flow X13 to a GPU as powerful as a 150W laptop-grade RTX 3080 for more performance. We don't have the XG Mobile on hand just yet, but we're keen on seeing what it can do.
Testing Notes
For today's benchmarking we've tested the Ryzen 9 5980HS in two power configurations: the usual 35W, which long term limits to 35W but does have a boost period above 50W for a couple of minutes; and a 45W power overclocked mode which is accessible through Asus' Armoury Crate utility. This will be compared apples-to-apples with other laptop CPUs running at similar power limits. The full list of laptops we tested is here.
As we've noted in previous reviews, we test like this to remove OEM-specific configurations and bottlenecks as possible. This normalized testing allows you to see how much performance you can get out of a given laptop cooling design, and of course, more power hungry parts require larger, heavier and often louder coolers. Intel H-series CPUs in particular are often run well above their 45W default configuration in gaming laptops, which leads to higher performance at the cost of heat, size and noise. Because power is so important to laptops and almost everything is designed with this metric in mind, we normalize for power and compare performance that way.
Benchmarks
Let's kick things off with our usual look at Cinebench R20 multi-thread performance. AMD's 8-core Ryzen H-series already dominated this chart, with Intel's best CPU, the Core i9-10980HK, only delivering a score in the mid-3000s at its default power level. With Ryzen 5000, AMD is sitting in the mid-4000s, and approaching 5000 when we run the 5980HS at 45W.
What we end up with is a modest 10% performance improvement for the 35W configuration versus the Ryzen 9 4900HS, although at 45W there is a gain in the 17% range versus the 4800H. That's probably not as good as some people were hoping for, given the 5800X on desktop is about 22% ahead of the 3700X in the Zen 3 to Zen 2 transition, but mobile CPUs are a more constrained platform.
The key improvement that has been made to Ryzen 5000 though isn't really the multi-thread performance, but single-thread performance.
With the Zen 3 architecture, the Ryzen 9 5980HS is able to match and slightly outperform Intel's Tiger Lake Core i7-1185G7, which is a 28W processor. Both top out at a maximum boost frequency of 4.8 GHz, so in this workload at least, AMD's Zen 3 cores and Intel's Willow Cove deliver similar IPC. It's just that the 5980HS overall gives you double the core count, which makes it significantly faster for multi-threaded workloads.
In Handbrake, which is a long term test that is mostly run in the long term power state, the 5980HS is just 7% faster than the 4900HS, which is lower than we expected. At 45W in comparison to the 4800H, the gap widens to 12%, however in both instances we are only seeing a modest improvement. With that said, Ryzen still destroys Intel's H-series processors by a significant margin, offering 30 percent more performance in the same power class.
Blender is a similar situation to what we just saw: with just a 5-10% performance improvement on last-generation parts, the 5980HS isn't that much faster for multi-threaded performance in heavy workloads like this. It is faster, just not heaps faster.
AMD CPUs remain a very solid choice for code compilation, with the 5980HS at 45W delivering 17% more performance than the 4800H, making it the fastest mobile processor we've tested. This allows Ryzen to maintain its nearly 30% lead on Intel's flagship processors in a core-for-core, power-for-power battle.
The gains aren't as significant in the Chromium compilation, a test that takes nearly two hours to complete and is a great stress test for thermally soaked laptops. Still, we're getting around that 10% performance improvement we've seen in a few benchmarks now.
Matlab was one of the weaker workloads for Ryzen processors in the past, as Intel's higher single-thread performance assisted in this test. However that's not the case anymore, with a handy 15 percent performance improvement allowing the Ryzen 9 5980HS to leapfrog the Core i9-10980HK into a leading position. AMD is only 8% faster here than Intel's best processor, but that's enough to make the 5980HS the better CPU for this application.
Thanks to a doubling of L3 cache, our Microsoft Excel workload shows massive gains for AMD's new Zen 3 APU. At 35W we're seeing a 13% gain on the 4900HS, but at 45W, the 5980HS is over 25% faster than the 4800H, which is a substantial improvement. The end result is that Ryzen is now neck and neck with Intel in this workload.
Ryzen 5000 does not provide a significant performance gain in PCMark 10's essentials test. The 5980HS is slightly faster than last-generation processors, but not significantly so, and regardless of which top-end processor you get, you'll have a similar experience for basic applications.
In the applications test, Ryzen shows a 10-15% gain on Zen 2 CPUs. This takes AMD into a position where they are delivering roughly the same performance overall than Intel's H and U-series processors. The 5980HS is still beaten by Intel's 1185G7 in the overall score, but on a per-test basis it trades blows with Intel winning some tests, and AMD others.
7-Zip compression was another workload where Ryzen was unable to match Intel's best H-series processors and we suspect that had to do with Ryzen 4000's smaller cache. With 16MB of L3 at the helm and better overall IPC, the Ryzen 9 5980HS is now able to match the Core i9-10980HK in this workload, at a lower boost power level.
