AMD Athlon 64 for Socket AM2 (part 2)
Benchmarking
To compare the performances of Socket 939 and Socket AM2
platforms, we assembled the following test systems which, apart from
the processors, differed in motherboards and RAM.
Test configuration for Socket 939:
|
Motherboard
|
|
|
Memory
|
2õ DDR-400 512 MB Hynix PC3200
(3-3-3-8) |
|
HDD
|
Samsung HD080HJ (80 GB, 7200 rpm, 8 MB, SATA-300)
|
|
Video Card
|
|
|
Power supply unit
|
SuperPower 480X, 480 W
|
|
Operating system
|
Microsoft Windows XP SP2
|
|
Drivers
|
NVIDIA ForceWare 91.31 WHQL
|
Test configuration for Socket AM2:
|
Motherboard
|
ASUS M8N-SLI Deluxe (nForce 570 SLI )
|
|
Memory
|
2õ DDR2-800 512 MB GEIL PC6400 (5-5-5-15)
|
|
HDD
|
Samsung HD080HJ (80 GB, 7200 rpm, 8 MB, SATA-300)
|
|
Video Card
|
|
|
Power supply unit
|
SuperPower 480X, 480 W
|
|
Operating system
|
Microsoft Windows XP SP2
|
|
Drivers
|
NVIDIA ForceWare 91.31 WHQL
|
Prior to comparison of Athlon 64 Socket 939 versus Socket AM2,
we decided to investigate how sensitive the latter are to the memory
operating frequency. To this end, using the BIOS settings we turned
DDR2-800 into DDR2-667, DDR2-533 and DDR-400 (with the latency timings
set as per SPD) and investigated how the performance changes in so
doing.
GEIL DDR2-800 in
the DDR2-667 mode
GEIL DDR2-800 in
the DDR2-533 mode
GEIL DDR2-800 in
the DDR2-400 mode
Here are the results which we produced:
Judging by these results, the acceleration of memory from
DDR2-400 to DDR2-800 in most tasks gives a 10-20% performance boost, in
2-5% increments between each DDR2 standard.
We leave the most powerful mode, i.e. DDR2-800, and compare it
versus the "old" Socket 939 platform equipped with DDR-400 (3-3-3-8).
To start with, here are the results for synthetic tests:
Since the CPU core has not undergone changes, the performance
does not change much, even under substantial acceleration of the
memory. Judging by the results of synthetic tests, for Socket AM2 we
can expect a small performance boost in only resource-intensive
applications which are demanding primarily for the speeds of the memory
subsystem whose increased clock speeds are eaten up by the raised
latency and probably some shortcomings in the memory controllers. Now
move on from synthetics to practice:
Among the typical multimedia tasks, only the video encoding is
the winner, albeit at the minimum positive result as much as +1%.
But using CAD/CAM systems will be a bit more comfortable which
can be felt in not only the tests – we found one of the areas
where Socket AM2 and DDR2 will prove useful not only theoretically.
The gaming synthetics by Futuremark was unable to enjoy all
the pleasures of DDR2, so it is unlikely to expect essential
performance boost in games, unfortunately.
We had a surprise in Quake 3 which proved very sensitive to
the memory latency and revealed imperfection of the memory controller.
The test became a smooth transition from synthetic tests to the results
produced in modern games.
The Socket AM2 platform disappointed us a bit by the drop of
performance at games – although the result is not much worse
and in some ways the same, but not that as we expected.
Final Words
As our tests have shown, AMD Athlon 64 processors after
acquiring support for the faster DDR2 memory have not only added
performance in most tasks but even lost it a bit. Therefore, it makes
no sense to recommend moving to the new platform. But while assembling
a new system you will have to think twice and answer the question:
«Is that the final system configuration or am I planning for
upgrade in some time?». If in some time there rises the need
to replace the processor, say, with a dual-core CPU. and add more
memory, then the Socket AM2 system will be more promising - it will be
not only cheaper but easier to upgrade. Besides, already now Socket AM2
has given a performance boost in some tasks.
We appreciate LLC
PF "Servis" (Dnepropetrovsk, Ukraine) for the processors and
other hardware presented for tests.
 |
Top Stories: |
 |
 |
|
MoBo:
| |
|
|
VGA Card:
|
|
|
|
CPU & Memory:
|
|
