DO this means L2 cache in Duron is USELESS???

[zane]

I just install a computer with Duron600. When running a program I got same score with 'L2 cache Enable' and 'L2 cache Disable'. I feel it is odd. Then I do test using Winbench99 and SiSoft Sandra test, got same score again. (in fact, less than 0.5%, sometimes L2 Enable is faster and sometimes L2 Disable is faster). I feel odd and confused because I originally think L2 cache is very important to the speed of CPU. The main difference between Duron and Tbird is the size of L2 cache. DO this means L2 cache in Duron is USELESS??? Anyone know something about this? Thanx.

 

[BurntKooshie]

I don't recall what winbench 99 tests....is it integer or FP information? If its floating point, then it could be that winbench99 isn't very cache dependant. and .5% is well within margin of error.

But the answer is no, the 64k L2 is not useless. It helps it destroy the celeron 🙂

 

[MadRat]

The 128k L1 cache on the K7-family is pretty hard to fill up. The 64k L2 cache is more or less for overflow.

 

[Shudder]

Especially with most benchmark programs, they're probably written with Intel in mind. Meaning, you have 32k of L1 in a Pentium and 128k in an Athlon.

The benchmark would easily use L2 for the pentium but might not even go to the L2 on the Duron/Athlon because it wasn't written for that... 128k would be plenty for that particular benchmark.

I hate benchmark programs because they're only good at determining how much better one speed grade is over another, that's what I personally believe. Duron vs. Celeron is a weak argument. Duron Vs. Athlon is ok. Celeron vs. PIII is ok as well.

Do your own benchmark with UT or Q3 and disable/enable the L2 and see what kind of scores you get. I'm sure then it will be noticeable.

 

[MadRat]

Are you cracking RC5 for the TEAM ANANDTECH? Your Thunderbirds and Durons at any speed would be a welcome addition! 🙂

 

[jinsonxu]

That's the spirit MadRat! Join TA RC5!