Responding to an IBM claim of double performance per core, questions and concerns are raised about:
- Scaling is not free
- Choosing the denominator radically changes the picture
- SPEC Fair Use requirements
- Substantiation and transparency
- T5 claim for "fastest processor"
Occasional discussion of computer performance, especially performance of CPU, Memory, and Compiler (the components measured by SPEC CPU)
I contributed to the development of SPEC CPU2006 and SPEC CPU2000, and serve as run rules editor for the CPU subcommittee.
In the picture, I am the one on the right. The file name for the picture says it all - I need to get out on that bike more often.