Mobile computing is the next big goal at AMD with its new series of Ryzen 4000 APUs. The new processors will be available as the U series for low-power and ultraportables and as the H series for high-performance laptops. Today we have the first Ryzen 4000 retail laptop on the market that can be evaluated. Therefore, the focus is on the benchmarks so that we can see for the first time how Zen 2 is transferred to mobile stacks compared to Intel’s proven product range.
All Ryzen 4000 APUs are based on the Zen 2 architecture, which was built using 7nm technology with a monolithic chip that also contains a Vega-based GPU. Earlier this month, we released a full breakdown of the Ryzen 4000 APUs, which lists some of the architectural features and specifications. We'll skip most of these details in this review, but you should definitely check this out for an overview of what's to come in 2020.
Ryzen 4000 parts do not use AMD's Zen 3 architecture. The naming here is a bit confusing – Ryzen 3000 for Zen 2 desktop chips and Ryzen 4000 for Zen 2 APUs – but this is simply a continuation of what AMD started in 2017 when it launched the first Zen APUs as Ryzen 2000 Market. Ryzen 4000 APUs are therefore based on Zen 2 cores, just like the latest Ryzen desktop CPUs.
It is no coincidence that the Ryzen Mobile U and H series reflect what Intel offers for different types of laptops. With Ultraportables still pending, today’s all about the high-performance H-Series and how AMD is evolving for content creation, productivity, and gaming.
The Ryzen 4000 H series from AMD consists of three different SKUs: The Ryzen 5 4600H delivers 6 cores and 12 threads for the mainstream, the Ryzen 7 4800H bumps with up to 8 cores and 16 threads as well as a higher clocked 8-core variant at Ryzen 9 4900H. All of these APUs are equipped with a 45 W TDP and are supplemented by a 35 W version with low power consumption for laptops, which also complies with the other design guidelines of the AMD-HS series.
You will also find different GPU configurations for each SKU, using the updated Vega design with up to 8 processing units and higher clock rates. Interesting to mention, but we expect that most H-series laptops also include a discrete GPU, like our test system. Although the cache is listed at 12 MB, it is actually a combined L2 and L3 number: all parts have 8 MB L3 cache, Ryzen 7 and 9 get 4 MB L2 and Ryzen 5 3 MB.
The chip we're measuring today is the Ryzen 9 4900HS, a 35 W SKU that offers slightly higher base and boost clock speeds than the Ryzen 7 4800H, which is 45 W below, a bit of binning magic . It's not the absolute APD flagship in the AMD range, but it will give us a really solid insight into the performance of Ryzen 4000. We'll also give you a sneak peek at the performance of Ryzen 7 4800H using a technical example as we try to check the rest of the processors in the product range when new laptops come onto the market.
Our benchmark test area is the Asus Zephyrus G14, a decent portable 14-inch gaming laptop that contains Ryzen 4000 HS APUs and up to GeForce RTX 2060 Max-Q graphics. We received the model with the highest specifications for testing, with the RTX 2060 Max-Q and the 4900HS, 16 GB DDR4-3200 memory and a 1080p 120Hz display.
The focus of this test is on pure CPU performance, so we won't go into the design and other features of the Zephyrus G14. However, we will say this, after a week of use, we think it is a very well built laptop with a great keyboard and, as we will see shortly, a very convincing performance in such a small form factor.
In this test you will also see a number of different laptop CPUs in the diagrams and sometimes GPU configurations. The data in the diagrams are an average of the laptops that we tested with the specified hardware. Testing laptop components is of course somewhat more difficult than testing desktops, as each configuration can vary in terms of cooling and other hardware. These averages are therefore intended to illustrate the performance of a "typical" system. The averages do not include single-channel storage systems or other situations in which the existing components are severely throttled. We did our best to create data from apple to apple whenever possible.
Let's start with a classic performance benchmark: Cinebench R20. In this test we see the total dominance of the red team. The 4900HS is not only the fastest laptop processor we tested in the multi-threaded workload, but also the fastest in single-thread performance. The 4900HS is 35% faster than the 8-Core i9-9880H when all cores are used and 7% faster in the single-core. There is the Core i9-9980HK that we haven't tested and may start a fight, but the 90W numbers of the 9880H suggest that this could be a difficult task. Also keep in mind that these results come from the 4900HS with lower TDP. The 4900H at full 45W should be another step ahead.
