mas,
you are not getting the many factors difference in games you are getting with those subtests so they are not applicable or relevant. It's 6 years old and apps and processors have moved on.
Are you saying that, for example art is cracked beyong the fact that it fits in the cache, or just because it does fit in the cache?
Fitting in the cache can produce a many factor difference for an app, since cache access vs. memory access can have a factor of 10 or so.
As far as games, there is a multitude of algorithms in a typical game, and as cache goes up, progressively more of them fit in the cache. So the impact may not be as dramatic as if there are no memory accesses at all, but still, it is there. NGA receives further benefit in apps that fit in the cache from out of order loads, since the effective cache latency is reduced by the out of order loads.
The bottom line is that a processor, to be considered high end, needs to have high end size of cache, and it is about time for AMD to get on with the program. While Rev F may not have capability of adding L3, I very much hope that Rev G does. Waiting for Rev H to bring L3 would not be a very happy wait for AMD.
Joe |
