Benchmark results, food for thought

As i said, it should say threads and not cores.
It must be a typo from Kunos.

 

AC VERSION: 1.3.4 (x64)
POINTS: 14149
FPS: AVG=96 MIN=3 MAX=108 VARIANCE=0 CPU=48%

LOADING TIME: 46s
GPU: NVIDIA GeForce GTX 760 (1920x1080)
OS-Version: 6.1.7601 (Service Pack 1) 0x300-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:2X AF:8X SHDW:2048 BLUR:0
WORLD DETAIL: 5 SMOKE:3
PP: QLT:3 HDR:1 FXAA:1 GLR:3 DOF:3 RAYS:1 HEAT:1

 

Running Ultra everything but motion blur

AC VERSION: 1.3.4 (x86)
POINTS: 8438
FPS: AVG=57 MIN=23 MAX=116 VARIANCE=29 CPU=80%

LOADING TIME: 27s
GPU: AMD Radeon R9 200 Series (2560x1440)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:4096 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:1 FXAA:1 GLR:5 DOF:5 RAYS:1 HEAT:1

 

I did run it only once, maybe you get little better results with 2nd try. Does pp make any difference? I don't remember which I'm running but it is not the default. Would be lovely if my min value was little higher.

AC VERSION: 1.3.4 (x64)
POINTS: 20478
FPS: AVG=139 MIN=59 MAX=198 VARIANCE=34 CPU=87%

LOADING TIME: 17s
GPU: NVIDIA GeForce GTX 970 (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:2X SHDW:1024 BLUR:0
WORLD DETAIL: 3 SMOKE:0
PP: QLT:0 HDR:1 FXAA:0 GLR:0 DOF:0 RAYS:0 HEAT:0

 

PP and weather makes a big difference. Bigger the resolution you are rendering, more difference the PP makes.
Without post processing effects I get almost double FPS @5760x1080.

 

AC VERSION: 1.3.4 (x64)
POINTS: 10598
FPS: AVG=72 MIN=28 MAX=100 VARIANCE=0 CPU=52%

LOADING TIME: 23s
GPU: AMD Radeon R7 200 Series (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:2X AF:16X SHDW:512 BLUR:0
WORLD DETAIL: 3 SMOKE:3
PP: QLT:3 HDR:0 FXAA:0 GLR:5 DOF:0 RAYS:1 HEAT:0

pp are off, not sure if this is true for the bench but it is off in my settings.
pretty pleased with ac performance as of 1.3, haven't driven the latest build due to time constrains though.

 

PP = On (default weather clear, modified default PP.ini)
AC VERSION: 1.3.4 (x64)
POINTS: 9255
FPS: AVG=63 MIN=48 MAX=73 VARIANCE=0 CPU=39%

LOADING TIME: 12s
GPU: NVIDIA GeForce GTX 770 (5760x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:2X AF:8X SHDW:1024 BLUR:0
WORLD DETAIL: 4 SMOKE:1
PP: QLT:3 HDR:1 FXAA:0 GLR:3 DOF:3 RAYS:0 HEAT:0

PP = Off
AC VERSION: 1.3.4 (x64)
POINTS: 13523
FPS: AVG=92 MIN=66 MAX=110 VARIANCE=0 CPU=56%

LOADING TIME: 17s
GPU: NVIDIA GeForce GTX 770 (5760x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:2X AF:8X SHDW:1024 BLUR:0
WORLD DETAIL: 4 SMOKE:0
PP: QLT:3 HDR:0 FXAA:0 GLR:3 DOF:3 RAYS:0 HEAT:0

Differences are even greater when you drive your self, instead of using the benchmark, and have more AI on the track. I presume same goes for MP.
If you put on a "heavier" weather, you will go below the minimum 60fps with PP = on.
Appearence differences are minute, and for ~10fps you can turn on CSAA x8 from nvidia inspector if you like, and still average plenty over 60fps constant.

 

Wait.... there is a in-game benchmark feature??

 

Yes! Just because the framerate goes above most peoples monitor refresh rate, doesnt mean the unstable framerate wont be noticeable. Its even more apparent when playing FPS games with jumpy framerates, but it still alters the input lag you get wildly, even if its a small amount at such high framerates.

I run a 144hz screen here, so the visible window (outside of feeling something is wrong) is bigger for me, especially since my framerates run around 70-110 (trying to have as much eyecandy as I can without impacting gameplay)
Only smoke generation has given me problems since 1.3. I had it on High or Normal before, and whenever there is dust coming up I got major FPS drops. Putting it on Low has made that impact smaller but its still there. Im not sure why it happens but its not a big deal visually.

 

Yes, it's called frametime. At that moment it dips down to 80 from 180, if there is any 'frame skipping' then you will see stuttering no matter if the fps is still relatively high. To avoid this, a good practice is to put fps limits on games that support it, which should be close to what your GPU can keep stable. That way you keep a solid target for your fps and the frametime remains stable at most. Which is why console games that are capped at 30fps, while not being 60+fps smooth, they don't really get any stuttering either (unless we're talking about games that just can't run well on the archaic console hardware, even at 30 capped)

 

I want these PP off FPS values! Main difference of our rigs is a superior CPU on your end. But then again I never saw my CPU go anywhere above 70 %.
I always thought you couldn't run triples in FS mode so I really hope this will bring the difference for me (always ran windowed mode). Everyone else seems to be running triples in FS.

