|
July 2011 |
|
Our previous understanding that the CPU is the central processing unit is going to change. More and more software applications respond to GPU more than to CPU these days and the trend may follow an exponential curve. GPU is graphical processing unit. It was previously a subsystem of a computer system solely responsible for the graphical display function of a computer. That is not true now.
The above change was brought about by the advents of CUDA and Stream. CUDA is an Nvidia creation and Stream belongs to AMD. Nvidia and AMD are the 2 leaders in GPU technology. Some twenty years ago, they would be competing among 100 vendors to produce graphics adaptor cards and fight for market share. Today, they have created this duopoly for themselves. CUDA is by far more popular than Stream as CUDA is taught in over 350 universities across the globe. CUDA stands for Compute Unite Device Architecture. It is a scheme for uniting the processing capabilities of CPU and GPU and the implementation is through hardware and software.
For any application to take advantage of hardware with CUDA cores, it must have CUDA application programming interface. Furthermore, the interface must be consistent with the version of the hardware cores. Many commercial off the shelf (COTS) software packages have been tuned to CUDA but far more have not been. Those packages not tuned to CUDA will continue to respond to CPU capabilities and not to GPU. Those that are fine tuned to CUDA will respond to GPU better than CPU.
We have done a lot of performance tests in our workshop on various COTS packages with CPU and GPU variations. Some packages jumped in performance when we replaced Quadro Fermi 600 GPU with 2000. These packages responded slightly more when we replaced a Core i3 CPU with Core i7, and did not respond at all if the main memory was increased from 4GB to 8GB (on Windows 64bit).
At the same time, we have found many COTS packages that responded to CPU and not to GPU.
A good tool for revealing the above is called SPECviewperf11 but this tool is limited to 8 popular CAD or rendering software packages only. See this page to learn about it. http://www.compucon.co.nz/content/view/847/254/
AutoCAD from AutoDesk and Premiere Pro Creative Suite from Adobe are popular software packages. They have their own tools for benchmarking hardware performance. We have done tests with these tools for our internal reference. We are able to tell which hardware responds and which does not.
Feel free to contact
This e-mail address is being protected from spam bots, you need JavaScript enabled to view it
on any issues you may like to discuss.
|
|
|
July 2011 |
|
Surveillance camera technology has continued to improve and we will soon introduce a new model of cameras with a resolution of 4 megapixels to extend the capabilities of Internet Protocol (IP) based video surveillance systems. We must not see this camera capability independently and must consider it in a system perspective.
Many webcams have 1 to 2 megapixels. Most still cameras have 4 to 10 megapixels. We have heard of 50 megapixels too. So what is special about 4 megapixels for a surveillance camera? It is special and indeed a challenge for such a camera to transmit live footage over wire or wireless to a different location for display and storage, and for such footage to be ready for review after 7, 14 or 140 days. The challenge is to preserve the clarity of the footage over distance and time.
Suppose the camera transmits live views at 10 frames per second and we assume 1 pixel requires 8 bits to define its luminosity and colour, we will get 320 megabits or 80 megabytes per second (MB/s) of traffic for handling. Suppose there are 10 such surveillance cameras on the campus, we will be dealing with 800 MB/s of video. One day has 24 hours, one hour has 60 minutes, one minute has 60 seconds, and we will be dealing with 7TB of data per day or 49TB per week. This numerical workload is a challenge but it is not yet the real challenge.
The real challenge is the generation of footage with a reproducible resolution of 4 megapixels. This is a combination of camera lens, sensor, digital processing, compression, and transmission issues. System A with Brand B of 4 megapixel cameras is totally different to System C with Brand D of 4 megapixel cameras.
Some people think Megapixel IP cameras have over promised on their clarity advantage over standard analogue cameras that are about half a megapixel in resolution. This is understandable because 4.0 divided by 0.5 should give us a factor of 8. Our biological eyes are not as numerical. We may say IP is clearer but are reluctant to say it is 8 times better.
Furthermore the above scenario has assumed that the camera and the entire process of opto-electrical conversion, digital processing, compression, and transmission have taken place at 100% fidelity before the footage arrives in the computer. We can assume so at a lower megapixel level but the accuracy will lose as the megapixel increases due to imperfection of design and manufacturing processes.
