AMD Ryzen 9 5980HS Cezanne Review: Ryzen 5000 Mobile Tested
by Dr. Ian Cutress on January 26, 2021 9:00 AM EST- Posted in
- CPUs
- AMD
- Vega
- Ryzen
- Zen 3
- Renoir
- Notebook
- Ryzen 9 5980HS
- Ryzen 5000 Mobile
- Cezanne
IGP: 720p Gaming Tests
Testing our Cezanne sample for integrated graphics is a double-edged sword – AMD fully expects this CPU to be paired with a discrete solution in almost all notebook environments, whereas mini-PC designs might be a mix of integrated and discrete. The integrated graphics on this silicon is more geared towards the U-series processors at 15 W, and so that is where the optimizations lie. We encountered a similar environment when we tested Renoir at 35 W last year as well.
In order to enable the integrated graphics on our ASUS ROG Flex X13 system, we disable the GTX 1650 through the device manager. This forces the system to run on the Vega 8 graphics inside, which for this processor runs at 2100 MHz, a +350 MHz jump from the previous generation based on the improved power management and minor manufacturing improvements. We did the same to the other systems in our test suite.
Integrated graphics over the years has been built up from something barely useable in a 2D desktop environment to hardware that can competitively run the most popular eSports titles at good resolutions, medium settings, at playable framerates. In our recent review of AMD’s Ryzen 4000G Desktop APUs, we noted that these were the best desktop APUs that money could buy, held back at this point mostly by the memory bandwidth, but still enabling some good performance. Ultimately modern day integrated graphics has cannibalized the sub-$100 GPU market, and these sorts of processors work great in budget builds. There’s still a way to go on performance, and at least mobile processors help in that regard as more systems push to LPDDR4X memory systems that afford better memory bandwidth.
For our integrated graphics testing, we’re using our lowest configuration for our game comparisons. This typically means the lowest resolution and graphics fidelity settings we can get away with, which to be honest is still a lot better visually than when I used to play Counter Strike 1.5 with my dual core netbook in the late 2000s. From there the goal is to showcase some good graphics performance tied in with CPU performance to see where the limits are – even at 720p on Low settings, some of these processors are still graphics limited.
Integrated Graphics Benchmark Results | |||||
AnandTech | Ryzen 9 5980HS |
Ryzen 9 4900HS |
Ryzen 7 4800U |
Core i7 1185G7 |
|
Power Mode | 35 W | 35 W | 15 W | 28-35 W | |
Graphics | Vega 8 | Vega 8 | Vega 8 | Iris Xe | |
Memory | LP4-4267 | D4-3200 | LP4-4267 | LP4-4267 | |
Frames Per Second Averages | |||||
Civilization 6 | 480p Min | 101.7 | 98.9 | 68.4 | 66.2 |
Deus Ex: MD | 600p Min | 80.7 | 76.5 | 61.2 | 69.1 |
Final Fantasy XV | 720p Med | 31.4 | 31.3 | 29.1 | 36.5 |
Strange Brigade | 720p Low | 93.2 | 85.2 | 75.7 | 89.3 |
Borderlands 3 | 360p VLow | 89.8 | 93.6 | - | 64.9 |
Far Cry 5 | 360p Low | 68.0 | 69.5 | 60.0 | 61.3 |
GTA 5 | 720p Low | 98.9 | 80.7 | 80.0 | 81.9 |
Gears Tactics | 720p Low | 86.8 | - | 87.8 | 118.2 |
95th Frame Time Percentiles (shown as FPS) | |||||
Civilization 6 | 480p Min | 69.0 | 67.4 | 45.7 | 43.8 |
Deus Ex: MD | 600p Min | 45.6 | 57.3 | 38.1 | 44.1 |
Final Fantasy XV | 720p Med | - | 26.6 | 24.6 | 26.5 |
Strange Brigade | 768p Min | 84.2 | 77.0 | 68.6 | 73.0 |
Borderlands 3 | 360p VLow | 63.6 | 73.8 | - | 48.9 |
Far Cry 5 | 360p Low | 50.3 | 62.3 | 43.8 | 49.8 |
GTA 5 | 720p Low | 66.8 | 52.8 | 56.0 | 55.7 |
Gears Tactics | 720p Low | 67.5 | - | 78.3 | 104.5 |
Despite the Ryzen 9 5980HS having LPDDR4X memory and extra frequency, the performance uplift against the Ryzen 9 4900HS is relatively mediocre – a few FPS at best, or losing a few FPS at worst. This is except for GTA, where the uplift is more ~20%, with the Zen 3 cores helping most here. In most tests it’s an easy win against Intel’s top Xe solution, except in Gears Tactics, which sides very heavily with the Intel solution.
