Till startsida
Sitemap
To content Read more about how we use cookies on gu.se

Per Stenström constructs the computers of the future

Per Stenström’s research area is computer architecture, which is about the principles for computer design. Per has a big interest in the optimal architecture for multi-core microprocessors, with a focus on the memory system architecture. By looking more closely into computer architecture, the challenge is to create smarter computers that save a lot of energy at the same time as you get a faster computer. We are now seeing the biggest challenges for the field so far - in the future computers have to be built from completely new design materials and there are still no satisfactory solutions in sight.

Faster computers – how do we deal with the energy consumption?

- A cell phone is an example of a product where you need a powerful processor to handle a lot of parallel activities, says Per Stenström. You might want to play games on the phone, at the same time as you are answering a call, sending a photo or something else. But: making the processor faster also means higher energy consumption - which is a key challenge for a cell phone with a limited battery.

In addition to problems with limited energy availability in the case of a battery, the energy consumption in the computerized world is getting a bigger and bigger environmental problem.

You can take the searching activities at Google as an example; they all end up in a huge data-processing centre where the processors consume an immense amount of energy. In addition to the energy needed for the computer activities, the data centre also requires large and energy-consuming cooling systems that have to take care of the heat dissipation in the processors.

- This is where my area comes in, says Per Stenström. By looking more closely into computer architecture you can design computers that save a lot of energy. “Green computing” is no longer just a cosy buzz word, today the energy consumption caused by the computerized world starts to become a really big problem.

- The challenge within the area and the driving force for me is to see how to handle the need for faster and faster computers, says Stenström. We have to find technology that is optimal and sustainable at all levels, when it comes to computer capacity and from an energy saving perspective. When I say “computer” I don’t just mean a laptop, I also mean the huge amount of embedded microprocessors we have all around us; in cell phones, dish washers, cash dispensers etc. A new car contains 50-100 microprocessors for example.

Multi-core microprocessors

The microprocessor is the "brain" in the computer and it is responsible for all the processing of information. A microprocessor is built on a silicon chip, with the size of approximately 1 square centimetre. Today a chip contains over one billion transistors.

Ten years ago a microprocessor had just one computation unit, one core. To make the computation unit more powerful, the number of transistors per chip was doubled every second year – and the transistors were used for building parallel and fast computation structures directly on the chip. Computer designers did not bother to find out which operations could be performed simultaneously, these processes ran on the chip, so to speak "under the hood".

In the end there was no possibility of making one computation unit more efficient, since the power and the heat dissipation got too big in the unit. The solution was to increase the number of computation units or cores on the chip – and the programmer had to split the computation on different cores. The microprocessor became "multi-core" and this technology is used in all new computers today.

But the multi-core processor technology caused a new problem: the computer could no longer find which processes that could be performed at the same time and the full capacity of the computer could not be used the way we wanted to. In order to achieve efficient parallel processes, the software must be rewritten – a complicated process that could easily go wrong.

Per Stenström and his graduate students have been working a number of years to find methods for how to set up parallel processes that can be performed on different CPUs in a multicore computer. In 2006, Per started a company in Lindholmen to develop a product that makes it easier for programmers to customize software for multi-core technology. The product is designed to shorten the lead time to develop software-based products designed for multi-core and to lower the cost of testing software.

Focus on the memory systems within multi-core

Per Stenström started already as a doctoral student in the late 80’s to look further into multi-core processors and his focus has been on the memory system, especially the cache memory.

A computer is in fact a rather simple thing, says Per Stenström. The computer fetches a "note" over and over again, with instructions on what information to fetch from the memory and what to do with the information. The computer performs a few billion of these simple operations per second. Since the computers are working faster and faster, a bigger and bigger main memory is needed. But a bigger memory also means a slower memory. Per has been doing research on how to build cache memory hierarchies. These consist – to put it simple – of a small and fast memory placed close to the processors, a larger and a bit slower memory placed a bit further away, then another one outside the second one, and outmost we have the biggest and slowest memory. The four memories are built together in such a way as to be perceived as one large and fast memory by the user.

New building material for the processors is needed around 2020

Meanwhile, the number of transistors on a chip will double every two years and to meet this computer manufacturers have doubled the number of units (cores) every two years. Transistors are becoming smaller and smaller and the wires connecting the transistors on the chip are now so thin that they can be counted in the hundreds of atoms. In the near future we will see wires with a thickness of a dozen atoms. We have then reached Quantum Physics, which means a huge challenge for the constructors. The radiation in the surroundings can be enough to disturb the traffic in the extremely thin wire and a zero could sometimes change to a number one, and vice versa. When the risk for errors in the system increases, more and more of the building material in the computer will consist of control systems to eliminate the errors.

- The technology development suggests that the computer building material we have today will be sufficient until about 2020, and then it will get harder and harder, says Stenström. We are now looking for a completely new technology for computer construction, but so far there has been no breakthrough. We are in the same position as the environmental field – we desperately need alternatives to the systems we have today, but what we have found so far has not been good enough compared to what we already have. Constructing computers in alternative building materials means a huge challenge and since we are somewhat groping in the dark the research has not started in a big sense yet, says Stenström. There is always a risk for the researchers when a completely new technology has to be invented - you have to be aware that the research track you are following could always be a blank track.

Network of excellence in computer architecture

Per Stenström has been engaged in many things besides the research in recent years. Per has spent some time on building a network of excellence within computer architecture called HiPEAC. HiPEAC is financed by the European Union and consists of 200 members from both industry and academia. The computer architecture field has previously been very fragmented and by gathering the skills with the leading researchers in Europe the computer architecture area got completely new financing opportunities which come with the new visibility.

Per has also become interested in how to get a researcher to take the plunge and make innovations out of his/her research and Per Stenström is a member of the board in Chalmers Innovation.

- But, says Per Stenström, what I am most passionate about in my job is to educate doctoral students. It is when I meet my doctoral students and discuss things with them that the new ideas occurs – that is really creative!

The latest distinction is a membership in Academia Europaea

There have been some distinctions for Per Stenström recently. In 2007 Per was appointed IEEE Fellow and the year after he was appointed ACM Fellow for his overall research within memory systems. In 2009 Per was elected as a member of the Royal Academy of Engineering Sciences in Sweden and in September 2010 Per got the announcement that he is invited as a member in Academia Europaea, an organisation that wants to promote scholars to take responsibility for the societal benefit from research.


October 2010


Written by: Catharina Jerkbrant

 



Contact information:

Per Stenström, Professor
Department of Computer Science and Engineering
Tel: +46 (0)31-772 17 61

Per Stenström

Per StenströmPer Stenström, professor at the Department of Computer Science and Engineering, wants to find the optimal technology for the computers of the future. To design computers in alternative design materials, with higher performance AND a sustainable energy comsumption is a major challenge for the researchers in the computer architecture field.

Page Manager: Catharina Jerkbrant|Last update: 10/6/2010
Share:

The University of Gothenburg uses cookies to provide you with the best possible user experience. By continuing on this website, you approve of our use of cookies.  What are cookies?