Are u running triple rendering?

 

Many things, which i can't understand why, change "variance". See below..


CPU @ 4.7 RAM @ 1333 GPU @ defaults 1000/1500
AC VERSION: 1.3.4 (x64)
POINTS: 15118
FPS: AVG=103 MIN=24 MAX=176 VARIANCE=15 CPU=67%

LOADING TIME: 28s
GPU: AMD Radeon R9 200 Series (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:512 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:0 RAYS:1 HEAT:1


CPU @ 4.7 RAM @ 2133 GPU @ defaults 1000/1500
AC VERSION: 1.3.4 (x64)
POINTS: 15121
FPS: AVG=103 MIN=27 MAX=177 VARIANCE=10 CPU=66%

LOADING TIME: 28s
GPU: AMD Radeon R9 200 Series (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:512 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:0 RAYS:1 HEAT:1


CPU @ 4.7 RAM @ 2400 GPU @ defaults 1000/1500
AC VERSION: 1.3.4 (x64)
POINTS: 15061
FPS: AVG=102 MIN=23 MAX=186 VARIANCE=5 CPU=61%

LOADING TIME: 27s
GPU: AMD Radeon R9 200 Series (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:512 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:0 RAYS:1 HEAT:1


CPU @ 4.7 RAM @ 2400 GPU @ OC 1100/1600
AC VERSION: 1.3.4 (x64)
POINTS: 16380
FPS: AVG=111 MIN=23 MAX=189 VARIANCE=16 CPU=67%

LOADING TIME: 16s
GPU: AMD Radeon R9 200 Series (1920x1080)
OS-Version: 6.2.9200 () 0x300-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:512 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:0 RAYS:1 HEAT:1

As you see, i get the min variance=5 with ram @ 2400 and no overclocked gpu.
If i overclock gpu variance goes up to 16....

 

Ignore the resolution, I took these shots with single screen.
Those benchmark results are with the following video settings.

 

Actually it is physical + virtual cores

 

Stock 8320@3.7 and 280x@1080 and 6500.
Mirrors kill my FPS

AC VERSION: 1.3.4 (x64)
POINTS: 14363
FPS: AVG=98 MIN=30 MAX=174 VARIANCE=28 CPU=73%

LOADING TIME: 34s
GPU: AMD Radeon R9 200 Series (1920x1080)
OS-Version: 6.1.7601 (Service Pack 1) 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:8X SHDW:2048 BLUR:0
WORLD DETAIL: 4 SMOKE:3
PP: QLT:4 HDR:1 FXAA:0 GLR:4 DOF:2 RAYS:1 HEAT:1

 

I'd like to share my results, because there are some interesting things and it's a nice setup for this kind of hardware.

Spoiler: Hardware

CPU: Q6600@3.6Ghz; GPU: HD7950@980Mhz/1450Mhz 3GB Vram; RAM: 8GB ddr2@800mhz, MONITORS: 1x 1280*1024, 1x 1920*1080, 1x 1680*1050 (lol 2 monitors I had around, works fine with Eyenfinity, the wheel isn't aligned to the center of middle screen but it's acceptable), with Eyenfinity resolution @ 4880x1024. DRIVES: SSD, 840 Evo for OS and HDD, WD Blue for games

Spoiler: OS

Win7 x64 SP1, all services and processes to minimum possible, classic theme. Latest graphic driver.

Spoiler: In game settings (shared by all the next benchmarks)

DISPLAY: Fullscreen ON; AF: 4x; AA: 4x; World details: Medium; Shadows: High (with a mod in the .ini to hide the zones between different shadow map sizes); EFFECTS: Smoke: Very Low; Smoke visible in mirrors ON; REFLECTIONS: Mirrors Resolution: High; HQ Mirrors ON; Reflections Quality: Low; Reflection frequency: Static. The other settings will vary with the benchmarks.

I've spent some time testing with different settings, my aim was to obtain at least 60fps average and with these settings I'm near it. I've decided to put the reflection on static because I like a better quality image overall, reflections only affect cars' paint so they're expendable for my objective. However, It's nice to have the possibility to set them static so at least there's something there. I don't like 3D grass very much, so World Detail is at Medium, does it fire some mechanics from the box too? Well, I can't afford them.

