|
The labs publishing these ratings are averaging the number of frames per second and offer an average value of performance. In reality the card is processing hundreds of frames per second, sometimes thousands, depending on the graphics, and should the card stall on any particular frame, say 250 or 350 milliseconds when all other frames are being rendered at roughly 40 milliseconds, that would produce a noticeable stall for the video game player.
However the lab is not publishing how many frames, or how many instances a frame experiences a stall, it is publishing an average value for an aggre-
gated total of frames per second. The difference in measuring this metric does not seem like much however it is critical for action-oriented video games where response time determines if
you win or lose. The other issue is that repeat stalls on a particular type of frame can occur making the video graphics card and the game useless for
the video game player. Yet in averaging the FPS value these issues are ignored until after the customer installs the card and starts playing.
Scott Wasson published a lengthy and very insightful article on this problem, “Inside The Second” more than a year ago. Scott works at a rating facility
and himself realized the issue after years of using the FPS benchmark for testing graphics cards. The video gamer needs the OEM to publish a summary of histograms stating the frequency of stalls on any individual frame beyond a threshold duration
in milliseconds. Buyer beware and happy shopping.
|