With all that being said, as mentioned, the Ryzen 9 parts here are more likely to be paired with discrete graphics solutions. The ASUS ROG Flow X13 we are using today has a GTX 1650, whereas the ASUS Zephyrus G14 with the 4900HS has an RTX 2060. These scenarios are what really dictate the cooling solution in these systems, as well as how they are both used in workloads that requires CPU and GPU performance.
For any users confused as to why we run at these settings; these are our low 'IGP'-class settings in our CPU Gaming test format. As mentioned in our new CPU Suite article in the middle of last year, our CPU Gaming tests have four sets of settings: 720p Low (or Lower), 1440p Low, 4K Low, and 1080p Maximum. The segment above our lowest this in our suite is 1440p, which for a lot of these integrated GPUs would put numbers into the low double digits, if not lower, which something we've done in the past to massive complaints about why even bothering with such low framerate numbers. The point here is to work from a maximum frame rate, see if the game is even playable to begin with, and then detect where in a game the bottleneck can be; in some of these tests we're still dealing with GPU/DRAM bottlenecks. I've played CSS1.5 and other games at a Lan party on dual core AMD netbooks in the late 2000s, having to use low resolution texture packs to get it even 20 FPS playable. I still had masses amount of fun. From these numbers you can see the best possible frame rates for a given title and engine, and work down from there. It provides a starting point for further directions. These processors more often being paired with discrete solutions anyway, making discussions about IGP performance almost somewhat trivial compared to the rest of the data/
218 Comments
View All Comments
Deicidium369 - Wednesday, January 27, 2021 - link
No time soon... market already not buying the 64C - the 32C outsells the 64C by huge marginsikjadoon - Tuesday, January 26, 2021 - link
Sigh, this is just a plain confused commenter: are you confused that 1T benchmarks exist? Why do people get so worried / panicked when the M1 comparisons start? What on Earth does EPYC have to do with a fanless laptop SoC?Yes, the M1 has faster 1T so it naturally has faster JavaScript performance.
Not complicated.
lilmoe - Wednesday, January 27, 2021 - link
Just buy your M1 Mac and leave us peasants alone. Let us worry what single threaded performance. Arguing with you apple fans is really getting old, and tiring. Live and let live.Deicidium369 - Wednesday, January 27, 2021 - link
I had a guy keep trying to sell me on Apple - like I am new to the computer game... finally had to put him on block...GeoffreyA - Wednesday, January 27, 2021 - link
As a Windows/x86/AMD fan, I admit it, the M1 is faster.Deicidium369 - Wednesday, January 27, 2021 - link
Faster on Apple Software - and doesn't run PC softwaremax - Wednesday, January 27, 2021 - link
Do you realize, that any Apple personal computer is a PC (Personal Computer)? So what You wrote above is pure nonsense. I know, that you mean Windows OS and MacOS. Just be precise if you want to say something properly.Sailor23M - Sunday, January 31, 2021 - link
Finally bit the bullet and bought my first Mac, M1 MBA. I must say that its very well built and everything feels faster than my Ryzen 4800U notebook.perch101 - Tuesday, January 26, 2021 - link
Sureeeeee, let's bring up the only benchmark where Safari's built in Javascript engine can cheat at, and call it a "magnitude" of improvement!I swear to god, Apple sycophants pretty much focus on the two use-cases in benchmarking software that are extremely friendly for wide-issue CPUs like theirs (JS and Geekbench) and then dishonestly assume that performance elsewhere is bad/broken for x86.
ikjadoon - Tuesday, January 26, 2021 - link
Oh, geez. M1 is notably faster in Kraken, Jetstream, etc. Any web benchmark you look at, yes: M1 has a sizable lead.Y'all get so emotional when the M1 is brought up and it makes zero sense: it is *no surprise* that one of the fastest 1T CPUs in the world also does well in JavaScript.
Pro-tip: 1T performance and JavaScript performance are quite closely correlated. :)
Sources: https://www.notebookcheck.net/Apple-M1-Processor-B...
https://arstechnica.com/gadgets/2020/11/hands-on-w...