Then when moving to decompression, the new Ryzen line-up easily beats Intel's offerings. The 5980HS is a solid 13 percent faster than the 4900HS here, which allows the 35W processor to outperform the Core i9-10980HK by over 30 percent. When pitting the 5980HS up against Intel's best Tiger Lake processor, the 1185G7, it's not even a contest because of the much higher core count.
In cryptography we're switching to a single-thread test as the multi-thread workload we used previously was becoming very memory limited on modern CPUs and we're more interested in how well the CPU core itself can perform AES crypto work.
With Ryzen 4000, AMD held a strong lead on Intel's H-series parts on a per-core basis, however was outperformed by both Ice Lake and Tiger Lake, Intel's newer architectures. Zen 3 is clearly much more competitive, and while not the faster CPU overall, it delivers competitive performance in most instances.
Acrobat PDF exporting has been a notoriously slow workload on AMD processors as the export is entirely single-threaded. However with the Zen 3 upgrade, the 5980HS is now able to jump into first position, with a 25% faster completion time than the Ryzen 9 4900HS. This allows the 5980HS to match Intel's Core i7-1185G7 at 28W, another indication that the single-thread performance on offer with AMD and Intel's latest processors is virtually the same.
We also see a huge performance uplift in Adobe Photoshop using the Puget Systems benchmark. There's up to a 30% generational leap in performance, taking the Ryzen 9 5980HS above the Core i9-10980HK and i7-1185G7 in this workload. This makes a new Ryzen 5000 H-series processor the best choice for Photoshop users.
Another great example of AMD's single-core performance is running a single instance of the warp stabilizer effect in Premiere. The 5980HS outperforms the 1185G7 here, and overall ends up 20 percent faster than the Ryzen 9 4900HS, a pretty decent generational improvement.
Next up we move into some accelerated workloads, starting with DaVinci Resolve. Unfortunately we had a few driver issues on the Nvidia side that prevented this test from running properly on the GTX 1650 Max-Q, so we just have integrated GPU data at the moment.
These are the perils of testing hardware before it becomes available on store shelves. Anyway, the Ryzen 9 5980HS which is GPU limited in this workload pretty heavily, is not substantially faster than previous Ryzen processors, and ultimately running Resolve on integrated graphics is a bad idea.
For accelerated Premiere exporting using integrated graphics, the Ryzen 9 5980HS is about on par with the Core i7-1165G7 - Tiger Lake's superior media encoding engine and faster GPU allow it to overcome its core deficit. We'd expect this to change when a discrete GPU is added into the mix, but you know, driver issues here as well.
One thing I was able to get running on this system with proper GPU acceleration was a 2 pass export using CPU encoding. Here Ryzen holds an advantage over previous generation processors, in line with other multi-threaded results we've seen so far. Ryzen when paired with a discrete Nvidia GPU should be very competitive, and we'll explore that further in future reviews when some of these early bugs are ironed out.
Gaming Benchmarks (iGPU)
Now let's move onto integrated GPU gaming to close this one out, just to give some sort of idea how much faster Ryzen Mobile 5000 is versus 4000 on the GPU front, given the GPU architecture is mostly the same. This will be much more important for U-series parts but we don't have any of those yet, so these are the 5980HS results.
In GTA V, the 5980HS and 4900HS deliver the same level of performance at 1080p using low settings. The clock speed gains don't seem to be fully utilized within the 35W power limit, so we're left with no improvement to speak of.
The Flow X13 overall is capable of gaming thanks to its discrete GTX 1650 Max-Q GPU, which allows it to outperform a similar Intel configuration.
In Civilization VI, the 5980HS is only about 3% faster than the 4900HS, so again this is not a substantial gain and points to Ryzen 5000 not having much of a GPU upgrade. This laptop is excellent when paired with the GTX 1650 though.
In CS:GO, the 5980HS fails to outperform Intel's Core i7-1165G7 on integrated graphics, only delivering around a 7% performance improvement on the 4900HS. However when GPU processing can be offloaded to a discrete GPU, the 5980HS is able to remove the GPU bottleneck and allow the system to shine.
Gears 5 is one of the most performance intensive benchmarks we run on integrated graphics, and it's fully GPU limited. Intel's 1165G7 with its 96 execution unit Xe GPU is able to outperform AMD's flagship Ryzen 9 5980HS; we're looking at a 13% advantage for Intel with LPDDR4X in both systems.
Intel is faster in F1 2019 with the 1165G7 holding on to a 10% performance lead at 28W over the 5980HS. With the Zen 3 APU providing the same performance as the previous generation of APUs, there's no reason to upgrade for integrated graphics performance because there's none.
Finally we take a look at Rainbow Six Siege running at 1080p with medium settings. Tiger Lake is faster again, although Ryzen shows capability with a discrete GPU attached. Asus have clearly made the right choice to include a GTX 1650 Max-Q as it's almost as fast as the full 50W GTX 1650 in some of these games.
