Z97 PCIe 3.0 x16/x16 CF/SLI?

[Skott]

Looking at the new Z97 mobos. Wanting one that does PCIe 3.0 16x16 but everything I'm seeing is 8x8 when running CF/SLI. Have any been released yet? Thanks in advance.

 

[singhh]

I think MSI's new one, the Z97 Xpower does it...although most of the Asuses don't. I don't know for sure, I'll have to check, and it's not available for a time.

 

[Skott]

ASUS just announced their new Z97 WS mobo that will do the job I'm thinking. I'm wondering witch slots I can use though. 1 & 2 or can I go 1 & 3 to get as much space between cards to get good air flow? I have a Saphire 290x Tri-X OC and would like to add a second one for CF.


http://rog.asus.com/322282014/count...tation-motherboard-with-4-way-graphics-power/

http://www.newegg.com/Product/Produc...82E16813132126

 

[zir_blazer]

What you want does not exist, nor will ever in current mainstream platforms. There are implementations that advertise themselves as 16x/16x, but they use a special switch. You may want to read this (And an old Thread here).

Haswell supports up to three PCIe Slots coming from the 16 PCIe lanes of the Processor itself, and can run in 16x, 8x/8x or 8x/4x/4x modes. Most Motherboards employs bridges that can simply route the lanes from one slot to another depending on if there is a card on it or not, I don't recall seeing any that instead of using bridges has them fixed wired. Fixed 8x/8x should be good and viable because is good enough for vast majority of users (I think still no Video Card saturates PCIe 3.0 8x Bandwidth, at least on previous generation they still weren't there) and removes some added complexity and cost due to the bridge chips.
Motherboards that advertise to have 16x/16x, uses a special switch between Slot and Processor that provides 16 lanes for each. HOWEVER, this switch is limited by the fact than Haswell still has only 16 lanes. The only thing than the switch does, is reroute them on-the-fly based on GPU needs. This means that if you needed at some moment 12 lanes worth of bandwidth at one slot and the other card is Idle, with this switch you would not have a bottleneck that you would on 8x/8x because it rearranges that on-the-fly. However, such scenarios are unlikely to happen, and you still have the 16 lanes limit from Haswell, so if you needed more, you would still have a bottleneck and the switch will be good for nothing. Actually, it is already good for nothing because it adds latency and overhead and no performance benefit, because as most Video Cards can't saturate PCIe 3.0 8x Bandwidth, the scenario where it could be useful shouldn't be possible on the first place.

 

[Galatian]

Hmmm I understood the PLEX chip a little differently:

Since each GPU get's the same data all off what the PLEX chip does is to multiplex the data. Hence it is not really rerouting anything according to the need.

 

[DigDog]

its 8x gen.3, equivalent to something like 32x. you cannot saturate it with any current GPU.

 

[zir_blazer]

Hmmm I understood the PLEX chip a little differently:

Since each GPU get's the same data all off what the PLEX chip does is to multiplex the data. Hence it is not really rerouting anything according to the need.

You could be right, I don't fully understand multiplexing and how the possible variations works. However, I interpret this as routing:

The PLX chip, in hardware, allows the CPU and memory to access the physical addresses of both GPUs. Data is sent to the first GPU only at the bandwidth of 16 lanes. The PLX chip recognizes this, and diverts all the data to the first GPU. The CPU then sends data from memory to the second GPU, and the PLX changes all the lanes to work with the second GPU.

 

[Skott]

Wow, that was some interesting info. Thanks for the links zir_blazer. Trying to wrap my head around all this. I understand some things but kind of confused still about the whole. So, in the end I want a mobo with this PLX PEX 8747 chip, correct? Or whatever its current version is?