Sunday, January 5, 2014

New Year's Resolution: 5760x1080

OK, so I have been rocking this new resolution for a while now, but I couldn't help the choice of title; it's an oldie, but it always makes me laugh (geek humour, eh?). Besides I only just got around to writing about this and it's now 2014, so I should be able to safely use the joke, right??? Anyway, as you can see from the above picture, I have been able to set up my primary gaming system with multiple displays, using three Iiyama ProLite E2407HDSD monitors. I swapped out my Iiyama X2377HDS B1 for the three 2nd hand displays, which meant downgrading from the lovely IPS panel, but I felt the gain in immersion was worth the drop in colour reproduction. However, the IPS panel hasn't gone to waste; it's now attached to my secondary system: Phoenix.

Each of three panels only have a VGA and DVI inputs, while my 6950 only has two DVI ports, one DisplayPort and one HDMI port. Looking into the requirements for Eyefinity, I discovered that I would need to purchase an active DisplayPort to DVI adapter for the three displays to work in conjunction with each other. My understanding of the reasoning for this is that there are only two timers present in the output circuitry of the card, which prevent the signal for a third display being synchronised with the other two. Interestingly, it's not possible to use the HDMI port in an Eyefinity display group; not a problem for me, as my HDMI port is connected to my flat panel TV. AMD actually provide a list of tested and supported adapters and in the end I opted for a Startech model, the DP2DVIS:

Here's a picture of the graphics card in my machine, with each output connected to a display: the two DVI and the DisplayPort connected to the Eyefinity displays and the HDMI to the TV:

Configuring the Eyefinity display group in the Catalyst Control Centre was fairly simple. However, I wanted to be able to use the displays in a non-Eyefinity configuration, as this means application windows can be maximised to any one single display, as opposed to across all three, which is the behaviour for an Eyefinity configuration. I also wanted to make switching to the TV a simple task as well, so for each possible configuration, I created a profile and a corresponding shortcut: ctrl+alt+d for the default configuration, ctrl+alt+e to switch to Eyefinity and ctrl+alt+t for the TV.

Performance Testing

Clearly, driving three displays puts more strain on a system; in fact, I noticed that Battlefield 3 became unplayable and I dropped the graphics from "high" settings to "medium" to restore playability. I thought I'd try to benchmark the difference between games running on a single display and three. Unfortunately, by the time I got around to benchmarking, I no longer had Battlefield 3 installed as I have moved onto Battlefield 4, so I opted for the following tests:

  • Unigine Heaven Benchmark 3.0
    • DirectX 11
    • 8x Anti-Aliasing
    • 16x Anisotropic Filtering
    • Shaders: High
    • Textures: High
    • Occlusion: Enabled
    • Refraction: Enabled
    • Volumetric: Enabled
    • Tessellation: Normal
  • Arma 2: Operation Arrowhead - I changed the settings I used for this test, raising them to make it a more difficult test. Instead of listing all the options, here's a screen shot of the settings I used, for both the 1920x1080 and 5760x1080 tests:

  • Crysis - here's another screen capture of the settings I used in the Crysis benchmark tool. When it came to running the tests for the Eyefinity display, I had to launch the actual game first and set the resolution to 5760x1080 before I could run the benchmark as the tool didn't have the appropriate configuration. So, I'm not 100% convinced that this test is entirely reliable, especially after seeing the results; 1920x1080 seemed to run for a shorter duration than 5760x1080. In fact, I noticed that the 5760x1080 benchmark seemed to be running noticeably slower, which made me think there's some issue me testing at that resolution.

  • The Elder Scrolls V: Skyrim - here's the settings used for the tests. Note I didn't install any of the Eyefinity fixes in order to benchmark the system. As I understand it, they are designed to fix UI elements that don't render properly at that resolution. The configuration below shows the 1920x1080 resolution; for the Eyefinity test I had to click "default" and let it automatically detect 5760x1080, as this isn't an option in the menu. After this, I simply ensured the rest of the options matched those used when testing 1920x1080.

The Results

First up, a GPU bound synthetic test the Unigine benchmark.

The average and maximum frame rates are almost exactly what you'd expect given there are three times as many pixels to render: the 6950 is around three times slower rendering at 5760x1080 compared to 1920x1080. However, the minimum frame rate is much worse at the higher resolution; performing approximately five times slower.

Moving onto the real world tests, we should hopefully see some variation in performance depending on the title:

Looking at the frame rates, Arma 2 at 5760x1080 only seems to exhibit half the performance of 1920x1080. However, the frame times and percentile graphs show how much greater the variation in performance is at the higher resolution. Moving onto Crysis, I decided to included these results, even with the reservations I had about them:

The average frame rate here is halved when jumping up to the high resolution, with the minimum and maximum frame rates dropping by a third and almost two thirds respectively. Interestingly, the frame times and percentiles show that the variation in performance at both resolutions is pretty similar. Finally, let's see the Skyrim results:

The frame rates show that the higher resolution provides under half the average and maximum performance, but the minimum frame rate drops almost a quarter of 1920x1080. The frame times and percentiles actually show both resolutions don't really exhibit much frame variation.

Conclusion

It's clear that driving three displays really taxes my system; I've had to drop graphical details to keep frame rates at a playable level. However, I think the extra immersion offered by having my peripheral vision filled by the additional screen more than makes up for it! In addition, I've found that being able to turn my head slightly makes up for the lack of free look in certain titles.

I'm upgrading my PC with a new motherboard, CPU, memory and case and hopefully this will see an improvement in frame rates across the board; particularly in Arma 2 and 3 (which I bought during the Steam Winter sale), which are both CPU intensive. I'm holding off from a GPU upgrade for now, as I'm trying to see whether I'll be staying with AMD or splashing out for an Nvidia card.