| Article Index |
|---|
| GeForce 8800Ultra Performance Comparison |
| Page 2 - 3DMark06 |
| Page 4 - Prey |
| Page 5 - Lost Planet Demo |
| Page 6 - Conclusion |
| All Pages |
The new GeForce 8800Ultra GPU from nVidia is their fastest ever, but it includes a price premium of
around $250 over their 8800GTX range, which is no slouch itself. Today at OzHardware we take a look at the new 8800Ultra and see if the extra cost is justified.
When we look at the specs of the 8800Ultra, it simply looks like an over-clocked 8800GTX, the core of the 8800 Ultra is clocked at 612MHz, up from 575Mhz for the GTX, and the same GDDR3 memory has been clocked up to 1080MHz, up from 900MHz of the GTX. It has the same amount of memory - 768MB, the same 128 stream processors and same 384-Bit memory interface. The additional clock speed of both the GPU core and memory do give a boost to memory bandwidth (103.7GB/s vs. 86.4GB/s) and texture fill rate (39.2billion/s vs. 36.8billion/s), but when comparing a $1200 video card against a $950 video card, is it really that big of a difference?
Today we’ll be testing the Gigabyte GeForce 8800Ultra video card and putting it through the standard DirectX 9 tests such as 3DMark06, Quake and Prey, as well as the new DirectX 10 benchmark – Lost Planet. Lost Planet is still in demo form for the PC, as it is currently being ported over from the XBOX 360, but at the time of writing this article it is the only DirectX 10 capable game that has the ability to be benchmarked.
We will be comparing the 8800 Ultra to the 8800GTX, as well as some mid range cards, the 8600GT and 1950Pro, to show real world performance differences. Going on the difference in clock speeds of memory and GPU between the Ultra and GTX, we expect to see around a 5-10% performance increase across the range.
The Test Setup:
- Intel Core 2 Duo E6600 CPU
- ASUS P5N32-E SLI, nForce680i Motherboard
- Corsair TWIN2X 2GB PC2-8500 Memory Kit (2 x 1GB)
- Seagate Barracuda 7200.10 250GB SATA-II Hard Drive
- Antec Basiq 500W ATX Power Supply
- Windows XP Pro SP2 for DirectX 9 Tests
- Windows Vista Premium for DirectX 10 Tests
The first test is 3DMark06:
From these results we see very little difference between the Ultra and GTX in the standard test, only around a 3% difference, however when they are stressed more with the inclusion of 8x Anti-Aliasing into the test, then the gap grows, out to around 8.5%, which is more in line with what we’d expect between these two cards.
{mospagebreak title=Page 3 - Quake 4}
Next: Quake 4:
Unlike the 3DMark06 test which was only run at a single resolution of 1280x1024, we will be running Quake 4 at resolutions from 800x600 up to 1600x1200, as well as 8x and 16x Anti-Aliasing modes. As a side note, the 1950Pro does not support above 6x Anti-Aliasing, so it was unable to compete in the AA tests in this case.
From the graphs shown above, we can see that in the standard (non-Anti-Aliasing) mode, there is very little difference between the cards, even the lower range cards are very close behind. This is more a limitation of our test bench’s CPU power (combined with a very intense demo) than anything to do with the video cards themselves. In the 8x and 16x Anti-Aliasing modes we really see the cards differentiate themselves from each other, with the 8800Ultra getting over the GTX by around 10% in the higher range tests.
Next: Prey
The Prey results show very little difference between the 8800Ultra and GTX cards, with the biggest difference showing up at 1600x1200 with 4xAA, and even then it’s only around an 8% difference, all the other tests are so close they’d almost be considered identical, which really doesn’t bode well for the 8800 Ultra.
Next: Lost Planet Demo
Lost Planet Extreme is a demo of an upcoming PC game release. The game itself was originally designed for the XBOX 360, and what is most important in this case, is as the developers have been porting it over to PC, they have been programming it for both DirectX 9 and the new Vista-Only DirectX 10, which is suppose to bring along with it some great new effects capabilities and finally something to test the capabilities of the 8000 series graphics cards from nVidia.
As this test requires DirectX 10 capability from the graphics cards, the 1950Pro was unable to compete. The benchmark has two stages, a segment out in the open with minimal game characters but lots of weather effects, explosions etc. and wide open spaces called “Snow” and another that takes place in an enclosed area with dozens or hundreds of game characters on the screen at any one time called “Cave”. We see the results below.
In this test we only see a difference in the “Snow” test of around 11%, so finally we see a significant difference between the two, however there is no difference at all in the “Cave” test between the GTX and Ultra, so you really have to start thinking if this new card is worth the price premium, it’s not looking too good.
Conclusion:
The test results we’ve seen today haven’t really impressed us at all. While it’s obvious that the 8800 Ultra is faster than the GTX, the real issue here is the price difference between the two and whether it’s justified by the gains you receive, and I just don’t believe it is. We’ve seen a speed difference of only a few percent, 10% at the most and even that was rare, measured against a 25% price premium. You could most likely get the same results by overclocking your GTX, or by buying pre-overclocked version or even certain Golden Sample versions, which most likely would still cost far less than the asking price of the Ultra.
If you’re after a high performance card today, I’d recommend sticking with the 8800GTX until the next generation of nVidia cards arrive, which will hopefully have a more significant speed increase than what we have seen here today.
The GigaByte 8800 Ultra we used for this review can be purchased online at www.techbuy.com.au (Direct Link)
