Multiprocessing Muscle
No RISC, No Reward
The multiprocessing power of Onyx graphics supercomputers begins at
the core, with the MlPS (r) R4400 (tm) 64-bit RISC microprocessor. The
combination of many R4400s working together gives Onyx users the kind
of computing performance typically associated with today's most
advanced supercomputers. With six internal CPU board slots (each
board contains either two or four R4400 processors), Onyx supports
configurations ranging from two to 24 R4400 microprocessors, all
operating together to provide balanced integer and floating-point
multiprocessing performance.
MIPS TFP Processor
- 64 bit 75Mhz MIPS TFP Processor: each TFP processor is a two-component chip set consisting of an Integer Unit with on-board 32KB
cache and a Floating Point Unit to deliver up to 300 peak Mflops per
processor
- 4MB Additional Cache: to achieve peak application performance
- One or Two TFP Processors per CPU Board: delivers up to 3.6
Gigaflops of Peak Performance, maximum of 12 processors in an Onyx
workstation
MIPS R4400 Processor
- 64-bit 150Mhz MIPS R4400 CPU: 32KB on-chip primary cache
- 1MB Secondary Cache: available for maximum application
throughput
- Two or Four MIPS R4400 Processors per CPU Board - maximum of 24
processors in an Onyx workstation
The MIPS TFP Processor: Floating-Point Punch
For floating-point intensive graphics applications,
Onyx offers processor configurations based on the TPS TFP
microprocessor. A single TFP processor is capable of attaining a peak
floating-point performance of 300 MFLOPS, making it the ideal choice
for compute-intensive applications such as computational fluid
dynamics, finite element analysis and image processing.
With a maximum of two TFP processors in a deskside model and up to
a dozen TFP processors in a rack configuration, users have access to
extremely powerful floating-point computing capabilities all tightly
integrated in a single platform with the world's most advanced
graphics engine, the RealityEngine2 (tm) graphics subsystem.
