The ACC GPU Benchmark Thread

In this other thread (don't know myself which one), someone said that RX5700XT beats RTX2060 Super in several UE4 titles. I assumed that it will be the same in ACC. So I looked up the 2060 result in this thread (88 FPS AVG) and added ~ 10 % - because the 2060 Super is some % faster and the RX5700XT is faster than the 2060 Super. So I assumed an AVG of 97 FPS for the RX5700XT. Comparing this to the fastest Vega 64 in our database - mine - that means

97/71 = 1.37 --> 37 % improvement of the RX5700 XT vs. the Vega 64.

As you can see, this is a result based on word-of-mouth and speculation. We need to wait for real results to see if it holds true.

 

I'm even afraid of switching to 1440)) So I'm sitting 25 centimeters away from the monitor. This is a feature of vision.

So I'll just be jealous of you

 

Yes real results will show but i think is a bit optimistic
This is from hardware unboxed 5070xt is on the level of 2060 Super and 15 % faster than vega 64
Which is not bad actually

Btw thanks about the thread is really helpfull bcs is very hard to find benchmarks for ACC with real optimized settings.

 

Thanks for posting those results. If you compare the Vega 64 to the 1070, there's a difference to the ACC results - in ACC, almost all 1070s are already better than the Vega 64s. So it appears as is ACC is extra bad for AMD :-( In Fortnite, Vega 64 beats the 1070 and is just so beaten by the 1070Ti.

You're welcome, btw.

 

My card is even slower (in AC) than yours. But I'll manage somehow.
Even if I only manage 40-50 fps in AC I'm fine, I've done much worse.

I've played GTA I on a 486 with about 15 fps. And I played it a lot. So I'm fine either way.
And last week did a race against 20 AI where my PC really freaket out. Was about 5-10 fps at times. Really had to guess where I'm driving in the next 1-2s.

 

Yeah, I know what you mean, but I'd like 60 FPS ))

I'll even say it like this - I want it the way it is now - ultra settings and 60 fps. All my life I was satisfied with 30-40 and minimum settings, so I was even tired.

Wouldn't want to lose what's on 1920x1080 right now.

 

So Fortnite is a best case scenario for unreal engine where AMD have actually done some driver optimization. It will be a shame if that good gpu underperform again in simracing.
I was interested in Radeon Image Sharpening which i think it will useful in the blury ACC but it runs only on DX12 and ACC is DX11

 

I saw a review where that was briefly addressed, it's rumored to later work on DX11 as well, but I'm not sure if that will really happen. It sounded like AMD didn't have the time to implement it yet to work with DX11.

 

More details if someone is interested

 

Check out the Generic Discussion thread - someone claimed that he got some new sharpening from AMD to work via ReShade.

And well...by now I also think those 37 % are not happening. To avoid disappointment, it's better to just assume a 15 % increase --> ~82 AVG FPS for a good 5700XT.

 

@LeDude83 did you figure out the settings I should use for the last release to have the same settings as the one in the OP?

 

No. Please give me a few more days - I will sort it out.

 

Yes i expect something like this. around 1070 ti level. Its so strange 5700xt is such a beast everywhere close to 2070 Super and in ACC 45% diffrence.
If i was not with 144hz monitor i will go for 5700xt ,save some cash and support Amd but now is hard decision

 

Ok, I'm going to change the rules a bit. Panky never replied to my questions on the settings so I'd have to speculate...which makes no sense. Also, manually adjusting the numerous GFX settings is a lot of effort and might scare people away. So let's make a clean cut and use the recently introduced presets for the benchmark.

So it goes like this:

- Session type/settings and test procedure as described in post 1
- choose GFX preset Epic and change res to 1920*1080
- In view settings (when in the car), reset to default

This should be a lot easier and faster than the previous approach. Post 1 will be updated, soon.

Here's my result with the new procedure:

2019-07-18 21:18:28 - AC2-Win64-Shipping
Frames: 4078 - Time: 60000ms - Avg: 67.967 - Min: 50 - Max: 102

The result with the new procedure is even in the ballpark with the previous procedure. Just a little worse.

 

Aren't the graphics setting in a json file?
If so, why not just upload it and tell people to place it there (and replace it after testing)?

 

I wanna keep it simple. It's just nice and easy to select a GFX preset from the in-game menu. Working with fixed json files means extra work for me to create and maintain those files and also extra work for the benchmark user by copy-pasting stuff around.

The difference from the new procedure to the old one is just 1.4 % on my system so it's even comparable to the old procedure results.

 

ok?

IMO, replacing a single file in the documents folder is way easier and quicker than comparing a long list of options on several screen shots. Especially if these move around, change names etc.

 

I don't know how to make my intentions to make the test easier any clearer. If you still think going to GFX options and choosing "Epic" preset is complicated, then be it so.

Edit: And I don't mean that in an agressive way. I think that we basically have the same intentions - avoid the back-and-forth comparison of my screenies with the ACC menu. By using the preset, this isn't needed, any more.
There seems to be a small communication problem...

 

If it's only setting to 'Epic', then you're right.
Thing is that the meaning of that setting may change, possibly even without us noticing.

 

This is correct but not an issue for me. I'm not after performance improvements coming from ACC, anymore:

- I have no support from Kunos in aligning the settings --> show stopper
- The performance has not changed due to new ACC versions ever since I started recording the data --> will most likely stay this way
- Possible performance changes in the Epic preset can quickly be checked and documented by me with a test run when a new version is out. Like I did with this version, already. So comparability to older builds kind of remains.

So we can now focus more on comparisons between GPU models.