« on: October 08, 2009, 09:16:18 PM »
Intel Corp. believes that GPU viruses are about to emerge and is taking steps to ensure that its forthcoming Larrabee is secure enough.
Full story at x-bit labs
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.
Intel Corp. believes that GPU viruses are about to emerge and is taking steps to ensure that its forthcoming Larrabee is secure enough.
Last Wednesday, AMD announced a partnership with Pixelux Entertainment to develop an open source physics initiative called Bullet Physics. The new API is being written in the vendor-neutral OpenCL and DirectCompute languages; that means games which use Bullet could run physics on ATI and NVIDIA cards alike.
Modern Graphics Processing Units (GPUs) are providing breakthrough performance for numerical computing at the cost of increased programming complexity. Current programming models for GPUs require that the programmer manually manage the data transfer between CPU and GPU. This thesis proposes a simpler programming model and introduces a new compilation framework to enable Python applications containing numerical computations to be executed on GPUs and multi-core CPUs. The new programming model minimally extends Python to include type and parallel-loop annotations. Our compiler framework then automatically identifies the data to be transferred between the main memory and the GPU for a particular class of affine array accesses. The compiler also automatically performs loop transformations to improve performance on GPUs. For kernels with regular loop structure and simple memory access patterns, the GPU code generated by the compiler achieves significant performance improvement over multi-core CPU codes.
While AMD and Nvidia battle for supremacy in the GPU computing arena, there's one obvious loser, Intel. AMD's 5870 appeared on schedule. Nvidia's Fermi is late, but its GTX 280 series still is competitive. Intel's Larrabee remains a no-show.
End users who buy their systems by the teraflop have discovered and validated an alternative approach that requires fewer x86 CPUs, less power, and less space. GPU computing is here to stay, and the market will punish those who lack a competitive offering.
Minimum system requirements
* Microsoft® Windows® Vista 32-bit/64-bit or Microsoft Windows 7 32-bit/64-bit
* Intel® Core™2 Duo or AMD Athlon™ X2 CPU or higher
* 1GB of system memory. (2GB is recommended)
* 100 MB free disk space
NVIDIA 3D Vision-Ready displays devices
* Desktop Displays
o Samsung® SyncMaster 2233RZ 120 Hz LCD display
o ViewSonic® FuHzion™ VX2265wm 120 Hz LCD display
o 100 Hz and higher analog CRTs
o Mitsubishi® 1080p DLP® Home Theater TV: WD-57833, WD-60735, WD-60737, WD-60C8, WD-60C9, WD-65735, WD-65736, WD-65737, WD-65C8, WD-65C9, WD-65833, WD-65835, WD-65837, WD-73735, WD-73736, WD-73737, WD-73833, WD-73835, WD-73837, WD-73C8, WD-73C9, WD-82737, WD-82837, L65-A90
o Generic DLP HDTV mode
o DepthQ® HD 3D Projector by LightSpeed Design, Inc.
o DQ-3120 by LightSpeed Design, Inc.
o ViewSonic PJD-6210-3D
o ViewSonic PJD-6220-3D
o ViewSonic PJD-PJD5351-3D
o ViewSonic PJD-PJD5111-3D
GPU3: Next generation GPU core, based on OpenMM. We have been making major advances in GPU simulation, with the key advances going into OpenMM, our open library for molecular simulation. OpenMM started with our GPU2 code as a base, but has really flourished since then. Thus, we have rewritten our GPU core to use OpenMM and we have been testing that recently as well. It is designed to be completely backward compatible, but should make simulations much more stable on the GPU as well as add new science features. A key next step for OpenMM is OpenCL support, which should allow much more efficient use of new ATI GPUs and beyond.
Release Notes - NVIDIA PhysX SDK 2.8.3
30th September 2009
What's New In NVIDIA PhysX 2.8.3
· This release enables clothing simulation. These new features are required to run the APEX clothing module
· The release also adds a hierarchical solver for cloth
· Introduction of split pair notification for cloth (existed for soft bodies only in previous versions)
· Improved statistics to track GPU memory utilization
· Updated with 64Bit Windows support
· The supported devices are the same as for PhysX 2.8.1 (see below), except that the AGEIA PhysX Processor (PPU) device support was removed
· The hierarchical cloth solver runs before the regular solver at each time step. This does not immediately improve cloth simulation performance. However, running the hierarchical solver achieves the same stiffness of the cloth with fewer solver iterations of the regular solver. Thus, the performance is increased by explicitly reducing the number solver iterations.
· Added support to run force field updates on the scene's worker threads. Can be enabled by raising the flag NxSceneFlags::NX_SF_MULTITHREADED_FORCEFIELD and by setting NxSceneDesc::internalThreadCount to the number of worker threads which will compute the force field updates besides the scene thread.
