The Finnish startup Flow Computing promises a kind of revolution in computer computing. It claims that by adding its companion chip, any CPU can instantly double its performance, and with the help of software tweaks and optimizations, efficiency can be increased up to 100 times.
Flow is a division of VTT, a Finnish research organization that resembles a national laboratory. The chip technology it commercializes under the Parallel Processing Unit brand is the result of research conducted in this laboratory (although VTT is an investor, Flow owns the intellectual property).
While CPUs have come a long way in evolution, in some fundamental ways they still haven’t changed.
Their main limitation is that as serial rather than parallel processors, they can only perform one action at a time. While modern CPUs are capable of quickly distributing tasks across multiple cores and threads, these are all ways of adapting to the single-lane nature of CPUs. The GPU, on the other hand, performs many related computations simultaneously, but specializes in certain operations.
«The CPU is the weakest link in»’s computing, — said Flow co-founder and CEO Timo Valtonen. «It’s not up to the task, and that will need to change».
CPUs have become very fast, but still do not optimally execute instructions due to a basic limitation – in simplistic terms, one task must complete before the next one can start. Flow claims to have eliminated this limitation by transforming CPUs from a one-lane street to a multi-lane highway. The CPU is still limited to running one task at a time, but Flow’s PPU essentially performs traffic management functions on the nanosecond scale on the chip to move tasks in and out of the CPU faster than was previously possible. The PPU does not increase the clock speed or otherwise push the system, which can lead to additional power consumption and heat generation. It simply makes more efficient use of the CPU cycles that are already occurring.
Importantly, there is no need to change any code or architectures to make the PPU work efficiently. In fact, arbitrary code can be executed twice as fast on any chip without any modifications other than integrating the PPU with the chip.
So, unlike a software product, Flow technology has to be integrated at the chip architecture design level. That is, it does not work retroactively with any processor. Flow has shown that its technology works in FPGA-based test systems, but chipmakers would have to dedicate quite a few resources to see the benefits in question. However, if, without major architectural changes, processor manufacturers can double the performance of their CPUs by simply implementing minor changes to the chip, this should be a significant incentive to do so.
Further performance gains come from refactoring and recompiling software to work better with the PPU-CPU combination. Flow says it has seen increases of up to 100x with code that has been modified (though not necessarily completely rewritten) to take advantage of its technology. The company is working on offering recompilation tools to make this easier for software developers who want to optimize their products for Flow-enabled chips.
Source: techcrunch