The Ryzen 7 5800U is AMD's answer to Intel's Tiger Lake and is their fastest available chip for ultraportable and ultrathin laptops. This is the first Ryzen Mobile 5000 U-series processor to land in our office even though it was announced way back in January, but like many releases this year, it's taken a bit of extra time to actually hit the market.
When Intel launched Tiger Lake last year, they were able to reclaim the performance crown in many key workloads for thin and light laptops. While they only had up to four CPU cores compared to eight in AMD's line-up, Intel was able to deliver class-leading single thread and graphics performance, topping the charts in many lighter applications you might expect to use on this kind of machines.
But with Ryzen 5000, AMD is back to nullify Intel's key advantage. The Ryzen 7 5800U retains its 8-core CPU design, double the cores of Intel's Core i7-1165G7 and 1185G7, but bumps up those cores to use the newer Zen 3 architecture. With a significant bump to IPC, a redesigned core complex with a single CCX, double the L3 cache, and power optimizations - AMD claims much higher single-thread performance versus Ryzen 4000 and multi-thread gains that will only extend AMD's lead on Intel.
However, not every area to Ryzen Mobile 5000 has been updated. The Ryzen 7 5800U features the same Vega GPU design, with 8 compute units, which may not be sufficient enough to beat Intel's Xe GPU. The multimedia engine, memory controller, PCIe layout and other features also remain very similar to their last-gen APU.
AMD's line-up is a little confusing in that not all Ryzen Mobile 5000 U-series processors use Zen 3 CPU cores. The Ryzen 7 5800U we're looking at today does, it's their new Cezanne die. But the Ryzen 7 5700U that sits directly below it remains a Zen 2 processor, using a refreshed die AMD are calling Lucienne. This is something to keep in mind (and further explained here) when buying a new Ryzen 7 5000 series chip.
The Ryzen 7 5800U packs 8 cores and 16 threads, with a base clock of 1.9 GHz and a maximum boost frequency of 4.4 GHz. There's 16MB of L3 cache and 4MB of L2, along with 8 Vega compute units clocked up to 2.0 GHz. This is all built on TSMC's 7nm manufacturing node with a default TDP of 15W. The 5800U ends up as a fully unlocked Cezanne die but isn't quite as highly clocked as AMD's H-series chips that, for example, push boost frequencies up to 4.8 GHz.
Despite hitting the absolute highest frequencies this silicon can do, the Ryzen 7 5800U does provide some frequency improvements over the Ryzen 7 4800U. The base clock is 100 MHz higher and the boost clock 200 MHz higher, which is in addition to the IPC gains of Zen 3 versus Zen 2. The GPU is also clocked 250 MHz higher although it uses fundamentally the same Vega design and compute unit count. We also get double the L3 cache as expected.
The system we're using for testing the 5800U is Asus new ZenBook 13, which is a very compact slim and light notebook with an impressive OLED display. I'm glad this new wave of ultraportables coming to market are using OLED displays, because this thing is seriously impressive in terms of contrast, black levels and colors. Even though it's just a 1080p screen, the benefits to using OLED are massive, including proper HDR functionality.
Aside from OLED, we're also getting the Ryzen 7 5800U in multiple power modes, 16GB of LPDDR4X-3733 memory, and in my unit 1TB of storage. The build quality is quite nice, I love how portable this thing is, it has a nice large touchpad and a decent keyboard. And that's without even talking about performance yet. Unfortunately it does lack Thunderbolt which is a common feature omission on AMD laptops.
Like our previous U-series processor testing, today we are benchmarking the Ryzen 7 5800U in both its 15W and 25W power configurations (long term power limits). However, as with all laptops tested, there's also a boost period, in the case of this laptop up to about 30W. The 5800U will be pitted up against other laptop processors also tested at 15W and 25W, although in the case of Tiger Lake, we've run Intel's design at 15W and 28W, as 28W is the highest power config available compared to 25W for AMD.
Our benchmark charts are an average of results taken from multiple laptops running at the same long term power configuration. Here you can see the full list of laptops we've tested. We test this way to give you a general idea of how laptop processors compare without variables like OEM configurations and cooling designs having much influence.
For this review, we should also note that we re-tested the Ryzen 7 4800U with boost behavior that closely matches the 5800U in this ZenBook 13, just so we could have the absolute best apples-to-apples comparison between AMD processor generations.
Benchmarks
In Cinebench R20 we get our first look at the Ryzen 7 5800U's multi-threaded performance. At 15W we don't see a significant performance increase over the Ryzen 7 4800U, just a couple of percent all up. However at the higher 25W configuration, the 5800U ends up about 5% ahead of the older design.
