risc said:
Then maybe toms hardware is not a good place for advice, the cpu charts
benchmarks on this site do show the d820 to be the better performer, they
should have included some benchmarks that reflect your opinion then.
The problem with the Tomshardware benchmarks, is they are tuned for
multithreaded performance. In other words, they emphasize dual
core processors, by only using programs that can use both cores
simultaneously. Toms was careful to use only the latest version
of the test programs, getting patches for dual core operation,
so that the benchmarks would no longer show a fast single core
in a good light.
I had to look long and hard at those charts, until I could find
a benchmark that only used a single core on that site.
There are other, smaller sites, that have compared single threaded
performance of dual core processors.
If you look at the Itunes benchmark on Tomshardware, that one seems
to be single threaded.
http://www23.tomshardware.com/cpu.html?modelx=33&model1=486&model2=448&chart=181
Notice how:
Pentium 4 630 3.0GHz single core
Pentium 4 530 3.0GHz single core
Pentium D 920 2.8GHz Dual core
Pentium D 820 2.8GHz Dual core
are all near one another at the bottom of the chart. That means the ITunes
benchmark is not using both cores at the same time. It also means the
ITunes benchmark is not cache sensitive, as those four processors show
cache differences, and it isn't making a difference. Within a few seconds
of performance, that is. That benchmark is only sensitive to equivalent
CPU core clock rate. And by "equivalent", that is why there is a
2.9GHz processor at the top of the chart, because it gives much more
than 2.9Ghz worth of work on a single core. That is called IPC or
instructions per clock -- some processors/families do that better than
others. So you could buy a Core2 Duo with a high clock rate, use only
one of its cores, and still win in a single threaded benchmark.
The Athlon64 4000+ is way up the chart on that one, because the Athlon64
has "close to" a 4GHz clock equivalent of performance.
What other anomalies can present themselves ?
SSE/SSE3/SSE3 issues. Some programs will use SSE on an Intel processor,
but not the equivalent flavor on an AMD processor. For some multimedia
applications, an application can be optimized for one family over the
other. If doing multimedia (rendering, transcoding) you had better find
a web page that reviews the program's behavior, rather than a CPU
performance page.
The designers of the Tomshardware web site, had as their objective, to
enhance the appearance of dual cores. If you have a lot of older software,
and don't plan on buying a lot of new software, it is possible a single
core is a better choice.
Benchmarks can be tilted to prove just about anything
If you are on a budget, and cannot afford a device from the top of the
chart, then you have to tune your purchase, to the software you are
using.
Paul