4 ms·

Hmm I never heard of the "Parallella". This one has 64 cores but never seems to have been released: http://www.adapteva.com/white-papers/parallella-64-cores/ ht

by esaym 11y ago

Hmm I never heard of the "Parallella". This one has 64 cores but never seems to have been released: http://www.adapteva.com/white-papers/parallella-64-cores/ http://www.adapteva.com/white-papers/parallella-64-cores/ Is there anything else like that on the market for cheap? I'd actually like to play around with a high core count chip. "Only" 4 cores is for kids :)

ChrisGranger 11y ago
I was one of the Kickstarter backers for the Parallella a few years back. The 64 cores are in the Epiphany IV coprocessor, while the board itself has two ARM cores and would behave a lot like the RPi and similar if the Epiphany wasn't taken advantage of, assuming you were able to get your hands on one of the early 64-core boards. The 16-core Epiphany III-based Parallella is available at Amazon.
wtracy 11y ago
As my sibling poster mentioned, 64 (or 16) cores is for the coprocessor. The coprocessor has an architecture halfway between a traditional CPU and a GPU. Linux does not run on the coprocessor itself, so your coprocessor code runs on the bare metal and has no access to the POSIX APIs, or any existing libraries, really. Finally (this was the deal-breaker for my project, after I bought a bunch of Parallellas!) the Epiphany has only something like 8mb (not gb!) of RAM. All 16 cores have to share that. (The main CPU has access to something like a gig, but the bandwidth bottleneck between the two processors is unacceptable for trying to randomly access main memory from the Epiphany.) If anyone is still not scared off, I have two extra Parallellas lying around still in the original packaging. Ping me at afishionado@gmail.com if you're interested.
wyldfire 11y ago
ODROID-XU4 uses Samsung Exynos5422 (8-core big.LITTLE) for ~80 USD - http://ameridroid.com/products/odroid-xu4 http://ameridroid.com/products/odroid-xu4 I'm happy with mine. Ridiculous core counts typically find workloads that saturate the memory controller or I/O, which is why so few are made. They typically bring with them a non-uniform memory space, cache-incoherent/rings, etc. All of that makes software teams groan and decide to spend a few thousand USD on a multi-socket Xeon or a GPU.

by esaym 11y ago

ChrisGranger 11y ago

I was one of the Kickstarter backers for the Parallella a few years back. The 64 cores are in the Epiphany IV coprocessor, while the board itself has two ARM cores and would behave a lot like the RPi and similar if the Epiphany wasn't taken advantage of, assuming you were able to get your hands on one of the early 64-core boards. The 16-core Epiphany III-based Parallella is available at Amazon.

wtracy 11y ago

As my sibling poster mentioned, 64 (or 16) cores is for the coprocessor. The coprocessor has an architecture halfway between a traditional CPU and a GPU. Linux does not run on the coprocessor itself, so your coprocessor code runs on the bare metal and has no access to the POSIX APIs, or any existing libraries, really. Finally (this was the deal-breaker for my project, after I bought a bunch of Parallellas!) the Epiphany has only something like 8mb (not gb!) of RAM. All 16 cores have to share that. (The main CPU has access to something like a gig, but the bandwidth bottleneck between the two processors is unacceptable for trying to randomly access main memory from the Epiphany.) If anyone is still not scared off, I have two extra Parallellas lying around still in the original packaging. Ping me at afishionado@gmail.com if you're interested.

wyldfire 11y ago

ODROID-XU4 uses Samsung Exynos5422 (8-core big.LITTLE) for ~80 USD - http://ameridroid.com/products/odroid-xu4 http://ameridroid.com/products/odroid-xu4 I'm happy with mine. Ridiculous core counts typically find workloads that saturate the memory controller or I/O, which is why so few are made. They typically bring with them a non-uniform memory space, cache-incoherent/rings, etc. All of that makes software teams groan and decide to spend a few thousand USD on a multi-socket Xeon or a GPU.