The fast and the furious: compare Cell/B.E., GPU and FPGA

For decades everyone was spoiled by Moore's law directly translating into an exponential speed increase, the CPU clock was going up exponentially to 3GHz which was reached in 2003, but in the last 5 years it seems to be stuck at that point. Instead, manufacturers try to pack multiple cores into a chip. People started to look for alternative ways to get faster computation (see MRSC 2008 conf.): Field Programmable Gate Arrays (FPGA), General Purpose computing on Graphics Processing Units (GPGPU) and most recently the Cell Broadband Engine (Cell/B.E.) from IBM-Sony-Toshiba.

Tony Williams over at the ChemConnector Blog has had a couple of people ask him for comments about which way to go and which one is better for a particular application. The writer of this article, though, just invested two man years of effort porting to the Cell/B.E. and not only does he have strong opinions he also have enough "hands-on experience" to comment.

[cahill]

good article there

[pixelsword]

...I saw it and thought "this really explains a good amount well" because I read a bunch of things and sometimes it seems like the people writing them are trying to get their money's worth out of the word-a-day calenders they bought.

This gets just as detailed without all of the unnecessary verbage.

[tangerine]

???!!??

[MNicholas]

The Cell is a beast.

[MNicholas]

The Cell is a beast.

Supports what I've been saying all along.

The Cell provides the highest performance/dollar both in terms of aquisition costs as well as operating costs.

[Fishy Fingers]

Developers or people genuinely interested in Cell development....

"The Cell/B.E Developers group aims to bring together Cell developer specialists world-wide for the purposes of collaboration, networking and to ease resourcing for projects utilising this unique architecture"

http://www.linkedin.com/gro...

Thought someone might be intrigued. Found it in the articles comment section.

[kyleg]

they need to make CPU's that are over 4 GHZ

[Scenarist]

they need to make them way past 4gz dude... ive seen overclocks past 4ghz years ago

[Oner]

Funny how there isn't much fanboyism going on in here...yet. Certain people on N4G can only "talk" and downplay the benefits, prospects, possibilities and untapped power of the Cell B.E. but these guys at Simbiosys.ca actually make stuff and DO things that PROVE otherwise.

http://www.simbiosys.ca/sci...

http://www.rapidmind.net/ca...

[Ju]

There is actually one interesting thing related to CellBE/Xi. The Xi being a DP version of the BE. They have expected the DP performance to explode with the Xi. And in fact it does (compare 16Gflops at first on the BE vs. 136Gf on the Xi). But somehow they managed to get DP working more efficiently to run on the BE (has something to do with some sort of dynamic scaling of SP and DP but getting the same precision eventually by using SP most of the time. There is a document about that somewhere). Well, but with that said, they could push the DP performance of the BE solely thru SW design/changes to close what the DP does. Now they have 109Gf for the BE vs. 136Gf for the Xi. Isn't that amazing what SW alone can do ? I just thought that's pretty interesting. (SP = single precision, 32 bit float values = 4x32 bit is one 128 bit vector, 64bit is double precision but requires a 256 bit SIMD register to store a DP vector, e.g.).