There are other brutal results in this Cinebench diagram. The 4900HS crushes the Core i7-9750H, the most popular CPU for slim and light gaming notebooks. The Zen 2 offer is over 60% faster in the MT test. We also see that the AMD part of the previous generation is somewhat humiliated. The 4900HS is so much faster than the Ryzen 7 3750H in the same performance case that it's not fun.
We also tested with the older Cinebench R15, which shows results from a larger number of CPUs that are more years ago. As expected, the 4900HS remains firmly at the top of the charts. For example, anyone upgrading a Core i7-7700HQ that was a popular H-series CPU in 2017 can more than double performance with a Ryzen 9 4900HS.
One area in which Zen 2 receives a particularly extensive upgrade is dealing with AVX-256 instructions. For operations with a wide floating point, it is simply much faster than before. So the Ryzen 9 4900HS gets a huge performance boost in our Handbrake x265 test, which uses AVX instructions. The 4900HS is 179% faster than the 3750H in this workload, which is simply unfathomable for a single generation to jump in performance.
On an AMD-against-Intel front, it is also very cheap for AMD. Not quite the same margins as in Cinebench, but we still see a 23% performance advantage of the 4900HS over the 9880H and a 45% advantage over the 9750H. Frankly, these are massive deltas for a laptop form factor that often gets single-digit improvements from gene to gene.
In order for Intel to combine the 4900HS with its 8-core offering, it has to blow its performance target out of the window and use a 90 W TDP instead, which is possible with some gaming laptops with a turbo mode or similar. The difference in wall power consumption for these two systems is incredible: the G14 with the 4900HS ran comfortably at around 66 W in the long term, compared to 150 W for the 9880H in our HP Omen 15 test system. This just shows how much more efficient AMD's Zen 2 design is with these long-term workloads.
Blender tells a similar story with tremendous performance increases ranging from ~ 35% for the 4900HS over the 9880H and ~ 65% for the 4900HS over the 9750H. Most Blender users will likely render on the GPU instead, as it tends to be much faster. However, this is another yardstick that illustrates the long-term multi-core performance of these laptops.
Regarding decompression, we see that AMD gets away with a significant win in 7-Zip. Ryzen processors are known to work very well in this workload. However, it lags ~ 98% behind the 9880H in terms of compression, although AMD outperforms lower core parts like the 9750H.
Adobe Photoshop's performance is interesting. In our Iris Blur test, which is largely limited to the CPU, the 4900HS manages to beat the 9750H, but falls behind the 9880H and sits comfortably between these two processors.
This is also the case in the larger Puget Photoshop benchmark, which goes through a series of tests. In both workloads, where different effects are applied to very high resolution images, the 4900HS is about 10 percent slower than the 9880H in a core-by-core battle. Granted, we're seeing 40% more performance than the latest generation of AMD APUs.
Now let's look at the PCMark 10 numbers. We look at the workloads for essentials and productivity because they are CPU-limited, while the rest of the tests are more GPU-based.
The Essentials workload includes things like loading apps, surfing the web, and video conferencing. AMD manages to achieve the 9880H's performance, which may not sound that impressive until you find that the last generation 3750H is slaughtered in this test. Now AMD is on parity for these everyday workloads.
We see something similar in productivity utilization as before: The Ryzen 9 4900HS surpasses something like the Core i7-9750H. These types of tasks were less than decent compared to previous Ryzen. Although Ryzen 4000 doesn't put Intel under pressure here, the performance parity is a good result.
However, there are situations with productivity loads where Ryzen 4000 is not as impressive. Our custom Excel benchmark offers a lot of number processing for a large data set, and here the 9880H outperforms the 4900HS again. The 4900HS is faster than the 9750H, so it's by no means a terrible result, but it seems that large amounts of data are a weak point for Ryzen.
With MATLAB we can confirm this with our ODE and FFT benchmarks. Again, lots of data crunch with large amounts of data, and the 9880H from Intel wins.
Another workload that Ryzen does well but is not enough to beat Intel's 8-core competitor is our Acrobat PDF export test, which is all one-thread. The 9880H is slightly faster in Acrobat, although we still see a significant increase in performance in the 4900HS compared to the 3750H, with a significant increase in single-threaded performance by 20%.
