Re: Performance improvement using X5355 over 5080

Torbjorn Lindgren wrote:
Bill Davidsen <davidsen@xxxxxxx> wrote:
Qu0ll wrote:
Can anyone give me a rough guide to the relative performance gains I would expect to get by swapping 2 x Xeon 5080 3.73GHz Dual-Core CPUs with 2 x Xeon X5355 2.66GHz Quad-Core CPUs? I would like to know if it is worth the upgrade.

Probably not. Look at it this way... you are trading two cores which are 40% faster, for four cores. So it depends on the number of cpu-bound

Except that it's silly to compare MHz/GHz that way. Yes, the clock
frequency of the cores in the 5080 is 40% higher but that doesn't mean
the performance is... Core 2 is a *very* different beast compared to
the Netburst (P4) Xeon DC.

Where does a P4 come in? If the 5080 isn't Core2 then my cheat *** is wrong about "multicore Xeons, 5000-series" CPUs. And O.P. probably has to change motherboards, which adds another level of guess. While m/b exist which support both, no one would actually configure one that way (would they)?

I would be rather surprised if the 5355 doesn't provide at least the
same per-thread performance as the 5080 and it could be noticeably
faster (15-25% is by no means impossible). It will also of course
provide twice the number of (hardware) cores.

Enabling HT on the 5080 changes the landscape a bit, you get more
"cores" but lower per-thread performance (the more threads the bigger
hit, but there's a small base hit). However, this really only helps
with large number of runnable threads (see below), and in those
scenarios the 2x5355 will be a MUCH better choice.

OTOH two 5355 are expensive enough that I'm not sure it makes sense to
"upgrade" that way, how much extra does it cost to just buy another
machine instead and use both :-)


threads in your system, if the number of runable threads is two or less on average, you lose big time. If the number is three you slightly better than break even, and only when you get to four do you win by about 40%. And at time when you have load, but ony a few threads running, you lose, even if you win under max load.

This is why I suggested that 2x5160 (2x3.0) might be worth looking
into. Cheaper, uses less power and a little bit better single-thread
performance (but only slightly).

And the number here are four or less, not two or less. Two sockets
means twice the number of cores.


If this machine is loaded to death most of the time then more is better, but your description of the load in a later post doesn't sound that way, it sounds like spike load when you are doing certain things, and only a few threads the rest of the time. Look at your video programs running and see how many threads are waiting for cpu. Ditto your games, if they are not threaded you will hate the slower individual threads.

The post processing stage of video editing (including compression) is
one of the things that can use the quads well. The actual editing
varies depending on what program is used and how big the files are.

A heavily loaded multi-user server or Web server is also good
examples of this, but most "desktop" and even many "workstation" tasks
doesn't have that many heavy threads (especially not with two sockets,
even with duals you'll have 4 cores).

Games isn't for now, benchmarks show some (fairly minor) benefits for
a single quad over a single C2 dual at the same frequency (but you pay
in $ and power/heat), but with two sockets you reach the same number
of total cores with the dual-core CPUs.

I'm not sure anyone has even benchmarked 8 vs 4 cores for any game,
but I'd be surprised if there was any benefit at all (unless you also
run a heavy-duty CPU burning program at the same time).


--
Bill Davidsen
He was a full-time professional cat, not some moonlighting
ferret or weasel. He knew about these things.
.

Loading