I saw a picture the other day of a GPU that was blue, had a reference cooler, and said intel on the side.
Further investigation showed me that this was, in fact, an Intel desktop graphics card and not some joke.
So if intel doesn't just make the shitty GMA chips and the ridiculously over-priced Iris chips all for laptops, tha the bloody hell is THIS thing for?
4K pixel art printouts?
Do you mean this thing?
No this was a forum post of stupid computer crap. It looked like that, but a GPU and with a reference cooler.
It could have also been a mock up but the image traced back to intel in some GPU folder somewhere.
The hell is that thing anyways?
(le AMD user)
EDIT: I think that's totally what I saw now that I am looking it up. WTF is that thing?
was it this version?
http://www.newegg.com/Product/Product.aspx?Item=N82E16819117541&cm_re=xeon_phi-_-19-117-541-_-Product
^^ Yup
yeah just your average 61-core coprocessor :)
WTF are these things for? I go to the intel site for it and this robot is making sexy eyes at a chinese worker.
On top of that aren't coprocessors used for overflow? Like in cameras and shit like that? I would say this seems like a supercomputer module but an actual supercomputer is a bunch of cells in a box on a high transfer rate network.
Yeah not a graphics card at all. I don't really know what people use them for, maybe scientists use them or something.
On November 12, 2012, Intel announced two Xeon Phi coprocessor families which are the Xeon Phi 3100 and the Xeon Phi 5110P.[46][47][48] The Xeon Phi 3100 will be capable of more than 1 teraFLOPS of double precision floating point instructions with 240 GB/sec memory bandwidth at 300 W.[46][47][48] The Xeon Phi 5110P will be capable of 1.01 teraFLOPS of double precision floating point instructions with 320 GB/sec memory bandwidth at 225 W.[46][47][48] The Xeon Phi 7120P will be capable of 1.2 teraFLOPS of double precision floating point instructions with 352 GB/sec memory bandwidth at 300 W.
The Xeon Phi uses the 22 nm process size.[46][47][48] The Xeon Phi 3100 will be priced at under US$2,000 while the Xeon Phi 5110P will have a price of US$2,649 and Xeon Phi 7120 at US$4129.[46][47][48] On June 17, 2013, the Tianhe-2 supercomputer was announced[5] by TOP500 as the world's fastest. It uses Intel Ivy Bridge Xeon and Xeon Phi processors to achieve 33.86 petaFLOPS.
An empirical performance and programmability study has been performed by researchers,[49] in which the authors claim that achieving high performance with Xeon Phi still needs help from programmers and that merely relying on compilers with traditional programming models is still far from reality. However, research in various domains, such as life sciences[50] and deep learning,[51] demonstrated that exploiting both the thread- and SIMD-parallelism of Xeon Phi achieves significant speed-ups.
Phi is part of High Performance Computing solutions by Intel. The evolution of the SCC (single chip cloud computing). The research has been going for like 5-6 years. The so called many-core (not multi-core) chips. AMD is preparing sth similar.
It is used for academic or industrial applications that require huge amount of computation. They are very much cheaper than having a full on server farm. Usually these are applications for data analytics, scientific computing, aerodynamic modelling etc. Our partner university for instance uses the PHI to run Biophysical neuron models that are quite computationally heavy.
High performance computing is very a big trend nowadays. Mostly because of data anaytics. The field managed to perfect how to collect data (scientific or otherwise) but not really how to process them.
Intel Yeon Phis are used for applications requiering high amounts of computational power. It is similar to AMDs S9170 as both serve as recources of lots of FLOPS
For some pretty pictures and a little explanatory text, look here: http://www.servethehome.com/supermicro-x10-gpu-xeon-phi-superserver-solutions-overview/
I liked the part about the Power supply of the 4U system. Up to 3,2 Kilowatts of power draw.
From what I understand they are not easy to program for. They actually run their own os and processes what is passed to it. In certain situations they are better that tesla gpus.
Oh yeah, I see the advantage. Get one of these instead of a set of servers. Lower buy in for smaller jobs. You might get a lot out of parallel computing, but a GPU with hundreds of smaller cores isn't optimal, because those programs may be quite serial in nature. Right?
They are great for cpu bound processing on BOINC. (never used one, but heard from other people who did - that its very good for its price)
Now thats an answer thats better than "Whatever a TFLOP processor does"
I was wondering if someone could adapt these for encoding video/steaming. But that would require a skilled programer that probably has better things to do.
The long and short of it.
The PHi is a competitor to the Nvidia TESLA SuperCompute Cluster cards.