One final workload before we look at GPU acceleration is AES-256 performance, as provided by SiSoft's Sandra benchmark. In the multithread test, we see the AES performance of the 4900HS is 15% higher than that of the 9880H, which makes it the fastest CPU we have investigated for cryptographic workloads. This gives Ryzen a neat advantage on two heavily used low-level tasks in the areas of decompression and cryptography.
Now let's go through our Premiere tests, which are mostly GPU-accelerated, starting with our 1-pass encoding that uses Intel's QuickSync technology. Premiere does not currently support hardware accelerated encoding on AMD processors. For those who prefer a quick export with slightly reduced image quality in the end, Intel is still the way to go. In particular, the 4900HS in combination with the GeForce RTX 2060 Max-Q from Nvidia is ~ 20% slower than a Core i7-9750H with an RTX 2060, since no hardware acceleration is available.
On the other hand, our Core i9-9880H system does not support QuickSync acceleration because the iGPU is completely disabled to support G-Sync via the Nvidia GPU. So we have a nice comparison of the software coding between the 9880H and the 4900HS, in which the 4900HS withdraws strongly with 26%.
Then we come to our 2-pass coding, which produces superior image quality and does not support hardware acceleration. In this scenario, performance falls back on what we've seen on most long-term workloads: the 4900HS is 30% faster than the 9880H and 38% faster than the 9750H, although our 9880H laptop has a much faster GPU. This test isn't limited to the GPU, but even with a moderately powerful discrete GPU, AMD's Zen 2 APU appears to be significantly faster for Premiere encoding.
But wait, there are more Premiere tests. Here we have a single instance of Warp Stabilizer, an extremely sophisticated effect that runs on a single thread per instance. The 4900HS manages to stabilize the footage 14% faster than the 9880H and 22% faster than the 9750H. This shows the performance of the high single-thread performance we saw with Cinebench for the first time.
And finally, we have the Puget benchmarks. The Ryzen 9 4900HS is the best laptop CPU for live playback in Premiere. It offers 12% better performance than the 9880H, which means we get better machining performance. Then we see numbers for your export benchmark similar to what we just talked about, with QuickSync acceleration helping with some of these workloads.
While we don't believe that many people will actually use the built-in GPU in these H-series processors, since the vast majority of H-series laptops also contain discrete graphics, it's worth taking a quick look at how that iGPU in a small selection of processors cuts off low-intensity games just to see what improvements AMD has made to the Vega GPU.
The Ryzen 9 4900HS has the fastest iGPU configuration in the series with 8 processing units that are clocked at 1,750 MHz. And it delivers impressive results. In Grand Theft Auto V, the 4900HS delivered 36% more power than the Ryzen 7 3750H with 10 Vega processing units at 1,400 MHz. On paper, both CPUs have similar GPU performance, but with all the advantages that AMD spoke of, such as increased memory bandwidth, this new Zen 2 APU can take the lead.
In Civilization VI, the 4900HS again offers 37% more power than the 3750H within the same 35 W power envelope. At CS: Go, this margin remains the same at around 37%. The 4900HS, on which CS Go runs exclusively on the integrated graphics, achieved an average of over 100 FPS with low settings, which is very nice.
Finally, we have our most powerful iGPU test in Gears 5 with medium settings. The 4900HS is around 31% faster than the 3750H in this workload, which makes it faster than a discrete low-end GPU offering like the MX250 from Nvidia.
It's not common to combine an H-series laptop with an MX-class GPU. Typically, OEMs opt for more powerful configurations like a GTX 1650 or higher, but with Ryzen 4000 you really don't have to deal with an MX250 or similar discrete GPU in an H-series design.
Sustainable watches and other questions
At this point in the test, we usually guide you through some comparison summaries between the Ryzen 9 4900HS and various other CPUs. However, we believe there are still some performance issues on the table. One is which clock rates the CPU actually runs in practice and how Boost behaves. And the other is why we see lower performance with some data-intensive workloads like Matlab, Excel and Photoshop. Let's tackle this second part first.
We have two working theories about why we see data-intensive workloads performed this way. The first is simple and is based on the specifications of these processors, especially the cache sizes. The Core i9-9880H has a decent 16 MB L3 cache, which with parts like the i9-9900K corresponds to what Intel offers on the desktop. However, the Ryzen 9 4900HS only has 8MB of L3 cache, half of what Intel has to offer, and well below the 32MB of L3 that AMD packs into its 8-core Zen 2 desktop processors like the Ryzen 7 3700X .
