Multi-core processors

Miguel Mendez flynn at energyhq.es.eu.org
Fri Apr 15 09:53:01 PDT 2005


On Fri, 15 Apr 2005 09:31:52 -0700 (PDT)
Danial Thom <danial_thom at xxxxxxxxx> wrote:

> I find it rather interesting that all of the
> intel-bashing "testers" are now reporting that
> there is "no performance gain" from dual-core
> processors which has always pretty much been the
> case with DP systems. It seems that its going to
> take quad-core processors to see real gains over
> "designed for UP" operating systems, how long do
> you think its going to be before such animals are
> available?

I think AMD is working on something like that, but not for desktop use.
Well, if those intel-bashing testers are running Microsoft Office on
Windows XP there's certainly little to be gained. What are those tests?

I've used an SMP box as my main desktop for some years now, and even
though I have more powerful boxes here I stick to this because it feels
more responsive and a runaway app doesn't leave you with a frozen mouse
pointer.

> Also, does anyone know if dual-core processors
> share a cache, which I suppose would provide for
> more efficient cache use, or are they just 2
> separate entities wired together?

AFAIK Intel's approach is more or less two dies on a single package,
i.e. two full processors each one with it's own L1 and L2 caches and I
think AMD's version uses a shared memory controller to make the CPUs
pin-compatible with existing mo-bos. Hyperthreading processors do share
a single cache among the two execution units, and that results in
degraded performance in some cases.

Cheers,
-- 
Miguel Mendez <flynn at xxxxxxxxxxxxxxxxxx>
http://www.energyhq.es.eu.org
PGP Key: 0xDC8514F1

Attachment:
pgp00017.pgp
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgp00017.pgp
Type: application/octet-stream
Size: 187 bytes
Desc: "Description: PGP signature"
URL: <http://lists.dragonflybsd.org/pipermail/users/attachments/20050415/c635bb25/attachment-0015.obj>


More information about the Users mailing list