By,
Dave Altavilla
March 22, 2004
We fired up
Sandra as a quick sanity check on synthetic performance
metrics. We ran three of the built-in sub-system tests
that comprise the SANDRA 2004 suite (CPU, Multimedia and
Memory). All of these tests were run with our Prescott and
Northwood Pentium 4 CPUs at default clock speeds of 3.4GHz
|
SiSoftware SANDRA |
Synthetic CPU and Memory Benchmarks |
|
P4 Prescott 3.4GHz
CPU TEST
|
P4 Prescott 3.4GHz
MULTIMEDIA TEST
|
P4 Prescott 3.4GHz
MEMORY TEST
|
P4 Northwood 3.4GHz
CPU TEST
|
P4 Northwood 3.4GHz
MULTIMEDIA TEST
|
P4 Northwood 3.4GHz
MEMORY TEST
|
As far as
Sandra's basic Integer and Floating Point performance tests
go, in their "CPU" and "Multimedia" modules, Prescott is
simply a small step behind the Northwood Pentium 4, clock
for clock. As expected however, Prescott leads the way
slightly in the Memory test, due mainly to its larger on 1MB
of on chip cache. Unfortunately for Prescott, Sandra's
code fits fairly efficiently in Northwood's 512K of on chip
cache, so the only saving graces for Prescott are its more
accurate branch prediction unit and improved Hyperthreading,
both of which aren't quite enough to make up for its deeper
31 stage pipeline, which on a branch miss adds significantly
to its overall latency.
From a different
perspective, however, we were actually impressed at how well
Prescott performs versus an equivalently clocked Northwood,
given the fact that its pipeline is so much deeper.
Intel has made great strides to compensate for the penalties
of a branch miss with its 31 stage Prescott core versus
Northwood's 20 stage design. Now all they have to do
is ramp clock speeds in order to take advantage of it.
As with many things in life and in silicon, it's always
easier said than done.
|
FutureMark PCMark 2004 |
Synthetic CPU and Memory Benchmarks |
|
Again, on a
purely synthetic performance scale, we have put together
some scores with FutureMark's PCMark04 benchmark.
We ran two of the modules in their test suite, the CPU and
Memory tests. First up is the CPU test and here's what
FutureMark says this test performs.
"The CPU test suite is a collection of tests that are run
to isolate the performance of the CPU. There are nine tests
in all. Two pairs of tests are run multithreaded each test
in the pair is run in its own thread. The remaining five
tests are run single threaded. These tests include such
functions as file encryption, decryption, compression and
decompression, grammar check, audio conversion, WMV and DivX
video compression."
The Pentium 4
team sweeps cleanly here, with our 3.4GHz Prescott CPU
holding its own fairly well, just a small percentage point
behind the 3.4GHz Northwood core. Interestingly
enough, the 3.4GHz Extreme Edition P4 drops in slightly
behind the Northwood core but by a completely insignificant
margin.
Details on
PCMark04's Memory Test are as follows:
"The Memory test suite is a collection of tests that isolate
the performance of the memory subsystem. The memory
subsystem consists of various devices on the PC. This
includes the main memory, the CPU internal cache (known as
the L1 cache) and the external cache (known as the L2 or L3
cache). As it is difficult to find applications that only
stress the memory, we explicitly developed a set of tests
geared for this purpose. The tests are written in C++ and
assembly. They include: Reading data blocks from memory,
Writing data blocks to memory performing copy operations on
data blocks, random access to data items and latency
testing."
Clearly the
Athlon 64 FX's integrated memory controller is giving the
FX-51 and FX-53 processors an edge in this test.
However, the Athlon 64 3400+ can't compete with even the
3.2GHz Northwood P4. In terms of memory bandwidth, the
new 3.4GHz Prescott based P4 delivers a potent offering,
coming within striking distance of even the high priced
Athlon 64 FX-51 and FX-53 processors.
Business and Content Creation Winstone 2004 and XMPEG |