JeGX has presented new 1.4.0 version of FluidMark, benchmarking and testing tool for GPU PhysX/Hybrid PhysX systems.

Release Highlights:

• New: added two benchmark presets (1080 and 720).

Preset:1080: with this preset, the settings are the following:: 1920×1080 fullscreen, duration of 60 sec, 60000 particles, heavy additional graphics load and multithreaded PhysX synchronized on the rendering.

Preset:720: with this preset, the settings are the following: 1280×720 fullscreen, duration of 60 sec, 30000 particles, moderate additional graphics load and multithreaded PhysX synchronized on the rendering.

• New: PhysX built-in CPU multicore option added.
• New: score submission with oZone3D.Net account.
• Update: the additional graphics load option covers now the whole screen and not only the upper-right corner.
• Update: revamp of the main startup dialog box (now a bit simpler to use).
• Update: compiled with PhysX SDK 2.8.4.6.
• Bugfix: fixed a nasty deadlock that hung FluidMark sometimes at the end the tests.

You can download PhysX FluidMark 1.4.0 from Geeks3D.com.

We remind you, that in case if you want to use FluidMark 1.40 with Hybrid PhysX Mod, you need to delete/rename PhysXDevice.dll file in application folder, due to usage of driverless PhysX SDK 2.8.4.

Jerome “JeGX” Guinot has updated his popular benchmarking tool, known as PhysX FluidMark, to version 1.2.2.

Release Notes:

Version 1.2.2 – 2010-09-20
————————–
! Change: new post processing (post-FX) effect.
! Change: scores submissions are limited to fullscreen mode + No AA + No PostFX.
! Update: compiled with PhysX SDK 2.8.4.4.
* Bugfix: the resolution 1920×1080 was not saved at the closing of FluidMark.

As you may notice, particles have changed their color from vibrant yellow to bloody red, but major improvements are related to usage of newest PhysX SDK 2.8.4.4.

1) FluidMark 1.2.2. does no require PhysX System Software to be installed, since SDK 2.8.4. is driverless.

When FluidMark 1.2.0. can only helplessly display error messages, version 1.2.2. starts without any problems.

2) Binary version of PhysX SDK 2.8.4. is compiled with SSE2 optimizations, and, thus, we can expect some performance gain while running FluidMark on CPU. But lets’s test this out:

System: C2Q 9400 CPU, GTX 470 + GTS 450 GPUs, 4Gb RAM, Win 7 x64, 260.63 GPU drivers, 9.10.0514 PSS. FluidMark Settings: 1024x768, Async mode Off, Multi-core PhysX On, 60000 particles, 60000 ms benchmarking time, 3 emitters.

Stable performance improvement – around 14 %. Of course, it may not be purely SSE merit, but related to internal SDK tweaks, but it is good to see a progress.

And what about GPU calculations ?

No actual difference between FluidMark 1.2.0 and 1.2.2.

Obviously, single GTX 470 is significantly faster than dedicated GTS 450, because FluidMark is not stressing system with graphics calculations. In actual games dedicated PhysX GPU will perform much better.

You can download FluidMark 1.2.2. from here

Our thanks to JeGX for pre-release version of FluidMark 1.2.2., provided beforehand

As everyone else, we have very little information about next major release of PhysX SDK – 3.0, which was rumored as complete rewrite of current SDK, full of new features and extended capabilities, currently kept under straight NDA.

UPDATE: PhysX SDK 3.0 has been released

However, few pieces of information are beginning to leak:

Today, as answer to all the hype about PhysX and SSE instructions, Nvidia’s senior PR manager Bryan Del Rizzo has stated in interview to THINQ.co.uk website, that new SDK 3.0 will feature “a task-based approach that was developed in conjunction with [Nvidia] Apex product to add in more automatic support for multi-threading“.

In generally, SDK 3.0 will automatically take advantage of however many cores are available, or the number of cores set by the developer, and will also provide the option of a “thread pool” from which “the physics simulation can draw resources that run across all cores“. – adds THINKQ.co.uk

We will keep an eye on all SDK 3.0 traces and post new info as we’ll find it.

New version of PhysX FluidMark, benchmarking and testing aplication for GPU PhysX systems, was finally released to public.

• New: added support of multi-core CPUs (PhysX simulation is spread over several threads). New checkboxes are also available to control this feature.
• New: added a checkbox to set the number of particles.
• New: added a checkbox to force CPU PhysX.
• New: added Heavy PhysX mode: to make modern systems with GPU PhysX happy
• New: added point sprite based rendering for particles (keys F2, F3 and F4).
• New: added keyboard-based camera control in stability test mode.
• Change: compiled with NVIDIA PhysX SDK 2.8.3.21.
• Change: updated with the latest version of ZoomGPU for graphics hardware detection.

Update: limitations of GPU PhysX simulation

Update #2: Async mode details

Update #3: FluidMark 1.2.2. available

We’ve already did more or less detailed features overview in our FluidMark 1.2 Beta preview article (in addition, more technical details are available from original post), so this time we will just point on certain changes in release version in comparison to beta one.

