pawihte said:
If all target features are met by different motherboards of
reasonably recent technology, does the particular chipset used
make any significant difference in performance? If the question
is too vague, please say so and I'll expand it. Thanks in
advance.
Absolutely it makes a difference. The problem is, you won't be
able to find online information to quantify it.
To give an example, I have a (relatively) fancy chipset by reputation.
Yet the USB mouse can stutter under heavy PCI bus load. When
benchmarking hard drives, I get a decent (averaged) line
drawn in HDTune for transfer rate across the disk surface.
But the line has too much moment to moment variation, when
compared to my previous $65 motherboard with VIA chipset.
The transfer rate curve on the VIA chipset board was
smooth and only a function of the head to platter transfer
rate.
So yes, there may be differences. In my case, it could be
a BIOS tuning problem, with things like PCI versus PCI Express
bus arbitration policy setting, max PCI Express packet size or the
like. (There have been some famous analogous cases, such as
a certain chipset where the PCI 32 bit bus could only manage
25MB/sec.) I don't understand what could be upsetting my SATA ports.
The SATA issue has no practical significance, except as a warning
that something isn't tuned the way it should be.
And there are other cases, such as low USB2 transfer
rates on a certain ATI Southbridge. That one is why Anandtech
was testing USB2 transfer in reviews for a while, to see
who was getting it right. 30MB/sec for USB2 would be "normal"
in sustained transfer rate, and when multiplied by 8 would give
240 in this megabit/sec bar chart. The SB450 here is running
at roughly half speed. (USB2 mass storage is protocol limited,
and will never hit 60MB/sec no matter what chipset is used.)
http://images.anandtech.com/graphs/uli m1575_101305111056/9447.png
Nobody reviews hardware in enough detail, to know which chipsets
are inferior. For example, have you ever seen a PCI Express transfer
rate test of any sort ? Have you even seen a reviewer verify that
a "Revision 2" slot actually runs in Revision 2 mode. I personally
would feel re-assured if I could see that. (This would also be handy
to know, for any new-fangled PCI Express x1 slots that are supposed
to support both modes.) If the test case used involved DMA to
system memory, the system memory speed could make a difference
to the results (when testing x16 slots). But checking some register
on the interface, should indicate whether it is in Rev.1 or Rev.2 mode.
Paul
