My journey from AM4 to AM5

I’m seeing the truth in that statement.

I’m also hoping that FS2024 will make the 5800X3D somewhat underwhelming, as the new code unleashes the power of CPU’s like mine, with their additional cores running at higher base/boost frequencies.

If it doesn’t, then I will admit that I just wasted a bunch of money. My 5800X3D system was a really good performer. I want eye-watering greatness from this one. It’s not there in FS2020.

I suppose it is because I turned it off sometime during the last 2 months of tweaking all of the settings and simply forgot to turn it back on. Here’s a screenshot with it turned on.

Curious as to how you even knew that.

1 Like

Perhaps @TenPatrol noticed that the bridges over the Hudson river had no “black shadows” under them? :slightly_smiling_face:

2 Likes

I’ve been watching Suits on Netflix so I’m basically an expert on how Manhattan should look. If there’s no fancy law firm with a view of the skyline, something’s off!

But seriously, the buildings in Manhattan with photogrammetry enabled look different and are darker than the default ones. In the screenshot I don’t see well-known issue with the bridges that @SmotheryVase665 mentioned.

3 Likes

:rofl:

Hey @TenPatrol, thanks for your response. I’m surprised that I get very similar frame rates with photogrammetry turned on and off. I would have thought that they would have dropped fairly significantly with photogrammetry turned on. Weird.

You’re an expert in how Toronto looks… Only the basic city shots are NYC, out the windows and and outdoor character scenes are Toronto :sunglasses:

1 Like

Haha, well, I guess that means I’m not only an expert on MSFS and Manhattan but also on the “Toronto version” of it! If only I could load up MSFS and fly over the real Manhattan - no green screens required!

Well, I think I’ve dialed things in as well as possible.
Note that this is with a clean Community folder. I expect performance to be affected as I add things back in. But I’m very happy with this baseline.

Bonanza G36 Improvement Project
Flying into and over NYC at 1500 AGL

Xbox Game Bar disabled, Windows 11 Game Mode disabled, HAGS disabled.
HPET disabled, replaced by ISLC
Riva Statistics Server running - no framerate cap.
Rbar and ‘Above 4G’ enabled in BIOS.
Hyperthreading and Virtualization both disabled in BIOS.
nVidia Profile Inspector setup per this video:
https://www.youtube.com/watch?v=E_FRjo7aGgE
DLSS DLL - most current version using DLSS Swapper.

  • A lot of per core undervolt tuning using CoreCycler, adjusting for max negative offset to achieve max boost frequency and remain stable for 12 hours.

  • A lot of tuning in Process Lasso, mostly following advice in this video.
    https://www.youtube.com/watch?v=p9xu8u_0Xpk

I’ve gone back and forth on whether to have Lasso set the Power Plan to Bitsum Maximum Performance or Ultimate Performance on startup. For a while I thought the former was better, but I’ve settled back on Ultimate.

Here’s what I started with. All flights throughout this journey were captured and analyzed using CapFrameX. I highly recommend this tool.

Here’s what I’ve ended up with:

It’s worth noting that I’ve discovered what seems to be a pretty repeatable bug. When I start the flight over NYC, I see around 40 FPS and 30ms latency. If I immediately hit ESC, then Restart, the flight starts up with around 65 FPS and 15ms latency, and I see the results in the second CapFrameX analysis.

Here are my General Settings:



And finally, my nVidia 3D Settings.

Knowing myself, I’ll probably do a little more tweaking, but I think the heavy lifting is complete. I’m quite happy hitting 65 FPS with 16ms of stutter-free flying over NYC.

BTW, temps during these flights are great. The air-cooled CPU is around 60-65°C, and the liquid-cooled GPU is around 70°C, both well within safe limits.

My UPS reports 770W power draw during these flights. Probably no problem for a 1000W P/S, but I think I make the right choice getting the 1200W unit.

I hope this journal has been of some help to anyone who chooses the 7950X3D. The main thing I’ve come away with is that Process Lasso is an absolute must. It’s hard work tuning it, but the results speak for themselves.