Main window haven’t underwent much changes..  (click to view full picture)

.. while benchmarking process was improved, to achieve results standardization. Every benchmarking sequence now starts with so called “Warming-Up” state, to ensure that all particles are emitted on the scene. In addition, default settings are now set to 60 000 particles and one minute timed run.

Moreover, new FluidMark 1.2 is now officially approved by Nvidia. According to JeGX, FluidMark developer, NV engineers have helped with bug fixing and overall optimizations, but have not influenced development process in order to give an advantage to GeForce GPUs or penalize CPUs.

And, of course, some tests (Note: since we are using different settings now, and other internal changes in FluidMark, these results can not be directly compared with our previous benchmarks)

System: C2Q 9400 @ 2.66 GHz CPU, Nvidia GTX 275 + GTX 260 (192 sp) GPUs, 4GB RAM, Win XP, PhysX System Software 9.09.1112. Default settings, # of emitters = 3.

In final version of FluidMark multi-core PhysX checkbox is related only to PhysX SDK simulation part, while main app (GUI) and rendering threads are separated all the time – thus  it almost does not affect GPU PhysX mode, like in Beta version.

As for much higher difference between GPU PhysX and multi-core CPU PhysX – it seems that beta version contains a bug, that results in abnormal low score for PhysX on single GPU (in general, it should be much closer to dedicated PhysX GPU).

Sum: Good old FluidMark, but now with true multi-core CPU PhysX support and enhanced benchmarking procedure. One of the best GPU PhysX testing tools becomes even better.

You can download PhysX FluidMark 1.2 from here (2.5 mb)

P.S. Thanks to Jerome “JeGX” Guinot for Release Candidate version of FluidMark, provided beforehand.

As we mentioned previously, upcoming FluidMark 1.2, next version of popular GPU PhysX testing and benchmarking application, will include support for Multi-Core CPU PhysX calculations, and overall multi-threading optimizations as well.

Jerome Guinot, FluidMark developer, was kind enough to provide us with latest beta-version of new Fluid-Mark 1.2, and we’ll try to answer finally, what is faster – GPU PhysX or properly optimized CPU PhysX.

But first, lets take a closer look at new FluidMark. (click to view full picture)

Main control panel now includes several additional options, like “Force PhysX CPU” – ability to switch between GPU and CPU PhysX, without necessity to use Nvidia Control Panel.

“Multi-core PhysX” checkbox enables all multi-threading optimizations, vital and most interesting part of new FluidMark.

“# of CPU cores” is used specify number of CPU cores dedicated to simulation (up to 32 in current version), however this option is no so transparent as it looks – increased number of cores adds additional fluid emitters to the scene (one emitter per core or two in general), and with equal number of particles, various number of emitters can affect performance.

Application window has also changed – benchmark is still based on SPH fluid simulation, built into PhysX SDK (latest version 2.8.3.21 is used), but scene includes additional static objects, particles appearance if different and, as mentioned earlier, several emitters can be used simultaneously. Nice addition – GPU temperature overlay, usefull for GPU stress testing.

Final Global score in benchmarking mode is calculated now in a different way, and can’t be compared with previous version of FluidMark. It consist of two components – GraphX score (graphics framerate per second) and PhysX score (physics simulations per second).

Thus, Global score = (GraphX_score * 0.3 + PhysX_score * 1.7) / 2

Now, lets do some testing.

The Wonder of Multi-Threading !

Take a look at the following graph:

[Three emitters were used (# CPU cores = 4) with fixed number of particles - 15 000. Timerange - 60 sec. 800x600 rendering window. System: C2Q 9400 @ 2.66 GHz CPU, Nvidia GTX 275 + GTX 260 (192 sp) GPUs, 4GB RAM, Win XP, PhysX System Software 9.09.1112]

When “Multi-core PhysX” option is off, PhysX simulation and scene rendering are done in the same thread and, more important, PhysX SDK multi-threading flags are not set.

But when “Multi-core PhysX” is enabled, all PhysX simulations are done in separate threads and since there is still a thread for the rendering, scene rendering is boosted because there is no longer PhysX in scene thread. Same situation with PhysX, one or several threads are completely dedicated for physics simulation.

While SPH fluid simulation is running on CPU with  “Multi-core PhysX” set to off, load is destributed through several cores (probably due to internal Windows threads management), but in sum that’s 26% – full one core.

But with multi-threaded optimizations enabled, application fully utilizes all four cores by 100%, what results in great speed boost.

In addition, one interesting detail was discovered – fluid simulation is running faster on GPU when one emitter is used, and opposite way – for CPU it prefers multiple emitters (with equal number of particles) – probably that’s peculiarity of PhysX SDK itself.

For example, with one emitter and multi-core PhysX switched to off, CPU simulation results in 36 global points (64 with 3 emitters – on graph above), while GTX 275 GPU – in 247 points (128 with 3 emitters). But since one emitter can’t utilize more than two cores,  number of emitters was increased to gain equality.

Therefore, bechmarking seems to be a little tricky in new FluidMark. We are curious if someone will come with solid method after app release.

