Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Stefan

Pages: 1 ... 128 129 [130] 131 132 ... 160
With vReveal, you can stabilize, brighten, and sharpen your videos with just one click, for free. Plus much more.

 Minimum System Recommendations:

    * Operating System: Windows XP, Vista, 7
    * Processor: Dual-core CPU or better
    * Memory: 2GB RAM
    * Hard Drive: At least 60MB of free space

Minimum System Requirements for Optional GPU Acceleration:

    * GPU: NVIDIA GeForce with CUDA: 8-series, 9-series, 200-series, 400-series
    * GPU Memory: 256MB or higher for processing up to 720p video, 512MB or higher for 1080p video
    * Driver: 197.13 or newer

 Version 2.0.6349 - April 13, 2010

    * Now accepts input of HD video (up to 1080p)
    * Replaced 30-day trial with a free version of vReveal
    * Made super-resolution enhancement available for videos with an input resolution up to 576p
    * Added new video fix: Auto White Balance
    * Added special effects to a new "Add Effects" tab under "Enhance," including Black & White, Sepia, Vivid, Glow, Grain, Vintage, and Colorize
    * Optimized CPU performance, including addition of support for SSE4 and improved multi-threading
    * Faster launch time and Gallery scrolling
    * Introduced support for Windows 7 and better support for 64-bit systems
    * Fixed various bugs
    * Implemented new, more-user-friendly e-commerce and product activation systems

I wonder why nobody used SSE4 for video encoding before   ::)

Quick test with Blender's "Bick Buck Bunny": 7-8 fps with CUDA enabled (8800GTX)

If you're coding for multicore, you shoot for maximum performance but hit barriers along the way. Arm yourself with the latest development and optimization tips from the experts: Download the Guide for Developing Multithreaded Applications: Part 1.
   Part 1 of this technical guide focuses on application threading and synchronization techniques. A compilation of 11 top articles, it features examples and how to's on measuring performance, removing artificial dependencies, using tasks instead of threads, mitigating synchronization impacts, and more. Here's just a sampling of the articles inside:

•       Loop Modifications to Enhance Data-Parallel Performance—Tackle loop challenges with techniques that minimize synchronization and make parallelization easier—such as loop fusion, loop interchange, and loop unrolling.
•       Load Balance and Parallel Performance—Avoid common memory and I/O pitfalls to achieve the perfect load balance by sharing workloads equally across all threads.
•       Managing Lock Contention: Large and Small Critical Sections— Efficient parallel code balances thread performance in critical sections and manages idle time of other threads. Learn the concept of critical section size and see when it is best to use large or small critical sections.

Watch for Part 2 of this Guide next week for 10 additional articles on varied memory management approaches for multicore and the latest programming tools to streamline parallel programming.

3D-Tech News Around The Web / Flam4CUDA 0.85 & Flam4OCL 0.14 available
« on: April 15, 2010, 10:14:24 AM »
GPU Flame Fractal Renderer for flam3 cosmic recursive fractal flames implemented on GPU.

Flam4CUDA v.0.85
-Updated to Cuda 3.0 final.  Be sure to update your drivers!
-Fixed RNG bug introduced in 0.83

Flam4OCL Version 0.14
-Finished adding the rest of the flam3 vars.  Flam4OCL is now complient with all the flam3 2.8 variations, though a few still aren't working properly
-Moved opacity from being xform_node specific to being xform_node link specific.  This means that it's now specified in all the node references while being unused in the trunk level nodes, just like weight.  See the examples for more details.
-Expanded the behavier of state.  Now push always means to feed the point into the next xform, whereas pop always discards the point.  You can push -5 to go down in the stack 5 entries but still save the current point.  Similarily, you can pop -1 to move one space up the stack, but still discard the current point.  This lets you create composites of flames, each with its own section of the stack, and using pop to switch back and forth.

Unfortunately Flam4OCL still refuses to run on 8800GTX, Flam4CUDA works fine

3D-Tech News Around The Web / ATI Catalyst 10.4 Preview available
« on: April 15, 2010, 01:03:30 AM »
The ATI Catalyst 10.4 Preview provides resolution for the following issue:

    * Long load times for new maps in Battlefield: Bad Company 2 when using an ATI Radeon HD 5xxx series graphics card

CUDA 3.0 is a major revision number release that delivers important benefits for debugging; C++; OpenCL; CUDA driver and runtime developers; CUBLAS and CUFFT users, and many other areas that make this a "must install" update. The expanded capabilities discussed in this brief two-part article series should provide outstanding food-for-thought how the 3.0 release creates new opportunities for code development and integration of existing software projects.

