home *** CD-ROM | disk | FTP | other *** search
- Andrew Hofman <andyh@erinet.com> writes:
- >>On average, it was
- >>only 1.75x faster than my 21064a clocked at 300MHz.
-
- >Check out the July 1995 issue of VTU, page 50.
- >Two of these scenes are Texture Examples and Space Fighters.
-
- The problem is, these are terrible scenes for benchmarking. I did all
- my tests with the scenes I render from day to day. They are complex and
- designed to be as render efficient as possible. When working with such
- scenes, you will find that the difference is much smaller. Simple
- or inefficient scenes can yield much more unpredictable data, and I
- did see some evidence of 2x increases in a couple tests (but none that
- used material representative of real work).
-
- >The 21064 system you tested must have been exceptionally good
- >to yield that kind of performance vs. a 21164/300.
-
- I will say that its one very quick and cost effective system.
-
- >By the way, exactly how much faster *was* the EV5?
-
- In my tests, worst case was 1.4x, best case 2x, and average was 1.74x.
- And generally, most of my real world scenes hovered very close to the
- above mentioned average.
-
- What I would really like to know is how the 266 MHz systems perform
- when you clock boost them to 300, and if they run reliably at that
- speed. My 275 MHz 21064a has been running rock solid at 300 for nearly
- a year now.
- *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*
- * Mark Thompson http://www.mv.com/ipusers/fusion *
- * Fusion Films, Inc. mark@fusion.mv.com *
- * Director of Animation and Special FX (603) 424-1829 *
- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-