I’d also like to thank each and every one of you who has offered your expert advice throughout the process. It is greatly appreciated, and as always…

Happy Flying!

8 Likes

What is the point of running RTSS if you’re not using the cap?

I definitely need to look into Process Lasso more. I don’t think I’ve set mine up in anywhere near as clever way as you seem to have done. I’m just using the free version though so I think some of those features are not available to me (like “Sets”).

My understanding is that it collates data for other apps, like MSI Afterburner.
It’s probably a very minor thing. But I’ve always used it, so why not?

ETA: I admit I never did a test to see whether turning it on or off changed my performance. I’ll do that and report back.

Also, if I’m seeing 70 FPS over NYC in the Bonanza, I may get 70 FPS in less demanding areas with more demanding planes (which would be awesome) in which case I’d use RTSS to cap frames @ 60 FPS. We’ll see.

I’m pretty sure the free version of Lasso has all the features of the paid version. I just don’t have to deal with the annoying 15 second wait when I launch it.

Also, I’m a big believer in financially supporting developers that add value to my sim experience. LittleNavMap and Addons_Linker are two other apps I’ve donated money to. I also made a donation to Savitarax, who did the Lasso video.

New info dump.

@Baracus250 got me thinking about RTSS. I’ve always used it to cap my framerate @ 60 FPS. Of course, in most cases I’m not hitting 60 FPS anymore. Demanding scenery, airports, aircraft - all conspire to stress our systems pretty hard. He pointed out that not using it to cap framerate wasn’t doing anything for sim performance.

So I decided to do a couple of tests. The results are interesting, to say the least.

These are results with RTSS on, with a 60 FPS cap. Note that Main Thread latency is way down, but overall system latency is higher. The GPU is definitely a hard bottleneck in this case. Also, look at the 1% and 0.1% Lows. Important metric, I think.

FPS_Bonanza_Sahara_XMP5_FCLK 2133_RTSS Cap 60FPS

Now let’s look at it with RTSS turned OFF. Main Thread latency is up, but overall latency is down, and Main Thread/GPU is nicely balanced. 1% and 0.1% Lows are much better.

Look at the ‘Sys Alloc / fr’. I don’t know what it is, but it’s WAY better (9 vs. 25) with no framerate cap.

FPS_Bonanza_Sahara_XMP5_FCLK 2133_RTSS OFF

I should also mention that I got some advice from the helpful folks over at overclock.net who looked at my RAM timings and made suggestions to tighten things up a bit. It may be confirmation bias, but Windows seems ‘snappier’ since I did that. But you can see that memory benchmarks have improved overall.

XMP-1 SPD timings, FCLK = 2000

XMP-1 6400_CL32_Normal Latency

Custom SPD timings, FCLK = 2133. (I tried FCLK = 2200, but the computer wouldn’t POST. I had to reset the CMOS and reload the BIOS settings. I was informed (after I reported the issue) that ~90% of Ryzen 9 CPU’s can run fine at 2133 MHz, but only ~50% will even POST at 2200 MHz. Mine’s in the bottom half. :stuck_out_tongue_winking_eye:

Of course I had some concerns that messing with RAM timings and Infinity Fabric frequency (FCLK) might lead to instability, but the system so far is incredibly stable. And temps are as good as I could wish for.

The Aida64 stress test loads all CPU cores @ 100%, and loads RAM at 100% as well. This one ran for an hour. RAM temps (40 and 44°C) are very good, as is CPU temp of 70°C (thanks to the PBO negative offsets.)

I’m almost done… LOL.

3 Likes

You forgot to mention the exact model of your memory.
Please also include a screenshot from the ZenTimings.

edit

You added a link to the homepage but could you please provide a direct link to your topic? The suggestions regarding PC memory settings might help someone in the future.

1 Like

Excellent news!! :partying_face: :+1:

I know where you are coming from as I updated my bios the other day and started tweaking my memory and wow what a difference the new bios made. I believe what we are seeing are the collateral benefits of AMD tweaking their AGESA to improve performance of their 9000 series cpu’s that landed with a thud when introduced a few weeks ago. Note: AGESA is the “mircro code” AMD provides to all of the motherboard manufacturers that get incorporated into their bios.

