I just received confirmation from Futuremark that 3DMark03, in its latest form, renders a number of effects differently from run to run. According to Patric Ojala, Futuremark's Senior Manager for Benchmark Development, the following objects are affected:
That a number of different objects are rendered slightly differently in each benchmark run explains many of the supposed image quality anomalies associated with NVIDIA's 52.16 drivers and the latest build of 3DMark03. The only screenshots I've seen with image quality differences highlight the above objects.
- The smoke puffs when the doors open & close in game test 2
- The magic flame of the sword in game test 3
- The girl's hair in the face close up camera of game test 3
- The biggest explosion in game test 1 when the two bombers collide, or actually one single particle system there
NVIDIA's optimizations still violate the company's own internal optimization guidelines, and indeed Futuremark's optimization rules for 3DMark03. However, it looks like the optimizations are at least breaking the rules without compromising image quality.
|Geil lights up its Evo X ROG-certified RAM||4|
|Google Compute Engine is now powered in part by Pascal||10|
|EVGA slaps 12 GT/s memory on the GTX 1080 Ti FTW3 Elite||14|
|G.Skill unleashes AMD-ready Trident Z RGB kits up to 3200 MT/s||14|
|Asus' ZenFone 4 Pro offers high-end photography and networking||21|
|Radeon 17.9.2 drivers put the pedal to the metal for Project Cars 2||4|
|ROG Strix X299-XE Gaming motherboard is rather groovy||4|
|Miniature Golf Day Shortbread||18|
|GeForce 385.69 drivers are Game Ready for a ton of titles||2|