We’re running preparatory tests for our AMD R5 2400G & R3 2200G Ryzen APUs, codenamed “Raven Ridge,” and that means testing low-end GPUs and CPUs for the APU counter-test. Historically, this has resulted in very clear, well-defined conclusions for gaming applications: Depending on present market climate, APUs can be an exceptional deal or, well, no better than a cheap CPU and dGPU. This time is different, though: This time, AMD is shipping its APU with the new Ryzen architecture – significantly more competitive than Bulldozer – and its Vega IGP component. Together, we anticipate this could change the game to push favor toward APUs. We’ll see in our APU test, of course. That’ll go live in the next few days, as we’re awaiting arrival of our CPU.
For when the coverage does go up, here's what we know of the AMD APU specs:
AMD R5 2400G & R3 2200G Specs
AMD R5 2400G | AMD R3 2200G | RX 580 | |
Cores/Threads | 4C/8T | 4C/4T | N/A |
CPU Base/Boost | 3.6/3.9GHz | 3.5/3.7GHz | N/A |
Compute Units | 11 | 8 | 36 |
Streaming Processors | 704 | 512 | 2304 |
TMUs | 44 | 32 | 144 |
GPU Boost Clock | 1250MHz | 1100MHz | ~1266MHz |
CPU L1/L2/L3 | 64KI/32KD 512K 4MB |
64KI/32KD 512K 4MB |
N/A |
TDP | 65W | 65W | ~185W |
DRAM Support | DDR4-2933 | DDR4-2933 | N/A |
Transistors | 4.94B | 4.94B | 5.7B |
Die Size | 209.78mm^2 | 209.78mm^2 | 232mm^2 |
Price | $170 | $160 (?) | $250-∞ |
For now, however, we can get started with low-end GPU and CPU testing. This will allow us to establish a baseline for what’s acceptable under various gaming workloads, then apply that to the forthcoming 2200G & 2400G APUs.
For testing, we made the following significant decisions for the test plan:
Intel: We’re going with DDR4-2400MHz memory, as opting for Z-series boards and 3200MHz memory seems wasteful in this particular application. Realistically, a user is probably testing at 2400MHz maximally (pursuant to platform compatibility sub-Z-class).
AMD: We’re going with 2933MHz memory. Again, diminishing returns given current memory prices make 2933MHz palatable in both price and performance. We plan to test the APUs at various frequencies, including 2933MHz and above. We are also running the R3 1200 at 3.9GHz, a trivial overclock.
Other than these components, we’re using an ASUS Crosshair VI Hero motherboard for the AMD platform and a Gigabyte Z270 Gaming 7 for Intel – again, we’re classing memory down on this board, despite using an actual Z-series for testing. Cooling is handled with the Kraken X62 CLC, and is no problem. We’re using our NZXT Hale90 v2 for the PSU. Memory is configured to 2400MHz CL14 on the G4560, and 2933MHz CL16 on the R3 1200.
Before beginning, just a reminder that, as usual, you can’t really compare results cross-site or cross-benchmark. We will return with APU numbers once we’ve run them in-house.
Rocket League Benchmark: GT 1030 & G4560/R3 1200
Our goal was 60FPS in Rocket League, as we figure these eSports titles would do best with lowered settings in exchange for higher FPS. We configured the game to High settings, with the World Detail drop-down configured to “Quality.”
The result was a tie – but it’s clear that we can achieve 60FPS AVG and reasonable lows on both configurations. For $160, the G4560 and GT 1030 do well here – we’re holding 60FPS somewhat easily, even with the cheaper memory, and the R3 alternative fares just as well. We’re bottlenecked by the GPU more than anything. The real test will be to see how AMD’s Raven Ridge 2400G performance. We have one of those coming in within a few days.
For now, we’ve set the stage for what the APU has to beat. These parts – both CPUs and the GT 1030 – are all readily available, which means you’re not contending with the GPU availability issues higher-up in the chain. This is direct competition to the APU.
DOTA2 Benchmark: GT 1030 & G4560/R3 1200
DOTA2 is next. For this one, we tested using high quality settings and 1080p at 100% resolution scaling.
