带着重重顾虑，我一头扎进茫茫网海，经过几番曲曲折折，终于搜到这则评述，以下是原文：
Intel vs. AMD
PIII or Athlon,
That is the question.
Well Shakespeare, I'm here to tell you that in many cases it just doesn't make a damn bit of difference. I'll pull out some numbers in a moment, but the bottom line is that both the Intel PIII Coppermine and AMD's K7 Athlon make exceptional DAW CPUs.
Lineage
Since the birth of the PC DAW, Intel CPUs have been synonymous with top-performance and stability. Having no serious competition over a several year period, end users were forced (for better or worse) into a Wintel world. Windows applications were optimized and tested for use with Intel CPUs - and those Intel CPUs provided FPU (Floating-Point Unit) performance that was heads and shoulders above units from Cyrix and AMD. And so it was…
Here and Now: Serious Competition
After losing the FPU battle for years, AMD went back to the drawing board. The result is their K7 Athlon which packs some serious FPU muscle. Early reports of the Athlon having 2-3 times the FPU performance of an equally clocked PIII were a wee bit exaggerated, but it easily holds its own.
Finally! Intel's competition has a CPU that's a real competitor!
So… how do they stack up???
Stability
I've logged many hours behind both PIII and K7 based systems and find them to be equally stable.
I'll share a couple of tips. If you choose to run a PIII system, use a BX Chipset Motherboard. They have been in use for quite a while now, so they're mature and refined to near perfection.
The AMD K7 Athlon can reportedly be a bit finicky about power-supply and RAM. For what it's worth, I've never experienced any problems, and I've built quite a few machines. If you choose to run an Athlon system, use a UL Listed power-supply and good quality RAM. These are built to tighter tolerance and will reduce the odds of related problems.
Flexibility
The first generation of K7 Athlon motherboards were little more than prototypes (ie: the original FIC SD-11). This board would not allow both 'master' and 'slave' devices (on the same IDE channel) to function in DMA mode. You had to choose one or the other. Also, the 'Super Bypass' option was not available on first generation motherboards. This option results in about a 10% speed improvement. Newer 2nd and 3rd generation motherboards (ie: Asus K7m) don't suffer these limitations.
The UDMA Hard-Disk controllers on (PIII) BX chipset motherboards use about 2% less CPU than (K7) VIA chipset motherboards. For example, a 40 Gig 7200 RPM Maxtor UDMA Hard Drive will sustain 28MB/Sec on both BX and VIA chipset motherboards. A BX board will use about 2.5% of the CPU for disk I/O (as opposed to a VIA board using about 4.5%).
Thus… the PIII has a minute disk I/O advantage.
Audio Performance: to Each One's Own
I wanted to create the definitive Intel vs. AMD audio test using real-world applications. To make the tests as equal as possible, I used identical hardware (other than the CPU and motherboard) in both machines.
The following hardware was present in both systems:
Matrox G400 video card
256 MB PC100 RAM
40 Gig Maxtor UDMA Hard Drive (dedicated to audio files)
Frontier Dakota audio card
I ran what I call a 'stress-test' in several audio applications running under Win98SE (Cakewalk Pro Audio v9.03, Logic Audio Platinum v4.2.2, Cubase VST/24 v3.71r2, and Samplitude 2496 v5.57a.
The 'stress-test' consisted of:
- 24 solid 24 Bit channels of 44.1k audio (32Bit Float files in Samplitude 2496 and Sequoia)
- 48 bands of Waves' Renaissance EQ (12 instances - each with 4 active bands)
- Timeworks' 4080L reverb (Aux with 4 channels feeding)
- TC Native Reverb (Aux with 4 channels feeding)
- As many Waves' Renaissance Compressors as remaining CPU would allow
Cakewalk Pro Audio v9.03
PIII 600e – ran above stress-test with 4 additional Waves' Renaissance Compressors
CPU @ 71-78%
Disk @ 10-44%
Athlon 600 - ran above stress-test with 4 additional Waves' Renaissance Compressors
CPU @ 71-76%
Disk @ 10-45%
In both cases, latency was set to 104ms (using Dakota's MME driver) and I/O buffer size was 256.
Cubase VST/24 v3.71r2
PIII 600e – ran above stress-test with 7 additional Waves' Renaissance Compressors
CPU @ 95-98%
Disk @ 90%
Athlon 600 - ran above stress-test with 7 Waves' Renaissance Compressors
CPU @ 95-98%
Disk @ 90%
VST was able to sustain the above load using the Dakota's MME or ASIO drivers.
When using MME: latency was set to 464ms and Disk-block size was set to 256k
When using ASIO: latency was set to 12ms and Disk-block size was set to 48k
In both cases: Audio was given highest priority and the number of output channels was set to 64.
Logic Audio Platinum v4.2.2
PIII 600e – ran above stress-test with 2 additional Waves' Renaissance Compressors
CPU @ 95-98%
Disk @ 30-33%
Athlon 600 - ran above stress-test with 4 additional Waves' Renaissance Compressors
CPU @ 95-98%
Disk @ 30-33%
In both cases, Latency was set to 60ms (using Dakota's MME driver). ASIO isn't quite up to snuff in Logic at this time.
Samplitude 2496 v5.57a
PIII 600e – ran above stress-test with 12 additional Waves' Renaissance Compressors
CPU @ 77-78%
Disk @ 31%
Athlon 600 - ran above stress-test with 12 additional Waves' Renaissance Compressors
CPU @ 62-64%
Disk @ 27-28%
In both cases, Latency was set to about 250ms (using Dakota's MME driver).
Pay the Price… Count the Cost
Athlon 600 w/heatsink-fan is about $170
PIII 600e w/heatsink-fan is about $250
Athlon 800 w/heatsink-fan is about $320
PIII 800 w/heatsink-fan is about $569
Note: Celeron 533a @ 800 MHz is about $189 and will very closely approximate the PIII on all audio applications.
As you can see by the numbers… both the PIII and Athlon have strengths and weaknesses. There isn't much difference between the two CPUs when running Cakewalk, Cubase, or Logic. However, if you are running Samplitude 2496 (or Sequoia), the Athlon is about 14% faster than the PIII.
It was interesting to note that Samplitude2496 was able to run the stress-test (with 32Bit Float audio files) AND *12* instances of Waves' Renaissance Compressor at 62-64% CPU use. That leaves 40% CPU left for more tracks and effects… and this is on a 600MHz CPU. Move up to an Athlon 800MHz or 1GHz… and open the floodgates.
NOTE FOR ATHLON USERS: If you run an Athlon system… avoid installing DirectX 7.0 (stay with 6.1 – already a part of Win98SE). Installing DirectX 7.0 will reduce system performance by 5% when running the stress-test.
<IMG SRC="http://www.prorec.com/prorec/editors.nsf/editors/Jim+Roseberry/$FILE/jimr.jpg" border=0>
Jim Roseberry
Product Review Editor