Live data from Hacker News

A Cheap 12-Core, 30-Watt Ubuntu Cluster

phoronix.com

31–35 of 35 posts

Re: A Cheap 12-Core, 30-Watt Ubuntu Cluster

#31

I like the idea of independently replaceable boards, but the drying rack isn't the answer. It may be great for heat dissipation, but you'd be surprised how many dangers are present in an apparently empty room (dust, ceiling leaks, your hipster keychain, etc.).

Why isn't there a market (equal to the one for boards like these) for enclosures, "home racks" or something similar to house these things? Or is there a market but this guy just wanted to save a few quid?

Re: A Cheap 12-Core, 30-Watt Ubuntu Cluster

#32
post #10

For comparison, you can now get http://ark.intel.com/products/65703/Intel-Core-i5-3470T-Proc... http://ark.intel.com/products/65735 http://ark.intel.com/products/65714/Intel-Core-i7-3517U-Proc...

the wattage measured here is not just the processor, its also the entire boards. Ive done some benchmarks testing Intel Atom processors Atom330's and D525's against Tegra3 processors. On average the Terga3 processors outperformed the Atoms 4x-8x in Flop/s per Watt. Admittedly that is not the only important metric, but gives an initial performance comparison between ARM and x86 performance

I think the consumption of the rest of the system is fairly negligible, especially considering how inaccurate the TDP figures are to begin with. Eg this[1] system with a 95W TDP CPU draws 92 watts from the wall. And that is with off-the-shelf parts. Swap in a low-power CPU and overall optimize for power consumption and I wouldn't be surprised if you'd get below 50W.

Re: A Cheap 12-Core, 30-Watt Ubuntu Cluster

#33
I can't imagine that the 96 core cluster would be either power efficient or cost efficient compared to Intel offerings. Assuming 3 watts and $100 per core (which align fairly well to the figures presented in the article), 96 core cluster would draw nearly 300 watts and cost nearly $10000.

Contrast that to eg Intel E5-2650L[1]: 8 cores, 16 threads, 70 watts and $1200 price. Obviously the whole system cost would be much higher, but I find it hard to believe that you couldn't get/build two 2-socket systems with such CPU for $10000. The CPUs alone would draw 280 watts, which would put the whole system power consumption in the same ballpark as the ARM cluster.

Remaining question would be performance. ARM cluster would have 50% more (logical) cores, but I'd imagine that Intel would have far higher per-thread performance. And for lots of tasks the 100M Ethernet of the Beagleboard as an interconnect could become a bottleneck.

[1]http://ark.intel.com/products/64585/Intel-Xeon-Processor-E5-...

Re: A Cheap 12-Core, 30-Watt Ubuntu Cluster

#34
post #30
post #13

Earlier quoted context omitted.

I was seriously considering trying something like this with either the Raspberry Pi or the APC " rel="nofollow">http://apc.io/> . It looks like one PandaBoard can roughly match the CPU horsepower of four APC boards at a slightly lower price, so that's good to know. The Raspberry Pi might still beat out the PandaBoard in terms of purchase cost-to-Megahertz ratio. We'll see. :-)

Don't just look at MHz/$. The Raspberry Pi has a ARM v6 compatible core vs the PandaBoard which has a v7-A core, i.e., Raspberry Pi MHz < PandaBoard MHz.

Do you know of anyplace that quantifies the actual performance difference of the different ARM versions? Just want to see the general improvement jump over time

Re: A Cheap 12-Core, 30-Watt Ubuntu Cluster

#35
post #29
post #23

Earlier quoted context omitted.

That being said, the Atoms themselves are nor a good benchmark in performance / Watt. I'd rather be interested in a comparison vs. current gen. xeon or interlagos systems. Sounds silly, but ARM has been making progress and I could see them being used in the future as companions to GPUs in computing clusters. With current GPGPU computing models like OpenACC it does not really make sense to put 16 race horses (Interlag…

Well, I think Intel wants Atom to compete with ARM for that market, so it may be relevant. There's also Nvidia's Project Denver, which will probably come out in 2014. It's based on the 64 bit ARMv8 architecture, it's a custom CPU made in collaboration with ARM, and I think they want to pair it with their next-gen GPU architecture Maxwell. It's intended for servers and supercomputers.

I've heard about that Nvidia project. I think they are on the right track. The only thing that's missing is enough programmers (und thus software) for this model. As an example (and I'm saying this as a layman in terms of databases) I think that DBMS might be able to profit a lot from the GPGPU based model. For high read traffic databases you could scale a system with n GPUs, based on how much "storage" the database needs (the storage being the GPU ram, continuously mirrored to harddisks when writes occur).
Post reply on HN