ChoMar wrote:Now, im talking about AAA title and so on, not factorio:
Except in some rare use scenarios CPU rarely matters in gaming.
The CPU only limits you in very high Framerate scenarios. Like >100 fps. Now, E-Sports needs this framerates. The rest of us is better of with looking at beautiful games in high resolutions. In high resolutions, even a Titan X or stuff like that will limit your performance long before the CPU does.
Conclusion: Unless you are a professional FPS-Player, the Ryzen will have no downside and you can use those massive cores for a lot of stuff. That said, if you DONT need those cores for other stuff, there ARE good and cheap Intel CPUs out there that will, due to the same logic, allow a good gaming experience for less money.
To a certain point I agree with the notion that with certain games the CPU really doesn't matter because the game might not be that demanding and you get >100 FPS (which only professional E-Sports players may need) ... OR they are memory-bound (like the Factorio-devs stated they are for example) ... OR because the game is a true graphics monster (like Crysis used to be back in the days) where the GPU is going to be the limiting factor long before the CPU becomes the bottleneck.
But that said CPUs do matter in various categories of games... mostly older games that aren't well optimized to take advantage of multi-core CPUs because there were no multi-core CPUs around when they were developed, or even in newer games where the developers just don't have the time, money, know-how or just not the will to optimize for multi-core CPUs or because the stuff they are doing can't really be parallelized that well for some reason.
What those games will do is push 1, maybe 2 cores to 100% while the others are idle and bored to death. If I take a good look around on the games I'm playing mostly then that's what still happens more often than not, even with newer games (though it got better the recent years, especially at the AAA sector or cutting-edge indie games that are really trying to impress).
So when a good portion of the games you are playing fall into that category then that's when you find out that an expensive 8 core (350 bucks is expensive in my opinion) is essentially a waste of money... which is a downside. Because that money could have been spent on a faster quadcore CPU or a better GPU for example... especially if most of the time you don't take advantage of what 8 CPU cores are capable of.
Also the 8 core CPUs often don't perform that well on those kind of ill-optimized games/applications for the reason that the 8 cores face a problem with their TDP. They reach their TDP limits much earlier due to the excessive amount of cores/cache and other features that still draw some idle power and therefore can't be clocked any faster (even with advanced Power states and Turbo and whatnot) without exceeding the TDP. So they often tend to suck at single/dual-core performance due to generally/usually lower default clock rates. Basically they sacrificed single-core performance to have that many cores. Which is another downside.
On-top of that after a certain amount of cores the latency between cores/cache/IO/memory controller becomes more noticeable anyways due to the way there are limits on how they are all interconnected to maintain the symmetric multiprocessing aspect. Which also seems to be more noticeable especially on the Ryzen 8 cores due to the way they are designed with 2 x 4-Core modules as
Koub already pointed out earlier in the thread.
So knowing that I've always been picking a quadcore CPU the recent years... because with the same TDP (for example 95W) they usually should have much higher default frequencies and turbo clock rates to increase single/dual-core performance, while still being able to use 4 cores for the few games that actually take advantage of more cores. It's the best of two worlds in my opinion.
Which is why I initially wrote that I wonder who is in charge of AMD's market research department because if they really wanted to target gamers then they should know by now that if they want to score on that sector that single/dual-core performance is still quite important and that can also be achieved with high clock rates (Increasing Instructions per Cycle is fine and it's awesome how much they improved on it in comparison to Bulldozer, but there's still an obvious micro-architecture-based limit to how much you may gain that way).
So what they probably should have done first is to release a really fast 95W TDP quadcore with high clock rates and large Turbo margins for the gaming/performance desktop to target Intel's 7700K (which is probably also one of the most sold CPUs and also sets bar on that sector) and delay the release of the 8 core for workstation/server/enthusiast sector a few months until the yield is satisfying and they know how Intel reacted to the quadcores and then adjust from there where necessary.
ChoMar wrote:Factorio? Im still wondering how Factorio will perform on Ryzen.
Probably equally or worse than on a comparable Intel CPU, or that's at least what I expect since the game is memory bound as the Devs often stated... so you'd waste money with an 8 core anywars, even if a Ryzen R7 1700 is available for 350 bucks.
ChoMar wrote:Im ALSO wondering if the DEVs compile Factorio for Ryzen. Seeing how much optimization went into the game so far, im pretty sure they already tested if CPU Optimization has any impact and if yes, they already compile for Intel.
If I would have to take a wild guess then they don't have a computer with a Ryzen CPU in their office yet... and they will only get one once/if people start reporting/complaining about weird behavior of Factorio on such a CPU... that or when they really need another workstation.
Knowing AMD it will probably also take a while for them to release the development kit that actually helps profiling/optimizing software for Ryzen CPUs, so I wouldn't wait for that miracle to happen... and if AMD really wanted developers to optimize for their CPUs then they should have thought of that optimization gap beforehand (since I'm sure they pretty much knew of the gap already months ago due to lab tests) and co-developed such tools during the development of Ryzen so that at least major game development studios have a patch ready when Ryzen hits the shores (which they didn't)... instead now they are rather whining that all recently popular games used in all the benchmarks are too optimized for Intel CPUs... as if that wasn't an obvious logical consequence of their long absence from the competitive market.
Also in the end both Intel and AMD are going to push out a new generation of CPUs with new micro-architecures/improvements/manufacturing processes anyways next year so optimizing for a particular CPU is not really that profitable for smaller companies that don't have the manpower... so I bet even if there were development kits for optimization to squeeze out the last few percent of performance... a lot of software companies will still not take advantage of it.
Which is why I wrote... if AMD wanted to perform on the gaming sector then they should have done it through raw brute-force performance because that's the only way they may be able to get back into the competitive business.