GPU Vs. CPU
Next, we installed all of our test cards on a platform powered by AMD’s Athlon II X2 250. The objective was simple: using such a setup, does it make more sense to upgrade your processor (to an AMD Phenom II X4 955 BE) or a graphics card instead? We've limited testing to just two filters for this test (Chinese Painting and Kaleidoscope).
Offloading the filters and the encoding to the GPU really makes a difference. Accelerating the filters does offer some savings, but the most significant improvement comes from offloading the encoding task.
For those interested, you can see the differences in running these cards on the two tested platforms from the table below.
|Athlon II X2 250 Chinese Painting||Athlon II X2 250 Kaleidoscope||Phenom II X4 955 BE Chinese Painting||Phenon II X4 955 BE Kaleidoscope|
|Radeon HD 5670||313||240||298||211|
|Radeon HD 5770||283||222||265||190|
|Radeon HD 5870 1 GB||231||205||202||167|
|Radeon HD 5870 2 GB||233||205||203||169|
So, is a new GPU a better choice than a quad-core processor for these sorts of apps? The difference between using the two processors with the Radeon HD 5670 and HD 5770 is about 15-18 seconds for the Chinese Painting filter and 30 seconds for the Kaleidoscope filter. Both Radeon HD 5870 cards show a difference of about 30-40 seconds. We would say that it’s worth investing in both upgrades. It would be interesting to see how a more affordable quad–core chip like the Athlon II X4 fares in this test.
These numbers represent total power consumed during the test. They paint a very positive argument for offloading tasks that are well-suited for the GPU. Below, you can see those results compared to those taken with the system equipped with the AMD Phenom II X4 955 BE.
|Athlon II X2 250 Chinese Painting||Athlon II X2 250 Kaleidoscope||Phenom II X4 955 BE Chinese Painting||Phenom II X4 955 BE Kaleidoscope|
|Radeon HD 5670||10||7||12||9|
|Radeon HD 5770||10||8||12||9|
|Radeon HD 5870 1 GB||9||8||10||9|
|Radeon HD 5870 2 GB||11||10||12||10|
Only 2 to 3 Wh separate these cards. Now, let’s look at the average power used for these two tests. We threw in the results for the AMD Phenom II X4 955 BE for comparison's sake.
|Phenom II X4 955 BE Chinese Painting||Phenom II X4 955 BE Kaleidoscope||Athlon II X2 250 Chinese Painting||Athlon II X2 250 Kaleidoscope||Phenom II X4 955 BE Chinese Painting||Phenom II X4 955 BE Kaleidoscope|
|Radeon HD 3300 (GPU not used)||147||133|
|Radeon HD 5670||111||112||147||153|
|Radeon HD 5770||125||126||168||164|
|Radeon HD 5870 1 GB||141||139||185||185|
|Radeon HD 5870 2 GB||169||167||210||210|
Unless you have very tight restrictions on maximum power draw, the differences are not that great. If you're already considering or already using a quad-core processor, such as the AMD Phenom II X4 955 BE, using a GPU like the Radeon HD 5670 still roughly falls within the same power envelope. Though we can’t know for sure without generating the benchmark data, pairing an AMD Athlon II X4 with the Radeon HD 5670 seems to be a very good idea from the perspective of balance. Our experience with an AMD Athlon II X4 620 indicates that you could see savings around 20 W compared to the AMD Phenom II X4 955 BE. The Radeon HD 5770 and both Radeon HD 5870 cards are a different matter, though.
If you prefer to use a dual-core processor, you can employ the Radeon HD 5670, 5770, or even the Radeon HD 5870 1 GB and still remain within the same power draw as the system with an AMD Phenom II X4 955 BE and integrated graphics. The only exception here is the Radeon HD 5870 2 GB. System power draw with an Athlon II X2 250 and the Radeon HD 5870 2 GB is higher than a Phenom II X4 955 BE without a discrete graphics card.