[PLUG] Real-life hardware comparison.

Derek Loree drl at drloree.com
Sat Apr 17 17:49:01 UTC 2004


On Sat, 2004-04-17 at 13:02, Russell Senior wrote:
> >>>>> "Derek" == Derek Loree <drl at drloree.com> writes:
> 
> >>  I sent this to Jeme last night, but thought I'd post it here for
> >> general information.  I have a dual P3-1GHz with a gig of PC133 RAM
> >> and a P4-2.53GHz with a gig of 533MHz DDR RAM.
> 
> Derek> What chipsets are on the boards, is this single channel or dual
> Derek> channel DDR?
> 
> The dual P3 is a VIA chipset (Apollo Pro266, a SuperMicro
> P3mumble-mumble-mumble).  The P4 is an Intel 845GEBV2.  
> 
> RE single-vs-dual channel DDR... dunno, how do I tell?.

By the chipset, that one is a single channel, and will only run memory
at either 266MHz or 333MHz, the memory bus itself is either 133MHz or
166MHz.  The processor bus is run at either 400MHz or 533MHz.  The 865
and 875 chipsets will run dual channel,  The 875 has a memory bus of up
to 200MHz, so the DDR comes out looking like 400MHz, dual channel means
the processor can access two sticks of RAM at the same time (but the
processor must be designed for it).  The processor bus on the 875 can
run at 800MHz.
> 
> >> For an allnoconfig 2.6.5 kernel compile (for pippin-to-gravenstein
> >> comparison), here's what I found:

I ran a similar compile with a 2.6.4 kernel on a P4 2.8Gig with an 875
chipset, 512Megs of DDR (dual channel).  My results were a different
format:

real 1m11.670s
system 1m2.922s
user 0m6.011s

For some reason, I can't convince it to be more verbose.
> >> 
> >> P4: $ time make 69.84s user 10.48s system 95% cpu 1:24.01 total
> >> 
> >> P3: $ time make -j4 159.04s user 20.73s system 181% cpu 1:39.19
> >> total
> 
> Derek> I should have known that make would fork out tasks, it has been
> Derek> a long time since I've compiled a kernel on a dual
> Derek> proc. system.  What kernel was running the compile?
> 
> 2.6.4 on the P4, 2.6.5 on the P3.
> 
> Derek> What happened when you didn't specify -j?
> 
> You saw the no-j for P4.  For the P3, I get:
> 
>   $ time make 
>   149.86s user 17.87s system 96% cpu 2:54.42 total
> 

Interesting, not the behavior I would expect from reading the man page
for make.

Thanks,

Derek Loree





More information about the PLUG mailing list