I’m not sure they meant it that way. What they stated was there goal would be to use DX12 to add more visual fidelity but I certainly wouldn’t rule out better performance too!
MSFS gets a newer much better optimized version of DX12 called DX12 Ultimate.
This brings much better performance then DX11, and yes, raytracing for those who have RTX cards…
it brings better posibility for CPU utilization, if the developer know to use that 
please stop getting your hopes up, like stated before, the devs already pointed out that we shouldn’t expect any meaningful performance upgrades from DX12, mostly some new visual effects and polish.
And better performance.
All graphical intensive games that got an upgrade from DX11 to DX12 performed much better.
That’s true but they have to say that, with variety of systems out there. I seem to remember Laminar Research not making any bold promises either and pretty much everybody had a nice performance boost there.
Hey, why can’t we be optimistic, it’s nearly Christmas!
XP11 10.50 uses the super api Vulkan.
And yes, that was really an incredible speed boost, especially for AMD gpu’s.
Unfortunately, this will be unthinkable for MSFS…
The good news is my OC has held, and it gave me more oomph than I expected. Now I have to tweak it to see if I can get even more. But experience tells me if I do, and it doesn’t work, I’m gonna have to turn it ALL off and start over again, so I’m hesitant to do so. Of course, I’ve gotten pretty good at at least getting back to where I am now.
Yes of course Vulkan will be unthinkable for MSFS but why is DirectX12 Ultimate a bad option, DirectX Ultimate and Vulkan are both based on the same low level threading principles laid out by AMD’s Mantle API.
I believe if Asobo make the best use of it, we’ll have a performance improvement using DirectX12 Ultimate.
Sounds like an AMD issue here. If you read the internet everyone says “buy AMD cause Intel sucks in 2020”… but speaking from personal GAMING experience, this is WRONG. I had an Intel processor, upgraged to AMD, then upgraded again to Intel i7 10700k. The AMD gave me so many gaming problems (lower FPS, stutters) that I had to go back to Intel. The 10700k is just so much better at gaming than the high core-count AMD chips.
I have the EVGA 3080 FTW3 and i7 10700k running very well on Ultra settings 1440p.
You really got be thinking which was best Vulkcan or DirectX12 for MSFS. 
This review is interesting in RDR2 (although Digital Foundry have stated there are issues with the games built-in performance test), the conclusion here is that although Vulkan offers higher average frames, DirectX12 has considerably faster frames for minimum 1% lows which would be vital for smooth stutter free performance in a sim (although X-Plane didn’t do too badly).
I’m seeing the same exact issue with a 3700x and a 2070S. I also have 32GB of RAM and a SSD. What I’ve noticed is that when I fly with the C152, the Savage Cub, or the DR400, the sim runs smooth like butter on my machine. I get 60-80fps and no stuttering, none. The developer FPS counter shows all green and the experience is smooth and immersive.
As soon as I switch to using a plane with a glass cockpit, such as the Baron, the stuttering starts. I still get 50-60fps in the same situation at a big airport, but the stuttering becomes quite apparent, ruining the sense of immersion. In the developer FPS monitor, my counter turns yellow and all I see is Limited by Main Thread.
Turning the refresh rate of the glass cockpit down does not eliminate the stuttering. Even when I keep the avionics/glass cockpits turned off, I still get the stuttering. So, I’m not sure if it’s the glass cockpit that’s causing the stuttering or just the fact that these planes are somehow otherwise more CPU intensive.
Can anyone here confirm a similar experience?
It won’t, don’t listen to him, he doesn’t understand how things work.
By increasing resolution you put more load on GPU which leads to GPU bottleneck. So if the game is bottlenecked by CPU and if you increase the resolution at some point you will reach a resolution where your GPU becomes a bottleneck.
So switching to 4k you’ll get less fps, more load on GPU and less load on CPU because there will be less frames to process.
For now it’s CPU limited and you can’t do anything. Cores don’t matter, only single thread performance matters.
Switch to ryzen 5900X or wait for DX12 update which might help with CPU utilization.
I would suggest running some benchmarks and see if your PC performs how it should. I have 2070s and 3070x and my performance is better while having terrain detail at 200 and most settings at ultra.
Yeah, I can also confirm this. I get horrendous FPS in the A320 cockpit view, but my CPU and GPU are only ramped up to about 40-60% usage. Seems like there’s some bizarre self-imposed bottlenecking at play here.
