Some other threads and my own curiosity, had some time last week to mess around, got me wondering as to whether I could completely cripple my system with the most taxing title I have available. Note PCARS is still officially in pre-alpha working towards a 2014 retail release. See http://www.wmdportal.com/ for more.
Turns out I could quite easily
You can find screen shots of my graphics settings, reported mem usage on my webspace. Obviously this game is far from optimized yet so performance may well improve as it nears release, although on the other hand proper triple screen support (rendering separately to each screen) hasn't been implemented yet which may cancel out improvements from any optimizations. On to the numbers, now updated with single screen results.
PCARS Build 528 - Max Settings, Single car, Belgium Forest, Rain
5900X1080 TS | i7 [email protected] | 2x GTX 780 @ 1100 boost | 16GB DDR3 @ 1663Mhz | measurement tools: FRAPS, EVGA Precision X
PCARS Build 529 - Max Settings, Single car, Belgium Forest, Rain
1920X1080 SS | i7 [email protected] | 2x GTX 780 @ 1100 boost | 16GB DDR3 @ 1663Mhz | measurement tools: FRAPS, EVGA Precision X
Build 529 had a few minor fixes and some physics updates, nothing that materialy affected the results of my testing.
No AA
Min 37 | Max 97 | Avg 67.687 | Mem 1261 | SS
Min 33 | Max 72 | Avg 58.141 | Mem 1731 | TS
High AA (I think this is 4x MSAA)
Min 35 | Max 93 | Avg 67.412 | Mem 1266 | SS
Min 33 | Max 63 | Avg 53.063 | Mem 2032 | TS
dsx2 (In engine downsampling - low)
Min 33 | Max 88 | Avg 66.387 | Mem 1262 | SS
Min 32 | Max 54 | Avg 45.994 | Mem 2018 | TS
dsx4 (In engine downsampling - medium)
Min 35 | Max 75 | Avg 56.398 | Mem 1481 | SS
Min 19 | Max 32 | Avg 26.493 | Mem 2682 | TS
dsx6 (In engine downsampling - high)
Min 29 | Max 57 | Avg 40.951 | Mem 1696 | SS
Min 12 | Max 22 | Avg 17.588 | Mem 3009 | TS
dsx9 (In engine downsampling - ultra)
Min 23 | Max 42 | Avg 31.203 | Mem 2019 | SS
Game would not load | TS
High AA at normal settings
Min 32 | Max 79 | Avg 59.362 | Mem 1808 | TS
Percentage differences TS vs SS
No AA | Avg -16.4 | Mem +37.2
High AA | Avg -27.0 | Mem +60.5
dsx2 AA | Avg -44.3 | Mem +59.9
dsx4 AA | Avg -112.8| Mem +81.0
dsx6 AA | Avg -132.8| Mem +77.4
Conclusions:
As you can see it's easy to go past the 2GB memory mark which has been a hot topic recently, but only at my triple screen resolution. However the game starts becoming a little choppy at high AA/max settings and properly unplayable at dsx2/max settings and above. Even no AA/max settings and high AA/normal settings have far lower min frame rates then I'm comfortable with and none can average over 60FPS. Bear in mind this was a single car on track, the game supports up 64 either AI or online! Start adding cars and the frame rates drop even further, I did try running with some for these tests but at the moment the AI have a propencity for causing large pile ups which doesn't make for repeatable test conditions.
At single screen resolution the game didn't go past the 2GB memory mark, even with ridiculous levels of in engine down sampling, off course the game became unplayable well before I reached that point. dsx4/max settings was OK but above that was slide show territory. Again none of the resolutions had the greatest minimum frame rates, the old start/finish straight after la source hairpin is a real killer, but we do see averages above 60fps for the first time. With no AA/max settings it looks as though I may be CPU limited as GPU usage was struggling to get above 80%
There is no correlation between frame rates and memory usage as AA levels are increased and neither scales linearity with the resolution increase from single screen to triples. In this particular title it looks likely you'd need three 780's or higher in a Tri SLI configuration before you could start making practical use of that extra GB of VRAM these cards have over the GTX 770 and below.
