Both PC seem to have the same specs, why one is slower?

[jamewoong]

Hello! Me again! hehe

Here's the comparaison of the PCs:


Both use 1GB of RAM (2 X 512 MB). Look at the speed of the encode, the 1st one is like 3X faster than the 2nd!

1st PC (Celeron): took about 572 minutes to encode movie.
2nd PC (Pentium 4): took about 875 minutes to encode movie.

Is this possible? How come it is 2-3X faster?

[tsehou]

Celeron D is clocked slightly higher and has a faster FSB.

Celeron D supports SSE3, which helps to further speed up multimedia encoding.

[willy]

I agree ... I think SSE3 is playing a big role here ...

[jamewoong]

Hein? But the speed is like 3 X faster than the PIV. 2X could be okay, but it is 3X, so it is exagerated, no?

[K10]

Hein? But the speed is like 3 X faster than the PIV. 2X could be okay, but it is 3X, so it is exagerated, no?

Your Celeron D is more likely to be running on newer chipset, and I suspect given the age of your p4 2.0, you might be running i845 chipset or equivalent class that has no dual channel support. Might be 4.2GB/s vs 1.6GB/s bandwidth difference.

You should have just give us all those mainboard info as well.

[Amourek]

Many encoding programs take advantage of SSE3. I bet if you tried a different benchmark like SuperPI you will get very different results. Different applications take advantage of different features of the CPU. Taking one test and saying one CPU is 2-3x faster than the other is asinine.

[K10]

SSE3 is a very minor boast over SSE2 when it comes to encoding.
http://www.behardware.com/articles/5...celeron-d.html



Look at the minor difference between P4 3.0C vs 3.0E.

[Vinman]

Possible reasons:

1) Newer core - streamlined, faster gates, etc. and 16K L1 vs. 8K L1 makes more difference than you think.

2) Faster CPU bus, and therefore much higher CPU and memory bandwidth.

3) Newer platform, faster/better memory support, dual channel, etc.

3) SSE3

[crazdefool]

on top of that the celeron is most likely running the ram at 133mhz and the p4 at 100mhz..
i guess you covered that under the bandwidth point

[Aske001]

Many algorithms are critically dependent on exactly how much cache memory you have. The fact that the Celeron has 16 Kbytes of L1 data cache instead of 8 might just make a huge difference in this one encoding algorithm.