This small performance uplift allows the 5800U to extend its lead on Intel. In the higher power class, AMD holds a 75 to 80 percent performance lead, which is absolutely massive when comparing two processors at a similar price point. At the lower 15W power spec, AMD's design appears to be more efficient again, pushing their lead up to over a 2x advantage. This sees AMD's CPU sit in the region of H-series processors like the Core i7-10750H and Core i9-10980HK for multi-thread performance.
The big question around Ryzen Mobile 5000 though isn't multi-thread performance, but single-thread performance, and how that compares to Tiger Lake. The results are promising: the 5800U is 14% faster than the 4800U in Cinebench R20 single-thread, and it remains a situation where single-thread performance is virtually the same whether testing at 15W or 25W. That's because a single Zen 3 core operating at up to 4.4 GHz tends to only consume about 15W of power, giving us no gains at the higher power spec.
When compared to Tiger Lake there's a few interesting things at play. Intel's design does benefit a lot from operating at 28W, even for single-thread performance, as clocking a Willow Cove CPU core up to as high as 4.8 GHz consumes more than 15W. This gives AMD a small performance advantage in the lower power class of around 5%. However when each chip is given more power and allowed to run in the 25W range, the 5800U fails to match Tiger Lake and ends up 5% behind the Core i7-1185G7.
A few people have been asking for Cinebench R23 numbers in our previous reviews so here you are. While the scores are different using a 10 minute run, the margins between each processor don't change substantially. And we find similar things when looking at the single-thread results as well.
Handbrake is a really interesting benchmark for Ryzen 5000. At 15W, we actually see a performance regression for the Ryzen 7 5800U compared to the 4800U. Not by much, but it is 3% slower in this test, which will disappoint those expecting performance gains in this long term, heavily multi-threaded workload. However, when given the extra power allocation of 25W, the situation flips and now the 5800U is faster, about 5% faster in fact.
What appears to be happening here is that Ryzen Mobile 5000 has a higher performance lead over Ryzen Mobile 4000 the larger the power budget is. It's roughly equal or slower at 15W, about 5% faster at 25W, and then in the H-series with the 5800H ends up about 8% faster at 45W. In other words the margin between 15W and 25W has grown this generation, and the sweet spot for Ryzen now appears to be in the higher power class.
While a lot of this discussion has centered around the 5800U versus 4800U, ultimately either of these processors destroys Tiger Lake at CPU video encoding with margins of at least 70%. It's just not possible for a quad-core design to be competitive with an eight-core design for these sorts of workloads.
In Blender, again nothing overly different from what we just showed in Handbrake. The Ryzen 7 5800U is still in a class-leading position, it just doesn't fundamentally change much compared to what the 4800U was providing.
In GCC code compilation there is a grouping of Ryzen processors at 15W, which only begin to separate themselves at 25W where the 5800U is fastest. The 5800U ends up between 60 and 70% faster than Intel's Core i7 Tiger Lake processors, and we see similar margins in Chromium compilation as well. If you are a coder that does a lot of compilation work and you want a slim and light system, then the last two generations of Ryzen are going to be a pretty good choice.
MATLAB's built-in benchmark is a good mix of multi-threaded, single-threaded and cache heavy tests. It's here we see some of AMD's IPC gains come to play, allowing the 5800U to easily outperform the 4800U in both power classes. The 5800U is 23% faster, and I suspect a lot of that is down to the doubling of L3 cache. It also puts the 5800U at least 11% ahead of Intel's Tiger Lake processors, so with the design improvements AMD have made, they've been able to go from slower than Intel, to faster than Intel.
Our Microsoft Excel workload is also heavy on processor cache, so the 5800U sees a performance gain of 23% again compared to the 4800U. As this is a multi-threaded workload, Ryzen goes from being only slightly faster than Tiger Lake in the prior generation, to being substantially faster. The 5800U sees a performance lead of 72% at 15W and over 40% at 25W.
In PCMark 10's Essentials workload, which measures basic things like web browsing and app loading, it's another good result for the Ryzen 7 5800U. Thanks to IPC improvements, the 5800U leapfrogs Tiger Lake into the leading position for U-series processors, with about a 5% lead on the 1165G7, or an 11% gain over the 4800U. This isn't a groundbreaking difference looking at AMD versus Intel but it does point to similar or slightly better basic app performance on a top-end Ryzen configuration.
