Intel i9-14900KF or Ryzen 9 9950X3D for a new DCS VR rig? (RTX 4090)

[darkman222]

Since my PC died yesterday, I am thinking of switching to AMD coming from Intel.

I had the i9 13900KS before. Using VR ( Aero and now the MeganeX) together with a RTX 4090.

Depending on the DCS online server or complex single player missions I had severe stutters due to long CPU frame times 12 to 19 ms = CPU bottleneck.

As I need a replacement PC soon, I dont want to dig too deep and too long.

 

So the basic questions are:

1) If I compare the single thread rating here: https://www.cpubenchmark.net/compare/5684vs6549/Intel-i9-14900KF-vs-AMD-Ryzen-9-9950X3D both CPU seem to have pretty much the same performance. BUT the L3 cache is much higher for AMD with 128 MB (Intel only 36 MB) will this potentially give less stutters in VR? Also the clock speed is 1,1 Ghz faster on the AMD.

2) Is there any issues running a RTX 4090 (upgrading to RTX 5090 later) on an AMD system for VR (Aero and MeganeX) use?

Thanks in advance for the quick help!

 

[Markus77]

If gaming is the most important application: AMD Ryzen 7 9800x3d, 96GB DDR5 6000 with low (really low) latency and max. 2 (in words: two) memory modules and a really fast 4TB SSD (ideally PCIe5.0) for DCS and an additional fast 1TB SSD for the operating system

[diego999]

8 minutes ago, Markus77 said:

If gaming is the most important application: AMD Ryzen 7 9800x3d, 96GB DDR5 6000 with low (really low) latency and max. 2 (in words: two) memory modules and a really fast 4TB SSD (ideally PCIe5.0) for DCS and an additional fast 1TB SSD for the operating system

 

This. For gaming the difference in performance between 9800X3D and 9950X3D is minimal, while the price is not.

[Qcumber]

8 hours ago, darkman222 said:

Is there any issues running a RTX 4090 (upgrading to RTX 5090 later) on an AMD system for VR (Aero and MeganeX) use?

No issues. This should be fine. 

[darkman222]

Thanks guys. Would you say, with the AMD, I will see improvements in CPU frame time aka CPU bottleneck stutters compared to the Intel CPU?

Edited May 9 by darkman222

[Aapje]

We had forum members report a much improved situation with regard to stuttering, after going to the 9800X3D.

[lefuneste01]

Hello. I saw some people complaining on core parking issue with 9950x3d that are not existing with 9800x3d. I replaced my 12800K by a 9800x3d and got  great fps boost.

[darkman222]

Thanks for the opinions. So the money I will spend is not the main consideration. Improving stutters is. Seems like abandoning Intel and going AMD is the way.

From the youtube tests people do with the 9800x3d vs the 9950x3d the fps gain is minor. But the tests are done in 1080p resolution. Which tends to reduce the advantage of a fast CPU.
We also have the 9900x3d nobody seems to talk about for unkown reasons.

Edited May 9 by darkman222

[Aapje]

The 9950X3D has two compute chiplets, with 3D-cache on one of them. The game should entirely run on the chiplet with 3D-cache, because otherwise things actually get slower/more stutter, since sending data between the chiplets is very slow. That is why the 9800X3D and 9950X3D perform almost equally, even though the 9800X3D has only one compute-chiplet. For gaming, only one chiplet tends to matter. 

The 9900X3D has two chiplets of 6 cores, again with one chiplet having the 3D-cache. So when running games, that would run on the chiplet with 3D-cache with only 6 cores. If you get a 9800X3D, then you actually run on a 8-core chiplet, and 8 is more than 6.

So we don't talk about the 9900X3D because it's a CPU with a very limited use case: a person that runs both games, and productivity software that can take advantage of many cores and benefits little from the 3D-cache, but can't afford the 9950X3D which is good at both, so they get the 9900X3D, which is worse at both.

I strongly suspect that most buyers of the 9900X3D lack technical knowledge and buy the wrong product for their use case.

