What I am going to be interested in is this versus EPYC parts. I think the higher clocks are mainly to achieve some of the more insane (and useless) FPS counts for games. If you are willing to ramp down the FPS to a number that your monitor can actually display, it may be better to find a general purpose EPYC MB and chipset, and use that. Especially if homelab / big data / compiling linux/ occasional gaming is you cup of tea.
For regular gamers I totally agree with your comment. But, since there are Twitch streamers out there the Threadripper might be a good fit if you want to stream since encoding must be done and preserve a high FPS (60 FPS or more). Also, it would future proof you maybe for VR streaming or 4k and higher streaming.
Ryzen 7 and 8 targeted the streamer market. That platform has reproducibly breezed through game streaming workloads.
Threadripper is for something else. "Professional work" most likely: developer workstations, 3D modeling workstations, etc. Given that all Ryzen SKUs support ECC [1], this is probably a good platform for individuals who have historically sought out server platforms.
[1] So far as I understand it, they haven't disabled ECC in hardware but don't promise that it works.
Ryzen 7 already struggles with two streams (2 separate x264 encoders with different settings for different streaming services, and I think that was tested at 1080p). Really serious streamers might pick up Threadripper :)
IME, Ryzen's very happy with a 1080p60 stream with fast or even medium quality, leaving headroom for CPU spikes because of rough encoding patterns. Really serious streamers are generally more advised to just split the existing stream at that point; 3.5K/medium (Twitch) and 5.5K/medium (YouTube) are pretty similar in motion.
Resolutions has nothing to do with it. To the contrary: You need less cpu power since you will get your FPS limit from the gpu, not the cpu. Similar for VR: Current cpus are strong enough to get the 90FPS current VR headsets want. 240Hz maybe, but those processors won't be faster in games than a Intel Core i7-7700K, as games don't use that many cores.
I think that is is comparing Threadripper to other workstation cpus, or it rather explicitly does compare to EPYC. Since Threadripper has a high clock, it might work better in games as well. Parent is right in saying that when adapting expectations, that additional gaming performance is unlikely to be necessary.
Interesting argument, do you know of any CPU that, combined with an FE VEGA or GTX 1080Ti, reaches > 144fps in modern games at max settings and 4K resolution?
The GTX 1080 Ti does not reach that many FPS on 4K, it mostly strives to get 60. It does not matter which cpu you run. When you are on 4K, you could be happy with a small Intel i5.
My only concern with 4 core/4 thread CPUs like the majority of i5 models is that the video game industry is already showing some signs of optimizing for having 8 threads on the PS4 and Xbox.
Watch Dogs 2 on PC, as an example, has widely documented CPU bottlenecks on 4 thread CPUs even at 1080p. I just replaced an overclocked i5 4690k last week with an i7 to solve this issue, and is the first game I've played that was meaningfully CPU bound with my 1080 ti on 1080p/1440p displays. I think the days of an overclocked 4 core i5 being a great value choice in high(ish) end PC gaming are probably coming to an end soon. I'd certainly think twice on a new build.
Watchdogs 2 also runs surprisingly well on FX cpus, as seen in that benchmark. But one has to be careful with console ports anyway, Batman Arkham City for example would paint a very strange picture of PC performance (the port was and is a disaster)
But there are other valid examples, like Battlefield 1 in multiplayer. The 4 core era will end, and consoles might make that happen faster, but especially for 4K gaming I would not worry yet. Till gpus are fast enough to make cpu the bottleneck in those there will be a few more cpu generations to come. At least based on current performance and how gpus normally develop. We only just reached 60 FPS on high settings there, and that with the most expensive consumer gpu available.
But on 1080p and 1440p, that's a different story. Being more future proof for that development is one of the appeals of the Ryzen 5 1600 (6c/12t).
It's already ended. I have an 8-"core" AMD FX CPU, and I decided to underclock it to 1GHz to see how games reacted. You'd be surprised how playable many games are. Mostly, I guess, because the thread communicating with the GPU does not need all that much power, so frame rates stay semi-playable.
Anyway, with an underclocked 1GHz FX 8350, The Witcher 3 saturates about 5 cores in cities and nibbles on the sixth core. Dragon Age: Inquisition uses up to 8 cores. RiME uses 4 cores. TrackMania Nations uses 1-2 cores.
Generally, AAA games have been post-quad-core for years now.
I'd say it's already all but at an end. The Witcher 3 is probably the most intensive game I currently own. On my CPU (i7-4790k, 4 core/8 thread) standing still in a crowded city environment shows about 50% load across all 8 HT cores. Once you start running through the city use across all cores rises to a steady 60-70% with regular (every 2-3 seconds) spikes to 90+% on all cores. That's a two year old game. I have a feeling that when the next big wave of major AAA titles starts to hit, many of them will be severely limited by 4 core/4thread.
If your 1080 Ti "mostly strives to get 60", your game is rendering way too much heavy effects :) If you actually tweak game settings (mostly turn off Screen Space Reflections, Ambient Occlusion, crap like that), you can play many games at 4K60 with much cheaper hardware.
Well, OP talked specifically about max setting. But you are not wrong. That's a problem in general currently. Ultra/Max setting are incredibly heavy, but don't change the optics that much, especially not on a resolution that high.
Not too sure how stable your FPS will be on cheaper hardware, but not running max has to help.
