After our test of AMD's new Ryzen Mobile 4000 laptop CPU and after we covered productivity performance and gaming with the integrated GPU in our first test, it's time to tackle the other main use case for these processors take discrete GPU.
This is especially important for parts of the Ryzen H series, since gaming laptops work almost exclusively on these 45 W processors. We know that AMD has a very compelling part in its hands, with much better efficiency and mostly better multi-thread and single-thread performance, but gaming can often be a story of its own. Latencies, increased performance and frequency can play a role. So today we're going to do our best to find out how the Ryzen 9 4900HS performs in games.
The biggest challenge for this test was to get an apple-to-apple platform that could compare the 4900HS to Intel processors. Our Ryzen 4000 test bench, the Asus Zephyrus G14, has a GeForce RTX 2060 Max-Q with a power limitation of 65 W that we have not seen in any other laptop.
We wanted this to be a completely fair fight with the same GPU, but at least we managed the next best. We put the 4900HS against the closest platform we could find, namely a Core i7-9750H laptop with an RTX 2060 at 80W. Unfortunately, we couldn't get an 8-core Intel laptop with the RTX 2060, but that will prove to be easier with the upcoming 10th generation Intel product line with the 8 core i7-10875H.
As part of today's testing, we're going to look at some 1080p results – CPU and GPU-limited scenarios to see what the situation is there – and then dive into some heavily CPU-limited 720p games. On the way, we tried to do our best to examine the effects of the CPU in games. 720p causes a CPU bottleneck with the RTX 2060 in some titles, which allows us to isolate the effects of the CPU.
We know that most people will play at 1080p. So let's look at this first …
We decided to start with some very GPU-restricted environments so that we could get a basic overview of the differences between the RTX 2060 Max-Q and non-Max-Q models. The control at 1080p high settings is very GPU demanding and here we see that the RTX 2060 is on average 9% better than the 2060 Max-Q and 7% better in 1% lows.
Similar story in Metro Exodus: The non-Max-Q model lands about 7% faster. It has a 23% higher performance limit and is generally up to 10% higher in the situations we've seen. So this is the limited performance of our GPU baseline, so we're all only aware of the differences we're dealing with.
With Grand Theft Auto V, let's go into some less GPU-restricted scenarios that can be limited with higher-performance RTX GPUs and 1080p CPU. However, the RTX 2060 is not powerful enough to achieve a consistent CPU bottleneck, so the Intel system with non-Max-Q graphics is 3% faster on average. That's not a bad result given the differences in GPU performance, but let's continue.
Watch Dogs 2 is one of the most demanding titles in our test series and hits both the CPU and the GPU hard. What we find here is that the Ryzen 9 4900HS with RTX 2060 Max-Q, unlike our GPU-limited situations, is slightly ahead of the Intel system with the slower GPU. We're talking 1% faster here, which is a margin of error, but this is definitely a promising result for AMD's Ryzen APU in gaming.
In Star Wars Battlefront II we see an interesting phenomenon with the Ryzen APU compared to the 9th generation Intel Core i7 offering. The RTX 2060 system achieves 6% better average frame rates, but with the Intel processor loses against Ryzen at 1% lows. The Ryzen 9 4900HS offers a 2% better minimum performance and a more stable frame rate, which indicates that the 4900HS can keep up in sections of the game that require more CPU.
Far Cry 5 and the Dunia engine are reasonably CPU-bound at 1080p with high-performance components, not 100% CPU-limited, but benefit from a faster CPU. Again, the average frame rates with the Intel configuration, which includes a faster GPU, were 6% higher, but Ryzen performed better by 1%: 2% better, as we've seen a few times.
And we will see situations like this again and again as we move through the rest of the games tested. Assassin's Creed Odyssey is pretty demanding and here we see the biggest performance advantage for Ryzen in 1% lows. AMD's gaming APU option is 12% better here, which is a considerable margin, although again it loses average frame rates as we're not entirely limited to the CPU during this benchmark run.