We saw more of the same: Performance bound at 63FPS AVG, with the GPU likely contributing to some limitations here. The 0.1% low performance looks a bit better on the R3 1200, but we are also within margin of error. For all intents and purposes, these results are functionally the same. What is noteworthy is that the GT 1030 (and its low-end CPU partners) are fully capable of high-quality eSports gaming at $160 to $180 combined. Not bad at all. The Raven Ridge APU is the price-point challenger of this. We’ll soon know how that does.
CSGO Benchmark: GT 1030 & G4560/R3 1200
Counter-Strike: GO is next. For CSGO, we tested with a mix of high, medium, and very high details, pursuant to previous optimization tests we’ve done for graphics-versus-performance. The configuration is in the article.
For this one, the G4560 and GT 1030 ranked at about 111FPS to 120FPS AVG, with lows functionally equivalent between the two configurations. As expected, CSGO is almost laughably easy to run on any modern hardware configuration, even a $300 box.
Overwatch Benchmark: GT 1030 & G4560/R3 1200
Overwatch provides another low-load FPS eSports title. With Medium settings, we appear to be bound again by the GPU. Both CPU configurations rank at roughly 61-64FPS AVG – within variance for this title – and lows are functionally equivalent. We’re able to sustain about medium settings on this El Cheapo PC – not bad overall.
Ghost Recon Wildlands Benchmark: GT 1030 & G4560/R3 1200
For a non-eSports title with higher quality graphics, we can look at Ghost Recon: Wildlands at 1080p and low settings. This puts into perspective how much a GT 1030 will struggle with higher-fidelity games. We’re at 38FPS AVG when running low quality settings, which subsequently look pretty bad.
Ghost Recon at Very High is obviously going to be completely unplayable on this configuration, but let’s get a chart on-screen anyway. The point of this is to demonstrate relative performance scaling versus our standardized GPU review bench, which uses a 7700K and the GPUs listed. The G4560 doesn’t matter much here, as the GT 1030 is choking the entire pipeline. For perspective of scaling, we get about a 68% performance uplift by stepping up to the next worst configuration.
Total War: Warhammer Benchmark: GT 1030 & G4560/R3 1200
Total War: Warhammer is one of the games where we leveraged our existing data pool just for some perspective. Although not really that playable or enjoyable on the low-end platforms we’re testing today, benchmarking at 1080p/High still allows us to establish some relative scaling versus other CPUs with our standard 1080 Ti test GPU.
Again, all the other CPUs were tested in our CPU reviews, so had a 1080 Ti with them. With these low-end platforms, we end up at about 21FPS AVG for the G4560 and GT 1030, or about 30FPS AVG for the R3 1200 overclocked CPU, also using the GT 1030. This establishes just how much vertical room there is in the components.
Moving on to 1080p/Medium, a bit more realistic, we observe a GPU bottleneck at about 36FPS AVG, with both CPUs roughly equally constrained by the $80 GPU.
Timespy Benchmark: GT 1030 & G4560/R3 1200
For one final relative performance identifier, TimeSpy shows us that the overclocked R3 1200 performs significantly better in its physics processing for this test, leading the G4560 by about 50%. GPU scaling shows us as roughly equal in both configurations, expectedly, with the RX 460 4GB leading by about 60%.
Conclusion: 2400G & 2200G APUs In-Bound
Our AMD R5 2400G & 2200G are on the way, and we’ll soon come to a full conclusion (in a separate, full benchmark piece – or series thereof). For now, we can conclude this: It is possible to play these lower-load “eSports titles,” like DOTA2, CSGO, Overwatch, and Rocket League, at reasonably high framerates with a GT 1030 and ~$100 CPU (or less). Higher fidelity titles like Ghost Recon struggle significantly more, and although technically “playable,” they are poor enough in visual quality and framerate to be devoid of fun. We’d recommend a higher-end set of componentry for such titles.
For the eSports games, the GT 1030 and G4560 or R3 1200 are both sufficient. What matters, of course, is whether AMD’s new APUs can surpass the framerates achieved by these dGPU + cheap CPU combinations. The price totals to about the same, ignoring complications arising from DRAM configuration on APUs, and so the comparison will be a relatively price-linear one.
We’ll revisit this in a few days.
Editorial, Testing: Steve Burke
Video: Andrew Coleman