Full story at DDJ

An extreme gaming computer with the Intel Core i7-980X Extreme Edition -- (codenamed Gulftown) featuring six cores and 12 threads -- was demonstrated at the Intel Developer Forum in Beijing in April 2010.

Watch video at Youtube


This video shows DirectX 11 and Tessellation features in popular games, and how DirectX 11 these technologies can dramatically improve the quality of the image! Learn more about the latest and fastest DirectX 11 hardware at EVGA.

Games/applications shown:
*Battlefield Bad Company 2
*Metro 2033
*NVIDIA Supersonic Sled
*Unigine Heaven 2.0
*Bitsquid Stone Giant
*Aliens vs. Predator
*Dirt 2

3D-Tech News Around The Web / OpenGL Samples Pack released
« on: April 13, 2010, 08:00:28 PM »
The OpenGL Samples Pack is available and it catches up with some of the main features of OpenGL: Uniform Buffer, transform feedback and instanced array.

Full story at g-truc

3D-Tech News Around The Web / Free-to-use CryEngine plans emerge
« on: April 13, 2010, 04:14:02 PM »
Frankfurt-headquartered Crytek may be about to compete with both Unity and Epic Games on the emerging battleground of free-to-use engines

Full story at develop

3D-Tech News Around The Web / Yet another GTX 480 test incl. Furmark
« on: April 12, 2010, 05:41:23 PM »
HT4U built some custom hardware to measure the power drain of the GPU only. Actually they tested a bunch of contemporary graphic cards.

GTX 480 power drain:
Idle: 48.71 Watt
Tom Clancy's HAWX DX9:  230.4 Watt
Stalker CoP: 239.86 Watt
Heaven 2: 242.6 Watt
NVIDIA SuperSonic-Demo: 262.70 Watt
3DMark06: 283.16 Watt
Furmark: 317.84 Watt

According to HT4U NVIDIA is aware that Furmark needs much more than 250 Watt and calls it "not representative".

It appears that the "Palit GeForce 9800 GT Green" cannot stand Furmark load.

The company is announcing the launch of the three products on Monday at a Visual Studio developer conference in Las Vegas. Visual Studio 2010 and the updated .Net Framework are now broadly available, Microsoft said, while Silverlight 4 will be made available for download later this week.

Full story at cnet

Dr. Dobb's Journal tells you What's New in Visual Studio 2010 & .NET 4 for Web Developers

ComputeMark is freeware. This app benchmarks your DirectX 11 graphic card Compute Shader performance.

Based on Fluid3D demo by Jan Vlietinck

- Windows 7 32/64-bit
- DirectX 11
- DX11 graphic card

Download: ComputeMark (version: 1.0) - hosted at CzechGamer

Made by Robert Varga (c) 2010


"I think graphics matter a lot," Yerli asserts. "It depends on the type of game. If graphics and technology, AI and physics provide you a better experience but a challenge, then ultimately it does matter. You can get an experience out of Crysis 2 you just don't get anywhere else."

Full story at Gamasutra

Crisys 2 official website has a new trailer, also on Youtube

3D-Tech News Around The Web / Jitter - Fully managed physic engine
« on: April 09, 2010, 11:19:49 AM »
Hi, this are some screenshots of a physic engine for .NET Iam working at for some month now - called 'Jitter'. It's my first attempt to make a physic engine (JigLibX was my first engine, but it's just a port of a c++ engine).

It uses accumulated impulses for the solver and mpr (xenocollide) for collision detection. It currently uses some code of Bullet for raycasting - I didn't had the time to implement that myself.

It's entirely written in C# and does not depend on the XNA framework - using it's own math classes. (The demo uses XNA). It's highly optimized for C# and pretty lightweight.

At the moment it does all the basic things a physic engine should do (Octree based trianglemesh collisions, constraints, stable simulation..). I hope to implement vehicle and cloth-simulation later.

