Reply to post: Re: The problem is not this generation but the *next* generation

Re: The problem is not this generation but the *next* generation

@PowerMan

This is silly. You are speculating that a 32-socket SPARC M7 server will be comparable to a 16-socket POWER8 server. How do you know that? Isnt this just wishful thinking, or do you have access to 32-socket SPARC M7 servers and have benchmarked them before they reached the market?

Besides, informed people say the SPARC M7 which is 4x faster than SPARC M6 - should be more than twice as fast as POWER8. If not, I will despise Oracle for bad engineering. Yes, truly. I mean, SPARC M7 has 10 billion transistors, and POWER8 has half of that - and if SPARC M7 is not much faster than POWER8, then Oracle has failed miserably with their transistor budget. And I despise bad engineering. So, I promise you, I will despise Oracle for SPARC if the M7 is comparable to a POWER8. Promise.

But, everything points in the direction of SPARC M7 is more than twice as fast than POWER8. For instance, I would like to see one POWER8 cpu doing SQL queries at 120GB/sec - which SPARC M7 can do. How many POWER8 sockets does it take to do 120GB/sec queries? Four? Six? Anyway, hoping for SPARC M7 being as slow as a POWER8 is quite optimistic thinking, dont you think?

So, there is no way a SPARC M7 with 32-sockets and 64TB RAM will be as slow as a 16-socket POWER8 16TB RAM. The SPARC M7 cpu is at least 2x as fast as a POWER8 cpu. And the SPARC server has twice the amount of sockets. So in total, the SPARC M7 server should be at least 4x faster than the largest POWER8 server.

Its a far stretch to hope for SPARC M7 server being as slow as a 16-socket POWER8 server. The SPARC M7 monster will crush the POWER8 server. Trample it in the dust.