Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Parallella: A Supercomputer For Everyone

kickstarter.com

Post reply on HN
Print view View original

71–80 of 89 posts

Re: Parallella: A Supercomputer For Everyone

#71

Post by ricksta »

Parallel computing is limited by Amdalah's Law. Having more core does not mean you can have have more speed because it's not easy to use all those cores. Most imperial languages are not designed with running codes on multiple core and few programers are taught how to design their algorithm for using a handful of cores.

I can see this platform being a good tool for students and researchers to experiment with algorithm speedups by making their sequential code, parallel.

In my parallel programming class, our teacher had to rig together a computer lab to connect the 12 quad core computers to simulate a 64 core cluster. Then again, 64 core cluster of Parallella would cost like $7000. You can get the same 64 core setup by buying 8 x 8 core consumer desktop computer for under $3000, which will still be more cost effective and probably have ten times more computing power because of the x86 architecture.

http://en.wikipedia.org/wiki/Amdahls_law

Re: Parallella: A Supercomputer For Everyone

#72

Post by adapteva »

microtherion wrote: post #63
adapteva wrote: post #58

Earlier quoted context omitted.

Yes, agreed! We definitely got carried away with the marketing lingo and we apologize! This was our thought process: We have received a lot of negative feedback regarding this number so we want to explain the meaning and motivation. A single number can never characterize the performance of an architecture. The only thing that really matters is how many seconds and how many joules YOUR application consumes on a specif…

As one of the HN contributors who mocked the GHz performance spec in an earlier discussion thread, I welcome the fact that you're engaging in debate about this. That said, I still think that the GHz stat is just about as BAD a metric as any (I suppose "pin count times # of cores" would be worse :-). About the only positive inference I can draw from this is that you have the thermal situation in your system under cont…

Thanks for posting a very valid concern! Does the following FFT based image processing demo address the concern about communication bottlenecks in the approach?

http://www.adapteva.com/white-papers/using-a-scalable-parall...

Corner turns for 2D FFTs are usually quite challenging for GPUs and CPUs.[ref] Yaniv, our DSP guru, completed the corner turn part of the algorithm with ease in a couple of days and the on chip data movement constitutes a very small portion of the total application wall time.(complete with source code published as well if you really want to dig).

It's hard to market FFT cycle counts to the general audience:-)

Re: Parallella: A Supercomputer For Everyone

#73

Post by jamieb »

ricksta wrote: post #71

Parallel computing is limited by Amdalah's Law. Having more core does not mean you can have have more speed because it's not easy to use all those cores. Most imperial languages are not designed with running codes on multiple core and few programers are taught how to design their algorithm for using a handful of cores. I can see this platform being a good tool for students and researchers to experiment with algorithm…

"Pledge $199 or more: 64-CORE: You get everything in the SUPPORTER reward and a 64-core Epiphany-IV based Parallella board"

Re: Parallella: A Supercomputer For Everyone

#74

Post by AustinGibbons »

ricksta wrote: post #71

Parallel computing is limited by Amdalah's Law. Having more core does not mean you can have have more speed because it's not easy to use all those cores. Most imperial languages are not designed with running codes on multiple core and few programers are taught how to design their algorithm for using a handful of cores. I can see this platform being a good tool for students and researchers to experiment with algorithm…

If you like Amdahl's law you may also like... http://en.wikipedia.org/wiki/Gustafsons_law

It is a more powerful expression of the benefit of scaling with parallelism. Principally, instead of scaling speed with respect to a fixed data size, you scale the data size with respect to a fixed speed.

Having more cores means you (sometimes) can have more data. You still need those parallel programmers with their parallel algorithms though :-)

Re: Parallella: A Supercomputer For Everyone

#75

Post by IsTom »

AustinGibbons wrote: post #74
ricksta wrote: post #71

Parallel computing is limited by Amdalah's Law. Having more core does not mean you can have have more speed because it's not easy to use all those cores. Most imperial languages are not designed with running codes on multiple core and few programers are taught how to design their algorithm for using a handful of cores. I can see this platform being a good tool for students and researchers to experiment with algorithm…

If you like Amdahl's law you may also like... http://en.wikipedia.org/wiki/Gustafsons_law It is a more powerful expression of the benefit of scaling with parallelism. Principally, instead of scaling speed with respect to a fixed data size, you scale the data size with respect to a fixed speed. Having more cores means you (sometimes) can have more data. You still need those parallel programmers with their parallel alg…

As to parallel algorithms, there's probably not that many of them. https://en.wikipedia.org/wiki/P-complete#Motivation

Re: Parallella: A Supercomputer For Everyone

#76

Post by adapteva »

wtracy wrote: post #67

Since I thought I saw an Adapteva person posting here earlier: If the Kickstarter falls through, what options could you still make available to hobbyists? Is there some version of your current prototype setup that you could sell, even if it's not one convenient board?