Performance Breakdown
When looking at a performance breakdown, we have a limited set of comparisons between the Ryzen 9 5980HS and Ryzen 9 4900HS, a flagship vs. flagship battle between generations. With the 5000 series, AMD is able to provide substantial single-thread performance gains between 20 and 25% in the best cases, while multi-threaded performance improvements are more modest.
The 5980HS in a powered up configuration running at 45W also gives us a good idea of what to expect for the regular H and HX processors. Compared to the 4800H, the 5980HS is impressive, with slightly higher gains than we saw at 35W. In the best cases we saw up to a 30 percent advantage for the Zen 3 design.
While not a totally fair comparison as these two CPUs are not in the same power class, comparing the 5980HS to the Core i7-1185G7 at 28W gives us a good idea of where Intel and AMD sit in terms of single-thread performance with their current architectures. And the answer is: it's a match, with each part trading blows in lightly-threaded apps. Of course, in multi-threaded workloads the core disparity means Ryzen is faster.
The Ryzen 9 5980HS is also now universally faster than the Core i9-10980HK. AMD is able to outperform Intel's flagship H-series processor by anywhere from 2% to 20% in single-threaded benchmarks, and beat it by upwards of 30% when heavy multi-threading comes into play. AMD also has a much faster integrated GPU in the 5980HS, although it's rare to see this iGPU actually used in an Intel H-series design.
What We Learned
Arriving one year after Ryzen Mobile 4000, AMD was facing a tough challenge to iterate with a new processor series that was as impressive as what they delivered last time. Naturally, we aren't getting the same sort of gain multiples with Ryzen Mobile 5000, but the shift to faster Zen 3 CPU cores and extra cache deliver a straightforward boost.
With Ryzen 5000, and by extension the Ryzen 9 5980HS, AMD is delivering much better single-thread performance. The 5980HS provides a ~20% gain versus the 4900HS, which is significant for an architectural improvement on the same process node. This takes the 5980HS above Intel's 10th-gen H-series in all workloads, and allows Zen 3 to trade blows with Intel's Tiger Lake featuring Willow Cove CPU cores.
This is similar to what we saw on the desktop: Zen 3 closes the IPC gap significantly to the point where Intel no longer can claim an advantage. We suspect this will have implications for gaming laptops, something we'll have to look at in a later review with more powerful discrete graphics.
Multi-thread performance hasn't improved in line with single thread. Here we see gains more around the 10% mark. This is in contrast to what we saw with desktop Zen 3, where in a lot of multi-threaded applications we saw larger gains following a similar trend to single-thread numbers.
Having that said, Ryzen 5000 still holds a rather dominant position in H-series multi-thread performance, extending whatever lead they already possessed. The Ryzen 9 5980HS is by far the fastest mobile processor available in the 35W power class, and easily offers 20-30% more performance than Intel's best chip, the Core i9-10980HK, at a lower power level. Tiger Lake 8-core designs aren't too far away by all reports, so this situation may change soon, but for now AMD is in a comfortable position.
If we're talking pure CPU performance, there isn't anything better than a Ryzen 5000 processor like the 5980HS, although we'll have to wait and see how that holds up for other CPUs in the line-up. What we can say for now though, is that the R9 5980HS is either equal to, or faster than, any other equivalent mobile CPU available today across all types of workloads, single or multi-threaded. The Ryzen 9 5980HS is going to give you a well-rounded experience with excellent CPU performance across the board.
Another conclusion we can draw in this mobile CPU generation, from testing just one Ryzen Mobile 5000 APU is that AMD has lost the lead for integrated graphics performance. The Core i7-1165G7 with Intel's Xe design is simply the faster iGPU in most games and GPU workloads, despite the overall chip running at a lower power level, 28W versus 35W. Intel Xe isn't available in the H-series just yet, but it will be shortly.
In the performance segment, iGPU performance is not as significant when most H-series processors are paired with discrete graphics. We plan to comparing Intel and AMD gaming laptops with dGPUs in the coming weeks, however this could be more problematic for AMD's Ryzen 5000 U-series.
Perhaps, most positive than all, in this generation AMD is set to receive several high-end design wins from laptop manufacturers. No more topping out at RTX 2060-level graphics. There are several laptop makers that will be pairing Ryzen H-series with up to the flagship RTX 3080 GPU from Nvidia. That alone will make Ryzen much more attractive to laptop buyers, not just for gamers, but also for productivity workloads that need GPU acceleration. Pairing this sort of CPU performance with a powerful GPU should be a killer combination for content creation.
Even by just looking at the ROG Flow X13 that we tested today, it's clear that Asus have taken their AMD designs to the next level. The Zephyrus G14 was impressive last year but the Flow is a great ultraportable on its own right with killer CPU performance out of the box.
How the laptop platform battle evolves beyond performance will be interesting to watch, as it's clear some laptop makers haven't bothered to consider Ryzen 5000 this generation. Intel still holds some advantages here, like widespread use of Thunderbolt, a better media encoding engine with support for newer codecs, and PCIe 4.0 support.