|
Untitled Document
Performance
We already went into gaming performance a little in our PCI
Express article, and showed how little difference it currently makes. When
you remove PCI-E and 64-Bit processing from the equation, what we basically
have here is a dual Prescott system. This means performance should be similar
to a Prescott with DDR-2 memory. For those of you who don't want to wade through
our PCI Express article, here are the results of the nVidia 6800 Ultra card,
compared using a 3200+.
| |
3200+ |
Dual
Xeon |
% Increase (61.45) |
| 3DMark 2003 |
11,709 |
12,058 |
+3% |
| Aquamark |
64,960 |
64,960 |
0% |
| % Increase |
|
|
+1.5% |
| Far Cry (Low) Average FPS |
63.95 |
63.04 |
-1% |
| Far Cry (Medium) Average FPS |
55.81 |
58.80 |
+5% |
| Far Cry (High) Average FPS |
42.65 |
47.42 |
+11% |
| % Increase |
|
|
+5% |
| Halo (Low) Average FPS |
72.75 |
93.25 |
+28% |
| Halo (Medium) Average FPS |
73.34 |
82.76 |
+13% |
| Halo (High) Average FPS |
58.62 |
64.73 |
+10% |
| % Increase |
|
|
+17% |
| Call of Duty (Low) Average FPS |
105.2 |
105.1 |
0% |
| Call of Duty (Medium) Average
FPS |
99.3 |
99.6 |
0% |
| Call of Duty (High) Average FPS |
91.2 |
90.6 |
0% |
| % Increase |
|
|
0% |
| Average % Increase |
|
|
+6% |
The results above can be attributed to a combination of the
PCI Express/ Dual Xeons or DDR-2 technology. Note that this is a 3200+ (1024k)
compared to the 3400+ (512k) used below.
Considering this is not the sort of machine you are going to be using for
Gaming, we felt some video encoding tests should be performed. We used XMPEG
and 20,000 frames from "The Mask" DVD. We encoded this using DivX
PRO 5.1.1 at 780Kbps Video and 128Kbps MP3 Audio. We used the fastest iDCT which
on both test systems was SSE2.
The second test system was an MSI K8T Neo with an Athlon 64 3400+ (512k) coupled
with two 512MB Corsair CMX512-3200XLPRO memory modules. We removed one of the
memory modules from the Nocona system so they both had 1GB of memory.
We ran the tests 3 times and they were very consistent with very little variance.
The results are in FPS (higher is better).
|
Athlon 64 |
Single Nocona |
Dual Nocona |
36.26 FPS |
35.92 FPS |
41.77 FPS |
The results are slightly surprising. Not only were we expecting the Xeon to
perform faster, we were expecting the DDR-2 memory to give a boost in performance
as well. The Athlon kept up surprisingly well, more than likely helped by it's
low latency memory (compared the CAS4 DDR-2). XMPEG doesn't seem to be particularly
SMP aware, having only a 16% increase in performance when running in Dual (only
35% CPU usage was shown in task manager). We expect to see more of an increase
in performance with other applications.
After seeing the performance in our PCI Express article and the video encoding
performance shown here, it really does show how good the Athlon 64 processor
is. A dual Opteron system could well be a better performer, considering its
dual channel memory and theoretically better 64-Bit performance.
One thing the Opteron doesn't have just yet, is PCI Express, so this really
does give Intel a head start. For now.
Conclusion
The Xeon platform has made quite a jump. From 533MHz, DDR and AGP to 800MHz,
DDR-2 and PCI Express in one go. Considering workstation and server users
aren't going to be upgrading as regularly as enthusiasts, it is a good idea
to do these changes all at once. Looking at this system does give us a glimpse
at what to expect from a future 64-Bit desktop CPU. It is a pity we couldn't test the 64-Bit performance. Hopefully we can follow up with this in the future.
<< Back | 1 | | 2 | | 3 | Next >>
|