Edited May 10 by Aapje

[darkman222]

Thanks for the explanation about the 9900x3d. Makes sense why nobody talks about it.

I found this video of a spanish speaking youtuber. He is testing the 9950x3D and the 9800x3D in Vr in this video. Unfortunatey he seemed to have bottlenecked his test with the GPU. He has a lot of setting to high. I thought the RTX 5090 would handle it.

But he has a CPU frame time chart in the end, which shows that the 9950x3d gives some kind of improvement to the CPU frame time.

Here is the video at 6:00 min: 

Or see this screenshot:

[Raven (Elysian Angel)]

19 hours ago, darkman222 said:

But the tests are done in 1080p resolution. Which tends to reduce the advantage of a fast CPU.

Quite the opposite: CPU benchmarks are done in 1080p for very good reasons.

[darkman222]

Thanks for all the useful hints here. I went with Ryzen 9950X3D. The system is running now. And wow.

On servers where I had around 6ms CPU frame times, spiking up to 12ms even up to 19ms with the Intel -> On these servers I now have between 2,5 ms to 4 ms. No spikes at all, no more stutters.

I will try to give you guys a better comparison looking into my stutters video archieve. I cant compare directly any more as the Intel system is dead.

But its a huge improvement!

[Hootman9104]

In addition, I have found that no matter what hardware you plan on using, make sure and limit your FPS to something like 60fps, which I find comfortable.

The main issue remains CPU bottleneck, and to be honest (again, just my opinion/experience) I've found that GPU's cant' really be fully utilized at this time due to the CPU bottleneck. If the graphics were just tied to GPU we'd all be in dreamland, but they aren't. 

But your graphics settings are going to be the main thing you're concentrating on. Learn what each one does and how it affects the CPU usage and find your balance. So for me, limiting to 60 fps allows me to dial up those graphics a bit more, while taking it easier on the CPU. 

Then again, there are others who claim they just let their FPS run wild and things work great. Good for them, but I have never found that to be the case.

[Qcumber]

7 hours ago, Hootman9104 said:

I have found that no matter what hardware you plan on using, make sure and limit your FPS to something like 60fps, which I find comfortable.

That might depend on what headset you are using. If I limit the FPS to 60 I do not get a smooth experience. I find that it is better to aim for an FPS which matches refresh rate. 72Hz most of the time although in light missions I can use 90Hz. I avoid using reprojection. 

[darkman222]

Its the exact same configuration. i9 13900KS replaced by Ryzen 9950x3D together with a RTX 4090 and the Varjo Aero. DCS settings set to VR-preset.  90 Hz was no problem before either. But the CPU occupation was higher in general. Now I barely get any CPU frame time spikes. Its like night and day.

From what I read about the differences of the two CPUs it could be that the huge L3 cache of the Ryzen makes a big difference in my setup.

[Raven (Elysian Angel)]

9 minutes ago, darkman222 said:

From what I read about the differences of the two CPUs it could be that the huge L3 cache of the Ryzen makes a big difference in my setup.

Based on my understanding, games in general and flight simulators specifically have a lot of tiny data packages being sent back and forth between the CPU and the memory, very repetitively.

Thanks to the V-cache of X3D CPUs, those data packages can stay in the cache of the CPU itself instead of being sent in and out all the time. This causes a massive reduction in latency, so even in scenarios where the average frame rates are not that different compared to other CPUs, the 1% lows will see a huge improvement with X3D.

[Dawgboy]

14 hours ago, darkman222 said:

Thanks for all the useful hints here. I went with Ryzen 9950X3D. The system is running now. And wow.

On servers where I had around 6ms CPU frame times, spiking up to 12ms even up to 19ms with the Intel -> On these servers I now have between 2,5 ms to 4 ms. No spikes at all, no more stutters.

I will try to give you guys a better comparison looking into my stutters video archieve. I cant compare directly any more as the Intel system is dead.

But its a huge improvement!

I, too, went with the 9950X3D, and it's smoking.  Glad to hear your results are similar.  I added the 5090 to ensure I was adequately future-proofed, and now I'm just waiting for my Pimax Super.  Very excited about the prospects because my old system struggled with DCS.

