The Microprocessor: Revolutionary New Architecture: Dream or Reality?

Each processor was benchmarked using Ziff-Davis Publications' Winstone 98, Winbench 98, and id Software's Quake 2.   Each test was run a total of 3 times and an average of the three scores was taken and recorded as the final score.  Each test system used 1 x 64MB Mushkin PC100 SDRAM DIMM, a 5.1GB Western Digital UltraATA HDD, a Matrox Millennium G200 (AGP) Video Card, and the latest drivers for all of the devices installed in the test system running under Windows 98.  The Quake 2 tests were conducted in software mode and used no OpenGL or 3DNow! drivers.  No processors produced any seemingly erroneous results and no crashes were experienced during the test7ing process.

The Results

Explaining the Results

Shocking results once again wouldn't you say?   Once again Cyrix steps up with the best overall Business Application performer with their M-II that consistently beat the competition in both a core-to-core comparison and a clock for clock comparison.  Removing the L2 cache from the Cyrix test system left the computer relatively unaffected compared to the competition which is why the M-II still remains strong in the tests without L2 cache.

For raw FPU performance, the two Intel chips still bring the numbers to the table.  Without L2 cache, the Pentium II still stands strong with the most powerful FPU which is why the cache-less Pentium II, the Celeron, has been so successful among gamers.  The Pentium MMX isn't too far behind the Pentium II as far as FPU performance goes, however the difference between the two processors grows to a more noticeable degree once you remove the L2 cache. 

The AMD K6-2 is the most surprising chip in this roundup, as its performance varies greatly courtesy of its incredible reliance on the performance boost the 100MHz FSB gives its L2 cache.  With the L2 cache enabled, the K6-2's benefits from the 100MHz FSB are quite visible as it produces a CPUMark32 score of 631, even greater than that of a Pentium II.  Because of this the K6-2 also gives the M-II a fair challenge under Winstone 98, falling behind by a total of 1.2 Winstone points.   What happens when you take the L2 cache away from the K6-2?  Then the system is no better than the older K6 system used in the original Evolution Article.  The performance drop is incredible between a K6-2 with and without L2 cache, great enough to fully illustrate the need for the K6-2 to be run at a 100MHz FSB.  A surprising result is the performance of the K6-2 under Quake 2 without L2 cache enabled and  without 3DNow! drivers, the 7.9 frames per second the K6-2 cranked out was easily beat by the 8.6 the M-II test system managed to pull.  Even with L2 cache enabled the K6-2 only offers a 0.2 fps improvement over the M-II illustrating the second reason for the K6-2's success, its 3DNow! instructions which boost its performance under supported games to Pentium II levels.

Conclusion

Luckily you don't have to run your processors without L2 cache, and luckily all of the above processors aren't locked at 200MHz.   With the Pentium II roaming around the 450MHz mark, and the K6-2/M-II available in 300MHz (or PR2-300) flavors these tests do nothing more than illustrate how processors change over time, very little.  What do you think the difference between a Pentium II 400 and a 450 is?  Not much.  Once you approach a new processor generation then you begin noticing differences, however until then most of the architectural differences among CPU's are negligible.