Archive for the ‘Benchmark’ tag
Metro: Last Light PhysX Benchmarks roundup
PC version of recently released Metro: Last Light title features not only vivid DX11 based graphics, but also hardware accelerated PhysX effects.
In the following article we’ll try to gather the most reliable and accurate GPU PhysX benchmarks and tests for this game.
[14.05.2013] 
Metro Last Light – GPU Test by GameGPU
Sufficient amount of NVIDIA GPUs was tested in this article with the help of Metro’s built-in benchmark. However, since heavy graphics options (like SSAA) were used, it is hard to determine actual PhysX performance.
GPU PhysX in Metro: Last Light
Metro: Last Light, a post-apocalyptic first person shooter with survival horror elements, is joining the family of PhysX enabled titles by offering a support for GPU accelerated physics effects.
Update: Metro Last Light PhysX Benchmarks roundup
Update #2: Metro: Last Light – GPU PhysX Profile
First game in the series – Metro 2033 – was also featuring a GPU PhysX content, however, it was limited to basic particle effects.
Was the Last Light able to improve the results of its predecessor? Let’s find out.
Multithreaded performance scaling in PhysX SDK
Recent “The Evolution of PhysX” article has unvealed the current situation with performance improvements among various PhysX SDK vesions, however, one interesting case has remained outside the coverage – performance scaling in multithreaded environments.
It is known that, while PhysX SDK 2.8 has rather limited multi-threading capabilities (mostly working on per-scene or per-compartment basis), PhysX SDK 3.x can distribute various tasks across worker threads much more effective, and thus offer better support for multi-core CPUs.
But how well does multi-threading actually work in PhysX 3 (we’ll take the latest 3.3 version)? Using the same PEEL (Physics Engine Evaluation Lab) tool to the record the performance metrics, we will try to shed the light on this question.
Scene #1 – random dynamic primitives in a box
Static container filled with 256 random primitives (sphere, box, capsule).
The Evolution of PhysX SDK, performance-wise
A quite interesting, unexpected and a little emotional article – The Evolution of PhysX – was published today by Pierre Terdiman, senior software engineer in NVIDIA and one of the developers of the original NovodeX engine.
Update: Multithreaded performance scaling in PhysX SDK
The article provides in-depth performance comparison between various versions of PhysX SDK (2.8.4, 3.2 and 3.3 Beta), using well-known open-source Bullet physics engine as as a reference point.
The performance tests were performed using PEEL – Physics Engine Evaluation Lab, a specialized tool that is using within NVIDIA to research behaviour and performance of various physics engines using a set of standartized scenes.
GPU PhysX in Hawken
GPU PhysX integration in Hawken, free-to-play mech shooter from Adhesive Games, has a long and yet ongoing story – preliminary PhysX effects have emerged in Alpha and Closed Beta versions, then they have undergone a significant overhaul in Open Beta release.
Update: Hawken – GPU PhysX Profile
Update #2: Upcoming PhysX features in Hawken – Destructible Environments
And now, recent “Raider” update has added a set of new, APEX Turbulence based particles effects, and we think that GPU PhysX support in Hawken has grown up enough to be reviewed.
Catzilla: a new cross-API benchmark
A Beta version of the Catzilla benchmark, developed by Plastic Demo in collaboration with Polish post production company, Platige, is now available for download.
Catzilla is a cross-API (OpenGL 4.0 and DirectX 9/11) benchmark designed the Windows platform, and is also featuring a parallel graphics engine that can take advantage of multi-core CPUs.
Benchmark consists of the main demoscene-style dubstep-heavy giant cat fight scene, which stresses both CPU and GPU, and a set of smaller benchmarks – CPU rigid body physics, GPU smoke simulation, fur rendering, etc.
As interesting note, Catzilla uses PhysX SDK 3.2 as physics engine, however, it is running purely on CPU (as confirmed by one of the developers).
Borderlands 2 PhysX Benchmark roundup
Borderlands 2 is latest and one of the best GPU PhysX titles so far, but what system do you need to handle it efficiently ?
In following article we’ll try to gather all reliable PhysX benchmarks and tests, published on the web, to determine the GPU and CPU performance patterns of this title.
[19.09.2012] Borderlands 2 – GPU Test by GameGPU
One of the first articles with PhysX performance comparison, includes wide range of NVIDIA GPUs tested. At the same time, scene, choosen for benchmark, is too simplistic and does not fully represent load during actual gameplay.
Borderlands 2: is CPU capable of handling the PhysX effects?
Borderlands 2, latest and probably one of the greatest games with support for GPU accelerated PhysX effects, is a HOT topic these days.
Update: GPU PhysX in Borderlands 2 – PhysX review and comparison video
Update #2: Borderlands 2 PhysX Benchmark Roundup
Usually, extra PhysX effects are meant to be executed on compatible NVIDIA GPUs, so even if one can force his CPU to do the work, it is not very effective – massive slowdowns and fps drops during scenes with intence physics are make the games hardly enjoyable. This is valid for titles like Batman series, Alice: Madness Returns, Mafia II and others.
Said matter was a tough topic over recent years, even resulting in claims that NVIDIA “hobbles” the CPU PhysX performance by purpose, to make their GPUs look more advantageous.
However, recently we saw many reports (mostly from AMD users) that Borderlands 2 shows surprisingly good performance, while running with all PhysX effects enabled even without a NVIDIA card in the system.
Thus, we decied to perform a little investigation to answer the question – can a CPU handle all the extra PhysX effects in Borderlands 2 ?
A boss fight against “Boom and his brother Bewm” is a good candidate for PhysX testing – scene contains a lot of particles (also, particles are constantly generated over time) and some cloth objects.
Gun in our hands has explosive rounds - this produces even more particles
Passion Leads Army: DX 11 and GPU PhysX Benchmark
Multiplayer online shooter “Passion Leads Army” (PLA), currently in development by Giant Interactive in cooperation with the Chinese military, was firstly showcased (as a real-time benchmark demo) during Jen-Hsun Huang Keynote at NVIDIA Gaming Festival.
Featuring full DX 11 support (for the first time – in Chinese game) and intense GPU PhysX effects, this UE3 based title has drawn much attention of the audience.
Now, public release of PLA Benchmark gives us the opportunity to try it out by ourselves.
Update: PLA Benchmark overview at GeForce.com
Following DX11 effects can be seen in the benchmark:
- Tesselation.
- Horizon-Based Ambient Occlusion (HBAO).
- Bokeh-DOF effect.
As for GPU PhysX content, it includes:
- GPU accelerated rigid bodies (destruction scenes).
- APEX Particles effects (leafs, sparks, debris, dynamic fog).
- Interactive and tearable PhysX cloth objects.
GTX 680: PhysX Benchmarks roundup
Finally, NVIDIA has officially presented their new GPU – GTX 680, built on 28nm GK104 chip, which itself is based on next-gen Kepler architecture.
Many rumors were floating around Kepler and its physics acceleration capabilities, some were endowing new GPU with dedicated PhysX blocks, others were claiming ability to run CPU PhysX games in hardware – none of this happens to be true. Nothing special (like NVENC video encoder) for physics calculations, just general improvements to chip architecture, SM clusters, etc.
Nonetheless, we will try to gather all available information regarding PhysX computation performance of the GTX 680, in comparison to previous generation GPUs.
NVIDIA Kepler GPU GeForce GTX 680 Video Card Review by HARDOCP
Interesting results – in this case, GeForce GTX 680 performance is higher than Radeon HD 7970 even with GPU PhysX effects enabled. However, we can not directly compare GTX 580 and GTX 680 due to different resolution settings.