Doom Eternal Graphics, CPU Testing: id Shows How to Optimize for Performance

[Admin]

We've tested Doom Eternal on 20 GPUs and 10 CPUs to find out what sort of PC it requires to run it at 60 fps or more.

Doom Eternal Graphics, CPU Testing: id Shows How to Optimize for Performance : Read more

 

[dbrees]

So it seems that for 1440p/144hz gaming the best combo is R5 3600, 5700 XT. Pretty solid.

 

[SteveRNG]

This is the kind of data the CPU "elites" need to look at. My Ryzen 5 3600 has plenty enough power to play at 1440 and 4K. I just need to spend my money on a good GPU.

 

[JarredWaltonGPU]

This is the kind of data the CPU "elites" need to look at. My Ryzen 5 3600 has plenty enough power to play at 1440 and 4K. I just need to spend my money on a good GPU.

How much CPU you need does vary by game, as well as what GPU you're using. Doom Eternal is incredibly light on CPU needs compared to something like Assassin's Creed or Red Dead Redemption 2 for example. But at 1440p and especially 4K, GPU is by far the most important aspect. If you're on Ryzen 5/7/9 3000 or Core i7/i9 8000 or better, there's no game that won't run well enough and it will probably be years before we reach the point where something faster would be required.

 

[dalauder]

I think we missed a chance to talk about the PCIe 4.0 advantage. Notice that the Ryzen chips with PCIe 4.0 16x all perform at the top. It was mentioned that the 3400G with PCIe 3.0 8x only was slower, but was that the reason the Intel chips were slower as well?

If that's the case, then NO Intel chips can be listed at the top on gaming performance CPU lists anymore because you'll buy an RTX 2080 and get bottlenecked by PCIe bandwidth in high-resolution, high-spec situations.

Clearly this needs extensive testing...

 

[JarredWaltonGPU]

I think we missed a chance to talk about the PCIe 4.0 advantage. Notice that the Ryzen chips with PCIe 4.0 16x all perform at the top. It was mentioned that the 3400G with PCIe 3.0 8x only was slower, but was that the reason the Intel chips were slower as well?

If that's the case, then NO Intel chips can be listed at the top on gaming performance CPU lists anymore because you'll buy an RTX 2080 and get bottlenecked by PCIe bandwidth in high-resolution, high-spec situations.

Clearly this needs extensive testing...

Lower resolutions with high fps are certainly a use case where PCIe 4.0 could be beneficial ... except the RTX 2080 Ti isn't a PCIe 4.0 part. So the CPU testing has Ryzen in the lead for some other reason. I suppose I could have tested RX 5700 XT instead of the RTX 2080 Ti, but even at 1080p low that doesn't match the FPS of the RTX 2080 Ti at 1080p ultra.

 

[dalauder]

You need to share these results with Paul Alcorn: https://www.tomshardware.com/reviews/best-cpus,3986.html#xenforo-comments-3398759
His top two suggestions aren't consistently tops at gaming.

EDIT: Or maybe they're tops everywhere else? I'm probably wrong thinking that DOOM is representative of upcoming titles because it's optimized very differently from how many other games will be.

 

[dalauder]

Lower resolutions with high fps are certainly a use case where PCIe 4.0 could be beneficial ... except the RTX 2080 Ti isn't a PCIe 4.0 part. So the CPU testing has Ryzen in the lead for some other reason. I suppose I could have tested RX 5700 XT instead of the RTX 2080 Ti, but even at 1080p low that doesn't match the FPS of the RTX 2080 Ti at 1080p ultra.

Oh yeah, I forgot the 2080 is PCIe 3.0. I wonder what that "other reason" could be, especially when the 3950X was slower.

 

[dmorisette]

Interesting that at 1440p when the Intel chips pop up on top they are led by the I5, 6 core with no hyper-threading. Not the first time I have seen this. Usually the non ht part is the 8 core 9700k. Does this happen with the AMD parts also if you turn off smt? Better HT may be desired?

 

[JarredWaltonGPU]

Interesting that at 1440p when the Intel chips pop up on top they are led by the I5, 6 core with no hyper-threading. Not the first time I have seen this. Usually the non ht part is the 8 core 9700k. Does this happen with the AMD parts also if you turn off smt? Better HT may be desired?

Yeah, as you reach the point where a game becomes almost fully GPU limited, Hyper-Threading and SMT tend to reduce performance. Given the Ryzen 9 3950X underperforms compared to the other AMD chips, I'm sure running without SMT would help it a lot -- it should beat every other AMD chip. It's a bit weird that Doom Eternal struggles with scaling to 16-core/32-thread but doesn't appear to have any issues with 12-core/24-thread scaling, though perhaps there are other factors affecting performance. I'm also still curious about why the AMD chips did better at lower settings -- in real-world IPC, Zen 2 is pretty close to Coffee Lake, and CFL has higher clocks. More cores doesn't appear to matter beyond about six, so what's helping AMD out?

Intel's 9700K is probably technically Intel's fastest "stock" chip for Doom Eternal right now (even though I didn't test it), given it has higher clockspeeds and two extra cores compared to the 9600K. But even overclocking likely doesn't matter too much, considering I tested 9900K (4.7GHz), 8700K (4.3GHz), 9600K (4.3GHz I think?) and 9100 (4.0GHz). Only the 4-core/4-thread chip is even moderately behind the others.

 

[dalauder]

Great discussion feedback. Thanks.