Test: AMD Eyefinity x 3

The Test

Following test setup used in this test:

• 3 x Dell 24" inch model 2408 WFP monitors from AMD with DisplayPort, HDMI og DVI-I+D input - 1920x1200 resolution
• 1 x HiS HD5870 2GB Eyefinity 6 Edition graphics card with 6 x MiniDP inputs
• Various cables and adapters composed in different ways for the purpose - 3x1 setup with and without Extended Desktop for BD-ROM on TV

• OS: Windows 7 Ultimate 64-bit
• Driver for graphics card: AMD Catalyst 10.9
• Driver for chipset: Catalyst 10.9 Driver Suite

Test software:

• Stalker:COP DX11 benchmark tool
• Unreal Tournament 3 Benchmark tool -  we used DM-Sanctuary map, 12 bots, sound, full screen, max. details, DX10 
• Furmark 1.7.0 OpenGL 3.2 benchmark tool 
• Unigine Heaven 2.1 benchmark tool DX11 
• Stonegiant DX11 benchmark tool
• AVP-4 benchmark tool
• TessMark 0.2.2 OpenGL 4 benchmark tool
• World of Warcraft
• TotalMedia Theatre 3 BD-ROM player - special version for Eyefinity
• Cyberlink PowerDirector 8
• Cyberlink MediaShow Espresso 5.5 
• Photomatrix Pro 64-bit
• AutoPano Giga 2
• DiRT 2 Demo
• COD:MW2
• CS:Source
• Half Life 2 - Lost Coast Performance Benchmark/Demo
• Adobe CS:5
• Office 2007
• RocketDock 1.3.5
• HD Tune Pro

Because this is a usability testing rather than a performance test, we therefore make things a lot differently than normal.

For example, I have no benchmarks or baselines I can compare with the graphs, since I only have one data base; the used graphics card performance in 3 Eyefinity solutions. This can not be compared properly graphically or otherwise, against anything else I've tested on the same platform. Therefore, emphasis on the use of the solution. Performance is secondary. The user of the solution, buy the hardware it takes to operate with the Eyefinity concept. The consumers limits the performance - not the hardware, other than what is maximally possible technologically at the moment.

First mount the graphics card and connect the monitors properly:

Then select number of screens and the order they stand in purely physical:

Then refine the setup of the screens as the displayed graphics match up with any crevices, etc..:

And the AMD Eyefinity solution is ready!

Live demonstration of AMD Eyefinity at LAN-party (NPF#11):

The demonstration begins:

First volunteers play WoW PVP test in the Battleground:

Next "victim" is a CS:S clan member - he exclaimed spontaneously: "WOAW - I was just 300% better for CS:S with the visual field!"

As seen here, a group had difficulty to detach themselves from the demonstration:

AMD Eyefinity Video Simulator:

AMD has launched a video simulator, where you can see the Eyefinity x 3 experience in different games, versus in only one screen.

Try this link!

Performance results using Eyefinity solution on this PC's hardware:

FurMark and TessMark chosen both because it first uses OpenGL 3.2, while the other uses OpenGL 4.0, but with tesselation effects. Therefore TessMark comparable with a DirectX-11 Tesselation Benchmarks - but tests the graphic solution's performance in OpenGL 4.0 API.

The following settings are used in FurMark and TessMark. Both tests are run over 10 minutes 3 times for the validity fault:

Overclocking & Stability Testing - and why "NO" to "full blast of OC-volt - max. PowerShot!"

TessMark was used during the stability test with the card overclocked 50 MHz on the GPU and 100 MHz on the RAM. Normally running the card factory set at 850/1200 MHz.

There was overclocked to 900/1300 using AMD ATI Catalyst Control Center ATI's Overdrive feature. It is not much compared to what I normally overclock AMD graphics card in my torture tests to find the maximum stable OC border. I have not wanted to squeeze max, since it is borrowed hardware, and simultaneously costing an annual salary if we burn it.

The following screenshot shows the card status at time of writing (eds: note that the OC is disabled):

The system did not crash and the graphic card had no error. The system detected stable.

Power consumption, price, number of screens, and choice of graphics card:

Very important for this article is precisely the power consumption - how much does it cost to use Eyefinity?

3 good screens that can run 1920x1200 on LED technology costs of course, as mentioned, only approx. 4.500 DKK, and 1800 DKK for a HD6870 Eyefinity graphics card, and you are up running.

The couple versions I've seen of really "extreme graphics" (as example, Sapphire Toxic HD5970 4GB Six Edition) costs 9,318 DKK. But you need to go all in then. Check the power consumption below:

The conclusion is that Eyefinity solutions use a lot of power. But my measurements are based on several years old monitors have long been discontinued and specially made for just graphic design, 3D modeling etc. With the use of LED screens I've seen you can get as far down as 30-40 watts per display, where my two monitors use 60-75 watts, while my TV in power "save mode" can be as low as 55 watts although it is the whole 32 inches.

Nobody says you have to run all 3 monitors at the same time, right? I actually only spent two screens mostly under the notification. One for data and one for being productive. Number 3 actually showed only Windows Live Messenger and Outlook. But I would anyway drag Outlook onto the 3rd screen in the middle of answering and checking e-mail. Get the point? :) And with an HDMI switch for a Series-5000 graphics card it is obvious that to TV should be connected to BD-ROM or the media player of any kind. Normally you will not play WoW or CS:S tournament while watching BD-ROM at the same time, right?

The conclusion is crystal clear: A cheap card and 3 cheap monitors can easily handle the task. And since HD6870 1GB to only approx. 1,800 DKK supports the 6 screens, why not give it a try? The possibility is already there - you just need grab it.