If there is less cache, less data can be stored in superfast memory and retrieved immediately. If you have a high number of cores, powerful CPU cores, but not enough cache, this can be a bottleneck in some cases. And while this amount of cache on Ryzen 4000 has doubled compared to previous mobile series, the number of cores has also doubled. It's probably not the whole story, but definitely part of it.
The other is the storage system. Yes, AMD offers higher memory bandwidth than Intel because it supports DDR4-3200 speeds. Intel only offers DDR4-2666 with the 9th generation. In a benchmark like Sandra, we see that AMD offers around 35% more memory bandwidth. However, Ryzen 4000 appears to have lower memory latency. Like cache size, this can become a performance constraint. The Core i9-9880H has a memory latency of approximately 30 ns for data records larger than 32 MB, while the Ryzen 9 4900HS has a memory latency of 46 ns. That is a considerable gain for Intel.
It's hard to say for sure whether these factors are part of the cause or the whole cause, but when we look at the low-level benchmarks, these were the two things we noticed. If we rate more Ryzen 4000 APUs, we get a clearer picture of where these bottlenecks are.
Let's look at clock speeds and boost behavior. After a cold start in our handbrake AVX workload, the Ryzen 9 4900HS has constantly increased to about 65 W of power for a few seconds, achieving 4 GHz boost clocks throughout the core, before going to 54 W with watches for a longer period of time dropped by 3.7 GHz. Finally, the CPU is set to 35 W to provide an all-core 3.2 GHz, just above the 3.0 GHz base clock of this processor. Boost time may vary depending on how warm the system is. On a cold start, however, we generally saw at least 2 minutes of boost of ~ 53 W, which is generous. Temperatures were actually very well regulated on the Zephyrus G14, with its air cooler maintaining a tick above 70 ° C, but this will of course vary between laptops.
Ryzen 9 4900HS vs. Core i9-9880H
The big one here is the Ryzen 9 4900HS over the Core i9-9880H. These are processors of a similar class, 8-core versus 8-core, and should give us a pretty solid hint of what we can expect from the 10th generation if we get slight improvements in clock speed from Intel, as is rumored.
On most long-term, multi-threaded workloads, we see the Zen 2 chip perform 35% better. We also get impressive numbers for video coding. Single-threaded workloads are usually faster or, in the worst case, with low productivity just like with Intel. However, the performance falls behind in data-intensive benchmarks such as Excel, Photoshop and Matlab.
The astonishing thing about most of these results is that AMD can achieve at least 25% better multithreaded performance on many workloads while at the same time achieving lower sustained power consumption at 35W versus 45W. So what happens if you limit Intel's 9880H to only 35W PL1 with Intel's XTU software?
Well, the margins are growing even more. While we let these processors do what they normally do in the boost phase, some benchmarks don't change too much. Long-term workloads now speak strongly for AMD. In some situations, over 50% better performance is possible, and a 40% improvement in the handbrake is impressive. This means that an OEM developing a slim and light system with limited cooling will generally have the best experience with Ryzen by far.
Amazingly, Intel can't beat the 35W AMD processor on long-term multithreaded workloads, even if we throw Intel's 45W power limit out of the window and increase it to 90W forever. With an astonishing gain in efficiency, the 4900HS either matches this 90 W processor configuration or is up to 15% faster. We noticed a difference of 80 W in these two systems compared to the wall. So if Intel wants to balance AMD with 14nm CPUs by just raising the performance limits, laptops need much bigger coolers to cope with it.
Ryzen 9 4900HS vs. Core i7-9750H
If you compare the 4900HS with the 9750H, there is only one case where the 9750H system can be faster, with the QuickSync-accelerated Premiere encoding. In every other benchmark, including those in which the 9880H used to be faster, the 4900HS is now faster. This is a good sign for future battles in the Ryzen 7 / Core i7 series, where AMD has a decisive advantage in offering only 8-core parts.
Ryzen 9 4900HS vs. Ryzen 7 3750H
Now let's see how far AMD with Zen 2 vs. Zen + has come in a 35 W mobile design. The difference with certain workloads is amazing. AMD not only achieves a 30% performance improvement with single-thread workloads, but can also achieve 2.5 times the performance with the same power consumption for long-term multithreaded tasks such as Blender, Handbrake and Cinebench.
