I have heard from various (semi-reliable) sources that most motherboards that are sold with a x16 PCI Express slot specifically intended for graphics adapters will not work at full width with other PCI Express devices.
Has anyone heard of this? For example, if I want to build a server-style system where graphics performance is not a concern at all, but storage system throughput is, I would install a PCI graphics card (or use an on-board one) and put a x8 RAID controller in the x16 Express slot. Let's say this board, for example.
I have been told that many boards of this type will negotiate down to a x1 Express link when there is not a graphics adapter in the slot. (and that the write buffers are only implemented in the downstream direction of the Express switch in the MCH) However, I have not seen this documented anywhere on the motherboard manufacturer's site or in Intel's chipset documentation. Also, perusing through the PCI Express 1.0a spec does not specify any method for doing this.
Anyone know anything specific about this, or have a source that will definitively tell me one way or the other?
Has anyone heard of this? For example, if I want to build a server-style system where graphics performance is not a concern at all, but storage system throughput is, I would install a PCI graphics card (or use an on-board one) and put a x8 RAID controller in the x16 Express slot. Let's say this board, for example.
I have been told that many boards of this type will negotiate down to a x1 Express link when there is not a graphics adapter in the slot. (and that the write buffers are only implemented in the downstream direction of the Express switch in the MCH) However, I have not seen this documented anywhere on the motherboard manufacturer's site or in Intel's chipset documentation. Also, perusing through the PCI Express 1.0a spec does not specify any method for doing this.
Anyone know anything specific about this, or have a source that will definitively tell me one way or the other?