MSI GTX 580 Lightning Xtreme SLI
A week or so ago I tested the awesome MSI GTX 580 Lightning Xtreme, that review can be found here. Luckily I had 2 of these cards so the obvious next test was to look at SLI.
Would I be able to achieve the same crazy overclocks with two cards as I could with one and would that additional 1.5GB of VRAM play a more important role in SLI. Read on to find out…
With the 2 cards installed and using MSI’s longer flexible SLI connector (a fixed one would not work due to the design of the card) I booted into Windows to begin some default clock testing. The default clock for the cards are as follows (Core Clock: 832Hz / Memory Clock: 4200MHz (Effective)).
After looking at the cards at stock speeds it was time to move on to the overclocking. Luckily MSI’s Afterburner allows you to overclock both cards at the same time by only changing one value via the ‘synchronize settings for similar graphics processors’ setting.
With both cards running stable at the same overclock (Core Clock: 940Hz / Memory Clock: 4400MHz (Effective)) that I got during the MSI Lightning Xtreme review it was time to start the benchmarks and put some facts and figures together to get some idea of the performance on offer with £1000 worth of GPU power…
The first set of results gives you some idea of the additional performance gained by having 2 of the these cards in your rig. Testing was done at 2560×1440 & 1920×1080; if you are running at lower resolutions than this then you probably don’t need 2 of ’em! 😉
| Setup | Resolution | Score | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Single Card | 1920×1080 (Extreme) | X2234 | 21.5 | 65.0 | 43.5 |
| Single Card | 1280×720 (Performance) | P6767 | 21.5 | 65.0 | 43.5 |
| Dual Card (SLI) | 1920×1080 (Extreme) | X4321 | 22.0 | 69.0 | 47.0 |
| Dual Card (SLI) | 1280×720 (Performance) | P11330 | 21.5 | 71.0 | 49.5 |
| Setup | Resolution | Score | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Single Card | 2560×1440 | 497 | 22.0 | 67.0 | 45.0 |
| Single Card | 1920×1080 | 750 | 22.0 | 67.0 | 45.0 |
| Dual Card (SLI) | 2560×1440 | 977 | 21.5 | 71.0 | 49.5 |
| Dual Card (SLI) | 1920×1080 | 1481 | 21.5 | 71.0 | 49.5 |
| Setup | Resolution | FPS | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Single Card | 2560×1440 | 17.67 | 21.5 | 67.0 | 45.5 |
| Single Card | 1920×1080 | 29.0 | 21.5 | 67.0 | 45.5 |
| Dual Card (SLI) | 2560×1440 | 31.67 | 21.5 | 71.0 | 49.5 |
| Dual Card (SLI) | 1920×1080 | 51.0 | 21.5 | 71.0 | 49.5 |
The second card is providing, on average, an 85% performance increase. The only test letting it down is the 3D Mark 11 Performance test where the performance increase is only approx 67%. But this is mainly down to the low resolution (1280×720) of the test…
This next set of results shows the performance gain after overclocking both cards.
| Setup | Resolution | Score | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Dual Card (SLI) | 1920×1080 (Extreme) | X4321 | 22.0 | 69.0 | 47.0 |
| Dual Card (SLI) | 1280×720 (Performance) | P11330 | 21.5 | 71.0 | 49.5 |
| Overclocked SLI | 1920×1080 (Extreme) | X4801 | 21.5 | 77.0 | 55.5 |
| Overclocked SLI | 1280×720 (Performance) | P12167 | 21.0 | 78.0 | 57.0 |
| Setup | Resolution | Score | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Dual Card (SLI) | 2560×1440 | 977 | 21.5 | 71.0 | 49.5 |
| Dual Card (SLI) | 1920×1080 | 1481 | 21.5 | 71.0 | 49.5 |
| Overclocked SLI | 2560×1440 | 1072 | 22.0 | 77.0 | 55.0 |
| Overclocked SLI | 1920×1080 | 1616 | 21.5 | 78.0 | 56.5 |
| Setup | Resolution | FPS | Ambient Temperature | Max GPU Temp | Delta Temperature |
| Dual Card (SLI) | 2560×1440 | 31.67 | 21.5 | 71.0 | 49.5 |
| Dual Card (SLI) | 1920×1080 | 51.0 | 21.5 | 71.0 | 49.5 |
| Overclocked SLI | 2560×1440 | 34.33 | 21.5 | 77.0 | 55.5 |
| Overclocked SLI | 1920×1080 | 55.0 | 21.5 | 77.0 | 55.5 |
Increasing the clocks (core & memory) yields even more performance from the cards, with the cards now delivering an additional average of approx 9% over the stock performance.
So the performance looks good, but how does it compare to my old (almost stock) GTX 580s found in my rig. Well as you can see the figures are very interesting, one figure is of particular interest…
- Overclocked MSI Lightning Xtreme 3GB SLI vs Palit GTX 580 1.5GB SLI (@ 2560×1440)
| Card | Benchmark | Score |
| Palit GTX 580 (1.5GB) | 3DMark 11 | X3967 *(1920×1080) |
| MSI GTX 580 Lightning Xtreme (3GB) | 3DMark 11 | X4801 *(1920×1080) |
| Palit GTX 580 (1.5GB) | Unigine Heaven | 938 |
| MSI GTX 580 Lightning Xtreme (3GB) | Unigine Heaven | 1072 |
| Palit GTX 580 (1.5GB) | Metro 2033 | 22.67 |
| MSI GTX 580 Lightning Xtreme (3GB) | Metro 2033 | 34.33 |
Overall the performance improvement at high res (2560×1440) is good but the performance improvement in the Metro 2033 test is almost shocking! The data shows a 51% performance increase. This is mainly due to the fact that at a resolution above 1920×1080 the 1.5GB of VRAM of the old cards becomes the bottleneck. The additional VRAM of the MSI Lightning Xtreme allows the cards to power on further, utilizing all of the GPU horsepower the cards offer.
So if you like your gaming at higher than 1920×1080 resolutions and you want what is probably the highest graphics performance (2 cards only) on the market today, then maybe a pair of MSI Lightning Xtremes is the answer. I know it is for me, these Bad Boys are staying in my rig!
