Open1Your1Eyes0
02-01-2010, 09:39 PM
***The Ultimate SLI Test***
http://www.slizone.com/docs/IO/27574/SLI_Art_Contest_Scott_Sarbe.jpg
Introduction
Many people have been wondering about a few questions regarding SLI and required cabling. It has pretty well known that people use an SLI cable in order to run SLI setups but few have ever wondered about whether SLI would even work without a cable. Apparently there have been some sources that claim this is possible and others that deny it. In an effort to provide everyone with factual up-to-date information on it's possibilities. This is exactly the test I will be conducting for everyone.
I have also been noticing a few people running SLI using a Tri-SLI cable (dual connectors) and I was interested to see if there is any chance that running such a setup will increase the performance any. Therefore I will also be testing on whether it makes any difference to use the Tri-SLI cable on a regular SLI setup as theoretically it should increase the performance due to both connectors being in use simultaneously, but it is of course possible that NVIDIA never implemented that ability, therefore I am here to answer all these questions and hopefully anymore that may arise will also be answered.
Testing will include a 3DMark Vantage Professional Edition benchmark run with a Single, SLI (with no cable), SLI (with SLI cable), and SLI (with Tri-SLI cable) test. All test will be conducted using the following system.
Test Setup
Motherboard: EVGA nForce 780i
BIOS Version: P09 (Latest)
CPU: Intel Core 2 Duo E8500 (4.27 GHz)
RAM: OCZ Vista Upgrade 4GB 800 MHz DDR2 (5-6-6-18-2T Timings)
GPU: XFX GeForce 9800GTX (Single + SLI)
HD: Samsung 80GB SATA
PSU: Delta Electronics 800 Watt
CPU Heatsink: Xigmatek Dark Knight HDT-1283V
OS: Microsoft Windows 7 Ultimate 64-Bit
GPU Drivers: NVIDIA ForceWare 186.18 64-Bit
Benchmark Settings
3DMark Vantage Professional Edition: Default Performance Settings
Testing and Results
Single
http://img42.imageshack.us/img42/7494/dscn2093d.jpg
http://img705.imageshack.us/img705/7728/singleq.jpg
SLI (with no cable)
http://img705.imageshack.us/img705/7671/dscn2094.jpg
http://img42.imageshack.us/img42/1702/slinocable.jpg
SLI (with SLI cable)
http://img709.imageshack.us/img709/8167/dscn2096.jpg
http://img705.imageshack.us/img705/369/slislicable.jpg
SLI (with Tri-SLI cable)
http://img709.imageshack.us/img709/6295/dscn2097e.jpg
http://img705.imageshack.us/img705/5357/slitrislicable.jpg
Conclusion
After conducting these tests I was able to deduce that apparently using the cable does in fact make a difference but a very minor one. Obviously compared to a single card there is a huge difference but that's expected. Adding the single cable on gave a minor boost in performance and changing the Tri-SLI one gave no performance increase at all. This means that the Tri-SLI is absolutely useless in an SLI setup. As a summary, I was quite surprised that the cables made such a minor difference but apparently they aren't as important as one would think.
http://www.slizone.com/docs/IO/27574/SLI_Art_Contest_Scott_Sarbe.jpg
Introduction
Many people have been wondering about a few questions regarding SLI and required cabling. It has pretty well known that people use an SLI cable in order to run SLI setups but few have ever wondered about whether SLI would even work without a cable. Apparently there have been some sources that claim this is possible and others that deny it. In an effort to provide everyone with factual up-to-date information on it's possibilities. This is exactly the test I will be conducting for everyone.
I have also been noticing a few people running SLI using a Tri-SLI cable (dual connectors) and I was interested to see if there is any chance that running such a setup will increase the performance any. Therefore I will also be testing on whether it makes any difference to use the Tri-SLI cable on a regular SLI setup as theoretically it should increase the performance due to both connectors being in use simultaneously, but it is of course possible that NVIDIA never implemented that ability, therefore I am here to answer all these questions and hopefully anymore that may arise will also be answered.
Testing will include a 3DMark Vantage Professional Edition benchmark run with a Single, SLI (with no cable), SLI (with SLI cable), and SLI (with Tri-SLI cable) test. All test will be conducted using the following system.
Test Setup
Motherboard: EVGA nForce 780i
BIOS Version: P09 (Latest)
CPU: Intel Core 2 Duo E8500 (4.27 GHz)
RAM: OCZ Vista Upgrade 4GB 800 MHz DDR2 (5-6-6-18-2T Timings)
GPU: XFX GeForce 9800GTX (Single + SLI)
HD: Samsung 80GB SATA
PSU: Delta Electronics 800 Watt
CPU Heatsink: Xigmatek Dark Knight HDT-1283V
OS: Microsoft Windows 7 Ultimate 64-Bit
GPU Drivers: NVIDIA ForceWare 186.18 64-Bit
Benchmark Settings
3DMark Vantage Professional Edition: Default Performance Settings
Testing and Results
Single
http://img42.imageshack.us/img42/7494/dscn2093d.jpg
http://img705.imageshack.us/img705/7728/singleq.jpg
SLI (with no cable)
http://img705.imageshack.us/img705/7671/dscn2094.jpg
http://img42.imageshack.us/img42/1702/slinocable.jpg
SLI (with SLI cable)
http://img709.imageshack.us/img709/8167/dscn2096.jpg
http://img705.imageshack.us/img705/369/slislicable.jpg
SLI (with Tri-SLI cable)
http://img709.imageshack.us/img709/6295/dscn2097e.jpg
http://img705.imageshack.us/img705/5357/slitrislicable.jpg
Conclusion
After conducting these tests I was able to deduce that apparently using the cable does in fact make a difference but a very minor one. Obviously compared to a single card there is a huge difference but that's expected. Adding the single cable on gave a minor boost in performance and changing the Tri-SLI one gave no performance increase at all. This means that the Tri-SLI is absolutely useless in an SLI setup. As a summary, I was quite surprised that the cables made such a minor difference but apparently they aren't as important as one would think.