We can apply the TV Lines counting technique to measure the clarity of footage, but it is a laborious process that can be done mainly by camera vendors and well equipped laboratories. For the rest of the population, we have to rely on trust and track records. The purpose of this article is to inform as well as to set the level of expectation. 4MP does not fit all situations. Feel free to contact the
This e-mail address is being protected from spam bots, you need JavaScript enabled to view it
for discussion.
|
|
|
July 2011 |
We will look at the results of 2 surveys. The first one was carried out by New Zealand Statistics department in 2011 on our ICT industry. The second one was carried out by Australian Reseller News at the turn of the year on the IT reseller channel in Australia. We can learn from these surveys what our business customers expect from us and how we can prepare our operation better. However, we have to be aware if our learning is based on illumination or is just a reinforcement of our prior views. A drunken person uses the lamp post for support more than for illumination.
We will attempt to explain that most of us are small players. Small players have natural advantages and limitations. We quoted 3 well known concepts for illumination. The first one was based on the E-myth books written by Michael Gerber. He pointed out that a successful business needs 3 personalities. The second one was based on the book "Inside the Tornado" written by Geoffrey Moore. He prescribed the best approach for dealing with technologies at various phases of their adoption lifecycle. The third one was based on a book called the 7 effective habits of successful leaders written by Steven Covey. The key point is that only mature people know that we are inter-dependent.
The slides attached here were extracted from the seminar held on 29 June 2011. Feel free to contact
This e-mail address is being protected from spam bots, you need JavaScript enabled to view it
for your feedback or any discussions.
Presentation:
|
|
|
July 2011 |
Computer system platforms for professional and serious users are still our bread and butter for the foreseeable future. We have planned to evolve the DXT series of Dual Xeon servers and workstations from Tylersburg to Romley in October timeframe. Similarly we will evolve Superhawk Plus from X58 chipset to X79 skipping Z68.
AMD will introduce Llano Accelerated Processors in July. We will explain how AMD positions these newly designed processors against Intel. We will evolve Thunderbird from 880G chipset to A75 chipset for Llano processors.
We have a 2U Twin2 model in our server range for a while, and have planned to introduce 2 new models based on the same concept. The first new model has 8 Atom systems in 2U and the second new model has 8 Xeon systems in 3U.
Tesla high performance computing systems are also on our roadmap. It has propelled a HPC system called "Sky River" (our translation) based in China to be the Top Supercomputer in the world on 1 November 2010 and has again smashed a world computing performance record on AMBER for Bio-molecular simulation study in May 2011.
IPVS (IP video surveillance) is based on our belief that “IP is Everything” and is our entry to a separate blue ocean for end to end solutions. We have just won a new project for a honey farm and this is the second major win for our channel peers. We believe our level of competence will open this blue ocean for more peers over time. Our winning technologies are Storage System design and a high powered brick lookalike wireless bridge.
Please see the slides presented to Compucon channel peers on 29 June 2011. The slides are commercially sensitive and confidential and are not meant to be circulated to the public.
Presentation:
|
|
|
June 2011 |
|
This is one of the themes of the Tech Briefing event held on Wednesday 29 June 2011. "Inside the Tornado" is the title of a book written by Geoffrey Moore in 2004. It explains the behaviour of customers at different phases of a technology product adoption lifecycle and prescribes the best ways for vendors and suppliers to adjust their business approach to match. The concept provides useful hints for us all.
"Inside the Tornado" refers to the phase that we normally call “early majority” and it is the rising slope of the bell curve. "Main Street" refers to the phase that we normally call “late majority” and it is the declining slope.
There is a big gap between the early adopters and early majority. In order for a new technology product or service to be accepted in an en mass style, it must “cross the chasm” first. The book prescribes the bowling alley analogy for crossing the chasm. There are 10 pins to aim at. Consider the first pin to be the beachhead. Once we have acquired the beachhead, we shall attempt to acquire the next two pins. One pin is to sell a new product to the same customer. The other pin is to sell the same product to a new customer. It is a binary mode of business progression. This sounds logical and easy to put into action. In reality, we should always apply this Bowling Alley approach all the times.
The approach for operating in the Main Street is to open boutiques. In technical terms, we shall extend our products from general platforms to specific end applications.
The lot are good points for sharing. (Feedback to
This e-mail address is being protected from spam bots, you need JavaScript enabled to view it
)
|
|
|
|
<< Start < Prev 141 142 143 144 145 146 147 148 149 150 Next > End >>
|
| Results 1342 - 1350 of 2512 |
