View Single Post
Old December 29th 07, 06:40 PM posted to alt.comp.periphs.videocards.nvidia,alt.comp.periphs.videocards.ati,alt.comp.periphs.videocards.matrox
Benjamin Gawert
external usenet poster
Posts: 1,019
Default Fastest graphic card for Windows workstation use (2D, not gaming)

* Luca Villa:

Argh, I see that that page is protected. You've to read the Google
cache copy of it to see the full review with the 2D benchmark results:

I think you probably refer to the PC Mark 05 results. Well, it might
look to you that way but just reading some numbers without having a clue
what has been tested doesn't help.


- the "2D Graphics Memory" tests are testing the bandwidth of the gfx
memory for 2D copy operations. The relevance to real world applications
is *zero* because memory bandwidth isn't a limiting factor for 2D for
almost a decade now.

- The "2D WMV video playback" test has a mis-leading title: it doesn't
test 2D performance but the performance of video playback, done by
simply playing back a HDTV video (1920x1080) with Windows Media Player
at maximum possible frame rate. It doesn't tell you *anything* about 2D
application performance, it just tells you how well HDTV videos can be
played back by Windows Media Player. This test is just nonsense as
todays cards often support HDTV hardware playback with certain players
or additional software, so basically this test is useless.

- "2D Transparent Windows" creates 30 Windows with a sweeping "fading"
effect (alpha blending). The number just tells you how many of these
Windows can be created per second, it's not only affected by the gfx
hardware but also by the driver and even by what other processes are
running on the computer. While this test at least has some remote
relevance to real work (Window drawing) it also has no real world
relevance as you never ever see or notice the difference between a
system that can draw 3800 of these windows per second or "just" 2800.

Mind you, understanding hardware is no idiot's game where you just have
to compare some numbers. If you don't know what exactly has been tested,
how this stuff works and interacts and what also influences the results
you can't read anything from the numbers.