Only appears to be an issue when you’re in the cockpit. The avionics refresh-rate setting does not help much either.
Yes, forgot to mention that also. My CPU utilization hovers around 25% and my GPU is around 75%.
I’ve perused the thread and just want to offer my $0.02 here. This is not meant to be a textbook answer. There are way too many variables to go over here that make some of this back and forth. I’m trying to make it as relevant to FS and FS only as I can. If you want to know how things relate to non-FS tasks, the rules are similar, but different tasks may react better or worse.
First thing I see in the video is “coherentgtdraw” with severe steps when looking forward and at the same time, your FPS is poorer than average. This tells me you’re using the high instrument refresh settings, which is a heavy CPU burden. Given AMD clock speeds are behind Intels (and even on intel, I might only put it to medium), this should be on low to loosen up the CPU for other things.
AMD has 2 things different that they focus on different from Intel - IPC and more cores/threads. MSFS isn’t a multi-threaded workload like video processing or 3D CAD rendering, but is more of a single-threaded workload, which IPC can balance a bit, but clock speed is still king. IPC (instructions per clock) can make a 3.4ghz processor feel like a 3.6ghz processor, but it won’t make a 4.6ghz processor match a 5.3ghz processor. 4.9ghz AMD might equal a 5.2ghz Intel, but that depends a lot on the optimization done and the instructions being asked of the CPU. DX12 fixes SOME of these issues and allows some tasks that are linear in processing in DX11 to be done simultaneously on different threads in DX12, but that’s only SOME things, not all. You WILL NOT SEE 16threads at 100% just because you have DX12 in a game.
With a game like FS, there is one thread that is in charge of everything that is calculated and tasked. That thread is called the world thread. In FS terms, it’s the MainThread. The main thread asks for a flight model calculation, asks for files to be loaded, asks for memory to be copied, asks for data to be calculated from the multiplayer engine, tasks the ATC engine to make decisions and everything. Those other tasks can sometimes be delegated to other threads. You can load 10 files at once from the hard drive instead of waiting on one by one depending on some factors, but each file loading may use a different thread to do so. Until all the tasks are completed by the workers, the MainThread can’t move on to the next step.
I’m a metaphor kind of guy. Let’s say you’re making something like scratch waffles. You have 3 people in the kitchen. The head chef knows the recipe while the other 2 people just assist the chef. In a optimum position, you could have the 2 assistants each with their own assistants, but at some point there’s just too many people and they get in each others way.
So, first step is to get all the dry ingredients and mix them. So, head chef gives some ingredients to assistant 1 and other ingredients to assistant 2. They get the ingredients, measure them, then put them on the table in front of head chef. As they bring him the ingredients, he verifies the ingredients and measurements and adds them to the mixing bowl. There may be time spent waiting on someone to come back with an ingredient, so the head chef is idle. There maybe times where the assistants both come back at the same time, so the head chef can only deal with 1 at a time, so the second assistant is just waiting to hand off his ingredient and then get assigned the next ingredient.
Once, the dry ingredients are done, now it’s time to repeat for the wet ingredients.
Now the wet ingredients are done, it’s time to start the actual cooking. One assistant whips the egg whites while the other assistant sets the waffle maker to the proper temperature and keeps checking the progress. Head chef, mixes the wet and dry ingredients until it’s time to add in the stiffened egg whites. Once the assistant with the egg whites is ready, he informs the chef, who’s likely already done mixing the wet and dry. They mix the chef’s batter with the egg whites, but now have to wait on the waffle iron to heat up.
Once the waffle iron is hot, the chef just has to pour the batter in, wait until the iron says it’s reached the cooking temperature and is done, opens the iron and out comes the completed waffle.
Now, if the game were like this cooking situation, the loading of files from the hard drive to RAM and copying from RAM to VRAM would be like gathering ingredients. Then, positioning those in the 3d space, calculating the flight physics and where the objects will wind up after a particular amount of time has passed and figuring out what the ATC engine has cooked up for the various aircraft bumping about is like the actual process of cooking. Visually, you can’t draw anything to the screen until you know where each and every object must be located. You can’t know where the objects are supposed to be placed until you’ve calculated the movement and physics that determine the motion paths. So, things must be done in a certain order to accomplish the task.
