PCI Express Graphics Card Review

[Graeme*Kustom*]

Written by Mikael Strand

Introduction

The aim of this article is to give an overview of the performance of our current selection of PCI Express graphics cards. All results should be taken as a guide and only compared with the other results in this article. The results might differ from what you would get using the same graphics card in your system.


Benchmarks:

3D Mark 2005


This is a standard 3D Mark run with a resolution of 1024x768 and no AA or AF.

Battlefield 2, Patch 1.02


We have created a time demo on the Operation Clean Sweep map. The time demo starts with flying over the battle zone in an F15, then some on-foot fighting and finally driving in a buggy across the battle field.
This time demo is run in High detail and Medium detail.

Doom 3, Patch 1.3


We are running the included demo1 as a time demo. This time demo is run with Ultra and High detail settings.

Half-Life 2


We have created a time demo on the Canals_09. The time demo is of the airboat going on the canals and the helicopters flying over, pushing up water drops with their rotor blades. This time demo is run with High and Medium detail settings.

Far Cry, Patch 1.32


We are using the Volcano time demo included with Far Cry. For high end NVIDIA cards, we run this time demo in the High Dynamic Rendering (HDR) setting 1. For all other cards we run the time demo in Very High and High detail settings.

Need for Speed Underground 2, Patch 1.2 UK


NFSU2 does not have an ability to create time demos. We have used FRAPS to collect the frame rate while trying to drive as consistently as possible. We are using a quick race on the resort map for benchmarking. Half way through the second lap it starts to rain. Using nitro on the long straight in the rain put a big load on the graphics cards. Due to the inconsistent nature of this benchmark it should not be taken as an exact measure. The aim of this benchmark is more to see if NFSU2 is playable at the different settings.

Max Power consumption

The power consumption has been measured at the wall socket and is for the whole test system. To get the loads that the system put on the Power Supply Unit we need to take away the efficiency losses in the PSU. Assuming the PSU has an efficiency of 80% the calculation would be as follows:

(Load at the wall)* 0.80 = Load at PSU

Something to bear in mind is that this is a minimum system, 1 HD, 1 DVD and no case fans.

Test platform

Motherboard: ASUS A8N – SLI Deluxe
Power Supply: OCZ Technology ModStream 450W
CPU: AMD A64 3500+
Memory: 2 x OCZ PC3700 Platinum Edition 512Mb
Harddrive: Samsung SpinPoint P80 80gb IDE
DVD-Rom: Sony DDU-1613

Tested grapics cards:

Gainward Ultra 3500 PCX Golden Sample
Club3D Radeon X300 128mb
Gainward Ultra 1960 PCX GLH Edition
Club3D Radeon X700Pro 256mb
SLI 2X Club3D GeForce 7800GTX
Club3D GeForce 7800GTX
Gainward Ultra 1760 PCX GS
SLI 2 x Club3D GeForce 6600GT 128mb
Club3D GeForce 6600GT 128mb
Club3D Radeon X800XL 256mb
Club3D Radeon X850XT 256mb
Club3D Radeon X850XT PE 256mb
XFX GeForce 7800GT

Drivers

For all nVidia cards, except 7800GT, driver version 77.72 was used. For nVidia 7800GT 77.77 was used.

For all ATi, except X800Gt(Rx), 5.7. For X800Rx 5.8 drivers.

[Graeme*Kustom*]

Comments

Having the 7800GTXs in SLI gives a 31% increase, and the 6600GTs give an 82% increase.
The prices are correct at the time of publishing this article. In the higher end of the market the 7800GT and the X800XL give the best value for money. Interestingly the 6600GTs in SLI gives the best value for money in the top half of the 3D mark scores. It might be a good time to upgrade to SLI if you have one 6600GT and an SLI capable motherboard.

[Graeme*Kustom*]

Comments

The 7800GTXs in SLI is again the fastest system. The notable thing is that they are slower than the single 7800GTXs once AA and AF are turned off. I believe this is a result of the benchmark becoming CPU limited and SLI putting extra load on the CPU. The 6600GTs in SLI is between 20% and 140% faster than a single card.

[Graeme*Kustom*]

Comments

Similar results as for BF2. Having the 7800GTXs in SLI only really pays off when turning everything up to max. The 6600GTs give a performance rise of between 7% and 55%.

[Graeme*Kustom*]

Comments

SLI gives a slight advantage, but more interesting is that the Club3D 7800GTX and the XFX 7800GT is faster than the Gainward Ultra 3500 in some of the benchmarks with AA and AF turned on.

[Graeme*Kustom*]

Comments

The 7800GTXs are slower than the single cards across all the benchmarks. I don't know why this is the case, it might come down to CPU limitations, it might be problems with Far Cry not being optimised for GTXs in SLI. In contrast the 6600GTs in SLI give a performance advantage for all the tests.

[Graeme*Kustom*]

Comments

NFSU2 is surprisingly challenging even for the new generation of graphics cards.

[Graeme*Kustom*]

Comments

This shows that most systems would be ok with a high end 400W power supply. Only the GTXs in SLI might need more power than that if you want to overclock and use a higher end CPU.

[Graeme*Kustom*]

Something to consider is that a lot of gamers might be more limited by their screen than their graphics cards. Most gamers are probably gaming on 1280x1024, with only a minority using resolutions higher than 1600x1200.

The 7800GTX is the new king and the 7800GT his prince! They have more than enough power to deliver playable frame rates for 1600x1200 resolutions with everything turned up full in today’s games. Only if you are gaming on resolutions higher than 1600x1200 is it worth considering getting the SLI option. But then again, we don’t know how demanding the next generation of games will be. A system with a couple of 7800GTXs in SLI is as future proof as you can get.