I was able to OC my DDR5-7200 memory to 7600 MT/s with my FCLK at 2100. I didn’t know about hitting 2200 was ~50% so I’ll have to give that a try - thanx for that! :+1:

Enabling PBO and undervolting the cores also got me a ~3% bump in performance. Enabling PBO on my old bios with the same undervolting really didn’t get me much.

I also got my best score ever on CineBench 23. AIDA64 got me results very similar to yours:
Memory Read: 89643 MB/s
Memory Write: 95358 MB/s
Memory Copy: 83434 MB/s
Memory Latency: 61.0 ns

My results with the sim over NYC got me render times similar to yours:


ZenTimings_7600-2100Screenshot

2 Likes

Double thumbs up WingWarper!

1 Like

I just did a NYC flight, and I can’t be unhappy with this, especially considering how disappointed I was when I began this journey (before I really realized how much work it really was to get this 7950X3D working well with the sim.)

All in the green and nicely balanced! :+1: :+1:

Btw - you look much closer to the ground than my screenshot above…

1 Like

I’m very tempted to upgrade my RAM. I saw a nice boost just tweaking the timings of my 6400 MHz DIMMs. My motherboard supposedly supports up to 8000 MHz. Maybe I’ll grab some QVL 7600 MHz sticks on Black Friday.

Good thoughts on the BIOS improvements, BTW.

LOL. I had to divert around the tallest building. :smile:

1 Like

My point was really that I did not know what RTSS was doing for you if you run it even when you are not using the frame cap feature. You wrote before that you DON’T use the cap, but here in this post you say you always used it?

I don’t use MSI Afterburner, so my only reason I started using RTSS was to cap the frames to half my Quest 3’s target of 72Hz.

As I wrote in the other topic (I think lol) I only discovered that other forms of frame limit (Nvidia Control Panel or the Virtual Desktop SpaceWarp thing) were not really capping the frames properly when I installed this VR Tablet that has it’s own FPS counter in the cockpit. https://flightsimulator.me/

I doubt you would want to invest in this “just to see” as you don’t use VR (…yet… muahahaha) but it WOULD be super interesting to see if you have the same result as I did in terms of discovering that the frame rate was all over the place.

It was after this, that I looked at installing RTSS and it really worked for me. Maybe it’s just a VR only thing?

I will try and do a video of this tablet readout with RTSS not running to show you what I mean.

Basically, for 2D OR VR I really don’t see the point in going above 60fps. It is more critical in VR I suppose, as the frame generation really needs exactly half all the time to work properly.

But either pancake or VR, by going over a capped frame, you are only putting more stress on the whole system, for no gain, only pain. It has to draw and calculate EVERYTHING that many more times per second, and it hits a wall that it cannot jump over at some point when everything in the scene is against you, even if it’s just for a moment, which breaks fluidity.

You seriously cannot see any difference above 60fps in a flight sim anyway IMO.

But your point about the Sys Allocs is interesting. I also noticed this goes into the red for me last time I had Dev mode panel open. I don’t know what it means, but I see it fluctuate a LOT depending on where you are looking. It seems heavily linked to the quantity of objects in view (be that buildings, trees, other aircraft, individual chunks of photogrammetry etc) so maybe the difference you are seeing between the 2 screenshots above is just for the moment you took the actual screenshot, there weas less going on. Or is it consistently lower?

But it comers back to I cannot see HOW it can be a benefit to latency on any front by forcing the CPU to output more frames. It’s weird… you would think by capping within a reasonable limit you know your PC can handle, that should give MORE headroom always. Perhaps it’s just a symptom of either the GPU or CPU having SPARE cycles and waiting? If so that is not really a bad thing, if it stops the whole system from going over a knife-edge when things in the sim get heavy?

I have been suffering from quite a lot of SSS recently though (LOSS: The Lexicon Of Sim Stutters) … can’t iron it out :cry:
I think it’s just the sim, or that plus some Mamu mods (power lines).