"Richard" == Richard L Hamilton <[EMAIL PROTECTED]> writes:

>> Antoon Huiskens <[EMAIL PROTECTED]> writes:
>> > ouch.
>> >
>> > assuming it all runs on sun boxes, I'd look at the
>> biggest and
>> > baddest: the M9000: 2TB of internal memory
>> currently. Likewise, a
>> > T5240 coolthreads machine maxes out at 128GB (but
>> then again, that in
>> > 2 rack units).  As for x64 systems, a quick
>> glancing through the
>> > various (sun) systems suggest 256GB.
>> >
>> 
>> As for CPUs, a current M9000-64 has up to 64 sockets
>> (256 cores, 512
>> strands), although there was a bug mentioned in the
>> announcement of the
>> SPARC64 VII support about > 256 strands per domain,
>> IIRC.
>> 
>> The x4600 supports up to 32 cores.
>> 

Richard> See
Richard> 
http://src.opensolaris.org/source/search?q=&defs=NCPU&refs=&path=&hist=&project=%2Fonnv

Richard> which I read as follows:
Richard> 32 max for x86
Richard> 256 max for x86-64 (aka amd64, em64t)
Richard> 32 for generic sun4u (although platform-specific support could 
increase that)
Richard> 64 for E10K
Richard> 256 for sun4v (ditto?)

No, we don't change NCPU at a platform level on sun4v.  We're trying
to stay away from platform specific changes like that.

One of the things I want to change in sun4v, however, is all these
arrays whose size is NCPU.  It wastes memory and possibly in some
cases processing time on the older sun4v systems like T1000/T2000 (32
"cpus" max) or even T5120/T5220 (64 "cpus" max).  See CR 6685223 for
one of these issues.

-- 
Dave Marquardt
Sun Microsystems, Inc.
Austin, TX
+1 512 401-1077 (SUN internal: x64077)
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to