Menu Content/Inhalt
Home

Compucon HPC for NAMD Print
November 2012

NAMD is a world class software program for molecular dynamic studies created in 1995 and maintained by the University of Illinois.  It is designed for parallel computation to study the interactive behaviour of as many as 100 million molecules.  NAMD stands for “Not Another Molecular Dynamic” program.  See Slide 1.  Nvidia listed NAMD among 8 other well know molecular dynamic programs on its website as in November 2012.  

(Hover Mouse over to Enlarge slides)




NAMD uses the GPU for non-bonded force evaluation only whereas energy evaluation is done on the CPU.   The CUDA code in NAMD is completely self-contained.

We tested NAMD v2.9 with its benchmarking sample in length in our technical service workshop in October 2012 with 4 desktop systems based on X79 and Z77 platforms, and verified the information given on a chart shown on Nvidia website.  The chart shows that one Tesla is equivalent to 4 Quad Core CPU for running STMV with NAMD.  STMV stands for Simulated Test Molecular.  See Slide 2.

In brief, one Tesla GPU is equivalent to 4 Quad Core CPU.  Our tests used Tesla C2075 with 448 cores, Pentium PG630 with 2 cores, Core i3 3220 with 4 Hyper-threads (2 cores), Xeon E3-1230V2 with 8 threads, and Core i7 3930K with 12 threads.  We have been able to confirm that the variations of CPU frequency, cores, and generation have some effect on the NAMD computation process; but the presence of Tesla has a much more significant effect than additional CPU cores or frequency or generation.   We ran NAMD on single precision floating point only and with Non ECC memory.  

The significance of our verification is that a Compucon HPC workstation fitted with a Tesla card can run NAMD like two Dual Xeon servers in terms of time to compute, and a Compucon HPC 4U workstation is equivalent to 8 Dual Xeon servers.  A research scientist can afford to have a Compucon HPC workstation at his desk and this was virtually impossible a year ago.

Like most other parallel processing application programs, NAMD runs on Linux and our tests were based on Ubuntu.  Visualisation of the NAMD process needs a separate program called VMD.  The visualisation we have obtained is simplified. See slide 3.

Please contact us for more info.  Click this: Contact Us

 
Edge Storage of IP Cameras Print
November 2012

Edge Storage or Edge Recording refers to the storage of camera video footage including snapshots and video clips in the same camera that generates the footage.  It is called edge as against the standard arrangement of central storage.  The storage medium for edge is Micro SD card with physical dimensions of 15mm x 11mm and a thickness of 1mm.  See Slide 1.  A camera with its own storage would not need an external storage for footage such as a Network Video Recorder (NVR) if the latter is not economical or feasible.  This article examines how Edge Storage operates and its real life applications and desirable arrangements. 

(Hover mouse over to enlarge)


   

Standard NVR uses hard disk drives (HDD) for storage of video footage.  HDD with a form factor of 3.5” and a SATA interface has a large storage capacity such as 2 Terabytes (TB) or more as of November 2012 and is inexpensive.  Micro SD card is small in physical size and rigid storage capacity.  We have 32GB as the largest capacity at present.  Fortunately, we can allocate the whole lot of 32GB to one camera whereas the HDD storage of 2TB will be shared by 10 or whatever count of cameras supported by the NVR. 

Micro SD card is not designed for high volume or high speed recording.  It is used for camera edge storage mainly for its small size and low cost.   As such, we have to reduce our expectation of the quality of the recording from edge storage.  For a 4 mega-pixel camera equipped with H264 compression, we should be able to achieve 2 frames per second on edge whereas it would be 12 frames per second with central storage. 

Video footage is meant for review.  In order to review the footage recorded at the edge, we will need to connect the camera to a PC or server via a standard IP network cable unless the camera has been equipped with wireless transmission.  This is the requirement at least for setting up the camera initially before deployment.  The footage files will be in a standard format for snapshots and could be in a proprietary format for video clips.  We will supply a software program to play back the video clips in a PC running Windows.  Files would have been set up to have date and time as part of the filename for ease of reference.   See Slide 2.  It is not possible to obtain live views of the camera without connecting the camera to an IP network full time.

While edge camera is easy to install and low cost, it does not enjoy the benefits offered by a centralised NVR.   Even if edge is connected with IP full time, we can review the footage on a file basis only and not in synchronization with other cameras with a graphical user interface as offered by NVR.  There are indeed situations where centralised storage is too expensive to achieve and edge is an acceptable compromise.

