Hi guys,
There is a lot of misinformation around the web with regards to using SLI on a Mac Pro. After looking up all the info on the web, I decided to just take a chance and get a second GTX660. The good news is, that it works!
My Machine Specs:
Early Mac Pro
2 x 3.0ghz X5872 Xeon Processors
16GB ram
2 x EVGA GTX660
My aim was to be able to play all games at around 100fps on a 120hz monitor, although this looks like it may be out of my reach, at least when playing on full graphics.
Here are the steps I took:
1 - Plugged in second GTX660
2 - Attached SLI Bridge
3 - Booted up OSX to make sure all was OK
4 - Booted up Windows
5 - Downloaded HyperSLI 0.96
6 - Followed the installation instructions
7 - You need to boot to the HyperSLI boot option in order to get SLI to work on the motherboard, set this to default.
8 - Once SLI was enabled and detected, I reinstalled the latest Geforce Drivers.
After that it should be working!
The first game I tested was Crysis, and unfortunately there was no performance increase. This freaked me out a bit, but the more I tested other benchmarks and games, it was confirmed that it was working. It appears that either Crysis is maxing out my CPU (it only uses one thread) or not using SLI, I am not sure which at this time.
All the benchmarks I ran were at the highest settings I could do at 1680x1050 (My 120hz monitor resolution). Here are the benchmarks that I have tested so far:
3DMark11
nSLI: 6489
wSLI: 12431
Scaling: 91%
Unigine Heaven
nSLI: 35
wSLI: 67
Scaling: 91%
Crysis 2 dx9:
nSLI: 63
wSLI: 84
Scaling: 33%
Crysis 2 dx11:
nSLI: 53
wSLI: 62
Scaling: 16%
Metro 2033 dx11:
nSLI: 36
wSLI: 56
Scaling: 55%
The scaling for the benchmark tools is amazing, it is a pity it doesn't translate as well into the games. However, Metro 2033 is very impressive, and Crysis2 dx9 is also pretty good. I don't know why the dx11 performance in SLI for Crysis 2 is so "bad". All of these FPS measurements are better than a GTX680, which is way more expensive. Looking at the benchmarks and comparing them to other benchmarks online, it looks like I get similar performance to newer Ivy bridge processors too, which was one of my worries.
There was one major problem. Due to the setup of the PCI slots and the card sizes, the temperature of the bottom card goes up a lot, so much in fact that I needed to change the fan settings (EVGA has a lovely fan speed curve you can edit). Before changing the fan settings the bottom card would go to 90 degrees, which is only a few degrees away from the danger zone. Applying the fan controls the max temperature I've seen has been 83%, and that was during Unigine, it doesn't go up that much during games. Anyway, previously the cards were completely silent, but they are now slightly audible if you are playing games without sound. If you put on your speakers or headphones you won't be able to hear it. When you aren't playing games they are silent as usual.
Anyway, just confirming SLI works and performance is pretty good
There is a lot of misinformation around the web with regards to using SLI on a Mac Pro. After looking up all the info on the web, I decided to just take a chance and get a second GTX660. The good news is, that it works!
My Machine Specs:
Early Mac Pro
2 x 3.0ghz X5872 Xeon Processors
16GB ram
2 x EVGA GTX660
My aim was to be able to play all games at around 100fps on a 120hz monitor, although this looks like it may be out of my reach, at least when playing on full graphics.
Here are the steps I took:
1 - Plugged in second GTX660
2 - Attached SLI Bridge
3 - Booted up OSX to make sure all was OK
4 - Booted up Windows
5 - Downloaded HyperSLI 0.96
6 - Followed the installation instructions
7 - You need to boot to the HyperSLI boot option in order to get SLI to work on the motherboard, set this to default.
8 - Once SLI was enabled and detected, I reinstalled the latest Geforce Drivers.
After that it should be working!
The first game I tested was Crysis, and unfortunately there was no performance increase. This freaked me out a bit, but the more I tested other benchmarks and games, it was confirmed that it was working. It appears that either Crysis is maxing out my CPU (it only uses one thread) or not using SLI, I am not sure which at this time.
All the benchmarks I ran were at the highest settings I could do at 1680x1050 (My 120hz monitor resolution). Here are the benchmarks that I have tested so far:
3DMark11
nSLI: 6489
wSLI: 12431
Scaling: 91%
Unigine Heaven
nSLI: 35
wSLI: 67
Scaling: 91%
Crysis 2 dx9:
nSLI: 63
wSLI: 84
Scaling: 33%
Crysis 2 dx11:
nSLI: 53
wSLI: 62
Scaling: 16%
Metro 2033 dx11:
nSLI: 36
wSLI: 56
Scaling: 55%
The scaling for the benchmark tools is amazing, it is a pity it doesn't translate as well into the games. However, Metro 2033 is very impressive, and Crysis2 dx9 is also pretty good. I don't know why the dx11 performance in SLI for Crysis 2 is so "bad". All of these FPS measurements are better than a GTX680, which is way more expensive. Looking at the benchmarks and comparing them to other benchmarks online, it looks like I get similar performance to newer Ivy bridge processors too, which was one of my worries.
There was one major problem. Due to the setup of the PCI slots and the card sizes, the temperature of the bottom card goes up a lot, so much in fact that I needed to change the fan settings (EVGA has a lovely fan speed curve you can edit). Before changing the fan settings the bottom card would go to 90 degrees, which is only a few degrees away from the danger zone. Applying the fan controls the max temperature I've seen has been 83%, and that was during Unigine, it doesn't go up that much during games. Anyway, previously the cards were completely silent, but they are now slightly audible if you are playing games without sound. If you put on your speakers or headphones you won't be able to hear it. When you aren't playing games they are silent as usual.
Anyway, just confirming SLI works and performance is pretty good