Battle at 500MHz: Athlon vs P3 vs Celeronby Anand Lal Shimpi on September 21, 1999 8:36 PM EST
- Posted in
You can easily make the argument that the performance difference between the three processors comes from the Athlon's 200MHz FSB versus the 100MHz FSB of the Pentium III. However we have also learned that, in a single processor system the bandwidth offered by the 100MHz FSB is not as great of a limiting factor as we might be made to think it is.
The earliest benchmark reports of the 133MHz FSB and the Pentium III don't indicate the stellar performance increase that some manufacturers were hoping for. While the situation is a completely different one when you take more than one processor into account, the performance differences among the three processors compared here cannot be attributed to the differences in FSB frequency alone (the part FSB frequency plays is a very small one in comparison).
With the two Intel CPUs sharing the same architecture, the superior design of the Athlon shines unsurprisingly clear among the three processors, especially in FPU intensive applications. While it is very odd to be praising AMD for a superior FPU, the truth stands that the P6 core is outdated in comparison to AMD's creation and thus the raw FPU power of the Athlon is greater than that of both of the Intel solutions.
Without getting into debates over architecture and design, there remains one basic factor that contributes to the performance or lack thereof with these three CPUs. That factor? The cache.
The two Intel processors are quite similar because are both based on the same core. However their differences grow out of their L2 cache configuration. The Pentium III (and the Pentium II) makes use of a 512KB L2 cache located off the processor operating at 1/2 of the clock speed of the CPU while the Celeron includes an integrated 128KB L2 cache running at the clock speed of the CPU. What this brings about is the classic debate of whether a larger but slower L2 cache is better than a small but faster L2 cache. The true answer to that question depends on the type of application that you're using to assess what is "faster."
For business applications that can fit in the smaller 128KB L2 cache of the Celeron (the two processors share the same 32KB L1 cache so we'll leave that comparison out of this), the Celeron will be the faster processor. This makes the Celeron the better processor for your basic set of business applications since they don't need the extra 384KB of L2 cache the Pentium III offers. At the same time, in situations where the application(s) do not fit into the 128KB L2 cache of the Celeron, they must go to the system memory which is obviously much slower than the L2 cache. If they do happen to fit into the 512KB L2 cache on the Pentium III, then those applications (whether it is 1 or more than 1 application) will experience greater performance on the Pentium III platform.
Outside of business applications, we have a number of high-end applications that can experience similar performance differences. As you'll be able to see by the benchmarks, there are some high-end applications that benefit from the larger L2 cache while others don't exhibit a noticeable difference in performance.
The most noticeable difference comes when you are running configurations of dual Celerons versus dual Pentium IIIs. What you will notice in those situations is that in situations when quite a bit of data is being manipulated (i.e. 3D rendering) the larger L2 cache of the Pentium III setup will be favored over the smaller but faster L2 cache of the Celerons. The severity of the performance difference is increased, in favor of the Pentium III, by the slower 66MHz FSB of the current generation of Celeron processors.
On the reverse side of things, if there isn't a lot of data manipulation taking place then there will be relatively no difference between the dual Celeron setup and the dual Pentium III setup.
So what happens when we add the Athlon to this Cache comparison? The Athlon is a unique creation in that it's L1 cache is four times the size of the Celeron/Pentium III and also features a L2 cache equal in size to the Pentium III. What we end up dealing with is a processor that has very little difficulty fitting an entire business application within its L1 cache. And for applications that cannot reside completely within that 128KB the processor also boasts a full 512KB of L2 cache operating at 1/2 its clock speed to rely on.
With the Athlon, we have the best of both worlds: a large L1 cache to help in business application performance and a large L2 cache to help in high-end application performance. Both of these pave the way for a strong multiprocessing platform with the Athlon.