Desktop or Personal supercomputers are exceptionally fast. This post takes a look at the Optalysys, Tesla, hardware scenario, future, promises, and how they work. We talk about desktop supercomputers assuming the current hardware scenario won’t change, except for making the computers more mobile.
What are Personal Supercomputers
Simply speaking personal or desktop supercomputers are personal computers with the exceptional power of computing. That is, to say, process several terabytes of data in a few seconds or so. As of now, the available desktop supercomputers are, Cray CX1, the Nvidia Tesla, and the Asus ECS 1000. These computers employ GPU processors heavily to run many tasks simultaneously and are surely high priced. I did not check out the prices, but you can search for them on the Internet.
According to Google CEO Eric Schmidt, current mobile phones could also qualify as personal supercomputers:
“A billion people on the planet are carrying supercomputers in their hands. Now you think of them as mobile phones, but that’s not what they really are. They’re video cameras. They’re GPS devices. They’re powerful computers. They have powerful screens. They can do many different things, and oh, by the way, you can talk on them too. That’s what the mobile phone of today is.”
Of course, desktop supercomputers are just small, portable, or standalone computers that have exceptional power of processing several terabytes in a second. But that is my own definition. Personally, I won’t support Eric’s statement, though he is true in saying that mobile phones are powerful computers. They are not really supercomputers. A supercomputer should be able to work at 1 teraflop at the minimum. In other words, they should be 10 times faster compared to a 64-bit processor, so that multitasking and such tasks, do not slow them down while processing a set of data.
Desktop supercomputers are not coming anywhere in near future, given that all mainstream operating systems are now focusing on mobility and most importantly, on the cloud.
The next section looks at the possible implementation of personal supercomputers as no hardware vendor could be preparing for RAM or CPU bus that could handle even 128-bit computing at this moment. There is no one processor that processes 128 bit at a time, as far as my knowledge goes.
Implementation of Desktop Supercomputers
I see only one method of processing 128 bits or more as of now. The method is to use multiple processors in parallel though no current mainstream operating system requires such a configuration. In the past, it happened that 32-bit processors were paired in parallel to process 64-bit data. Hence, if Microsoft or any other company comes up with a 128-bit operating system in near future, it would be pairing of 4 processors of 32 bit or 2 processors of 64 bit. Later, they can research and create a processor that can process 128-bit data at a time. Accordingly, bus width would have to be increased to help data travel in chunks in 128 bits.
It was this analysis that caused someone from Microsoft to comment that I am being a pessimist and that they are working with hardware providers. They did work with hardware providers, but to bring out tablets and convertibles for 32-bit and 64-bit operating systems and applications.
I had written that a 128-bit operating system, though pretty welcome, would be difficult to implement, as the current hardware cannot support it right away and hence the speed won’t be much different between a 64 bit and 128-bit operating system. I had and still have doubts on buses, etc. that are to carry 128 bits at a time.
In another story I read, a company called Optalysys is planning optical processors. That would use laser beams instead of flip-flop circuits to speed up the processing. However, the company accepts it won’t be as fast as a real supercomputer but would be faster than Tesla supercomputers. Such computers, with a fast processing speed, would be welcome and sought by scientific organizations as well as gamers if the costs are not very high. In the story, I can say there is still plenty of time before the computers actually hit the market. We will have to wait to see it in action.
Using IaaS for Supercomputing
Another idea that comes to my mind is to use cloud services. Because you can buy just the amount of infrastructure and platforms you need, you can use such services as IaaS from different vendors to achieve good speeds.
The best method could be to take up the services like the Google Compute Engine to process large quantities of data as and when required instead of investing in multiple servers and making them run in parallel to achieve a faster speed. However, the cost of Google Compute Engine is around 2 million per day approximately, so it would keep average users away from the Compute Engine. I am sure there must be other cloud companies offering similar services at a lower price. This part has to be researched to see how and which company offers a cloud platform that enables us to increase processing speeds. With such services, even mobile phones can be supercomputers as they too can connect to the cloud.
I hope I have cleared some basic concepts about desktop supercomputers or personal supercomputers. Please feel free to add your inputs.