Introduction

Right now both ATI and NVIDIA are gearing up for the release of Microsoft’s Windows Vista OS, set to debut early next year. The thinking among both camps is that consumers will be upgrading in droves in order to experience Vista’s 3D Aero Glass interface. Whether or not this actually ends up happening is anyone’s guess: many analysts expected the debut of Windows XP in 2001 would also drive hardware sales, and we all saw how that one panned out. What we can tell you however is that there is one game that’s driving a lot of new hardware sales today; that game is none other than Elder Scrolls IV: Oblivion.

While it’s been on the market now for less than a month, Oblivion has generated a huge following. NPD sales data has Oblivion commanding the top two spots in PC game sales (the Collectors Edition is ranked #2) while Bethesda has also announced that Oblivion is the fastest selling game for Xbox 360.

What all these gamers are discovering as soon as they boot up Oblivion is that the game is more demanding than anyone expected, as we showed you in our performance tests with the latest high-end cards last week, even the latest GeForce and Radeon cards can be brought to their knees if you don’t keep the eye candy settings in check. In hindsight we probably cranked the settings up a little too high for that article, we were running the cards with the highest visual settings available and in some ways that really isn’t feasible unless you’re running two graphics cards for SLI or CrossFire. For our second Oblivion performance article we’re running with more affordable graphics cards in the $120-$250 range. These cards don’t have the shading horsepower and memory bandwidth of the high-end cards we tested with last week, so we’re going to be more aggressive in tweaking the settings for optimal performance rather than image quality. We’ve taken a few comparison shots here:

Highest quality settings

Mainstream quality settings

Compare Images Interactively!

Highest quality

Mainstream quality, big difference here in regards to shadows

Compare Images Interactively!

Highest quality

Mainstream quality

Compare Images Interactively!

One of the chief optimizations we’ve made this time around is reducing the view distance. Oblivion is a wide open game filled with huge environments. It puts an enormous strain on the graphics card when you turn up the game’s view distance as the card literally has to render every strand of grass, pebble, and tree. As you can see in the screenshots there’s a huge price you have to pay in image quality by turning down this setting, but it can have a tremendous impact on performance.

Another setting we were tempted to turn down, but ultimately elected not to was the “iMinGrassSize” setting. One popular tweak that a lot of end users are doing is to turn this setting up from its default level of 80 to 100 or more. This can have a profound impact on performance, but it affects the amount of grass that is rendered. As you recall from our “foliage” demo last week, this puts a huge strain on the graphics card. Since we purposely designed this demo to stress the cards as much as possible we didn’t want to make foliage any less effective at stress-testing the graphics cards so we’ve left this setting at 80. We did however run a few quick benchmarks with a GeForce 6600 GT to illustrate the impact this tweak can have on performance as well as take screenshots:

Grass 80

Grass 100

Compare Images Interactively!

Grass 100

Grass 120

Compare Images Interactively!

Our optimizations for this article went much further than just turning down the view distance though, so rather than go point-by-point through the changes we’ve made, we’re instead providing our Oblivion.ini config file that we used for testing the cards.

One other topic that you guys asked us to take a look at is the performance impact (if any) 512MB of video memory can have on performance. With ATI and NVIDIA both selling 512MB graphics cards many of you are curious if the added video memory improves performance. Let’s take a look shall we?