View Single Post
  #25 (permalink)  
Old March 14, 2013, 10:28 AM
SKYMTL's Avatar
SKYMTL SKYMTL is offline
HardwareCanuck Review Editor
Join Date: Feb 2007
Location: Montreal
Posts: 13,622

Originally Posted by SugarJ View Post

SKYMTL, is it possible to stretch the charts shown on this page:NVIDIA GTX TITAN vs. SLI & Crossfire
What I'm asking is that instead of charting by frame number, you chart by time so that the comparison at any specific time is more immediately visible? It might make the charts easier to read, rather than have 3 different x-axis lengths on the same chart.
I think I know what you mean but I'm not 100% sure. The FRAPS output is in milliseconds and frame number. It doesn't attribute a time stamp to each data point. However, we also can't forget that most of those charts show upwards of three THOUSAND data points.

Originally Posted by beck2448 View Post
What do you think about a metric that measured how completely a frame is rendered? Some threshold below which a frame would not count in the fps number?
I think it would be helpful but you have to look at how benchmarks are currently being done. Personally, I find there should always be a clearly distinguished line between benchmarking an EXPERIENCE rather than measuring absolutes across a purely scientific spectrum. It begins a slippery slope which starts and ends with a simple question: what's being benchmarked?

I'll give you an example. If a solution is skipping frames in order to achieve a certain result and the end user has no way of seeing it, should it be counted as an ISSUE or a FEATURE? In a way it could be considered cheating but if it benefits gameplay while being completely invisible.

Basically, measuring how "completely" a frame is rendered is quite hard since in some extreme cases, simple screen tearing can be misconstrued as incomplete frame renders. In addition, a gamer typically doesn't see frame rendering issues unless they are accompanied with extremely high latencies.

Sure, you can inject a ton of technical testing into the equation in order to validate this testing method but in my roundabout opinion, if it isn't visible to the end user, it shouldn't be held against a given GPU's end results.
Reply With Quote