Ryzen 5 5600X + RTX 3080 Benchmark Videos

True. They're apparently enabling it in an upcoming driver soon and has it working on both Intel and AMD platforms. So it could possibly give a small perf boost for Ampere as well. In Valhalla it gave the 6800XT a 15% uplift in 4K so that's nothing to sniff at. It will depend on the title so I'm wondering what it will do for ACC. Maybe nothing, maybe something. Either way it's good that both AMD and nvidia will support it.

 

It's probably not that easy as it will also require a bios update. So don't expect it to work on chipsets and cpus that are eol.

 

3090 FE has been here since a couple days after launch. Reverb G2 arrived on the first day they showed up in the US. But still can’t find a damn 5900X to save my life...

When I do, I’ll get some benches up for the folks in here that have said they have interest.


Sent from my iPhone using Tapatalk

 

So i've had my 3090 for a while now and finally got my 5800X paired with it yesterday after much BIOS reflashing headache (but we got there)

Straight away, microstutters are way less in ACC VR, and I even put heavy rain, 12-15 visible opponents and 25 AI and mostly epic settings apart from volumetric fog and 50% car LOD. Track was Imola

In the Rift S with 55% render resolution and 2.0SS in Oculus Tray Tool, it managed to keep the above locked at mostly 80 FPS i believe but need to confirm.

Is there a way to Benchmark specifically in VR? Going to try and get some results posted or to compile in a YouTube vid for you guys.

Overall i'm really happy with ACC now, can't wait till my Reverb G2 arrives after the weekend. One last question, my Rift S settings above, would that equate to running same resolution on the Reverb G2?

 

Not sure if this is a vr only thread or not but just wanted to share that on my 2560 X 1080 monitor with resolution set to 200, all ultra settings, all cars visible, mirrors at highest settings and running 35 AI at Bathurst I was sitting in my capped limiter of 72 all race. Only dropped a bit in the first part of the replay where it always does. This is on a 3090.

 

Are you aiming for best visuals or best performance?

I would have Hoped you'd be able to hit a solid 90fps with that PC. I have a 3080 on order (probably won't see it this year) and I've been flirting with the idea of getting a 3090 instead (so I will see it this year). I don't know that there's that much of a jump in graphics between high and epic, is it worth the performance hit?.

At the moment I can get a pretty stable 90fps on the rift CV with medium(ish) settings. When I was setting up graphics the only concern I had was getting the rain effects.. I aimed for 90fps other than that. That's with a i7 8086K and a 2070 super.
I have a G2 on the way too (also won't see it this year) wasn't expecting to be able to set everything to epic, with the added resolution I'd be happy to keep the settings I have.

 

It may indeed hit a solid 90

I'm pretty sure it was capped at 80 as this is the refresh rate of the Rift S.

I'd say it was pretty solid 99% of the time with the tiniest microstutters on maybe the 1% Lows in the FPS which I'd say is always inevitably lower than the Average FPS.

 

Hadn't thought of the rift refresh rate.

 

The way I benchmarked with my Rift was by recording the frame rate with MSI Afterburner. You'll need to disable ASW in the Oculus Debug Tool to be able to see frame rates below your headset's nominal value. If you keep ASW enabled, retroprojection will kick in as soon as the nominal value cannot be reached and just target for half of it. But in any case, you will never be able to see frame rates higher than your headset's nominal value. If you want to check the headroom left, the easiest way is too look at the remaining GPU usage when the nominal frame rate is reached.
You can also have some precious information in the Oculus Debug Tool itself (check for Performance in the menu). There, I like to look at : FPS, number of frames lost, CPU render time and GPU render time. But I'm not sure how to interpret them yet. For example, render times seem they should be added together and never exceed 11.1ms to sustain 90fps. But they go a little higher for me even though I seem to have a constant 90 fps. I guess the CPU and GPU work simultaneously so it's not as simple as just adding these two values.
Also, Oculus Debug Tool statistics are updated with every new frame while MSI afterburner only updates every second by default.