· NxArray::insert() did not work properly if elements were inserted anywhere else than at the end of the array.
· Crash when fluid tries to access invalid convexes. Mirrored convexes can get corrupted when a heap allocation fails.
· GPU fluids: Force and deletion updates are performed on wrong particles in case the updates address particles that have been added since the last execution of NxScene::simulate().
· GPU fluid mesh collision kernel crashes on GT200 cards due to heavily unbalanced workload.
· GPU fluids: Releasing static primitive rigid body shapes has no effect on particle collision.
· Concurrency issues with multi compartment scene configurations.
· Wrong contact normals in the mesh vs. capsule collision case could occur, if the capsule hit an edge shared by two mesh triangles. Please note that this fix may change the simulation behavior compared to previous versions of the SDK.
· Contact reports: Under certain circumstances, incorrect contact normals in the capsule vs. mesh case were provided when continuous collision detection was enabled.
Known Issues and Limitations
· NxScene::cookFluidMeshHotspot(...) is only supported for fluids configured to run on the GPU. Calling the method on a scene with just software (windows) fluids just has a memory impact and no benefit. There is no effect on consoles.
· The 64 bit PhysX SDK has an extra limitation on Windows Xp64 due to an Xp64 thread local storage bug: The maximal number of scenes/compartments supported can be derived as follows:
#NxScene*4 + #NxCompartment(SW)*2 + #NxCompartment(HW)*1 < (65 - 25)
Note that the compartments can be inspected using NxScene::getCompartmentArray(...), including default compartments.
· Please also see the previous lists from 2.8.1 and earlier.
· Hierarchical cloth
o NxClothMeshDesc.numHierarchyLevels - used for cooking a mesh hierarchy into the cloth mesh
o NxClothDesc.hierarchicalSolverIterations - how many iterations of the hierarchical solver to run at each time step
o NxCloth::setHierarchicalSolverIterations(NxU32 iterations)
o The constrain positions and normals are set at each time step via
§ void NxCloth::setConstrainPositions(void* buffer, NxU32 byteStride = sizeof(NxVec3)
§ void NxCloth::setConstrainNormals(void* buffer, NxU32 byteStride = sizeof(NxVec3))
o The cloth vertices then stay within a certain distance to the constrain positions. The constrain positions are typically given by the animation of the clothing using skeletal skinning.
o Coefficients like the maximum distance from the constrain positions are set using
§ void NxCloth::setConstrainCoefficients(const NxClothConstrainCoefficients *coefficients, NxU32 byteStride = sizeof(NxClothConstrainCoefficients))
§ class NxClothConstrainCoefficients
o In a typical scenario, the coefficients are set only once after the cloth is created and remain constant during the simulation. They can also be updated between time steps.
o The clothing functionality is also available for soft bodies. Use NxSoftBodyConstrainCoefficients instead of NxClothConstrainCoefficients in connection with soft bodies.
o NxSoftBodyVertexFlags::NX_SOFTBODY_VERTEX_SECONDARY can be set per vertex in NxSoftBodyMeshDesc::vertexFlags before the mesh is cooked. This flag is used to make the softboy behave like cloth. Only the vertices without this flag set are simulated. The secondary vertices follow the primary ones keeping the tetrahedral mesh in shape for skinning of multi layered clothing.
o Hard stretch limitation. This feature is enabled by setting the flag NX_CLF_HARD_STRETCH_LIMITATION. When the flag ist set, the solver keeps the maximal amount of stretchiness below a factor given by NxClothDesc.hardStretchLimitationFactor. The featurer only works when constrain positions and coefficients are provided. It is useful in connection with clothing simulation to make the cloth inextensible (limitation factor = 1.0). The method to achieve this is non-physical. It works best when the cloth is attached at the top and hangs down such as with skirts or capes.
o Untangling. When self-collision handling is turned off (for performance reasons), the cloth can locally entangle and remain in this state. When this NX_CLF_UNTANGLING is set, the simulator tries to untangle the cloth locally. This is done non-physically and can introduce ghost forces! The flag can be turned on and off during run-time but it has to be set when the cloth is created in order for the necessary data structures to get initialized. The feature is supported for cloth only
o Added selfCollisionThickness parameter. Use selfCollisionThickness for for self collision, thickness for collision with other objects.
o Added NX_CLF_INTER_COLLISION cloth flag, to enable cloth to collide with other cloth. (not implemented)
· Soft bodies
o Added selfCollisionThickness parameter. Use selfCollisionThickness for for self collision, particleRadius for collision with other objects.
