The data were taken from a recent article entitled
“Half-Life 2 CPU Shootout: AMD versus Intel” from
FiringSquad, http://www.firingsquad.com/hardware/half_life_2_cpu_shootout/page4.asp
. The motivation of the review was to differentiate between Intel and AMD
platforms and evaluate how CPU performance scales with core frequency. To
stress processors differently, three video modes were used, one was low-end
800x600 with almost no special rendering features, the second one was
mid-quality 1280x1024 resolution found in most LCD displays, and the third one
was the top-line 1600x1200 with all bells and whistles.
Below is a typical way how benchmark scores and
performance trends are presented on hardware review websites.

There are two dominant designators on these charts, “Athlon 64”, and “Pentium 4”. Other than that, there is hardly any system in the presented data, . In fact, the “Athlon 64” family is comprised from at least four distinctive members: S754/512k, S754/1MB, S939/512k, and S939/1MB. It takes a while to sort out confusing “model numbers” and attribute them to the right processor family. Obviously, it was a much easier task to deal with Pentium-4 5x0J model numbers. Unfortunately, only one Athlon-64 family member (S939/512k) was represented sufficiently. The member set includes Athlon 64 3000+, 3200+, 3500+, and 3800+. The other family member was Athlon FX55 (S939/1MB), and the Athlon 64 4000+ was assumed to be the other member of the same set. NEXT PAGE will show how the same set of data looks in Time-domain coordinates, when “frames per second” are converted into “average frame rendering time”, AFRT = 1/fps, and processor frequency is converted into processor core clock period.