Ryzen 9 4900HS vs. Ryzen 7 4800H
We promised a little glimpse of the performance of Ryzen 7 4800H on a sample technical laptop that we could briefly access. With a small subset of benchmarks, performance looks the same, if not slightly better, than that of the 4900HS, which is very suitable for mid-range systems that use the Ryzen 7 part. Yes, the 4900HS is more efficient, but the 4800H should still deliver pure 8-core performance.
What we have learned
Ryzen 4000 provides the best performance for high productivity workloads that you can get in a mobile form factor. We can say that without testing the Ryzen 9 4900H with higher performance, because the 35 W variant of the Ryzen 9 4900HS already leaves Intel's 8-core competitors in the dust.
The Ryzen 9 4900HS is a good choice for core-intensive work: We are talking about video coding or transcoding, 3D rendering, file compression and the like. It is much faster and with less power consumption, which enables better performance with a smaller form factor. Even with single-thread workloads, AMD is generally ahead, which means that the Premiere editing experience, for example, is better with the Ryzen 9 4900HS.
AMD did a great job from a pure gene-to-gene upgrade perspective. The Ryzen 7 3750H was difficult to recommend and was clearly beaten by the mid-range parts from Intel. That has changed with the newer APUs.
Technically, Intel parts can still keep up in the performance range if you lift the 45 W long-term power limitation. We saw this with the Core i9-9880H and we expect it would be similar if we had tested the flagship Core i9-9980HK. However, to achieve this performance you need a larger laptop with a more powerful cooler, which is likely to run hotter too. The Ryzen 9 4900HS offers all this performance in smaller laptops – like this Zephyrus G14 – and runs cooler at the same time.
However, AMD's Ryzen 4000 CPUs are not the complete package. There are some limitations. One of them is data-intensive workloads like cracking massive spreadsheets in Excel, working with huge photos in Photoshop or executing Matlab scripts. Intel prevailed in these tests. Lighter loads are enough for AMD, but if you're a large number cruncher, Intel's core-equivalent processors may be the way to go.
The other is for workloads that use exclusive Intel technologies like QuickSync, with Premiere encoding being one of them. If you use this all the time, Intel is the way to go. It's worth noting that any other job in Premiere outside of QuickSync encoding on Ryzen 4000 is faster, including editing and applying intense effects like the Warp Stabilizer.
Outside of these two instances, these Ryzen 4000 8-core processors give you a better and faster productivity experience. This is new territory for AMD on a mobile PC form factor.
It also seems that AMD will offer this type of service at very competitive prices. The Zephyrus G14 with the Ryzen 9 4900HS and the RTX 2060 Max-Q is said to be commercially available for around $ 1,500. This is firmly the territory of the Core i7-9750H laptops with six cores. The Core i9-9880H laptop we bought for testing in this test cost $ 2,400 and was one of the cheapest we could find. Granted, there is a much faster GPU for gaming, but at the moment you won't find 8-core laptops in the $ 1,500 price range. While most of our data focused on Ryzen 9 and Core i9, they don't really compete directly for pricing, at least for the time being, as this could change with Intel's 10th generation.
A few additional notes to complete this review …
We haven't had the most stable experience with previous Ryzen laptops due to many software bugs and crashes that we've discussed a few times. Completely different experiences with the Ryzen 9 4900HS in the Asus Zephyrus G14. Even with a beta version of the GPU driver on this laptop, there were no stability issues, crashes, or applications that refused to function properly.
You may also be surprised at gaming performance with a discrete GPU. The Zephyrus G14 finally contains the RTX 2060 Max-Q. We'll look into that in a future report as we're still working to get the best comparison between apples and apples. I hope to have that soon.
Battery life will also be in people's minds, but we didn't want to test this because we didn't have enough data for different laptops to make a fair comparison.
Finally, we conclude with a few comments on Intel's 10th generation. The new laptop CPUs will likely be around the corner within a month. Rumor has it that Intel is holding onto 8-core parts for the H series and is still using 14nm technology. We just can't see how they can compete at performance or efficiency levels. If we only get a small clock increase with the same efficiency of 14 nm as in previous generations, Ryzen will easily win it. We'll see the fight unfold soon, but based on what we've seen so far, we don't have high hopes for the 10th generation.