Hi,

I have a question that has been bothering me for a while. I have a workstation with an AMD Athlon 2100+ processor (1.73 Ghz, 266 Mhz FSB, 256 k L2 cache) and a laptop with an Intel Centrino M (745) processor (1.8 Ghz, 400 Mhz FSB, 2Mb L2 cache). Both have 512 Mb DDR RAM, a hard drive running at 5400 rpm (udma 5) and a specifically compiled 2.6.9 kernel.

One would expect the laptop to be faster or at least as fast as the AMD processor when doing the same calculations but the AMD seems to have the overhand every time. For example, both have the same mySQL server installed with exactly the same database. I use the workstation for large queries that take a lot of time. If they both do the same query on the same data (2 copies of course) the AMD is a lot faster then the Centrino. I can give lot's of other examples ...

Does this mean something is wrong with the setup on the laptop (for example the speedstep technology which is compiled and seems to work) or is it well possible that the AMD workstation is just faster for some hardware reason?

thank you for your help,


maenho

AIUI, an Athlon 2100 is considered equivalent to a 2.1gig Pentium. That's where the numbering scheme comes from. So it should be the faster of the two.

ok that sounds reasonable,

I does make you wonder how AMD leaps over an Intel with 2.1 Ghz with an 1.73 Ghz processor though but I guess that would lead us off topic

Well I'm going to take it a bit off course with a history lesson.

The discovery that "megahertz don't matter" isn't new. In fact during the days when Cyrix was alive a 200mhz cyrix processor was just as fast as an Intel 300mhz processor. The thing is, that it wasn't until after the 1Ghz race that AMD found a creative way to express that their lower Mhz=a faster processor. That's what the whole x000+ scheme is about. Instead of posting the actual clock speed, they post the speed as it closely relates to the Intel equivelant speed. Also if you look at a list of AMD clock speeds as they relate to their name, a lot of times you'll have something like this (below is an example not an actual list:

AMD 2100+=1.7Ghz
AMD 2200+=1.8Ghz
AMD 2300+=1.6Ghz

No that's not a typo. A lot of times AMD will implement new technology that will increase the computer speed without increasing, and at times decreasing the actual clock speed.

A rule of thumb. Your computer is only as fast as its slowest part. Example: You could have the fastest processor but if you have slower RAM it'll bottleneck your processor.

AMD processors are very, very good when it comes to processing RAW data. When they are processing multimedia data, they are not that fast. Though the AMD Athlon 64 processors changed all that. Now AMD is the king in every data type you want it to process.

The bottleneck is the hard drive in every computer.

For database, more cache will speed up the transactions. If you have a controller with expandable cache, try increasing it. If you are using one hard drive, try to find models with 8 MB of cache, but do not mix it with controllers with on-board cache. Try to increase cache for your database server.

"For example, both have the same mySQL server installed with exactly the same database. I use the workstation for large queries that take a lot of time. If they both do the same query on the same data (2 copies of course) the AMD is a lot faster then the Centrino. I can give lot's of other examples ..."

The speed of large database queries is almost entirely dependent on disk access speed. The difference in speed between the two machines is most likely dependent on where the records are located in the data base and/or how efficiently the data base physical structure is laid out.

The fastest computer will be the one that has the least total disk arm movement during the comparison tests.

----------------------------
Steve Stites

I do have a question related to the disk activity. I'm running a large query on the AMD and it's been running now for 2 days. The SHOW PORCESSLIST gives status: Copying to tmp table. One would expect the hard drive activity to be high but gkrellm shows there's only sometimes a small peak of disk activity no heigher then 50K. CPU usage is constantly at 100% though. Does this mean something is wrong with the query?