We can also use edge as a back up of centralised storage.  If we store snapshots only in edge, we will have footage for a long period such as one or two months.  These snapshots could be valuable in case the NVR does not support such a long recording period or when the NVR breaks down when it is needed for police investigation.

As of November 2012, the Compucon stable has 4 models with edge storage.  They are K3911, K5211E, K7111 and K7311.   See Slide 3.

 

 

 
Edge Storage of IP Cameras Print
November 2012

Edge Storage or Edge Recording refers to the storage of camera video footage including snapshots and video clips in the same camera that generates the footage.  It is called edge as against the standard arrangement of central storage.  The storage medium for edge is Micro SD card with physical dimensions of 15mm x 11mm and a thickness of 1mm.  See Slide 1.  A camera with its own storage would not need an external storage for footage such as a Network Video Recorder (NVR) if the latter is not economical or feasible.  This article examines how Edge Storage operates and its real life applications and desirable arrangements. 

(Hover mouse over to enlarge)


   

Standard NVR uses hard disk drives (HDD) for storage of video footage.  HDD with a form factor of 3.5” and a SATA interface has a large storage capacity such as 2 Terabytes (TB) or more as of November 2012 and is inexpensive.  Micro SD card is small in physical size and rigid storage capacity.  We have 32GB as the largest capacity at present.  Fortunately, we can allocate the whole lot of 32GB to one camera whereas the HDD storage of 2TB will be shared by 10 or whatever count of cameras supported by the NVR. 

Micro SD card is not designed for high volume or high speed recording.  It is used for camera edge storage mainly for its small size and low cost.   As such, we have to reduce our expectation of the quality of the recording from edge storage.  For a 4 mega-pixel camera equipped with H264 compression, we should be able to achieve 2 frames per second on edge whereas it would be 12 frames per second with central storage. 

Video footage is meant for review.  In order to review the footage recorded at the edge, we will need to connect the camera to a PC or server via a standard IP network cable unless the camera has been equipped with wireless transmission.  This is the requirement at least for setting up the camera initially before deployment.  The footage files will be in a standard format for snapshots and could be in a proprietary format for video clips.  We will supply a software program to play back the video clips in a PC running Windows.  Files would have been set up to have date and time as part of the filename for ease of reference.   See Slide 2.  It is not possible to obtain live views of the camera without connecting the camera to an IP network full time.

While edge camera is easy to install and low cost, it does not enjoy the benefits offered by a centralised NVR.   Even if edge is connected with IP full time, we can review the footage on a file basis only and not in synchronization with other cameras with a graphical user interface as offered by NVR.  There are indeed situations where centralised storage is too expensive to achieve and edge is an acceptable compromise.

We can also use edge as a back up of centralised storage.  If we store snapshots only in edge, we will have footage for a long period such as one or two months.  These snapshots could be valuable in case the NVR does not support such a long recording period or when the NVR breaks down when it is needed for police investigation.

As of November 2012, the Compucon stable has 4 models with edge storage.  They are K3911, K5211E, K7111 and K7311.   See Slide 3.

 

 

 
AMD A-Series (A10) Processor Print
November 2012
AMD A-series processor features include: AMD A10
  • Manufactured on GlobalFoundries' 32 nm SOI process for Socket FM2
  • Richland (3rd gen) and Trinity (2nd gen) are upgrades to the Bulldozer microarchitecture known as Piledriver
  • Die Size: 246 mm2, 1.303 Billion transistors
  • Support for up to 4 DIMMs of up to DDR3-1866 (2133 with 6800K ) memory, with 6.4 GT/s UMI.
  • GPU (based on VLIW4 architecture) instruction support: DirectX 11, Opengl 4.2, DirectCompute, Pixel Shader 5.0, Blu-ray 3D, OpenCL 1.2, AMD Stream, UVD3
  • All models feature an integrated PCIe 2.0 controller, and AMD's Turbo Core technology for faster CPU/GPU operation when the thermal specification permits
  • All models support: MMX, SSE, SSE2, SSE3, SSSE3, SSE4a, SSE4.1, SSE4.2, AMD64, AMD-V, AES, CLMUL, AVX 1.1, XOP, FMA4, CVT16
  • Select models support Hybrid Graphics technology to assist a Radeon HD 7350, 7450, 7470, 7550, 7570, 7670 discrete graphics card.