We would rather not think of that option:-) if the ks project fails, we'll do our best, but seems unlikely that we could support selling kits to hobbyists and they would certainly cost thousands of dollars each due to a lack of volume.

Re: Parallella: A Supercomputer For Everyone

#77

Post by dfox »

wmf wrote: post #35
compilercreator wrote: post #30

Earlier quoted context omitted.

Prepare to be surprised. The die size estimate for the Epiphany IV is 10mm-sq according to Adapteva. It is more appropriate to compare it to embedded GPUs than desktop GPUs in die size, power and performance. For example, one particular embedded 40nm GPU design that I know about can deliver about 25 GFlops or so in the same die area.

Some of that GPU die area is used for graphics features that compute programs don't need. But a lot of it is providing performance even though it's not providing FLOPS. GPUs have caches and multithreading for a reason; if you could get better performance with an ultra-simple architecture then ATI/Nvidia would have done that already.

GPUs are primarily designed to be good for graphics, which implies completely different internal architecture. While GPUs have some graphics-oriented functional units the main factor is that all these cores have to access pretty large chunk of shared memory (textures, frame buffer...) and do that uniformly fast (and also support some weird addressing modes and access patterns). I suspect, that large part of die area of modern GPU is interconnect and that there really are few very wide cores (something like VLIW+SIMD+ possibly UltraSparcIV style hyperthreading, but that can be faked by compiler given sufficiently large register set) that are made to look like large amount of simple cores by magic in compiler (which seems consistent with CUDA programming model).

So: you can get large amounts of performance with simple architecture, but only for some problems, with graphics not being in set of these problems.

Re: Parallella: A Supercomputer For Everyone

#78

Post by jayhawk »

we got into a big discussion on super computers (the definition), the meaning of what a core is and a whole bunch of other issues... but the low power requirements of this are being completely ignored... as for applications... well portable and/or remote devices/sensors that need parallel computing capabilities and where high energy usage is prohibitive are possible applications. But the greatest asset of this is to spark the next gen of app developers and programmers to fully embrace parallel programming and truly make software scalable...

Re: Parallella: A Supercomputer For Everyone

#79

Post by driverdan »

runako wrote: post #36

Without commenting on the merit of this project, I'm alarmed to see a VC-backed making a Kickstarter pitch.

I'm of the opposite opinion. Companies that already have financial backing, have already put significant time and effort into a project, and already have experience running their business are much more likely to follow through on their campaign than some kid who build a new chair in his bedroom and thinks he can deliver it a month after his $100k campaign is finished.

Re: Parallella: A Supercomputer For Everyone

#80

Post by DeepDuh »

dfox wrote: post #77
wmf wrote: post #35

Earlier quoted context omitted.

Some of that GPU die area is used for graphics features that compute programs don't need. But a lot of it is providing performance even though it's not providing FLOPS. GPUs have caches and multithreading for a reason; if you could get better performance with an ultra-simple architecture then ATI/Nvidia would have done that already.

GPUs are primarily designed to be good for graphics, which implies completely different internal architecture. While GPUs have some graphics-oriented functional units the main factor is that all these cores have to access pretty large chunk of shared memory (textures, frame buffer...) and do that uniformly fast (and also support some weird addressing modes and access patterns). I suspect, that large part of die area…

Sorry, but I have to correct a little bit here. Today's GPUs are

- not simple SIMD. NVIDIA calls it SIMT (single instruction multiple thread), mostly since you can branch a subset of them, so for the programmer it does feel somewhat like threads.

- not just optimized for Graphics anymore. E.g. since Fermi, the Tesla cards have DP performance = 50% of SP - which has been specifically introduced for HPC purposes. They have also constantly improved the schedulers to go more into general purpose computing, e.g. Kepler 2 seems to support arbitrary call graphs on the device. Again, that's useless for graphics.

- suitable for pretty much all stencil computations. Even for heavily bandwidth bounded problems GPUs are generally ahead of CPUs since they have very high memory bandwidth. The performance estimate I use for my master thesis comes out at 5x for Fermi over six core Westmere Xeon for bandwidth bounded and 7.5x for computationally bounded problems.

HPC is all about performance per dollar, performance per watt - and (sadly) sometimes linpack results because some institution wants to be in the top of some arbitrary list. In all of these aspects GPUs come out ahead of x86, which has been very dominant since the 90ies. Which is why GPUs are now in 4 of the top 20 systems - each of those are hundreds of millions of dollars in investments. That wouldn't be done if they weren't suitable for most computational problems.

Post reply on HN