I did some serious finetuning in the bios myself long time ago, so I’m pretty sure I’m running my 5900X close to its sweetspot, in terms of clocks/temps/stability. I can’t recall the exact numbers right now, but it’s a combination of PBO + CO with custom power limits.
At the same time I’d say I have not been too lucky in silicon lottery, as for example I can’t run my 4000 MHz RAM kit on 3800 MHz at low latencies, which would be perfect for a Zen3 chip. The best it can do is 3733/CL16, but I’m runing it at 3600/CL14, which is still fine, but not exactly perfect. And it’s not RAMs limit, I had it tested with different CPU hitting 3800 with no problem. So its rather my CPUs memory controller not able to deal with it, I guess. Anyway I’m not saying it runs poorly, but we all know how CPU limited this sim is, so after almost 3 years I’m looking for ways to raise the CPU headroom up a notch.
Thanks for the reminder.
I’ve not dived very deep into the 2024 promo yet, so I’m missing on some of it’s features for sure. But there is a lot of time for me to catch up before it comes to light. Will be on it for sure.
If it ends up to be everything they promise, then I’ll have no problem switching to it. But to be completely honest I’m a bit sceptical mainly about the ability to massively improve multithreaded performance. There are things that can be multithreaded, but there are a lot of things that really can’t be. There are some details that seem contradictive in Asobo’s presentation of 2024 so far… such as ability to easily move add-ons from 2020 into 2024. This leads me to to believe that the very base of the sim is basicly identical, or very similar, for it to be a case.
This is also supported by the visual details from the trailer, matching some of the 2020s graphical “imperfections”, such as visible horizon line. All I can say for now is that I will not be in rush to buy the next edition until the detailed technical benchmarks show up, after its release.
the file format of 3d scenery and aircraft models (the visible models and textures)
the file format of airport and aircraft metadata (runway locations, taxiways, parking spots, etc; the layout of an aircraft and the coefficients and speeds and tables and other figures that control its aerodynamic behavior, engine behavior, and other configuration)
compatibility with code APIs for JavaScript and WebAssembly code running in the add-ons (avionics display and behavior, any additional custom coding for control the scenery and aircraft)
None of that is closely tied into how the engine performs parallel computations, like how clouds are processed or how aerodynamic lift across various portions of the wings might be performed in serial or parallel, or how it processes the many thousands of objects in the scene as you look around it.
A comparison might be to ask how a new version of Windows can be faster, or have new capabilities, if it can still run programs that ran on an older version of Windows. It’s Windows, not the program, that changes and increases, and it’s not the part of Windows that the program is interfacing with.
In general you should ABSOLUTELY ASSUME that most of the engine is the same – of course it is, that’s normal. But some portions will be massively changed and improved; that’s how software gets upgraded.
Fair enough, that’s your perogative.
Just another instance of where the Xbox showcase was abysmal – the visible horizon line was (allegedly) already fixed by the time the presentation trailer was shown.
Great post. Succinctly describes what is happening.
Great questions! Ultimately, the key reason is this: the AMD Adrenalin driver has an CPU auto overclock feature that will determine the optimal PBO settings. Press one button and bam, it’s overclocked! When I activated this feature and then went into my bios to see what had changed, I found the above parameters except for: A) I tuned the individual cores under the curve optimizer, B) I changed the max CPU boost from 200 to 75 to help minimize heat while allowing a slightly higher boost rate.
In digging around to get some background I found that the wildly high PPT limit of 1000 merely tells the motherboard to use the max stable value for this parameter. I still do not know what the end value is.
I use the scalar of 10x because it helps achieve what I want - longer boost times at a higher Ghz level than the stock 5900x. Normally, Ryzen CPUs manages the frequency based on the thermals and voltage and it will reduce your frequency so as not to damage the CPU. The bottom line is increasing the scalar can let your system sustain higher frequencies without being throttled.
Since AMD recommended 10x via auto overclock, I felt it was safe to use.
