|
Untitled Document
Results
All results are in frames per second with higher being better.
| |
1.8GHz, 256K (Sempron) |
1.8GHz, 512K (Newcastle) |
1.8GHz, 1024K (Clawhammer) |
| Doom 3 (6800) |
49.0 |
50.6 |
52.5 |
| Doom 3 (9800XT) |
27.1 |
27.1 |
27.2 |
| Video Encoding |
27.42 |
28.06 |
28.23 |
Performance difference across the chips is negligible, even when it came to
the Sempron. Most games now are limited by the power of the graphics processor
itself. This is shown by the meager 3.5fps difference across cores. Even when
it comes to video encoding, which is something that is traditionally kept for
high powered machines, the difference is still minor.
The Sempron processor is a lot cooler than the Athlon 64 and all of the chips
are based on the newer CG stepping. This should mean excellent overclocking.
We reached a speed of 2430Mhz (270x9) where we found our motherboard was the
limiting factor rather than the chip itself. With it's base voltage of 1.4V
there is plenty of leeway. We had to increase this to 1.575V to achieve this
speed. With the right board and memory, we can expect to see even higher speeds
possible. Considering you can run the memory at a lower divider, a voltage modification
to the north bridge may be all that is necessary to achieve some outstanding
speeds.
Overclocked Results
| |
2.25GHz (250x9) |
2.43GHz (270x9) |
| Doom 3 (6800) |
51.9 |
52.6 |
| Doom 3 (9800XT) |
27.2 |
27.2 |
| Video Encoding |
33.78 |
36.54 |
These results show how much Video Encoding relies on core speed more than cache.
For video encoding, this makes a 3200+ Newcastle more suitable than a 3200+
Clawhammer.
Conclusion
There is no doubt the Sempron is an excellent chip, with very little performance
difference. In testing done for another publication, we have even seen that
in a work environment it can keep up with the Athlon 64 with at very maximum
of 2% difference. On top of this, the Sempron is an overclockers dream, you'll
spend hours trying to squeeze the most out of them!
Level 2 cache has a different affect depending on each processor architecture,
but in this case it seems to be very minor. This could well be due to the memory
controller being integrated onto the CPU die. This coupled with high speed memory,
means there is less delay when using RAM as a replacement. It could well be,
when running in 64-Bit mode, 256k cache will have a more crippling effect but
this doesn't effect the Sempron.
If purchasing an Athlon 64, go for a Newcastle core. The extra 200MHz given
is a generous compensation for the loss of 512K cache. There is also a much
higher chance of receiving a CG stepping core, which overclocks better. Very
few ClawHammers are based on the CG stepping.
The question is of money. At £88 for an Athlon 3100+, the Athlon 64 2800+
is only £10 more and runs at 1.8Ghz. The Athlon 64 is obviously superior
with more cache and 64-Bit support. Unless they can bring the price down a little
more, the Athlon 64 seems the better deal. If losing 64-Bit support is not an
issue for you, just buy a Sempron and you won't notice the difference.
<< Back | 1 | | 2 | | 3 | Next >>
|
