|[Update: - Originally the PhotoShop version used for the reader's tests below was thought to be v5.5, but due to the very high Radial Blur filter test times for the Athlon, and results of the same filter test with a 1.4GHz (DDR) K7 system, I suspect that PS 6.0 was used. (The reader later verified that Photoshop 6.0 was used.) PhotoShop 6.0 has a bug that makes the Radial Blur filter performance many times slower than PS 5.5 at least on PCs. See the note below the table of results. If using PS 5.5 (or 6.01 I hear) the Radial Blur filter times would have been many times lower, which would have made the total times far less. (Since the PS 6.0 Blur filter time was nearly 50% of the total time for the 21 filter test]|
A reader sent results of PBench's 21 filter test with the 50MB test file option on an overlocked G4/867 Quicksilver system (running 1GHz) versus his 1.6GHz (overclocked 1.2GHz CPU) Athlon PC system (PC133 motherboard, not DDR) running Windows 2000.
"I wanted real world comparison tests, so I left the machines set up as I use them, not as they "should" be benchmarked. Both are overclocked, and both are noisy from the additional fans. There is no doubt that the Mac is faster for processing. But one thing I kept noticing is how fast the Abit KT7A RAID (Athlon MB) is. To bring the Mac loading and saving files up to that speed would mean adding a card and more heat.
Speaking of heat, the Athlon is made from an Antec case and it is heavily modified to keep things cool. The Mac has as good or better planning for air flow from the factory, but suffers from recirculation problems. It has the fans and they are well placed, but it then pulls the hot air right back in.
The Mac was $2400
The Athlon was $1300 (I built it from parts)
[he included the following comment in the email]
I replaced the stock fan this morning in the 1 Ghz G4, as it did not
make the night doing a burn in. Side fan (large one) was .32amp, new is
.5amp and a bit louder, but I want stability and speed, then I will work
I assume the larger fan is a higher RPM (higher CFM) and will ask if he will keep me posted on stability (long term). The table below shows the resuls of each filter test which is run 3 times. (Results in the table were not averaged into one figure for each filter.)
Note: Based on comparisons of a 1.5GHz P4 - the differences between PS 5.5 and 6.0 for the Radial Blur PSBench filter test is huge (appx 6 times slower in PS 6.0 vs 5.5 for that filter test). The Radial Blur test time is a major reason the Athlon score is so high as you can see from the table above (it's nearly 50% of the total time for all 21 filters). [Note - on a 1.4GHz Athlon w/DDR RAM (PC2100) running Win98SE, the 50MB Radial Blur filter test 3 runs were 101.1, 98.6, and 98.7 seconds (Avg = 99.47). If Photoshop 5.5 (or 6.01 I hear) had been used, the Radial Blue and total test time would have been far less than noted in the above table.]
Of course many will ask how the two machines compare in Quake3, Unreal Tournament (games) or other applications. (Photoshop has altivec-like support for the Pentium III and later CPUs but no specific support for the K7/Athlon. The next-generation Athlon due soon will have SSE/SSE2 support I hear.) I asked if he could also run Cinebench (3d rendering/OpenGL 3D cross-platform benchmark) on both systems. He sent the following results which also included a PIII 933mhz system.
Note: Since he had a Voodoo3 card in the 1.6GHz Athlon which has poor OpenGL performance, I included results with a home-brew 1.4GHz Athlon with GeForce3 which also had a DDR (PC2100) motherboard.