Spoiler: Triple Screen in Display Settings ON, PP OFF

AC VERSION: 1.3.4 (x64)
POINTS: 8361
FPS: AVG=57 MIN=31 MAX=88 VARIANCE=2 CPU=79%

LOADING TIME: 18s
GPU: AMD Radeon HD 7900 Series (4880x1024)
OS-Version: 6.1.7601 (Service Pack 1) 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 2 SMOKE:1
PP: QLT:0 HDR:0 FXAA:0 GLR:0 DOF:0 RAYS:0 HEAT:0

Spoiler: Single Screen in Display Settings, PP OFF

AC VERSION: 1.3.4 (x64)
POINTS: 9976
FPS: AVG=68 MIN=41 MAX=114 VARIANCE=3 CPU=72%

LOADING TIME: 18s
GPU: AMD Radeon HD 7900 Series (4880x1024)
OS-Version: 6.1.7601 (Service Pack 1) 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 2 SMOKE:1
PP: QLT:0 HDR:0 FXAA:0 GLR:0 DOF:0 RAYS:0 HEAT:0

OK, Variance is higher, but watching it go FPS are most of the time >59 (read the question in the very end of this post), even at the start.

Clearly using the Triple Screen mode will lower the FPS, so I'll leave it Single Screen in the next benchmarks:

Spoiler: PP ON, Global quality: Very Low, rest OFF

AC VERSION: 1.3.4 (x64)
POINTS: 7820
FPS: AVG=53 MIN=39 MAX=65 VARIANCE=0 CPU=58%

LOADING TIME: 18s
GPU: AMD Radeon HD 7900 Series (4880x1024)
OS-Version: 6.1.7601 (Service Pack 1) 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 2 SMOKE:1
PP: QLT:0 HDR:1 FXAA:0 GLR:0 DOF:0 RAYS:0 HEAT:0

Argh, it's a shame it taxes FPS so much, because AC's post processing is really nice.
However, I like it OFF too, it reminds me of GTR2 and I really liked that graphic.

Let's try with a tool setup to mimic the original post processing:

Spoiler: PP OFF; SweetFX with Bloom&HDR

AC VERSION: 1.3.4 (x64)
POINTS: 8317
FPS: AVG=56 MIN=41 MAX=71 VARIANCE=0 CPU=65%

LOADING TIME: 19s
GPU: AMD Radeon HD 7900 Series (4880x1024)
OS-Version: 6.1.7601 (Service Pack 1) 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 2 SMOKE:1
PP: QLT:0 HDR:0 FXAA:0 GLR:0 DOF:0 RAYS:0 HEAT:0

A bit better than the original in terms of absolute performance, but I assume the original works in a more complex way given its look. Or doesn't it?
I've never used SweetFX before, maybe its settings need more fine tuning.

I don't understand much why the reported CPU % goes down when PP or SweetFX are working. I'd thought the contrary.

Can we have more data in the benchmark results? Something like frametime %s. But it's nice to have a Variance at least and a benchmark tool, too: it makes getting these results fast and easy.

It would be nice to have the possibility to save the video settings in different sets, so that we could change all of them with a click, e.g. when we want full eye candy for practice alone.

 

The Benchmark is a nice tool but has raised an interesting question for me. I am running with a evga GTX970 SSC and a single 4K/60hz monitor. I have been running at it's max resolution at 3840 X 2160. I ran a benchmark on it.

3840 X 2160
AC VERSION: 1.3.4 (x64)
POINTS: 7998
FPS: AVG=54 MIN=43 MAX=99 VARIANCE=0 CPU=27%

LOADING TIME: 16s
GPU: NVIDIA GeForce GTX 970 (3840x2160)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 4 SMOKE:3
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

When I saw this result I was struck by the fact that the GPU was not able to keep the monitor fed at it's full capacity while the computer was loafing. So I tried 1080p for the output.

1920 X 1080
AC VERSION: 1.3.4 (x64)
POINTS: 22911
FPS: AVG=156 MIN=72 MAX=241 VARIANCE=25 CPU=78%

LOADING TIME: 12s
GPU: NVIDIA GeForce GTX 970 (1920x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 4 SMOKE:3
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

The GPU is now really pumping out the frames, the CPU is working a lot harder BUT the quality of the picture on the monitor is significantly less detailed. It is easy to the see the quality difference. So I selected an intermediate resolution between 1080p and 2160p with the same aspect ratio.

2560 X 1440
AC VERSION: 1.3.4 (x64)
POINTS: 15517
FPS: AVG=105 MIN=77 MAX=148 VARIANCE=0 CPU=53%

LOADING TIME: 11s
GPU: NVIDIA GeForce GTX 970 (2560x1440)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:4X SHDW:2048 BLUR:0
WORLD DETAIL: 4 SMOKE:3
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

I am going to be using this setting for a while as this seems to be in the sweet spot for my rig but has me thinking about a new GPU that can comfortable perform with minimums above 60fps at a full 4K resolution. Is anyone currently running a GPU on a single 4K monitor that meets this objective.

 

Its hard to understand in terms of fps, but much easier in terms of frame times. 60fps means a 16ms frame time. If you have 100fps then 50fps, that will mean that some frames are displayed for double the amount of time than other frames, that will give you microstutter at even high fps.

 

for some reason PP doesnt really make a difference on me, i get better fps a 1440 than 1080, or its not much different, like difference of 3 or 5 FPS, crossfire is the only thing that makes a difference on mine, i get double with crossfire on, crossfire makes variance like 30, 1 card makes variance 0