CPU Cache vs Ghz for database speed?

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Looking for a standalone workstation PC for a single user running heavy
database queries, likely Windows 32bit. Is it best to have a CPU with
plenty of cache or plenty of GHZ?

AMD X2 4800 with 2x1MB cache is double the price of X2 5000 2x512KB
cache. But would 1MB cache be much faster than 512KB cache?

Also would Intel be significantly better, usually Intel is a bit more

The database system might not have much use for RAM beyond 2GB on
WindowsXP, but cache or Ghz?

Re: CPU Cache vs Ghz for database speed?

Jan wrote:
Quoted text here. Click to load it

I probably don't understand your application all that well. In a
database situation, there is the client (making the query) and
the database itself (the server). The database itself does a lot
more work per query than the client. If both the client and server
were to live on the same machine, then the problem might be
more "interesting". Otherwise, I would expect the client to be
a pretty lightweight application.

To see the impact of cache in the real world, you could look at
some benchmarks. Here, for example, you can see that the 4800+
beat the 5000+, in converting MSword to PDF. For whatever reason,
the extra cache helped here, much more than the clock rate difference.


In other cases, it is just the pure MHz that is helping:


Doubling the cache might be worth several hundred MHz of equivalent
performance. But modern processors are pretty fickle when it
comes to collecting on that promise of performance. There
are some games, for example, where extra cache doesn't help.
Thus, the best insurance, is to select the processor with the
highest clock rate as the first priority (for a given architecture
and instruction per clock rating). If someone is "giving the cache away",
then take it.

I would choose the 5000+ unless I had a benchmark, like
the first (counterintuitive) benchmark result, to guide me to a
different choice.

Also, by looking at that table, you can see that the Conroe/Allendale
processors are faster than the processors you are considering.
That is, even though their absolute clock rate is lower than a lot
of other processors in the list. Conroe has better instruction level
parallelism (higher IPC), and you can only compare Conroe clockrates
to other Conroes. Using a benchmark table like the one on Tomshardware
is one way to discover how they all compare, in general computing

Is it possible to compare existing computers doing queries ?
Do you see any appreciable difference between them ?
Gathering data from your present computing platform may
also help you decide whether client processor speed is a
factor at all. Maybe your most demanding application on the
new machine, is running a web browser.

If the database server is also going to live on that machine,
then you would be best advised to find an entirely different
benchmarking site. Databases are more likely to be benchmarked
on server machines, with an emphasis on supporting multiple users.
You will only occasionally find someone benchmarking a database
on a desktop machine.


Re: CPU Cache vs Ghz for database speed?

Server and Client is the same machine as there's only one user. At the
moment I'm using MS-Access which is not a client-server system in the
same way like SQL, might consider SQL in the future when table volume

Paul wrote:
Quoted text here. Click to load it
Quoted text here. Click to load it
Quoted text here. Click to load it

Site Timeline