[Hootman9104]

12 hours ago, Qcumber said:

That might depend on what headset you are using. If I limit the FPS to 60 I do not get a smooth experience. I find that it is better to aim for an FPS which matches refresh rate. 72Hz most of the time although in light missions I can use 90Hz. I avoid using reprojection. 

You're correct. I really missed a bit here, with the conversation around VR. I was speaking to just flat screen play

VR wise, I recently got the Quest 3 (was using Reverb G2), and find that the base 72Hz works best, and using opencomposite, XRTookit to up resolution to 3100, and dialing the graphics settings down a bit, especially the scenery sliders. 

[C3PO]

5 hours ago, Dawgboy said:

I, too, went with the 9950X3D, and it's smoking.  Glad to hear your results are similar.  I added the 5090 to ensure I was adequately future-proofed, and now I'm just waiting for my Pimax Super.  Very excited about the prospects because my old system struggled with DCS.

What was your system before? Your new one will last years!

[Qcumber]

1 hour ago, Hootman9104 said:

You're correct. I really missed a bit here, with the conversation around VR. I was speaking to just flat screen play

VR wise, I recently got the Quest 3 (was using Reverb G2), and find that the base 72Hz works best, and using opencomposite, XRTookit to up resolution to 3100, and dialing the graphics settings down a bit, especially the scenery sliders. 

Have you tried using Quad Views Foveated Rendering? 

[Hootman9104]

1 hour ago, Qcumber said:

Have you tried using Quad Views Foveated Rendering? 

That's next on the "to try" list

[SharpeXB]

23 hours ago, Hootman9104 said:

In addition, I have found that no matter what hardware you plan on using, make sure and limit your FPS to something like 60fps

I’m not sure of a good reason to do that. You would just be limiting an otherwise strong system for no purpose. There’s nothing wrong with 60 FPS but if that’s all you want you could save the money on your components. Otherwise you aren’t getting what you paid for. If your display is capable of a higher refresh rate then use if. 

23 hours ago, Hootman9104 said:

The main issue remains CPU bottleneck, and to be honest (again, just my opinion/experience) I've found that GPU's cant' really be fully utilized at this time due to the CPU bottleneck. If the graphics were just tied to GPU we'd all be in dreamland, but they aren't. 

That’s not always the case. For example I am GPU limited in DCS with the specs below, running 4K at 120 Hz. It depends on the settings. Some are CPU dependent and some are not. 

[Hootman9104]

31 minutes ago, SharpeXB said:

I’m not sure of a good reason to do that. You would just be limiting an otherwise strong system for no purpose. There’s nothing wrong with 60 FPS but if that’s all you want you could save the money on your components. Otherwise you aren’t getting what you paid for. If your display is capable of a higher refresh rate then use if. 

That’s not always the case. For example I am GPU limited in DCS with the specs below, running 4K at 120 Hz. It depends on the settings. Some are CPU dependent and some are not. 

Interesting. My system is similar to yours and actually more in some aspects. I'd be interested in your ingame settings to get GPU bottlenecked vs CPU bottlenecked. I'd certainly like to dial up my graphics and utilize more of my 4080 Super

[SharpeXB]

15 minutes ago, Hootman9104 said:

Interesting. My system is similar to yours and actually more in some aspects. I'd be interested in your ingame settings to get GPU bottlenecked vs CPU bottlenecked. I'd certainly like to dial up my graphics and utilize more of my 4080 Super

I’m just running all the settings at the highest. At 2160x3840 and 120Hz. That will max a 4090

[Qcumber]

54 minutes ago, SharpeXB said:

That’s not always the case. For example I am GPU limited in DCS with the specs below, running 4K at 120 Hz. It depends on the settings. Some are CPU dependent and some are not.

To expand on this CPU or GPU limited is just a binary toggle in DCS. The reality is much more subtle. It's like 50 shades of grey. It's better to look at the CPU and GPU latencies and variance of overall frame time.