Hi there, I am planning to build a gaming PC and have done a lot of research on hardware to find out what is the best value for money. I want great performace but don't want to waste money on small improvements or simply to be able to say I have the best gear on the market. So with that in mind I'm looking at the following setup: Motherboard: Asus Sabertooth with P67-B3 Chipset and Mil spec components CPU: Intel Sandybridge i7-2600K 3.4GHz Quad Core, with overclocking RAM: 8GB (2 Pair 2GB) Kingston Hyper-X T1 DDR3 1866MHz,CL9,support XMP, Tall Heatsink PSU: Coolermaster Silent power Pro Gold 1000W, 80Plus Gold, Modular Cabling Case: Coolermaster HAF-X Nvidia Edition CPU Cooler: Zalman CNPS11X Extreme GPU: 2 x Asus ENGTX560 Ti DCii, 1GB 256bit DDR5, 384 cuda cores @ 1.8GHz My question concerns the motherboard and SLi setup. The Asus Sabertooth uses 1 PCIe in 16x mode, but when 2 cards are inserted it runs the 2 PCIe slots in 8x mode. Does this lack in PCIe bandwith cause a bottleneck and will it slow down the 560 Ti cards running in SLi? I could go for the Asus P8P67 WS Revolution board wich supports 16x + 16x + 8x + 8x SLi, but that board costs 30% more. Will I get a 20 to 30% FPS increase running the cards in 16x + 16x vs. running them in 8x + 8x or will there be a small and unnoticable performance increase only? I'm not looking at going for a 3-way or 4-way SLi setup so the extra slots will never be used. Your opinions on the setup and on the PCIe bandwidth issue wil be appreciated greatly! Thank you, Ernst Smit
You prob won't notice the difference tbh....afaik gfx cards still can't saturate PCI-e 16x .....what settings/res do you use normally? If you're on a tight budget then why aren't you looking at a single card setup? Btw if you want any decent help then get a mod to change the title of your thread or else just start a new one....it's a bit vague.
http://www.hardocp.com/article/2010/08/23/gtx_480_sli_pcie_bandwidth_perf_x16x16_vs_x8x8 http://www.hardocp.com/article/2010/08/16/sli_cfx_pcie_bandwidth_perf_x16x16_vs_x16x8 http://www.hardocp.com/article/2010/08/25/gtx_480_sli_pcie_bandwidth_perf_x16x16_vs_x4x4 EDIT: No, the Nvidia NF200 MCP of the Asus P8P67 WS Revolution (or of any other motherboard, for that matter) does not magically add any bandwidth. All it does is distribute, between multiple PCIe x16 graphics cards, the total of available bandwidth - by adding more PCIe lanes between it and the PCIe x16 slots to which it can connect. So no, a x16 link through the NF200 is not at all faster than a native x8 link that runs straight from the Sandy Bridge CPU to a PCIe x16 graphics card. However, since the NF200 adds latency, a x8 link through the NF200 is nevertheless slower than a x16 link through the NF200. On a side note though, the Asus Maximus IV Extreme uses PLX switch technology which makes it able to completely bypass the NF200 (and therefore, to bypass the latency which the NF200 adds) so that the NF200 on it will not (at least, not if you're following the recommendation in the product manual of the motherboard) be used at all until more than just two PCIe x16 graphics cards are installed on it. So basically, current Sandy Bridge on socket 1155 with P67 / Z68 has 16 PCIe lanes built directly into the CPU, which can be used to either run a single PCIe x16 graphics card in native x16 mode, or be split up into two separate x8 links to run two PCIe x16 graphics cards rather than only one. If you'd need more than two PCIe x16 graphics cards on this, that's when an NF200 would start to make sense. http://www.xtremesystems.org/forums...-x16-vs.-Intel-x8-x8-P67-Performance-Analysis