In the Applications test which measures Microsoft Office performance and the Edge web browser, we end up with another somewhat interesting situation between AMD and Intel. At 15W, the 5800U essentially matches Tiger Lake's performance, however in the higher power class, Intel is able to pull ahead, so at 25W the 5800U comes in 10% behind the 1185G7. While the 5800U does show a healthy 10% uplift over the 4800U this isn't enough to close the gap in performance to Tiger Lake.
In 7-Zip Compression we see one of the largest gains for the Ryzen 7 5800U over the 4800U: 25% faster at 15W. Compression was a weak spot for AMD's Zen 2 APU designs but the increased cache and IPC improvements have sent the 5800U up the charts, even ahead of the Ryzen 7 4800H. Tiger Lake processors are slower here due to their lack of CPU cores.
As for decompression, this was always Ryzen's strong suit, so we don't see a large performance gain comparing the 5800U and 4800U, the newer Zen 3 chip is only marginally ahead. As with compression, Intel's quad-core Tiger Lake designs can't keep up here and get destroyed by Ryzen in what is the largest margin between these two chips in CPU-bound workloads.
In AES-256 cryptography, AMD has massively improved performance with their Zen 3 CPU core, it's something like 65% faster when testing on a single-core, which allows us to remove memory bandwidth differences. However, Intel still holds the lead here with their Tiger Lake designs, and ultimately the 5800U is about 12% slower.
Acrobat PDF exporting has been the Achilles heel of AMD's single-thread performance for some time in laptop form factors, but that has changed with Zen 3. The Ryzen 7 5800U is now able to match Intel's Tiger Lake designs like the Core i7-1165G7, although the Core i7-1185G7 at 28W is still a bit faster overall. Like other workloads, the 5800U is 20% faster than the 4800U here.
Adobe Photoshop performance was a big reason to buy a Tiger Lake processor over a Ryzen processor in the previous generation. That's not really the case any more. Intel is still slightly faster than the 5800U in the higher power class, but the margins are within 5%. Meanwhile the 5800U is able to pull 8% ahead of the 1165G7 at 15W, so it's a bit of a much of a muchness and will depend on the exact laptop configurations you're looking at. I guess the takeaway here is in general AMD has been able to equalize performance in this application.
In DaVinci Resolve using the Puget Systems benchmark, ultimately there isn't much difference between the Ryzen 7 5800U and Core i7-1165G7 for video rendering. The 1165G7 benefits from its faster GPU, but the 5800U's faster CPU nullifies that advantage. In fact at 15W the 5800U is up to 15% faster as a result. However if you do plan on doing a lot of video editing on your laptop, I would still strongly recommend something with a discrete GPU as it will be significantly faster.
It's a pretty similar outcome in Adobe Premiere using Puget System's export test. The 1165G7 is slightly faster than the 5800U at 28W, but a bit slower at 15W. So again it's going to be a situation where the power capabilities of the laptop matter. These sorts of systems aren't too bad for video editing in Premiere in my opinion, although you should expect to only be capable of lighter stuff unless you buy a system with a discrete GPU.
Gaming Performance
Now it's time for some integrated graphics game benchmarks. To be honest the results here aren't very interesting so I'm only going to show three games otherwise you'll all get bored very quickly. Here's Grand Theft Auto V looking at 15W CPUs. The Ryzen 7 5800U delivers identical iGPU performance to the Ryzen 7 4800U. In this title, this allows the 5800U to be 23% faster than the Intel Core i7-1165G7 when using the same LPDDR4X memory. At 25W in the same title, again there is no performance difference. Here the Ryzen processor matches Tiger Lake.
In Gears 5 running at 1080p medium settings, which is heavily GPU bottlenecked, there is no performance difference between the 5800U and 4800U at 15W. The 5800U is 13% faster than the 1165G7 in this test. However when switching up to 25W, now the Intel design is faster, holding a 23% lead on the 5800U. Again, the 5800U delivers the same performance as the 4800U.
Then in Rainbow Six Siege surprise, surprise, the Ryzen 7 5800U and 4800U deliver the same performance whether we are talking 15W or 25W. In both configurations, Intel's Xe GPU in their 1165G7 delivers better performance, and in the best case, with up to a 35% performance lead. So everything that we talked about in our Tiger Lake review still holds true as AMD has not improved iGPU performance this generation.
Performance Breakdown
Before jumping into the conclusion, here are some head to head comparisons. Looking at the Ryzen 7 5800U up against the Ryzen 7 4800U at 15W, we can see some clear trends. In multi-thread performance, the 5800U isn't significantly faster than the 4800U, in some workloads it's faster, and in others it's slower. But in applications that are either lightly threaded, or can take advantage of Cezanne's specific design improvements like double the cache, performance can be 20 to 25% faster.