If I can work up the motivation (I did - see above), or if anyone else is in a position to repeat my tests, it would be interesting to see what the results are when dealing with lower resolutions. And if anyone has this along with some Titans I'd love to know what dsx9 does for VRAM usage! I suspect my measly 3GB of VRAM is why I can't even load the game at this setting using triples as part of the front end is rendered in 3D.
Turns out I could quite easily

PCARS Build 528 - Max Settings, Single car, Belgium Forest, Rain
5900X1080 TS | i7 [email protected] | 2x GTX 780 @ 1100 boost | 16GB DDR3 @ 1663Mhz | measurement tools: FRAPS, EVGA Precision X
PCARS Build 529 - Max Settings, Single car, Belgium Forest, Rain
1920X1080 SS | i7 [email protected] | 2x GTX 780 @ 1100 boost | 16GB DDR3 @ 1663Mhz | measurement tools: FRAPS, EVGA Precision X
Build 529 had a few minor fixes and some physics updates, nothing that materialy affected the results of my testing.
No AA
Min 37 | Max 97 | Avg 67.687 | Mem 1261 | SS
Min 33 | Max 72 | Avg 58.141 | Mem 1731 | TS
High AA (I think this is 4x MSAA)
Min 35 | Max 93 | Avg 67.412 | Mem 1266 | SS
Min 33 | Max 63 | Avg 53.063 | Mem 2032 | TS
dsx2 (In engine downsampling - low)
Min 33 | Max 88 | Avg 66.387 | Mem 1262 | SS
Min 32 | Max 54 | Avg 45.994 | Mem 2018 | TS
dsx4 (In engine downsampling - medium)
Min 35 | Max 75 | Avg 56.398 | Mem 1481 | SS
Min 19 | Max 32 | Avg 26.493 | Mem 2682 | TS
dsx6 (In engine downsampling - high)
Min 29 | Max 57 | Avg 40.951 | Mem 1696 | SS
Min 12 | Max 22 | Avg 17.588 | Mem 3009 | TS
dsx9 (In engine downsampling - ultra)
Min 23 | Max 42 | Avg 31.203 | Mem 2019 | SS
Game would not load | TS
High AA at normal settings
Min 32 | Max 79 | Avg 59.362 | Mem 1808 | TS
Percentage differences TS vs SS
No AA | Avg -16.4 | Mem +37.2
High AA | Avg -27.0 | Mem +60.5
dsx2 AA | Avg -44.3 | Mem +59.9
dsx4 AA | Avg -112.8| Mem +81.0
dsx6 AA | Avg -132.8| Mem +77.4
Conclusions:
As you can see it's easy to go past the 2GB memory mark which has been a hot topic recently, but only at my triple screen resolution. However the game starts becoming a little choppy at high AA/max settings and properly unplayable at dsx2/max settings and above. Even no AA/max settings and high AA/normal settings have far lower min frame rates then I'm comfortable with and none can average over 60FPS. Bear in mind this was a single car on track, the game supports up 64 either AI or online! Start adding cars and the frame rates drop even further, I did try running with some for these tests but at the moment the AI have a propencity for causing large pile ups which doesn't make for repeatable test conditions.
At single screen resolution the game didn't go past the 2GB memory mark, even with ridiculous levels of in engine down sampling, off course the game became unplayable well before I reached that point. dsx4/max settings was OK but above that was slide show territory. Again none of the resolutions had the greatest minimum frame rates, the old start/finish straight after la source hairpin is a real killer, but we do see averages above 60fps for the first time. With no AA/max settings it looks as though I may be CPU limited as GPU usage was struggling to get above 80%
There is no correlation between frame rates and memory usage as AA levels are increased and neither scales linearity with the resolution increase from single screen to triples. In this particular title it looks likely you'd need three 780's or higher in a Tri SLI configuration before you could start making practical use of that extra GB of VRAM these cards have over the GTX 770 and below.
If I can work up the motivation (I did - see above), or if anyone else is in a position to repeat my tests, it would be interesting to see what the results are when dealing with lower resolutions. And if anyone has this along with some Titans I'd love to know what dsx9 does for VRAM usage! I suspect my measly 3GB of VRAM is why I can't even load the game at this setting using triples as part of the front end is rendered in 3D.
Last edited: