- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -
- Information contained in the HLRS Wiki is not legally binding and HLRS is not responsible for any damages that might result from its use -

NEC Aurora difference

From HLRS Platforms
Revision as of 18:50, 8 February 2020 by Hwwnec5 (talk | contribs) (Created page with "'''How does it differ from SX-ACE and previous SX models?''' in many ways. * it is affordable and low power (VE card is below 300W and competitive to GPU pricing, it is affo...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

How does it differ from SX-ACE and previous SX models?

in many ways.

  • it is affordable and low power (VE card is below 300W and competitive to GPU pricing, it is affordable for workstations)
  • it tries to follow standards and de-facto industry standards, it tries to blend in, by being little endian, by using glibc and linux where possible
  • as easy to program as previous machines, but also usable as accelerator
  • much more open, ISA documentation is public, VEOS sourcecode is opensource, YOU can write your own compiler!
  • targeting also ML and data analytics, with some unique software available

How does it differ from favorite x86-64?

  • memory bandwidth, > 1TB/s in one card in <300W.
  • less but strong cores, 192 FP operations per core per cycle (AMD Rome 16, intel cascade lake 32)
  • less memory (48GB) but fast memory (HBM2)