In the higher power class, the Ryzen 7 5800U is almost always faster. Multi-thread performance transitions from about even, to about 5% faster on average. Meanwhile, we see the same single-thread performance gains as in the 15W class, which are very solid. And this is highlighted further when comparing the 5800U at 15W and 25W, where the 25W configuration is roughly even at single-thread but can be as much as 25 to 30 percent faster for multi-threaded workloads.
When the Ryzen 7 5800U is pitted up against the Intel Core i7-1185G7 in the higher 25-28W power class, there is a wide range of results. The 5800U is significantly faster at multi-threaded CPU workloads, typically in the 70%+ range. However for lightly threaded and single-threaded workloads, the 5800U can be 5 to 10% slower, although this isn't always the case. Again similar story with the 5800U versus 1165G7.
However at 15W, the 5800U is generally faster than the Core i7-1165G7 at all workloads. Multi-threaded workloads have a 2x performance advantage on AMD, while lightly-threaded tasks can be 5 percent faster or so.
What We Learned
The Ryzen 7 5800U is finally here, the benchmarks are in, and now we have a great idea of how AMD's Zen 3 APU performs in a low power class. Like we saw in the H-series for gaming laptops, when you compare this generation to AMD's prior release, Renoir, the major step forward AMD has taken is in single- and lightly-threaded application performance.
The rest of this design is largely an iteration on the last release, which was a major overhaul, and even in some cases like the iGPU, performance goes unchanged.
The gen-on-gen gains AMD is providing with the 5800U versus 4800U in the 25W power class are decent: 5% better multi-thread performance is a little disappointing, but single-thread gains anywhere from 15 to 25% put a good showing. Given that lighter workloads are more common on thin and light systems, single-thread performance is most important for U-series chips, so big gains in apps like Microsoft Office, web browsing, Photoshop and PDF exporting are going to be a boon for everyday users.
The improvements seen in these lighter apps do carry over to the lower power class at 15W, but I was less impressed with the 5800U here overall. Multi-threading performance is largely unchanged compared to the prior generation, and there were even a few performance regressions, which you never want to see.
But that's comparing AMD to AMD... and the real battle this generation is between Ryzen Mobile 5000 and Intel Tiger Lake CPUs.
When talking about ultraportable APU performance, I tend to break it down into three key areas: multi-threaded, single-threaded, and GPU performance. When we compared Tiger Lake to Ryzen 4000, Intel held the crown in two out of three, which is ultimately why I ended up recommending it to thin and light laptop buyers. But now that we have Ryzen Mobile 5000, the tables have turned and in my opinion AMD has regained its performance advantage.
What hasn't changed in this comparison is multi-thread CPU performance. AMD continues to hold a remarkably strong lead here, destroying Tiger Lake especially at low power classes. 8 cores up against 4 will do that. But what AMD's been able to achieve this generation is the single-thread performance to match Intel. This has removed AMD's weakness in CPU performance, making it harder to recommend Tiger Lake for casual laptop buyers. Surely, in the worst cases, depending on the matchup there are some workloads where Intel's Core i7-1185G7 is faster, but this doesn't make up for quite a heavy defeat in other workloads. I now feel the balance of CPU performance lies strongly with AMD.
Integrated graphics performance, on the other hand, is a mixed bag. It's a pretty even battle at 15W, but Intel is the champion in the higher power class where their Xe design is able to shine. This doesn't seem to impact mixed CPU and GPU productivity workloads too much, but Tiger Lake right now is the better choice for integrated graphics gaming.
With that said, there are a number of slim and light 13 or 14-inch machines being made these days with discrete graphics, whether that's an Nvidia MX350 or even a GTX 1650 Max-Q. So if gaming is a priority for your ultraportable laptop, honestly I'd try get something with a discrete GPU and the pairing of a Ryzen 7 5800U with an Nvidia discrete GPU would be a very tasty proposition.
While I do think AMD has the better overall low-power CPU option right now, this is merely a comparison of chip performance and doesn't factor in much else about the platform. And it's platform features where AMD has generally struggled in. Ryzen is used in fewer premium designs, for example. Supply has been an issue before and continues to be constrained today. You likely won't get features such as Thunderbolt, USB 4 or PCIe 4.0 storage with an AMD laptop, which depending on your use case may be important.
You can't blame AMD for all of these, as it's OEMs who ultimately implement the chips, but Intel gives you more choices and generally more platform features. A design you like just simply may not come with an AMD equivalent.