You can download a compiled demo version here:


3D-Tech News Around The Web / K-3D available
« on: April 09, 2010, 10:58:23 AM »
K-3D is free-as-in-freedom 3D modeling and animation software. It features a plugin-oriented procedural engine for all of its content, making K-3D a very versatile and powerful package.
K-3D excels at polygonal modeling, and includes basic tools for NURBS, patches, curves and animation. Give it a try!

  K-3D 0.8 Release Notes

    * Completely rewritten data storage for geometric meshes - good for rendering one million polygons at interactive rates.
    * Introductory support for the LuxRender render engine.
    * Experimental support for CSG operations.
    * Introductory support for NURBS.
    * Experimental material editor.
    * New / rewritten support for 3DS, Collada, OBJ, MD2, Ogre, PLY, STL, and SVG files.
    * Interactive Python Shell
    * New CMake-based cross-platform build system.
    * Over 300 regression tests run every night.
    * Nightly coverage testing.
    * New Mercurial-based version control.
    * New website.
    * New time-based release schedule - the next K-3D release will be October, 2010.

>Fast download mirror<

3D-Tech News Around The Web / GLEW 1.5.3 available
« on: April 09, 2010, 09:39:20 AM »
The OpenGL Extension Wrangler Library

Change Log

    * 1.5.3 [02-28-10]
          o New extensions:
                + GLX_INTEL_swap_event
                + GL_AMD_seamless_cubemap_per_texture
                + GL_AMD_shader_stencil_export
          o Bug fixes:
                + Correct version detection for GL 3.1 and 3.2
                + Missing 3.1 enumerants
                + Add glew.pc

NVIDIA® OptiX™ ray tracing engine beta download

Changes from version 2.0.0 B3
  Additional documentation for interop added to the OptiX reference manual.
  Programming guide has been updated.
  Slight modifications of the OptiX headers to remove any dependence on the CUDA run time if
using these headers.  You should not need to update your code.
  Added simpleGLTexInterop sample to demonstrate how to use the new texture interop
functionality of OptiX.
  Mouse interactions with sutil’s Mouse class will now ignore interactions that result is setting the
camera with NaNs or Infs.
  Added support for non-affine transforms in Transform nodes.
  Some fixes for the new GL interop functions.
  Fixed problem with memory fragment errors when using many acceleration structures.
  OptiX shared libraries now depend on 3.0 release version of the CUDA cudart library.

Known limitations with version 2.0.0 Beta 4:
  OptiX will choose to ignore older devices if a SM 2.0 device (i.e. “Fermi” class) is present in the
  The Lbvh acceleration structure builder is known to fail in single GPU Linux systems and while
using multiple GPUs in all OSes.  Please use a different builder for the time being.
  There currently is a concurrent texture limit of 128 textures. This limit will be either increased, or
entirely removed in the future, although large numbers of textures will always be likely to
negatively impact performance.  An error is returned by OptiX if this limit is exceeded.
  Texture arrays and mip maps are not yet implemented.
  Applications that use RT_BUFFER_INPUT_OUTPUT or RT_BUFFER_OUTPUT buffers on
multiGPU contexts must take care to ensure that the stride of memory accesses to that buffer is
compatible with the PCIE bus payload size.  Using a buffer of type RT_FORMAT_FLOAT3, for
example, will cause a massive slowdown; use RT_FORMAT_FLOAT4 instead.
  Linux only: due to a bug in GLUT on many Linux distributions, the SDK samples will not restore
the original window size correctly after returning from full-screen mode.  A newer version of
freeglut may avoid this limitation.
  Performance on Windows Vista and Win 7 should be expected to be somewhat slower than XP
due to the inherent nature of these operating systems 

There's no denying that NVIDIA needs to bring Fermi down to the mid-to-high-end space, and early rumours suggest it'll happen in less than two months.

Full story at Hexus

3D-Tech News Around The Web / GeForce GTX 480 Supports 4-way SLI?
« on: April 08, 2010, 04:24:15 PM »
Soon after EVGA laid its hands on stocks of GeForce GTX 480 graphics cards, it claims from an internal test that 4-way SLI is possible on the reference design GTX 480 graphics cards.

Full story at TPU

Pages: 1 ... 128 129 [130] 131 132 ... 160