…I recognize when there is a difference in fluidity which is not linked to the servers but due to Windows/PC/Components etc and effectively, Dx11 feels much better on keyboard/mouse, meaning it's smoother and it's like you're suddenly playing without V-Sync BUT cause more stuttering in my case than Dx12 (i'm not the only fan to have this experience on this game specifically)
Just to be clear, are you talking about just input lag on IO, or actual perceived stutter in rendering as a result from you using the keyboard/mouse?
Because if IO somehow affect frametimes, then that sounds very much like "poor" engine design to me, that different threads are somehow waiting for each other. And there are differences in how DirectX 11 and 12 dispatches rendering queues btw.
And even though I haven't played that game, it seems to me to be a multi-API game, which usually means an abstraction layer that is common to cause such things.
Do you observe the same tendency in games that are made purely for either DirectX 11 or 12?
I personally use the Bitsum power plans with ParkingCore software, it means that the CPU have always maximum frequency.
Arrow Lake will age better than x3D competitor but just because it have more clock speed and more cores or there is no relation?
Because, if we only look at raw performance (for gaming, due to 3DV Cache) x3D CPUs are the most powerful, right? so does x3D will age better?
Firstly, Arrow Lake does have higher IPC, so running it at locked clock speeds it will outperform Zen 5 in raw computational performance. (except AVX-512 loads)
As for L3 cache, having more of it doesn't add computational performance, it only helps to keep the CPU fed. As anyone proficient with low-level coding could tell you, adding more L3 primarily helps with bloated software, as cache-hits in L3 are primarily instruction cache hits. So in applications where the computational density is lower vs. software bloat, then the
relative performance gains will be higher. If you optimize the code to get rid of abstractions and make it denser, then the "advantage" of more L3 will actually dissipate. Many would think this is a bad thing, but on the contrary, code should be as scalable as possible. The perceived advantage of more L3 cache is usually the greatest in unrealistic edge cases, such as gaming at 720p or low with the most powerful GPU. But, they are not clarifying real differences, no, they are in fact introducing or amplifying bottlenecks that real users will never see. And especially, as new games tends to get more demanding, any such "imaginary" bottleneck will only become less relevant as your system ages. The tendency of absurd benchmarks isn't new though, and I've complained about it for over a decade. Along with emphasis on synthetic benchmarks and very cherry-picked outlier games (like
AotS for a while), this has skewed recommendations for buyers for a long time.
In all honesty, the real reason to buy Zen 5 would be AVX-512 support, which is massively powerful,
if your application benefits from it. It will unlikely be of significant use for games though. And such things are one of the few cases where the hardware will "mature with age".
It's enough that I have to keep a large spreadsheet to keep track of it.
And to answer your question better;
Up until about ~2020, I've used enough hardware to know that for my use case (which is mostly coding, some multimedia and gaming), Intel have consistently been much more responsive. While I suspect that may still be the case with Arrow Lake vs. Zen 4/5, I'm not going to claim that before I've compared it. The reason why I have both Zen 3 and Zen 1 too, is to verify software.
But one thing I would like you and others with access to Arrow Lake and relevant Zen 4/5 CPUs is the following;
- Run your browser with 5-10 medium-to-heavy tabs for at least a week.
- Then,
without closing the browser, test gaming and/or productive applications to see whether you can detect some kind of input lag or stuttering vs. a fresh boot with just the game or application.
This sort of test would probably be more useful than anything to determine a real world use of such systems. And pretty much any "prosumer" I know (incl. those I build systems for) continue to run their browser while doing other stuff.
