GPU Shoot-Out – Part II – Setting New Benches
Part II – GPU Shoot-Out
Setting New Benchmarks with our Top 4 Configurations:
-GTX280, HD4870-512MB, HD4870x2 and 4870 Crossfire-X3
Since our last benchmarking for Part One [here] of our GPU shootout, Catalyst 8-9 and new certified Geforce drivers, Nvidia 178.13, were released. In our previous article, we used Catalyst 8-8 vs. Geforce 177.41 as we examined the performance of five video cards and came to some interesting conclusions. The cards we tested were: HD4870-512MB, HD4870X2-2GB, GTX280, 8800 GTX, and 2900 XT. 8800GTX and 2900XT represented the top and midrange cards of the last generation. The resolutions we test at are fairly demanding for any PC – 1920 x 1200 and 1600 x 1050. We used a combination of ten benchmarks, including modern PC games’ built-in performance benchmarks and custom time demos as well as synthetic tests. For this Part Two, we added a couple of more games and refined our testing slightly. Our CPU for the last article is e8600 at 3.33 GHz which we now overclock for this one; and we always use the DX 10 pathway in Vista 32, whenever possible for this entire series.
-We saw our two older cards – GTX 8800 and 2900 XT – both struggle at 19 x 12 with all the details maxed – and especially with 4x/AA and 16xAF. The 2900 XT – a midrange card of the last generation – could not really handle even 16 x 10 with all the details, never mind adding antialising and anisotropic filtering. Even the venerable 8800 GTX has finally had its day in our opinion, after remaining on top for almost two years. So our new testing will be done with 4870-512, 4870×2, GTX280, and crossfire X-3 (CF-X3 = 4870 + 4870X2). We’re going to determine if 3.33 GHz is enough speed on the CPU and we are also going to use the next drivers, Catalyst 8.9 and Geforce 178.13. There wasn’t that much performance difference between driver releases from both vendors – barely a percentage or two improvement at best, if any, and only for a few specific games; the newer drivers appear to address bugs and 4870X2/crossfire performance in the 8-9 driver release. So we will start right out with a brand new set of benchmarks and compare the performance of Cat 8.9 to Geforce 178.13 and set a new standard. Of course, you should look back at our first article [link] and get a rough idea of what changing out an older e4300 CPU for e8600 [both at 3.33Ghz] did for the top 3 cards and also for Crossfire X-3 [CF-X3 = 4870X2 + 4870]. Since our older video cards, 2900xt and 8800GTX, havw been tested to death elsewhere and are found wanting at the resolutions and in the games we are choosing, we’re going to restrict ourselves to our top three cards and crossfire. We already know 8800 GTX and 2900 XT midrange cards from the last generations simply do not cut it with the latest DX10 games at 19×12 and even 16×10. So we will be testing with 4870/512MB, 4870×2, CF-X3, and GTX280.
This time we’re not going to directly compare performance with Catalyst 8.8 – although you can indirectly infer by looking back at our last charts - but now we are going to establish new benchmarks with Catalyst 8.9/Geforce 178.13, our newly overclocked e8600 and our P-35 motherboard. For our Part III article immediately following this one in our GPU Shoot-out series, we will switch out our PCIe 1.0 P-35 motherboard and began testing with our PCIe 2.0 X48 motherboard so that we will compare motherboard performance to each other using the identical set of drivers and hardware. This way we will see 4870X2 and Crossfire improvements by upgrading our platform and we can make continued comparisons with Nvidia’s evolving drivers for their top card, GTX280. We will be able to clearly see if the platform change will make a huge difference and of course we will need to clock our CPU identically on both boards. So for this article to be consistent with our series, we chose 3.99 GHz as it was the last completely stable overclock on the P35 motherboard with e8600. All other settings are identical to both motherboards and we will be able to directly compare the 1.0 PCIe specification to 2.0 as well as crossfire performance of P35′s crossfire 16x + 4x PCIe to x48′s full 16x + 16x specification.
As an interim, in this article – Part 2 – we’re going to show you the overclocked improvements with Catalyst 8.9 and AMD top configurations and also improvements with Nvidia’s top card, 280GTX over its older driver release. As we continue using Catalyst 8.9 in Part 3 we will show further performance improvements by switching out our motherboard to X 48. Future articles that are actually being written and benching that is taking place right now will include Cat 8-10 vs. Geforce 178.24, and then we compare the Nvidia Big Bang Drivers with Catalyst 8-11 and finally with Catalyst 8-12 – the new “Stream” drivers which was released Dec 12. Soon we hope to have this series completely updated. So this Part 2 will be quite a relatively short article and we will get right to testing:
Catalyst 8-9 vs. Geforce 178.13
As before, we continue to test everything at 1650×1050 and 1920×1600 resolutions. These are probably the most common resolutions that most gamers would be using for upper-midrange and top cards from this generation. As in Part I, we decided to test HD4870-512, HD4870X2-2GB [with both cards also crossfired together as crossfireX-3] and the fastest single GPU, Nvidia’s GTX280. As before, we took some of the very latest games and we benchmark them for you at the most demanding in-game settings, with Vista 32 and in DX-10 whenever possible with 4xAA/16xAF – if they can be set in-game and except for Crysis. These settings are probably the most commonly chosen and difficult settings for new games that a modern PC will run. The games we are testing as the same as the last review including, Crysis, Call of Juarez, F.E.A.R., STALKER, Half-Life2, Enemy Territory: Quake Wars (ET:QW), Unreal Tournament 3 (UT3), Lost Planet, PT Boats DX10 demo, as well as the synthetic benchmarks, 3DMark06, Vantage and Fur benchmarks. We added GRID and FarCry2 to our “shootout” series, also.