Test 3: Bezier Curves
This test paints the same picture as our line drawing test. The Radeon HD 5870 brings up the rear, reaffirming our suspicion that it isn't able to correctly handle 2D applications. This is completely intolerable, even for consumers.
With Aero graphics turned on and DWM activated, the GeForce GTX 285 delivers nine times the performance. Likewise, even the old on-board graphics chipset blows past the new ATI card. In comparison, the Radeon HD 4870's performance is quite interesting: it is capable of accelerating curves in hardware, even though its results fall behind those from both Nvidia solutions as well.
Test 4: Rectangles
Because the Radeon HD 5870 has clear issues with rendering lines (especially at heavier stroke weights), we saw that half of the results for our rectangle tests echo those observations, without any surprises.
Nevertheless, it’s quite interesting that performance doubles for the Radeon HD 5870 when hardware acceleration is turned on, versus when it’s turned off, even if performance still falls behind the GeForce 7050 GPU.
This rectangle test was the only one where we detected a measurable effect indicating hardware acceleration on the ATI cards, and where it merited such a designation. The Radeon HD 4870 benefited more from this effect than the 5870, even if it trailed the pack with acceleration turned off on this particular metric.
Test 5: Polygons
In this case, the win goes to the older on-board graphics chipset. Nvidia's nForce 610i bested all of the discrete boards by a surprisingly wide margin, both with 2D acceleration enabled and disabled. It was interesting to observe that, for both of the top 3D cards, polygon acceleration didn’t work at all.
The on-board chip, with Aero enabled, runs 10 times faster than the Radeon HD 5870. Without acceleration, the Radeon HD 4870 runs just a tad slower than the 5870. But with Aero turned on, it runs nearly twice as fast.
Test 6: Circles, Arcs, and Ellipses
These results are similar to what we saw above. Both of the high-end cards offer essentially no 2D acceleration, while the on-board graphics chipset scores again. The Radeon HD 5870 falls to the bottom of the pack here, while the older Radeon HD 4870 is somewhere in the middle.
We observed similar results, overall, when we used a Radeon HD 5750, which rules out some defect specific to the 5870. We also compared Catalyst 9.11 and 9.12, noticing a measurable performance increase from the older to the newer version, regardless of whether hardware acceleration was turned on or off. Our next exercise was to compare Windows 7 to Vista, but we don’t want to give away too much of part two. Suffice it to say that even here, we found a few surprises as we worked our way through those tests.
- Part 1: Laying A Theoretical Background
- Windows: Mouse As Window-Washer?
- The Limits Of 2D: One Space With Many Windows
- 2.5D: The Myth Of 2D Hardware Acceleration
- Windows XP: Old School 2D And The Limits of WM_PAINT
- Windows Vista: Real Progress And The Art of Omission
- Windows 7: Return Of The Prodigal Son
- Windows 7: Radeon HD 5000-Series Cards Lack 2D Acceleration
- Tom2D Benchmark: Radeon HD 5870 Vs. GeForce GTX 285 In Windows 7
- Tom2D Benchmark Results, Continued
- Tom2D Conclusions, Preview Of Part 2