As for your question about finding what supersampling value could equate to a G2, here's how I view it:
- The Oculus Rift's screen has a resolution of 2160x1200, which makes 233M pixels per second at 90 Hz. But all VR headset need to render more pixels than their actual screen to perform lens correction. The CV1 has a factor of 1.33 in each dimension (I saw that in a YT video and noted the information, I don't remember the original source). Which then makes 233*1.33*1.33 = 412M pixels per second (here's an official blog post that has about the same result : https://www.oculus.com/blog/powering-the-rift/)
- If the Rift S has the same factor, it needs 2560x1440x80x1.33x1.33 = 522M pixels per second
- WMR headsets have a lower coefficient of 1.1 (they need less power for the same amount of pixels but have a lower quality of lens correction). This gives 4320*2160*90*1.1*1.1 = 1016M pixels per second!

From what I researched, the HP Reverb G2 would then require 2.5x more power than the Rift and 1.95x more power than the Rift S at 100% SS. Since SS applies identically in both directions (SS at 200% would actually quadruple the number of pixels), you need to apply sqrt(2.5)=158% of SS on a Rift and sqrt(1.95)=140% of SS on a Rift S to have the equivalent of 100% on a Reverb G2.

PS : Also, you don't need to use Oculus Debug Tool for supersampling. There is a setting directly in ACC : VR Pixel Density.

PS2 : I worked on it all day, my video is coming soon.

 

According to MRTV at 9:50 the difference in Performance between G2 and CV1 is about 1.7

 

Hello everyone,

I finally finished my very first video. I'm not sure what to think about the end result though, your feedback will be well appreciated in PM or in the comments

 

Wow, moving from the Ryzen 5 2600 to the 5600X really changed everything. Hmm I have a ryzen 7 3700x and I am wondering if the ACC performance gets a significant boost when going from the 3700x to a 5600x. Looking at this other video, it seems so:
(11) AMD Ryzen 5600x vs 3700x in Sim Racing & VR (Wow! What a difference!) - YouTube

 

does it make sense to go for 5800x?
I play triples

 

No.

Based on Dan Suzuki's benchmarks the gap between the 5950X and 5600X at 3x 1440p (with a 3090) is 2 fps. It was 80.7 vs 78.7 respectively.
A later benchmark shows 90.2 fps (with a 3090) at 3x 1080p. The gap to a 5600X would increase a bit there, but I'm guessing the gap would not be more than maybe 5-10 fps. And the 5800X would fall somewhere in between that. So you would basically pay about $150 more for a few fps more.

Btw, in the same 3x 1080p benchmark the 3080 gets 86.3 fps. Which means that for just $800 more (3090) you get an astounding, phenomenal 4 (FOUR) fps more. Totally worth it!!!

 

this means that the CPU is already a bottleneck :-(

 

Keep in mind that this was with a 5950X, the fastest CPU available at this time.
I summarized these results in an Excel chart so we can see more clearly what's going on.
3090 is green
3080 is blue

(Btw, 1080p single screen results are not shown here, but he said in the video that they were identical to the 1440p single screen results)

 

It isn't, 3x 1080p is equivalent to 75% of the pixels of 4K. 8 229 400 vs 6 220 800 ≈ 75%. The 3090 is just terrible value.

 

Well, there's clearly a CPU bottleneck at 1440p as evidenced by the results in the video I mentioned above.
But dropping the resolution from 3x 1080p to 1440p single (with the 3080) gives a performance boost of 38%, so I can't see that there's much CPU bottlenecking going on at 3x 1080p.

This chart gives a clearer overview:

 

These are my results in VR

https://www.assettocorsa.net/forum/...hardware-for-acc-vr.67447/page-3#post-1175521

 

Never thought to turn SMT Off. Very interesting findings. I may have to do that too to grab a couple of frames!