Just wanted to post some benchmarks, results and "basic" info on SLI style systems because there seems to be alot of confusion regarding SLI performance, price, and configurations. First SLI stands for(3DFX terms) Scan line Interleave that was popularized by 3DFX with their Voodoo2 line of cards. SLI typically offers almost double the performance of the card being used in single form, this ranges from 50%-100% increase in rendering performance and 100% increase in fill rates. For example here is two 8500GT's in SLI configuration being benched against a single 8600GT: http://www.fudzilla.com/index.php?o... As one can see, the 8500GT SLI beats the 8600GT in every 3D benchmark. This is because SLI effectively cuts each GPU load in half. Figures like the shader mark in 3Dmark 06 clearly indicate an almost 100% increase in performance going from a mark of 1158 to 2171 in shader 2.0. While SLI is great for gaming, it will not (for example) turn an 8500gt performance into 8800GT performance. IE. if one receives scores of around 2000 in 3Dmark 06 expect a score of around 3000-4000 in 3Dmark 06 in SLI mode. Another thing is the cost. It costs ALOT of money to buy the SLI equipped board(currently only nvidia chipsets for nvidia SLI), and 2 video cards. HOWEVER! If you want the ultimate gaming experience and you have money to burn, nothing will beat two 8800GT's in SLI mode for gaming purposes. Plus it is fun to experiment with!

http://en.wikipedia.org/wiki/Scalab... nvidia SLI = Scalable Link Interface How does this technology differ from 3dfx's SLI technology? NVIDIA SLI technology differs in many ways. First, 3dfx’s SLI technology was implemented on a shared bus using PCI. The PCI bus delivered ~100MB/sec. of bus throughput, while PCI Express is a point-to-point interface that can deliver ~60x the total bandwidth of PCI. Second, 3dfx’s SLI technology performed interleaving of scan lines, and combined in the analog domain, which could result in image quality issues due to DAC differences and other factors. 3dfx Voodoo technology also only performed triangle setup, leaving the geometry workload for the CPU. This meant 3dfx’s SLI technology only scaled simple texture fill rate, and then used inter-frame scalability. NVIDIA SLI technology is PCI Express based, uses a completely digital frame combining method that has no impact on image quality, can scale geometry performance, and supports a variety of scalability algorithms to best match the scalability method with application demands. http://www.slizone.com/page/slizone...

I believe you tried to use that link to back your theory before. The 8500GT's used in that article have the GPU factory overclocked so stating that 2x 8500GT will beat a single 8600GT in all benchmarks isn't a fair statement unless a note is included about the cards used. A standard 8500GT using the "reference design" has the GPU clocked at 450MHz, but these cards are clocked at 700MHz. That's approx 55% over reference. "The Geforce 8500 GT cards are clocked at 450 MHz for the core and 800 MHz for the memory. However, we tested Zotac's 8500GT cards which are clocked at 700MHz, way higher than the reference board" If you compare two "normal" 8500GT's in SLi against an 8600GT, the 8600GT will win out. Plus it's cheaper, quieter & uses less power. If someone has the money to blow on an SLi config, that's their business but IMO, it's stupid to pair up two low end cards when a better single card is available. If you're gonna run an SLi config, high end cards should be used, otherwise you're wasting your money. The idea is to get better performance than ANY single card will offer, isn't it? Look at the results of two standard 8500GT's 512MB in SLi using Tom's Hardware "Overall Games FPS" chart...they score 636 FPS, not even as good as two 6600GT's 128MB (761 FPS): http://www23.tomshardware.com/graph... Unfortunately, Tom's didn't bother to test an 8600GT. I realize the 8600GTS is a better card, but using the same chart...a single GTS scores 1127 FPS: http://www23.tomshardware.com/graph... The 8500GT SLi config (636 FPS) is comparable to a single 7600GS (624 FPS) or X1600XT (633 FPS): http://www23.tomshardware.com/graph... As a side note, even the 6800GT (663 FPS) is better...LOL!

So i bought a 8600GTS-OC and my motherboard has the ability to support another PCI Express 16x card, but im not sure if the performance will really be worth it with a higher end card like that. Any thoughts? And btw, how exactly does it work when you install 2 cards? I'm guessing you only plug into one card, but both of the cards handle the load of the processing? And is the load evenly split? or does one work harder than the other? I apologize in advance if these questions seem silly, but i'm just trying to have it make sense for me.

8600gts's in SLI mode would equal roughly an 8800GTS in most benchmarks, some the 8800 pulls ahead, some the twin 8600's pull ahead. according to 3Dmark, the twin 8600's always pull ahead, but in actual gaming the 8600's can either lag behind the 8800, or pull ahead depending on the game itself. According to tweaktown, the 3Dmark score was around 10000 for the BFG OC'ed 8600's in SLI mode. around 9100 for non-OC'ed 8600's in SLI. An 8800 scores roughly around 9500-10100 in 3Dmark06 Core 2 Duo 1.86

Well the 3DFX method was ALOT better than the Nvidia method. The 3DFX drivers split the data up between the two cards, one card would process odd lines, the other even lines. The benchmarks spoke for themselves back in those days and no matter what game it was in, SLI mode worked great! Nvidia SLI mode either A: Splits the screen in half, top going to one card, bottom going to the other. or B method: one card renders a frame each, ie card 1 renders frame 1, card 2 renders frame 2, card 1 render frame 3, etc. Core 2 Duo 1.86