WAV -> MP3 (Lame codec 3.92, RazorLame Front-end)
The situation is identical to the 3ds max 4.26 test. I don't see a considerable
difference in the efficiency between these tests.
VideoCD -> MPEG4 (VirtualDub 1.4.10 + DivX codec v. 5.02 Pro)
The situation is close to the two previous tests, but the advantage
of the Pentium 4 is much brighter. Apart from the shining performance of
this processor (with the SSE2 instructions used) the memory throughput
has a decent effect on the final results, and the P4+PC1066 tandem becomes
unconquerable.
WinAce 2.2 archiving
This is one more sample of the MPEG4-like behavior. There are no severe
changes in the layout, only the results are more dependent on the CPU's
speed.
WinRAR 3.0 archiving
This is one more archiver of relatively slow compression and with a
large library. The Pentium 4 thrives, but the difference between the respective
pairs of the Intel's and AMD's solutions is petty and makes me think that
in both cases the performance is limited by the memory subsystem. The WinRAR
is able to enable entirely the fastest PC1066 RDRAM and the 533 MHz FSB.
Well, it seems to be a very good test for estimating the RAM's performance.
SYSmark 2002
The results of the Internet Content Creation show why AMD tears the
SYSmark 2002 to pieces :). It would be wrong to omit this fact, but it's
wrong twice to speak about an internal structure of the SYSmark subtests
without complete information on the test (or better, its source). AMD doesn't
like the new scripts of this version of the benchmark - it thinks that
the functions and filters used are oriented to those which work better
with the Pentium 4. Frankly speaking, we do not have an answer.
On the other hand, we must discuss not whether the scripts of the SYSmark
2002 are correct regarding certain processors, but whether they are correct
from a standpoint of operation of an average user. If they repeat the most
frequently used instructions, the "optimization for processor" turns into
the "optimization for real usage of software", and in this case the results
become more objective.
That is why from a standpoint of the programmers from SYSmark, the Pentium
4 is really much better than the Athlon XP in programs processing multimedia
contents. Time will show whether this assumption is objective.
SPECviewperf 7.0
Well, if one processor wins from the other by a small margin in several
subtests but loses to it considerably in one test, which one should be
acknowledged a leader? I think the first prize should be awarded to the
programmers developing the IBM Data Explorer code the DX-07 is based on
:). It's clear that one can make a program which will work too slowly on
one processor and quite promptly on another; and it's possible to make
such a program not purposely. But if it takes place, and it started a long
time ago, we should clarify why the IBM Data Explorer doesn't like the
Pentium 4 so much.
Nevertheless, at present we have no choice but to establish the fact
that there are applications which do not like the Intel's processor. Moreover,
they don't like exactly the Northwood core with its L2 512KB cache. That
is why the second prize is given to the AMD Athlon XP for the best compatibility
with one of the SPECviewperf's components. :)
3DMark 2001
The gaming applications will remain in the grip of Intel, to all appearances
(at least, if the Hammer makes no revolution).
The first thought that crosses my mind when looking at the diagrams
is a clear victory of the Pentium 4. Even the 2.53 GHz, which ceased to
be on the top with the release of the 2.8 GHz model, outscores the latest
Athlon XP 2600+. So, with the equal growth rate of frequencies (which are
even lower in case of AMD), the Athlon XP seems to be a hopeless loser.
Return to Castle Wolfenstein and Serious Sam: The Second Encounter
The conclusion will be drawn from the most revealing test conditions -
as the graphic quality level increases the situation doesn't change - it
just becomes more and more "blurry" as the resolution gets higher.
Well, there is nothing more to say after the 3DMark 2001 test. Even
the fastest Athlon XP is not able to catch up with the aging Pentium 4
2.53 GHz, not to mention 2.8 GHz. It proves again the assumption that in
games the palm belongs to the Pentium 4.
The Serious Sam: The Second Encounter changes the situation a little
:). Almost identical performance of the junior and senior processors from
Intel and AMD doesn't allow saying that the Pentium 4 has a sure benefit
in all games, just in the most of popular game engines. Therefore,
the Pentium 4 provides a higher performance only in those games which derive
benefits from its architecture, in particular, from its extended instructions.
Unfortunately for the AMD's fans, the number of such games grows very rapidly...
Conclusion
Before the release of the Athlon XP 2600+ we could state that AMD lost
the performance race. At present, when we have Pentium 4 2.8 GHz and Athlon
XP 2600+ the situation is more complicated, but taking into account the
overall scores AMD still loses. If the Athlon XP 2600+ had coexisted with
the Pentium 4 2.53 GHz for a decent period of time the parity could have
been noticed. But at present, from the standpoint of a pure performance,
i.e. without considering frequency and price differences, Intel with its
Pentium 4 2.8 GHz takes the lead.
But it's too early to bury the Athlon XP: given the slashing results of Pentium
4 2.53 GHz vs. Athlon XP 2000+/2100+ and the today's scores we can see that AMD
managed to make a strong jump ahead narrowing the gap. Well, new changes can be
expected only with new, faster models from both companies.
Vladimir Ribnikov (puree@ixbt.com)
Dmitry Mayorov (destrax@ixbt.com)
Write a comment below. No registration needed!
