The Inquirer - supercomputer hardware news and downloads. Visit the download keep in the present day.
Intel has rap elsewhere next to bunged principles taking part in in height performance computing (HPC), axiom with the purpose of it is a costly alternative route.
Kirk Skaugen, VP of Intel's architecture unit, told journalists with the purpose of Intel wants to "democratise highly comparable computing" and he was "urging the industry to duck costly detours down proprietary paths". Conversation up Intel's Many Integrated extract (MIC) architecture, which allows programmers to employ languages such in the role of C++ and Fortran to give somebody no option but to employ of its single coaching multiple data (SIMD) units, Skaugen criticised the employ of proprietary, bunged principles.
Skaugen alleged with the purpose of firms make not wish for to "spend all the money to seaport [applications]" to restricted hardware and with the purpose of solitary of MIC's biggest advertising points is the information with the purpose of it can run existing code and give somebody no option but to employ of SIMDs with relatively little code changes. Dr Eng Lim Goh, CTO of SGI backed up Skaugen's tip, axiom with the purpose of "a copious portion [of customers] are realising with the purpose of this [moving code to the GPGPU] is tough", tally "they don't wish for to splurge the same amount of money to seaport software in the role of on hardware".
In the role of in favor of what did you say? Exactly Skaugen was referring to while he alleged "proprietary paths", he did not say. However The INQUIRER asked Joseph Curley, director of MIC marketing next to Intel to process whether Skaugen was referring to Nvidia in Lg X120 Battery, Lg LB6411EH, Lg LB3211EE . Curley would not identify Nvidia by character but alleged with the purpose of several fixed with the purpose of simply allows code to run on its own hardware can be there classified in the role of proprietary.
Curley explained with the purpose of Intel wants to bring in height level encoding language portability to many extract encoding. With the purpose of puts Intel next to plain as the nose on your face loggerheads with OpenCL, a language driven by the Khronos association, of which Intel is a portion. Curley alleged with the purpose of "OpenCL is an exception" and it long for remain promote OpenCL in the role of it is an exposed standard with the purpose of long for allow code to run on multiple vendors' hardware, plus AMD's GPGPUs and accelerated doling out units (APUs).
Although Skaugen and Curley would not state Nvidia and its CUDA encoding language by character, it was pretty acquit with the purpose of Nvidia was the intended target. Curley did state with the purpose of other technologies such in the role of FPGAs and DSP chips plus fall into the "proprietary path", but pointed elsewhere with the purpose of Intel has not been taking part in competition with individuals chips in favor of a very protracted stretch.
Nvidia, which has had a relatively undemanding torment taking part in the HPC bazaar with its GPGPUs, looks frozen to be realistic not simply AMD but Intel in the role of a competitor to both
All right reserved!
No comments:
Post a Comment