PUBG Streaming Benchmark – R5 2600X vs. i5-8600K
For streamer-side FPS in PUBG, the i5-8600K’s baseline, stock performance permitted a 138FPS AVG framerate, which butts it up against the FPS cap – the average would stretch a bit higher without that consideration. Lows are well-timed insofar as PUBG, which has issues with spurious frametimes in the 0.1% low range.
The R5 2600X stock operated a baseline performance of 127FPS, with lows functionally equal and indistinguishably different from the 8600K’s lows. The 8600K technically maintains a lead of 8.7%, but again, this is only half of the equation – and arguably the less important half.
At 10Mbps and Faster H264, the 8600K only drops 10FPS, and the same is true at 12Mbps with the much more intensive Medium encode setting. This would indicate that scheduling is keeping the entire workload on Intel, with limited resources made available for streaming. We’ll soon check that.
The R5 2600X drops harder in framerate, so the loss as offset from baseline is a greater dip at 30% reduced from non-streamed performance, but we don’t know what that means just yet. What we know for now is that the player would likely be satisfied with performance on either CPU if live streaming, although Intel holds an objective streamer-side framerate advantage when left stock and without priority or affinity tuning.
Moving on to the viewer-side chart, here’s where those numbers manifest: AMD’s 30% drop for streamer-side performance proves a worthwhile sacrifice, because it’s able to successfully encode 100% of frames for the stream. The 8600K completely crumbles when under the load of even Faster settings for encoding. You could get this to work, but it’d require dropping the encoding quality further and tuning some affinities and priorities. Even at our reasonable, realistic quality of Faster and 10Mbps, the 8600K struggles to keep up when left stock and untuned.
For Medium, we’re looking at a synthetic test, as we noted, to illustrate outside of margins which CPU is superior. AMD claims that flag for this test, managing to deliver 26% of its frames under the strain of a Medium encoding speed, while Intel gets clobbered down to a handful of frames delivered. It’s not even a slideshow – it’s just a picture.
Now, as for how useful this is, that depends. What you’re seeing on the screen right now is a video playback of both streams at 10Mbps. This is about the quality we stream to YouTube when we’re being conservative, and the 6C 8600K has a hard time with managing both the stream and the game. Again, capping the game and prioritizing resources would help, but the 2600X doesn’t require that treatment. It’s good to go out of box. The 2600X has substantially more headroom to sustain a higher quality stream with PUBG. Both are capable of streaming, it just depends on how much quality you want and how much you want to be under the hood to work on the 8600K.
Power consumption while streaming is also sort of interesting. Clamped at the EPS12V rails, the Intel i5-8600K only ever draws between 37W for baseline non-streaming and 65W for streamed output. The R5 2600X baselines at 62W, with streaming at 111-114W. That’s a significant increase over the Intel CPU, but it’s also accomplishing the task that we asked it to do, so that’s fair.
The i7-8700K draws about 48W baseline, or 98 to 105W when streaming. The R7 2700X pulls between 68W baseline and 135W for a heavy stream. This is also why AMD’s TDP rating will confuse people – that’s not the same number as straight power consumption, as we explain in the review.
DOTA2 Streaming Tests – R5 2600X vs. i5-8600K
For DOTA2, streamer-side framerate positions the R5 2600X baseline at 141FPS AVG, led by the 8600K at 182FPS AVG. This isn’t uncommon for DOTA, which tends to favor frequency at least somewhat. On the face of it, Intel has a substantial lead – but again, we have to defer to the viewer experience chart.
Streaming on the 8600K doesn’t hurt our framerate too much, bringing us down 20FPS to 161FPS AVG.
As for the viewer experience, here’s what it looks like when left completely untuned by the player. It’s not so good for the 8600K, as illustrated plainly here, even though the streamer is getting over 160FPS AVG during this time period. The viewers get one frame every now and then – 1FENAT, as we affectionally call it. It’s our new metric. 1FENAT is equal to one frame every now and then. The upshot is that, when it does eventually cough-up a frame, it’s 58% likely to be within a 16.7ms delivery window.
The R5 2600X at 10Mbps Faster doesn’t have these problems, and delivers 100% of its frames. It may sustain a lower streamer-side framerate, but the stream itself is perfectly delivered, outside of a 10% variance on frame delivery timing. AMD’s frequency deficit contributes to the lowering of its hard FPS, but its thread advantage is what grants the outright lead in streaming.
For what it’s worth, we did try setting OBS to high priority of the 8600K. It helped, but you’re entering territory where the player has to do more hands-on tweaking to force something to work. It’s a bit ironic, really, as this is the situation AMD was in just one year ago, with Ryzen’s messy lineup of launch BIOSes and motherboards. Now, AMD’s platform is relatively mature and stable, but Intel’s i5s require some help to pass the tests.
Blender CPU Benchmark – R5 2600X vs. i5-8600K, R5 2600
We’ll start this one with Ryzen’s strong point, like the 2700X video. If you missed the 2700X review, absolutely go watch it – the biggest improvement in Ryzen+ is its ability to hold a similar clock at a significantly lower voltage than Ryzen 1, which also aids in power consumption. That doesn’t apply as you get to 4.2GHz, of course, but really applies at 4.0GHz. We won’t recap that here. Check the other video.
Anyway, starting with our in-house 3D rendering scenes that use Blender, we see the advantages of the R5 2600 and 2600X. The R5 2600X CPU completes the GN Monkey Head render in about 28-29 minutes, ranking it about 2 minutes behind the R7 1700X stock CPU. This is serious performance, considering the 1700X runs two additional cores – and that’s significant in the world of perfectly scaling render software.
Overclocking is, again, completely pointless here. XFR2 does the job for us, and we had no headroom to go to 4.3GHz.
For comparison, the 8600K at 5GHz barely manages to finish its render in 33 minutes, granting the 2600X a 13% time reduction. Not bad at all.
Perhaps more impressively, our GN Logo render applies a different type of load and stress to the CPUs, and allows an overclocked 2600X to surpass a stock 8700K, somewhat remarkably.
