Split: Benchmark

[Commander_X]

Some more results:

FPS: 162

(Lenovo ThinkPad T440s)
I7-4600U/12G RAM
Intel HD Graphics
Windows 7 Enterprise/64bit


FPS: 40

(Dell Latitude D830)
Intel Core2 T7600/4G RAM
NVidia Quadro NVS 110M (proprietary driver 304.88)
Slackware64 14.1

NB All the screenshots taken while facing Lave.

[Zieman]

Okay, I downloaded the test build (Trunk e855d90) and did a few tests.

Found out that Oolite seems to be CPU-bound on my system, single core performance matters most.

How did I arrive to that conclusion?

I left GPU-Z recording sensor readings in the background, and did a few tests with the task manager showing CPU utilization too.
GPU usage never went ofer 43%, even in full screen 1920x1080 with "anti-aliasing" = 1; - Core Temp showed CPU temps rising up to 55°C and task manager did show one core being used 80-95% & others 5-20% in those tests.

Got same FPS both in 800x600 windowed and 1920x1080 full screen, antialiasing on or off.
992 (stable) in Commander status screen prior to launch
~650 (fluctuating ±10, momentary drops to 590-620 when object count changing) front view facing planet, stopped right after launch
~810 (fluctuating ±10, momentary drops to 760-790 when object count changing) rear view showing Lave Station, stopped right after launch

[StathisR]

Running on my desktop



CPU i7 4790K 4.4 GHz (Turbo boost mode)
GPU Gigabyte GTX-970 G1 Gaming
RAM 16 GB DDR3 1600 MHz
Win 7 Home Premium 64 bit

[Tricky]

Lenovo C365
AMD A4-5000 APU 1.5 Ghz
AMD Radeon HD 8330
Windows 8.1 64 bit (with Classic Shell)

90 FPS looking at Lave.
100 FPS looking at space and nebulae.

I think the Catalyst driver limits the FPS to 100.

Re: nVidia Threaded Optimisation: The general consensous on many forums is to turn it off globally for all applications.

[UK_Eliter]

EDITED: I get 59-63 FPS, suggested - as Norby did - that perhaps somehow Vsync remains on. I get 187-90 when facing Lave. [Thanks to another_commander and Cody for getting me to have another look at my ATI settings. 'Off unless application specified' was not turning off Vsync in Oolite; I had to use the 'always off' setting.]

Windows 8.1, Intel Q6700, 4GB main RAM, AMD Radeon HD 5570 1GB (overclocked, but on PCI-E 1.1 rather than 2).

[another_commander]

I get 59-63 FPS, suggested - as Norby did - that perhaps somehow Vsync remains on.
.

Maybe v-sync is set to On at the driver level. You may want to see if you can temporarily set it to Application Controlled or Off for the purposes of the test.

[Cody]

Try disabling the Catalyst AI in the CCC?

[UK_Eliter]

I get 59-63 FPS, suggested - as Norby did - that perhaps somehow Vsync remains on.
.

Maybe v-sync is set to On at the driver level. You may want to see if you can temporarily set it to Application Controlled or Off for the purposes of the test.

Thanks. I've amended my original post.

[Fatleaf]

GPU GeForce GT 710M CPU intel(R) Core(TM) i3-3217U @ 1.80GHz Memory 6GB RAM

138 looking towards Lave


125 at station


Looks like I'm not setting things up correctly with such a low number. Either that or or my computer is not that great.

[Pleb]

GPU GeForce GT 710M CPU intel(R) Core(TM) i3-3217U @ 1.80GHz Memory 6GB RAM

138 looking towards Lave


125 at station


Looks like I'm not setting things up correctly with such a low number. Either that or or my computer is not that great.

I only get around 170 FPS when looking at Lave and around 180 FPS when looking at the station on my laptop and it's specs are kind of similar:

Intel Core i5-3317U Dual Core 1.7 GHz
8GB RAM DDR3 1600MHz
Nvidia Geforce GT 620M 1GB VRAM
Windows 8.1 64-bit

[cim]

If you're interested as to where (very broadly) Oolite is taking the time per frame, you can turn on "universe.profile" in your logcontrol.plist. This will generate a ridiculously large log file, so you probably shouldn't use it for anything other than a quick test like this.

The universe.profile.update lines are the game updating the simulation data (which is mostly CPU-intensive); the universe.profile.draw lines are the game drawing the result (which is mostly GPU-intensive). You can use the timestamps in the log file to then work out milliseconds per step.

I would guess with those specs that it may be your CPU which is the limiting factor here.

[spara]

Lave: ~270
Station: ~325

GeForce GTS 450
Pentium(R) Dual-Core CPU E6700 @ 3.20GHz
4GB RAM
Debian jessie/sid

[Cmdr Wyvern]

ASUS M5A99FX Pro R2 motherboard
AMD FX-8350 8-core CPU, with a large aftermarket heatsink and fan to match (Seriously, never go with the stock heatsink on a modern hot running CPU!)
16GB DDR3 ram
EVGA GTX 1070 SC (Micron memory, firmware patch applied)
Corsair 850 watt, Gold standard modular power supply
2 1 tb SSD drives
Large screen HD monitor, 1080P@60hz. Connected via HDMI
Windows 8.1 pro x64, with Classic Shell menu.
All hardware at stock clocks, only custom fan curves are used.
windows power management set to performance
HWInfo used to monitor system and GPU during test
Oolite 1.84 x64, buck naked with no oxps, Extra Details.

Geforce control panel, all settings default except:
Threaded = off
pwr management = optimal
// TB on and v-sync fast lets the GPU render as fast as it wants to, while maintaining monitor refresh on screen with no tearing artifacts. The downside is that excess frames are dropped, and it's a big waste of power. Good for benchmarking and troubleshooting, but little benefit otherwise.
Triple buffering = on
vsync = fast

Results in flight:

Minimum FPS = 320
Max FPS = 457

GPU clock peaked at 1000mhz, loading at 40%. It can boost to 1960mhz.
CPU load on two cores at 50% each, clocking to 3.9ghz (There was a 20% load on a 3rd core, but I suspect that was some windows housekeeping process in action and not related to the game.)

Getting here was an epic quest. See this thread for the gory details.

[Cody]

There was a 20% load on a 3rd core, but I suspect that was some windows housekeeping process in action and not related to the game.

Process Explorer would tell you exactly which process was using which core etc.

[Cmdr Wyvern]

There was a 20% load on a 3rd core, but I suspect that was some windows housekeeping process in action and not related to the game.

Process Explorer would tell you exactly which process was using which core etc.

It was prob'ly HWInfo creating a load, while it kept a running tally of the system loads. The irony does not escape me.