The faster you can get from start to finish, the faster you can send frames to the graphics card to be drawn to the screen. Sometimes, IPC allows you to tell two people things at the same time and speed up giving instructions, or maybe the head chef can tell an assistant what to do while he’s stirring the batter instead of focusing on one thing at a time himself. The faster the CPU clock speed is, generally the faster you can get from start to finish of the task list unless you’re waiting on something else to finish. The GPU drawing the frame would be like waiting on the waffle iron to heat up before you can cook the waffle.
When you load up task manager, you can see UTILIZATION of the entire CPU expressed as 0-100%. This includes every core of your computer and every thread if the cores can be multi-tasked into separate threads. If you right-click on the graph, you can then select “Change Graph To → Logical Processors” to see the individual thread utilization.
If you have a 4core 8thread system, with 50% UTILIZATION, that means that you are using 4x100% of 8x100% (100%+20%+80%+20%+80%+50%+50%+0% as an example). If you are running one of 8 threads at 100% and the rest are not being used, you’d see 12.5% or rounded to 13%. If the chef did everything himself without using the assistants and he was always busy doing stuff, but they were asleep waiting for the chef to tell them what to do, you’d have 33% UTILIZATION, 1/3 of 100%. On my system, FS rarely uses more than 1 core at 90-100%, 2 cores at 50-80% and 1 core at 20-80% regularly. I have a 4c8t system that usually sits about 50-60% utilization at best. On loading screens and when it loads scenery ahead of me, I get 100% all cores and 100% utilization for those brief moments, then it returns to normal. FS doesn’t care that you have 16 threads and I have 8 as it only uses 4 of them routinely and the clock speed matters most to get through the instructions to be carried out per frame.
Now, I’m hungry and Belgian waffles sounds pretty good, I’mma go to the kitchen. Hope this helps non-IT people get a sense of why AMD has been behind Intel in gaming and why the clock-speed increases AMD is proposing (along with IPC improvements again) might bring them up to par with Intel. If Intel can have several cores or even all cores at 5.5ghz boosted, it will still stay king for gaming as 4.9ghz AMD max boost (likely only one core boosted not all) even with IPC improvements won’t do much to remove Intel from being the best performance money can buy. Intel will have to get wicked in a hurry to not be dethroned, but I’m not sure AMD have done that yet, but they may have at least levelled the field if you want AMD in your system, you may not be losing anything to gain multi-threaded performance when you’re not gaming. Right now, the choice is multi-threaded loads, go AMD. Gaming and single-threaded performance, go Intel.
Also, of note, Hyperthreading (Intel name) or Simultaneous Multi-Threading (SMT) can take away some performance from a single core. On some systems (yours may or may not), disabling SMT will take away half of your threads, but it may boost the ability to process tasks more efficiently on a core or may generate less heat allowing less thermal throttling of the CPU (keeping the core clock speed higher longer). More modern CPUs don’t seem to be like this, but it has been known in the past that some games work better with SMT/HT off. Try it. If it doesn’t boost performance, turn it back on.
Just another data point that may or not have anything to do with this thread
Built pc just for this sim. Amd 2600x 16g 3200 ram, with old vid card. Stumbled my way thru getting the memory to 3200. Vid card then changed to 2070s. Windows now unstable. Long story short, dropped ram to 3000.
Once Cpu specs published, bought 2 more sticks of ram for 32gig.
Sim comes out, average 40 fps. A bit lower than some others.
Going over settings others are using, discover m-board set ram to 2133 with the 4 sticks. Spent hours on this. Solution was to bump speed up one step at a time. 2800 is the top stable speed I can get with 4 sticks.
I gained 5fps. Track IR so much smoother now.
G’day.
Tell me about your RAM.
I have found that RAM speed and configuration (XMP in BIOS) can effect performance by more than 30%
Also, AMD CPU’s are particularly dependent on a 1:1 speed ratio of the Infinity Fabric Clock.
Also, the CL speed of the RAM is critical with massive numbers of small memory reads and writes that are typical with a flightsim.
Perhaps you already know all this. I’m just curious and wanting to help if I can.
It would also help if you could load a known scenario eg: be sitting ready for takeoff at a particular airport so we can do some apples to apples comparisons to help you nail what’s going on.
Cheers
Same here…
They are not. If you look at HWINFO64, you can see the clock rates and states of the individual cores. This is true for AMD and Intel. For AMD CPUs ( I’m not sure about Intel), the bios can actually direct WIn to use the best cores.