Ed Light said:
A test that should not be the predominant basis of anyone's buying
decisions.
The supplies were 330-550W models, but Anandtech applied only 300W, at
most, in contrast to
www.silentpcreview.com and
www.tomshardware.com ,
which apply full rated loads or even more to the supplies they test.
Another problem is the "interference" test, where they tried to
determine the amount of interference generated by the supply by
writing to the memory and then testing the written contents 6 hours
later to find errors. It appears that they tested each computer over
an 18-hour period, running 3 different passes, but there's no way to
know if any variations were caused by temperature (power supplies were
mounted inside the case and could have made the memory run at
different temeperatures), AC line voltage or interference (no
filtered, regulated AC supply was used), electrical noise on the
output lines (no ripple or noise measurements were taken -- what
Anandtech referred to as "noise" is just DC voltage variation,
probably from load changes), or RF emissions (no field strength meter
and spectrum analyzer, despite Anandtech claiming that the Enermax
supplies were the most heavily shielded -- how did they know without
measuring the effectiveness of the shielding?). Anandtech isn't
exactly a shoestring operation, and even if it didn't have all the
needed test equipment, it should have enough contacts to be able to
borrow it. Or at least they could have checked for interference by
using AM and FM radios, TVs, and cordless phones.
About a 420W Turbolink supply, Anandtech wrote:
"The TurboLink 420's specifications struck us as
incredibly poor. Most noticeable is the 185W combined
rail specification. This is actually under the 200W
+5V specification so essentially, this TurboLink 420W
is never capable of obtaining the listed output."
This is outright absurd because at least one of those power ratings
had to be wrong, yet Anandtech ran no tests of its own to verify them
(or the ratings of any of the supplies).