P6 pentium 3 EB/S vs core 2 duo (clock for clock)

[noriseghir]

I wonder how much IPC improvement the core 2 arch accoplished against the old P6 arch (let alone the netburst)

how a p3s @ 1.4 Ghz would perform against a C2D @ the same clock speed (in single threaded bench)

 

[Markfw]

I don't have any real stats to back this up, but about 3 to 1. At least 2 to 1. I know the old P4 vs C2D is like 2 to 1.

 

[Adam8281]

When you say a P4 vs C2D is 2 to 1, you're meaning it takes the P4 twice as many clocks to do the same work, right? Not the P4s performance is 2x that of a similarly clocked C2D.

 

[Foxery]

Some hard numbers, stolen from a similar discussion I had on another forum:

FLOPS: FLoating-point OPerations per Second. The result depends on the type of Operations you are testing. Linpack is based on Linear Algebra, Folding@Home involves geometry. Popular ones these days are FutureMark, PCMark, Sandra, etc. Don't know what type of math they run, but the big numbers look exciting, right?

It might be more helpful to show the relative speeds of machines throughout the years, rather than fixating on some magical number to represent "performance." Here are a few common examples, using Intel's own Linpack tool as a benchmark that has been around long enough for me to find results for a ton of hardware.

(Source)
CPU....................MHz....MFLOPS
Pentium II...........450......62
Athlon................500.....180
Pentium III.........1000....316
Athlon-TBird...... .1000....372
Pentium 4...........1700....382 (Note the poor performance vs. Athlons/P3s)
Athlon-Barton.....1800....659 (Marketed as a "2500" vs. a Pentium4)
Opteron-?...........2000....753
Pentium 4...........3066....840
Athlon 64............2200....838
Core 2 Duo........2400...1315 (E6400? Result is from 1 Core only)
Core 2 Duo........3400...1844 (My own overclocked E6750)

Using this particular program as our basis, we find:
C2Duo is 72% faster clock-for-clock than the PIII/1000 result (Coppermine core)
C2Duo is 98% faster clock-for-clock than the last Pentium 4 (This is correct. P4 was a step backwards for Intel.)

Note that this site shows several P3 Coppermine cores as being twice as fast as the original Katmai P3s and Pentium IIs. Either something's fishy there, or this benchmark relies on fast caches - the original PII/P3s had slow L2 caches, if I recall correctly.

I can't seem to live 12 seconds without getting interrupted today, so I'll pick up this up again later

 

[Markfw]

My math based on that says a P3@2400 (if it could do it) would be 758, or 57% os the C2D (E6400 single core) result, so my 2-1 was pretty close. Thanks for that info Foxery.

 

[waffleironhead]

Originally posted by: Foxery


Note that this site shows several P3 Coppermine cores as being twice as fast as the original Katmai P3s and Pentium IIs. Either something's fishy there, or this benchmark relies on fast caches - the original PII/P3s had slow L2 caches, if I recall correctly.

Would fsb speed account for that as well as the cache? Original p3's were at 100 whereas they upped the fsb to 133 for the tulatins.

 

[Foxery]

Hmm, this is a big stretch on the ol' memory. I think there were Coppermines with both 100 and 133 MHz FSBs. The example given at 600 Mhz was *probably* running 6x100. The major upgrade to the Pentium III line was from a 512KB, half-speed, off-die (I think?) L2 cache, to a 256KB, full-speed, on-die L2 cache. i.e. On the original 500 MHz chip, the L2 cache only ran at 250 MHz.

Tulatin was a different beast, and a failed product line, (as in they didn't advertise it and no one bought it) which came out after the P4 was introduced.

Oddly enough, I have my own old P3/450 still in the house. Overclockable to 550. If I get some time, I'll run Linpack on it