The AMD Ryzen 7 5700G, Ryzen 5 5600G, and Ryzen 3 5300G Review
by Dr. Ian Cutress on August 4, 2021 1:45 PM ESTIntegrated Graphics Tests
Is 1080p Max possible?
My goal for integrated graphics is for it to be suitable one day for 60 FPS gaming at 1080p maximum settings. At these settings, we're asking a lot of the graphics solution to do lots of compute, sometimes with memory bandwidth that just isn't there. A normal suite does a few of these 1080p Max tests, and the results are almost always dismal:
The short answer is no. You're lucky to break 10 or 20 FPS in most cases.
eSports are the obvious differentiation point, with CS:Source (an old favorite, don't @ me with CS:Go) showcasing almost 60 FPS at 4K High.
Note that on this 1080p High graph, it is interesting to see the frame rates increase over several generations of AMD Ryzen APUs, increasing 33% in frame rate from the first generation Ryzen 5 2400G.
The 4K variation shows the previous generation coming out ahead, and this was fairly consistent. This might be a case of where the power is going between CPU and GPU and the algorithm that determines where the workload should be.
Sometime we get odd scenarios such as this, when the previous generation gets slightly better results. Each result has some level of uncertainty, but even at this value, the're pretty much both unplayable.
A full list of results at various resolutions and settings can be found in our Benchmark Database.
Facebook
Twitter
RSS
135 Comments
View All Comments
mode_13h - Tuesday, August 10, 2021 - link
> that hot, expensive Gen 4 M.2 NVMe SSD you want to use on your new> motherboard will not achieve the speed you paid dearly for.
None of the 1st gen PCIe 4.0 M.2 NVMe SSDs did, in fact. A lot of them still don't. And if you're not running it at PCIe 4.0, then it's probably also running a bit cooler.
alfatekpt - Monday, August 9, 2021 - link
Currently 5600G and 5600X are at the same price in my country. Should I get the 5600G? I already have a GPU so having an integrated one is only useful in case the GPU breaks or needs to go under warranty and I still can use the PC...mode_13h - Tuesday, August 10, 2021 - link
I wouldn't get the G. The X is faster in every single benchmark, and sometimes substantially! Plus, you get PCIe 4.0, in case that's ever of interest.If you just want a backup GPU, so you're not completely dead in the water, then maybe pick up a used low-end model (especially when GPU prices cool off, a bit). I'm seeing used RX 550's for < $100, which is roughly performance-equivalent.
If you don't care about performance, then you can go even older. I have a HD 5450 as a sort of last-resort fallback, and those are CHEAP! That's pre-GCN, but I know it still works on Linux. I think it shouldn't be too hard to find something a bit newer that's also cheap, though. Or, if you have some friends who would loan you an obsolete GPU in a pinch, that's also an option worth considering.
phoenix_rizzen - Monday, August 9, 2021 - link
The "Ryzen 5 APUs (65W)" table on page 1 lists the Ryzen 5 CPUs with 8 cores / 16 threads. Should be 6/12 instead.plonk420 - Tuesday, August 10, 2021 - link
thanks for the core to core latency tests! looks like RPCS3 will definitely benefit from it \o/Oxford Guy - Wednesday, August 11, 2021 - link
‘In our largest sub-test, the Intel processors crack on ahead,’Did I miss the stuff about performance-per-watt?
If an Intel chip needs a boatload more power to do the barely faster work, how is that a victory for Intel’s chip?
Performance-per-watt is important when we’re dealing with today’s 14nm vs. ‘7nm’ situation.
There should be an entire page devoted to performance-per-watt.
mode_13h - Thursday, August 12, 2021 - link
There is indeed a page on power consumption, but the most revealing charts only compared the three AMD 5000G-series processors to each other. That was a painful omission.Intel got included in the peak power chart, but we all know that peak power is hardly the whole story.
Oxford Guy - Thursday, August 12, 2021 - link
‘There is indeed a page on power consumption’Indeed, there is no page on performance-per-watt — and the article continues this site’s erroneous tradition of claiming that getting a slightly higher score in a benchmark whilst using a ton more power constitutes a victory.
Context is key. These articles should pay more mind to practical context, rather than things like pumping 1.45 volts into Rocket Lake and ignoring power consumption failure (vis-a-vis the competition) when examining a benchmark.
mode_13h - Friday, August 13, 2021 - link
FWIW, I was trying to agree with you. Their "Power Consumption" page had several key omissions.Oxford Guy - Sunday, August 15, 2021 - link
Regardless... peak power isn’t enough to constitute a page on performance per watt.