Looking at Hitman 2 and again similar results here. This is a very CPU-demanding title, and it seems like most laptop offerings will do roughly the same performance until we reach the 8-core i9-9880H core, which can assert itself with its RTX 2080 Max-Q . I don't want to read too much into these results from a CPU perspective because the GPU differences are quite large.
Shadow of the Tomb Raider behaves similarly to Assassin's Creed Odyssey in that our AMD laptop isn't really faster in terms of average performance, but clocks almost 10% faster in terms of 1% lows. This can be a surprisingly CPU-intensive title, so it's a strong result for AMD.
Then we come to Resident Evil 2 with the balanced preset at 1080p. This is one of the few games I've tested where the Ryzen system was significantly faster. As with some of the games we looked at today, it seems to be more CPU than GPU limited. Although the Zephyrus has a weaker GPU, the performance is actually better. This game thanks to the faster CPU power available.
And then in Star Wars Jedi Fallen Order, the performance between the two systems is very similar. Again, both options offer an equivalent experience in a game that sometimes has a decent CPU limit.
A short break
At this point we have a pretty good idea of the performance of these systems. The Intel combination is the leader in GPU-limited scenarios due to its faster GPU. While the average performance on the Intel side is better in many games, AMD delivers better performance of 1%, sometimes this is only marginally better in the range of 2-3%, sometimes the results are over 10% in favor of AMD.
There were also two games: Watch Dogs 2 and Resident Evil 2, in which the 1080p title was overall CPU limited enough to perform better on the Ryzen 9 4900HS system despite its weaker GPU option.
Although these results are a realistic representation of the performance of these combinations, in many of these benchmarks we are not entirely limited to the CPU. Laptops with 1080p are, depending on the game and hardware, a boundary configuration between a GPU or a CPU limit. So we wanted to test the 720p performance. Let's really limit these systems to the CPU and see which configuration works better when we encounter CPU bottlenecks.
720p testing to simulate CPU-limited scenarios
In Grand Theft Auto V, the 720p benchmark pass is completely CPU-limited, so we get into a situation in which the Ryzen 9 4900HS configuration is now 5% faster and 10% faster in 1% lows than the Core i7- 9750H is. This is a flip compared to what we had previously at 1080p where the Intel configuration was faster on average.
We previously knew that Watch Dogs 2 was CPU limited to 1080p, but the margins grow slightly to 720p, with the AMD configuration now being 6% faster on average at this resolution.
Star Wars Battlefront II is a big swing in favor of AMD at 720p. If we limit the game to CPU, the Ryzen 9 4900HS will be 10 to 15 percent faster. And the edges in Fary Cry 5? Very similar to Star Wars Battlefront 2 with the 10 to 15 percent performance advantage for Ryzen.
One of the biggest performance benefits for Ryzen that I saw benchmarking at 720p was Assassin's Creed Odyssey. In this title, the Ryzen 9 4900HS was over 25% faster when the CPU was fully CPU limited compared to the Intel Core i7-9750H. This is approaching some of the margins we have seen benchmarking multi-core productivity workloads.
Hitman 2 was a bit faster in our Intel configuration at 1080p, but this changes in favor of AMD at 720p. Here the Ryzen 9 4900HS is 8 percent faster on average with a similarly low output of 1%. Shadow of the Tomb Raider also chose AMD when the CPU was limited, with 6% higher average frame rates on the Ryzen 9 4900HS system.
What about Resident Evil 2, a game where AMD already performed better at 1080p? At 720p the margin only grows and gives results similar to Odyssey at 720p: over 25% better performance when the CPU is so limited.
In Jedi Fallen Order we see a performance advantage of up to 11% for the Ryzen 9 4900HS at 720p, similar to many other titles that we have gone through.
