We don’t care about ray tracing, DLSS or any other modern GPU feature, right?

[plane_crazy242]

I’m actually not trying to be an a$$, but just want to be sure I understand. 

For DCS ( in particular), all I should care about from a HW point of view is single core clock speed and GPU rasterazation performance, right?

DCS (and most games) aren’t taking advantage of DLSS or FSR or RT cores or Ray Tracing, correct?

If I can get the best single core speed, best rastorizing GPU, decent amount of memory, that’s about the best I can hope for . . .

I was thinking of going the 4090 route, but it’s got a lot of silicon I would just “waste”. If RDNA 3 is best in raster performance, what do I care if it sucks at ray traceing?

I’m VR 100% and also do IL2 and MFFS

Am I on the right track here or missing something ?

Ryzen 7 3700X, 64 GB memory, 3080 10gb, Reverb G2
 

TIA

Edited October 16, 2022 by plane_crazy242

[Supmua]

For the current DCS you mostly need rasterization power from GPU (brute force if you will) as DLSS or RT isn't supported, you can use upscaling tech such as FSR or NIS though.

In general, fast single core CPU, fast GPU, fast memory (can also make a big diff).  Knowing how to fine-tune the settings can help a lot too.

MSFS supports DLSS so it's a bit different, don't know much about IL-2 but it uses older engine so likely similar to DCS in terms of requirement.

[Voyager]

MSFS is also expected to support FSR2.0 come SU11. I'd expect it to be more likely for Il-2/DCS to end up supporting FSR2.0 than DLSS as well, since it does not require nVidia training on the game. So I'd only consider DLSS a factor is you want to play other genres. 

CPU single thread performance is a bit weird. As near as I can tell, CPU cache size and memory access speed/latency seems to be more important than pure single thread performance in Dx11 VR. Not as sure about Dx12 or Vulkan, but that's not today. 

As such, the 5800X3D seems to still be the king of VR, even against Zen 4 and what we're expecting to see from Raptor Lake, but the Zen 4 VCache chips should probably roflstomp the 5800X3D when they come out too. I sort of think the 5800X3D is a 60fps CPU, and the 7800X3D will probably be a 75 fps+ CPU. 

[Slammin]

OP said:

"For DCS (in particular), all I should care about from a HW point of view is single core clock speed and GPU rasterization performance, right?"

I say correct, in the context of your other statements.  Absolutely correct!

 

Then OP said:

"If I can get the best single core speed, best rasterizing GPU, decent amount of memory, that’s about the best I can hope for . . . "

And then things changed.  3D cache via AMD's implementation on the CPU kinda moves the goal posts already as far as current tech goes, so, as folks have responded; although it is basically as simple as the OP stated, future-looking is still the best approach IMHO.

But OP, you are correct, today, but I also detect leanings of a Red vs Green discussion in the making

Hope not, lol!

[Voyager]

@Slammin Well, that is going to have to wait until we see what AMD actually releases and how far AIBs are willing to push it first

i will say, I am *very* interested to see how AMD's second generation cache tech compares to nVidia's first gen. From what I gather, AMD's first gen cache gave great results up to 1440p, but did not hit often enough for 4k+, so frame times weren't as consistent.

Sounds like RDNA3 has done a lot of work on that cache, so it's supposed to be much better, with possibly twice the hit rate of the RDNA2 cache. And, if they decide to, they could stack it for 192mb of cache, but currently sounds like they don't think it's worth it. 

Given Lovelace is the first nVidia generation with GPU cache, it will be very interesting to see how they compare. And it will be very interesting to see how it all behaves in VR.