o Added NX_SBF_INTER_COLLISION cloth flag, to enable soft bodies to collide with other soft bodies. (not implemented)
· Split pair notification
o class NxClothSplitPairData
· Added NxSceneFlags::NX_SF_MULTITHREADED_FORCEFIELD to enable the multi-threaded version of force field for fluids
· Added NxSceneFlags::NX_SF_ALTERNATIVE_FLUID_TRIANGLE_COLLISION to enable an alternative fluids triangle-mesh collision CUDA kernel which could be significantly faster than the default one in some scenarios
We just released Irrlicht version 1.6. As always, this version contains a lot of bugfixes and corrections, but also some new features:
* New mesh and image loaders (.PLY mesh loader, .RGB, .RGBA, .SGI, .INT, and .INTA textures)
* New Material properties ColorMask, Mipmap LOD, Depth test function, AlphaToCoverage, selective Anti-Aliasing
* New console device and possibility to create different devices from the same library
* OverrideMaterial for changing material properties globally
* New Filesystem with tighter integration of different archive types
* Improved Software rendering engine
...STALKER: Call of Pripyat screenshots that highlight the game's DirectX 11 graphics. The first two screenshots highlight GSC's use of tessellation to add more detail to the DX11 character. The last 8 shots focus on the game's DX11 contact hardened shadows, which are compared to DX10.1 shadows.
In the DX11 contact hardened shadows screenshots (images 04, 06, 08, 10), you can see the sharper shadow edges where they contact the object that's casting the shadow. The shadow edge gets blurrier further away from the object.
NVIDIA today announced work with Microsoft to promote NVIDIA Tesla graphics processing units (GPUs) for high performance parallel computing using the Windows HPC Server 2008 operating system.
What’s New in v2.0-beta3
* ATI Stream SDK v2.0 OpenCL™ for x86 CPUs is now certified conformant by Khronos on September 3, 2009.
* 64-bit atomic built-ins are now recognized.
* Fixed some compiler assertion issues.
* Fixed some incorrect compiler warnings.
* Fixed some incorrect compiler assertion failures.
* Fixed some code generation issues.
* X86 CPU w/ SSE 3.x or later
We have a real-time task distribution engine. The frame buffer stream is being analyzed and checked on-the-fly. By approaching graphics on computational level we resolved frame dependency.
List of the new fonctionnalities :
* Add 3DXML V3 and V4 Ascii file format.
* Add Collada V1.4 file format (Mesh, structure and materials properties).
* Display axis to show view orientation.
* Posibility to chose between "Track Ball" and "Turn Table" navigation mode.
* Display the bounding box size in the view "Instance properties"
This is a beta driver for GeForce 6, 7, 8, 9, 100, and 200-series desktop GPUs and ION desktop GPUs.
New in Version 191.03
* Adds support for OpenGL 3.2 for GeForce 8, 9, 100, and 200-series GPUs and ION GPUs.
* Accelerates performance in several gaming applications. The following are examples of improvements measured with version 191.03 drivers vs. version 190.62 drivers (results will vary depending on your GPU, system configuration, and game settings):
o Up to 12% performance increase in ARMA 2
o Up to 8% performance increase in Batman: Arkham Asylum with GPU PhysX enabled
o Up to 50% performance increase in Call of Juarez: Blood in Bound with SLI enabled
o Up to 14% performance increase in Fallout 3 (indoor scenes) with antialiasing enabled
o Up to 10% performance increase in Far Cry 2 (DX9 version) with antialiasing enabled
o Up to 34% performance increase in Prototype with antialiasing enabled
* Adds SLI support for Darkfall, Dawn of Magic 2: Time of Shadows, Dreamkiller, Fuel, Majesty 2: The Fantasy Kingdom Sim, Need for Speed: Shift and more.
* Includes numerous bug fixes, including the following key fix (additional bug fixes can be found in the release notes on the documentation tab): For graphics cards supporting multiple clock states, 3D clocks correctly return to 2D clocks after exiting a 3D application.
New in Release 190/191 Drivers
* Installs PhysX System Software version 9.09.0814.
* Supports CUDA 2.3 for improved performance in GPU Computing applications. See CUDA Zone for more details.
* Supports single GPU and NVIDIA SLI technology* on DirectX 9, DirectX 10 and OpenGL.
* Supports NVIDIA PhysX acceleration on a dedicated GeForce graphics card. Use one card for graphics and dedicate a different card for PhysX processing for game-changing physical effects. Learn more here.
* Supports GeForce Plus Power Pack #3. Download these FREE PhysX and CUDA applications now!
* Supports GPU overclocking and temperature monitoring by installing NVIDIA System Tools software.
*Note: The following SLI features are only supported on Windows Vista and Windows 7: Quad SLI technology using GeForce 9800 GX2 or GeForce GTX 295, 3-way SLI technology, Hybrid SLI, and SLI multi-monitor support.
GPUs or graphics cards are specialized chips that run inside PCs to display 2D and 3D graphics. They tend to contain hundreds of floating point processors that are good at handling mathematically intensive and parallel processes such as Monte Carlo simulations