davis.anthony
Veteran
Current Generation Games Analysis Technical Discussion [2023] [XBSX|S, PS5, PC]
- Thread starter BRiT
- Start date
- Status
I don't know about ray tracing and 7700x but my 2700 at 3.7 GHz is pretty cool on this title without all that stuff.the 4090 is sitting on its laurels, I guess.
Wish Eurogamer also did a technical review.I am gonna miss that a lot, 'cos I like the native vs pixel reconstruction techniques comparisons and so on and wanted to know how this game behaves with XeSS and stuff like that.
I have a good ol' 3700X and I wonder what's with AMD not improving single thread performance that much on a new 7700X. Also wish there was a demo so I could test performance myself. According to this guy the game runs perfectly on a very similar CPU compared to mine -the 3900X- and the A770 (which is included in the recommended settings).
I can pretty much get an almost (almost, Zirael, almost) stable 60 FPS everywhere as long as I'm not GPU bound, including Hogsmeade with tuned settings. CPU sides of things are pretty calm on my end.
I'd really like to hear their takes on absurd RAM usage and potentially what texture settings do. As the other user posted, is there even a difference between low and ultra textures; if not, what is that setting doing? This is why I want more transparent setting explanations from devs. This is just shot in dark.
Sorry for multiple posts.
Sorry for multiple posts.
the 3700X ... only a 20% slower than the 7700X or a monster like the 13900K? Something doesn't compute here, the difference should be much bigger.
davis.anthony
Veteran
That's pure IPC at a fixed clock rate, don't forget in the real world these CPU's also clock 25%+ higher too.
Last edited:
7700x and 13900k clock higher5 GHz and above IIRC. Yours run around 4-4.2 GHz, IIRC. IPC improvements are indeed pathetic on the CPU side for the last 10 years. Even a mere %20 improvement is seen as "holy shit %20???". While GPUs get 2x improvements over mere 2 years, we can't even get 2x IPC improvements over like... 5-7 years. This is why I guess in the end NVIDIA said "damn it" and procured frame generation. They're aware that CPUs won't be able to keep up with their GPUs, especially in such a convoluted platform where CPU bound optimizations are non existent.
The real improvement here is also multi core performance.
The x3d performance is a bit odd. I really expected it to be much higher but maybe this will also only show in multithreaded applications.
Only wish more games exploited the vast power of multicore CPUs. Seems like we've almost regressed in that regard, with Gotham Knights and Hog Legacy being extremely single threaded and not very multi threaded. The argument of how it is hard to make an engine super multi threaded always pops up in such discussions, so I'm aware of the limitations as well. I wish whatever Doom Eternal was doing would be the standard. One of the rare AAA games that my 2700 can scale above 200 framerate average at low resolutions, while utilizing tons of threads gracefully. But then again, I'd like to see how that engine fares in an actual open world game too.
well, just let's say, Doom has almost no AI. Doom is filled with "random" enemies that just have one target (to kill the player). There is nothing in the world that is much more complex than this. They can purely concentrate on the graphics pipeline. In games like Hogwarts legacy it is a bit more complex, even if most NPCs, ... are just scripted. The open-world setting of that game also doesn't make it easier. But in the end, it has much to do with the development time and how experienced the developers are.
Making something single-threadded is much, much easier and less bug-intensive than something multithreaded. E.g. this is also one reason why nodejs is so successful it is just easy to not be aware that multiple threads can change a resource you are working with. I only develop business logic but even there multithreading can be a nightmare when you want to analyze a buggy behavior.
But maybe we see some improvement e.g. in the Hogwarts game when it is released for the last gen (including switch). Those CPUs are much less capable but they must only reach 30fps so this might that hard to reach when it already runs at 60 on current consoles.
davis.anthony
Veteran
I blame the mid 2000's when quad cores just hit the market and their PR was all about spreading everything over a 'x' CPU cores, it gave people the wrong expectations as to what can and can't scale across multiple cores which is why people now expect everything to fully load their 24 core CPU.
I think it's also worth mentioning that certain things just do not scale very well over multiple cores.
PS5 framerate in Hogwarts Legacy. I'd say the RT mode on a 120hz TV is running surprisingly well (for RT reflections + shadows on console obviously) and without big drops. Often around 40fps. Fidelity is useless, as always, and often barely at 30fps-ish. Performance seems the best choice as always at 60fps.
But we known the unpatched game has some frame-pacing issues in all 30fps modes and obviously short traversal I/O stutters in all others modes which can't be seen here (the FPS line would only see the big stutters that don't seem to happen on PS5 based on this video).
But we known the unpatched game has some frame-pacing issues in all 30fps modes and obviously short traversal I/O stutters in all others modes which can't be seen here (the FPS line would only see the big stutters that don't seem to happen on PS5 based on this video).
davis.anthony
Veteran
Why do people keep blaming DX12? It's just an API.
Did they ever think it's the developers who are responsible for ensuring it works as it should?
We do after all have some good DX12 implementations so it can't explicitly be down to just DX12.
Subtlesnake
Regular
Surely it depends ultimately on the engine, not the API, to appropriately schedule work?
arandomguy
Veteran
A problem I'm found is that there seems to be a perception that everything is either ST (single thread) or MT (multi thread) in the sense that MT effectively means it will scale out indefinitely. When in practice the reality is most applications including games at multi-threaded but there is inherent challenges in the amount of scaling and distribution you can do.
When DX12 was first being launched it was never presented that it would enable evenly distributing all CPU overhead across infinite threads, but for some reason there was a narrative that it being more MT compared to DX11 implied so.
DirectX 12
What’s the big deal? DirectX 12 introduces the next version of Direct3D, the graphics API at the heart of DirectX. Direct3D is one of the most critical pieces of a game or game engine, and we’ve redesigned it to be faster and more efficient than ever before.
devblogs.microsoft.com
Because DX12 demands optimization for every architecture and doing proper MT renderer on their own without the help of the DX11 driver. This is the job of the developer. But this should never be a topic for the PC plattform. What we are seeing now are these unoptimized console ports which run worse than Battlefield 5 from 2018. I can play it with over 170 FPS on my 13600K with Raytracing. So there is obvisously something very wrong with these games which cant even do proper raytracing while running so much more worse than than a four years old game which was the first to implement real time Raytracing.
The "API" is in the engine. That was one marketing point of low level APIs. The engine is driving the GPU.
Last edited:
davis.anthony
Veteran
What you've described is a developer issue and not DX12 itself.
No, it is a DX12 problem because this is enforced by DX12 itself.
Here is a comparision between a 4070 TI and 7900XT with a 5800X3D:
Calisto Protocoll has the same raytracing implementation with the UE4 engine. Only optimized for consoles and then it get ported to the PC without any adjustments.
- Status
Similar threads