What about some of the games we discussed earlier that were limited to 1080p more GPU? Well, that remains the case at 720p. Games like Metro Exodus and Control still offer better average performance on our Intel test laptop, as the GPU is tied to a high load of the 90s throughout the test.
We'll be releasing another game with strong CPU limits here before completing this feature …
CS: Run at 1080p with low settings. The margins between our Intel and AMD configurations weren't very high here, but the Ryzen 9 4900HS has an average 2% lead in this benchmark. It's no surprise that we have a CPU limit here if we work at well over 200 FPS. These results fit many of our other CPU-limited benchmarks.
This is a decent part of the benchmarks that deal with a number of test conditions. Well, the best test conditions we can handle given the limitations of testing with laptops that vary widely in hardware. Here are some interesting results that need to be broken down.
We didn't learn much when looking at GPU-limited games. Understandably, the Intel system performed better with its faster RTX 2060 GPU. This is always the case when comparing CPUs for games: if the title you play is not limited by the CPU or a bottleneck, the GPU you have is much more important and becomes the limiting factor. So you won't get better or similar performance with a Ryzen laptop with a weaker GPU if the GPU is limited. Makes sense.
In many games we compared to 1080p with ultra settings, the Ryzen 9 4900HS performed better in our Intel system than 1% of the Core i7-9750H. In many cases the difference was only slight, 2-3%. However, since our Ryzen laptop was paired with a weaker GPU, this suggests that the Ryzen 9 4900HS is more powerful in areas of our benchmark passes that have more CPU requirements.
This continues when we see very CPU-limited game scenarios. A few times this has been the case at 1080p in games like Watch Dogs 2, Resident Evil 2 and CS: Go. At 1080p, all were CPU limited, and in each of these cases the Ryzen 9 4900HS did better.
Then we saw a significant increase in favor of AMD at 720p. Since many of the titles we considered were completely CPU-limited at this resolution, the Ryzen 9 4900HS delivered an average of 5 to over 25 percent more performance in these games. Combined with the 1% low performance at 1080p, this suggests that the Ryzen 9 4900HS is the faster gaming CPU if the GPU is taken out of the equation.
These results are not too surprising when we look back at our productivity benchmarks. The Ryzen 9 4900HS was 5 to 15 percent faster in most single or multi-threaded applications, and we know that most games today are still lighter than multi-threaded threads. However, in some of the best scenarios, such as the 1% low performance of Resident Evil 2 at 720p, which was 39% higher at Ryzen, we are more in the area of these multithreaded results.
However, there are many limitations to these tests. We've stressed the differences in the GPU countless times, but we believe the other obvious one is comparing the Ryzen 9 4900HS to the Core i7-9750H. While these CPUs can be found in laptops with similar prices, Intel has 8-core Core i9 processors in the 9th generation and an upcoming 8-core core i7 option in the 10th generation. A part like the Core i7-10875H in particular produces higher single core frequencies than the 9750H, so these results may change if we can talk about benchmarks and 10th generation performance.
Ultimately, we learned that Ryzen 4000 is very suitable for mobile games. If we had two laptops that were otherwise identical apart from the CPU, the Ryzen 9 4900HS should either offer the same or a better gaming experience than the Intel Core i7-9750H, depending on how limited the GPU or CPU is. The more CPU is limited, the more advantages Ryzen has over the Intel option.
We find it quite impressive that AMD is entering this market segment and offering such a competitive alternative to Intel's most popular 9th generation gaming laptop CPU. Intel seems to be focusing on frequency, while AMD does the job while delivering excellent productivity performance.
It's also impressive to see this type of power at 35W, not 45W like the 9750H that we tested. This additional thermal headroom of 10 W is crucial when playing. This could result in an additional 10 W power allocation available for the GPU in a particular design, which could result in 5 to 10 percent higher frame rates in many situations.
All of this prepares us for an interesting battle between Ryzen 4000 APUs and the 10th generation Intel H-Series, including the new 8-core. Stay up to date as new systems will be tested again and again in the coming weeks.