P.S. Thanks to Jerome for beta FluidMark and detailed explanations

Recent articles about Metro 2033 have revealed so much technical details (about engine itself, and PhysX components specifically) so, one would think, nothing new can be added. However, recent interview with Chief Technical Officer Oles Shishkovstov by PCGamesHardware.com has something to offer:

PCGH: It could be read that your game offers an advanced physics simulation as well as a support for Nvidia’s PhysX (GPU calculated physics) can you tell us more details here? Does regular by CPU calculated physics affect visuals only or is it used for gameplay terms like enemies getting hit by shattered bits of blown-away walls and the like?

Oles Shishkovstov: Yes, the physics is tightly integrated into game-play. And your example applies as well.

PCGH: Besides PhysX support why did you decide to use Nvidia’s physics middleware instead of other physics libraries like Havok or ODE? What makes Nvidia’s SDK so suitable for your title?

Oles Shishkovstov: We’ve chosen the SDK back when it was Novodex SDK (that’s even before they became AGEIA). It was high performance and feature reach solution. Some of the reasons why we did this – they had a complete and customizable content pipeline back then, and it was important when you are writing a new engine by a relatively small team.

PCGH: What are the visual differences between physics calculated by CPU and GPU (via PhysX, OpenCL or even DX Compute)? Are there any features that players without an Nvidia card will miss? What technical features cannot be realized with the CPU as “physics calculator”?

Oles Shishkovstov: There are no visible differences as they both operate on ordinary IEEE floating point. The GPU only allows more compute heavy stuff to be simulated because they are an order of magnitude faster in data-parallel algorithms.

As for Metro2033 – the game always calculates rigid-body physics on CPU, but cloth physics, soft-body physics, fluid physics and particle physics on whatever the users have (multiple CPU cores or GPU). Users will be able to enable more compute-intensive stuff via in-game option regardless of what hardware they have.

Pay attention to last paragraph – Metro 2033 will feature true multi-core implementation of GPU PhysX content – feature that most PhysX titles are lacking currently ? We are curious to see if this will really work, and since game has already gone gold, we’ll learn that very soon.

We wrote previuosly about upcoming new version of FluidMark, that is going to include Multi-Core CPU PhysX support.

Recently, with help of David Legrand from PCinpact.com, Jerome “JeGX” Guinot was able to test new FluidMark 1.2 on setup with 16 CPU cores. (click to view full image)

And as you may see, CPU PhysX simulation with 48 000 fluid particles is 100 % utilizing all 16 Cores. Amazing work

New FluidMark is going to be available soon enough, after JeGx will catch and kill remaing bugs.

Update: JeGX is participating in “What Would You Do With 48 Cores ?” contest by AMD with new FluidMark (which already support up to 64 CPU cores). That’s going to be interesting 

Source: PCinpact | Geeks3D

Old PhysX fans will certanly recognize this name – Novodex Rocket (PhysX Rocket later on). This application combines two roles: demo physics playground (with large number of  preliminarily created and configurable scenes, ability to change and visualize SDK parameters) and physics editor, which can export objects data to COLLADA or NxuStream.

PhysX Rocket was often used by Ageia to demonstrate SDK features and PPU computing capabilities in year 2005, and was included in SDK package as PhysX tool (till SDK 2.7.3.).

John Ratcliff, PhysX Rocket creator, has made a nice present recently – he revealed an updated legacy version of Rocket, which is based on latest PhysX SDK 2.8.3 and includes advanced UI options (unavailable in previous public versions), but is missing some vital demo scenes and object files. Fortunately, we were able to merge this updated Rocket with old one, from SDK 2.7.3 Tools – this means all demos like in regular Rocket, but SDK 2.8.3 solver and additional UI options from new one.

PhysX Rocket remains a nice physics playground even today, so if you’re ready to spend half an hour by throwing physical bricks, bending physical grass and kicking physical ragdolls, use download link provided below:

Download PhysX Rocket for SDK 2.8.3 from our server (75 mb – download currently unavailable). RAR archive, no installation required.

Note: Requires latest PhysX Systems Software to run properly.

And, by the way, Rocket supports multi-threaded CPU PhysX

PhysX FluidMark is popular benchmarking application, that is often used to test stability and performance of GPU PhysX configurations. It performs PhysX SDK based SPH Fluids particle simulation, which can be calculated on CPU or compatible Nvidia GPU, however, only one CPU core can be used in first case.

After all those “Multi-Core CPU Support Is Disabled in PhysX” claims by AMD and following hype, JeGX (FluidMark developer) decided to leverage multi-threading capabilities of PhysX SDK and augment FluidMark with actual multi-core CPU support.

According to short preview, published today – task was successful.

As you may see, updated version of FluidMark, running on JeGX dev. machine with ATI Radeon HD 5770 GPU, is fully utilizing both cores of AMD X2 3800+ CPU.

Update: JeGX revealed another screenshot from upcoming FluidMark with multi-core PhysX support, running on all four cores of quad core Intel Core 2 Extreme X9650 CPU.

Previous version of multi-core FluidMark was able to load only dual-core CPU, but this, updated one – can utilize  quad-core or even n-core CPU.

Update #2: And now – FluidMark running on 16 CPU Cores.