 
"Trinity" (32 nm)

AMD A10-5800K processor feature improvements over previous generations include:

  • Quad-core accelerated performance
  • Up to 4.2 GHz Max Turbo
  • 384 AMD Radeon cores 2.0 for blazing fast computing
  • Unlock your potential by selecting the overclockable Black Edition
  • 45% more graphics performance than 1st Gen AMD A8 APUs
  • 26% better system performance than 1st Gen AMD A8 APUs

"Richland" (32 nm)

AMD A10-6800K processor feature improvements over previous generations include:

  • "Enhanced Piledriver" CPU cores
  • On-die Radeon HD 8000 Series (not Graphics Core Next) graphics chipset
  • Temperature Smart Turbo Core technology. An advancement of the existing Turbo Core technology, which allows internal software to adjust the CPU and GPU clock speed to maximise performance within the constrains of the Thermal design power of the APU.

Model Number
Step
CPU GPU
Memory Support
TDP
Released
Cores Freq.
Turbo
L2 Cache
Multiplier
Voltage
Model
Freq.
 Processor - Trinity
A10-5800K B0 4 3.8 GHz 4.2 GHz 2 × 2 MB 38-42× 0.925 - 1.475V HD 7660D 800 MHz DDR3-1866 100 W Oct 1, 2012
 Processor - Richland
A10-6800K RL-A1 4 4.1 GHz 4.4 GHz 2 × 2 MB 10-44x
0.9 - 1.4375V
HD 8670D 844 MHz DDR3-2133 100 W June 4, 2013


Click here to return
 
AMD A-Series (A10) Processor Print
November 2012
AMD A-series processor features include: AMD A10
  • Manufactured on GlobalFoundries' 32 nm SOI process for Socket FM2
  • Richland (3rd gen) and Trinity (2nd gen) are upgrades to the Bulldozer microarchitecture known as Piledriver
  • Die Size: 246 mm2, 1.303 Billion transistors
  • Support for up to 4 DIMMs of up to DDR3-1866 (2133 with 6800K ) memory, with 6.4 GT/s UMI.
  • GPU (based on VLIW4 architecture) instruction support: DirectX 11, Opengl 4.2, DirectCompute, Pixel Shader 5.0, Blu-ray 3D, OpenCL 1.2, AMD Stream, UVD3
  • All models feature an integrated PCIe 2.0 controller, and AMD's Turbo Core technology for faster CPU/GPU operation when the thermal specification permits
  • All models support: MMX, SSE, SSE2, SSE3, SSSE3, SSE4a, SSE4.1, SSE4.2, AMD64, AMD-V, AES, CLMUL, AVX 1.1, XOP, FMA4, CVT16
  • Select models support Hybrid Graphics technology to assist a Radeon HD 7350, 7450, 7470, 7550, 7570, 7670 discrete graphics card.

 
"Trinity" (32 nm)

AMD A10-5800K processor feature improvements over previous generations include:

  • Quad-core accelerated performance
  • Up to 4.2 GHz Max Turbo
  • 384 AMD Radeon cores 2.0 for blazing fast computing
  • Unlock your potential by selecting the overclockable Black Edition
  • 45% more graphics performance than 1st Gen AMD A8 APUs
  • 26% better system performance than 1st Gen AMD A8 APUs

"Richland" (32 nm)

AMD A10-6800K processor feature improvements over previous generations include:

  • "Enhanced Piledriver" CPU cores
  • On-die Radeon HD 8000 Series (not Graphics Core Next) graphics chipset
  • Temperature Smart Turbo Core technology. An advancement of the existing Turbo Core technology, which allows internal software to adjust the CPU and GPU clock speed to maximise performance within the constrains of the Thermal design power of the APU.

Model Number
Step
CPU GPU
Memory Support
TDP
Released
Cores Freq.
Turbo
L2 Cache
Multiplier
Voltage
Model
Freq.
 Processor - Trinity
A10-5800K B0 4 3.8 GHz 4.2 GHz 2 × 2 MB 38-42× 0.925 - 1.475V HD 7660D 800 MHz DDR3-1866 100 W Oct 1, 2012
 Processor - Richland
A10-6800K RL-A1 4 4.1 GHz 4.4 GHz 2 × 2 MB 10-44x
0.9 - 1.4375V
HD 8670D 844 MHz DDR3-2133 100 W June 4, 2013


Click here to return
 
<< Start < Prev 111 112 113 114 115 116 117 118 119 120 Next > End >>

Results 1027 - 1035 of 2511