Jump to content

Allocating precious PCIe lanes

I am planning a build with an i7 5930k CPU and an ASUS X99 A II mb, so together, I will have 40 available PCIe lanes. Tentatively, I plan to use 16 for a GPU for the display(s) and 16 for a GPU for programming with CUDA. I can think of several things to do with the remaining 8, but I don't yet understand enough to make a decision. Do M.2, U.2/NVMe, or SATA express use any of those 40 PCIe lanes? If so, is that lane occupancy with respect to the CPU, mb, or both? Also, are PCIe wireless adapters significantly faster than USB ones? Using PCIe for wireless seems silly because I expect the bottleneck to be transfer over a network.

Link to comment
Share on other sites

Link to post
Share on other sites

I believe most pcie lanes come from chipset while basically only graphics cards pull from the CPU. I am not on my PC to link you an x99 chipset layout to confirm but if someone else could.

Community Standards | Fan Control Software

Please make sure to Quote me or @ me to see your reply!

Just because I am a Moderator does not mean I am always right. Please fact check me and verify my answer. 

 

"Black Out"

Ryzen 9 5900x | Full Custom Water Loop | Asus Crosshair VIII Hero (Wi-Fi) | RTX 3090 Founders | Ballistix 32gb 16-18-18-36 3600mhz 

1tb Samsung 970 Evo | 2x 2tb Crucial MX500 SSD | Fractal Design Meshify S2 | Corsair HX1200 PSU

 

Dedicated Streaming Rig

 Ryzen 7 3700x | Asus B450-F Strix | 16gb Gskill Flare X 3200mhz | Corsair RM550x PSU | Asus Strix GTX1070 | 250gb 860 Evo m.2

Phanteks P300A |  Elgato HD60 Pro | Avermedia Live Gamer Duo | Avermedia 4k GC573 Capture Card

 

Link to comment
Share on other sites

Link to post
Share on other sites

GPUs will step down to x8 each no problem and you're talking about 0.000001% reduction in performance between x8 and x16 on dual GPUs. So you don't have to worry about lanes.

Current Network Layout:

Current Build Log/PC:

Prior Build Log/PC:

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Lurick said:

GPUs will step down to x8 each no problem and you're talking about 0.000001% reduction in performance between x8 and x16 on dual GPUs. So you don't have to worry about lanes.

For displays and rendering, sure. But for CUDA programming, bandwidth is the biggest bottleneck, so I think I'll end up wanting x16 for the non-display card.

Link to comment
Share on other sites

Link to post
Share on other sites

It's up to the motherboard manufacturer how those "up to" 40 pci express lanes coming from the processor are split to various parts of the motherboard. Some use 32 of those on two x16 ports, some use only 16 lanes and automatically configure the two pci express x16 slots to x8 speeds when two cards are inserted.

Some higher end boards use a 4 lanes for an onboard 10 gbit network card, some use 2-4 lanes for extra usb 3.0 ports on the back of the motherboard..

I would guess socket 2001 boards would use up to x8 for NVME (ssd on pci express), they may use some lanes (x2 or x3) for some m2 slot for SSDs , maybe even use a couple of lanes for an extra usb 3.0 sata controller.

 

Onboard audio and onboard gigabit network cards are usually connected to the chipset, which creates 8 pci express 2.0 lanes to which onboard audio, additional gigabit network cards, extra sata controllers, extra usb controllers  can be connected.

 

Chipset Block Diagram.png

 

Link to comment
Share on other sites

Link to post
Share on other sites

4 hours ago, mariushm said:

It's up to the motherboard manufacturer how those "up to" 40 pci express lanes coming from the processor are split to various parts of the motherboard. Some use 32 of those on two x16 ports, some use only 16 lanes and automatically configure the two pci express x16 slots to x8 speeds when two cards are inserted.

Some higher end boards use a 4 lanes for an onboard 10 gbit network card, some use 2-4 lanes for extra usb 3.0 ports on the back of the motherboard..

I would guess socket 2001 boards would use up to x8 for NVME (ssd on pci express), they may use some lanes (x2 or x3) for some m2 slot for SSDs , maybe even use a couple of lanes for an extra usb 3.0 sata controller.

 

Onboard audio and onboard gigabit network cards are usually connected to the chipset, which creates 8 pci express 2.0 lanes to which onboard audio, additional gigabit network cards, extra sata controllers, extra usb controllers  can be connected.

 

So then I should do my research and definitively find out:

 

1. Which PCIe sockets on the X99 A II connect to the PCIe lanes on the CPU

2. Which of the 40 mb PCIe lanes don't connect to the CPU

3. Which sockets just use PCIe-ish technology without counting as part of the 40 lanes on the mb.

 

Trouble is, the X99 A II manual doesn't have this information, and the PCIe schematics I've found online don't tell me how the lanes relate to the mb layout diagram. Where can I look?

 

Clarification: for any given slot/socket on the mb layout diagram in the manual, I need to know how many formal mb PCIe lanes are used, how many formal CPU PCIe lanes are used, and how many of those mb PCIe lanes connect to CPU PCIe lanes.

Edited by goof doofus
Link to comment
Share on other sites

Link to post
Share on other sites

9 hours ago, goof doofus said:

 

So then I should do my research and definitively find out:

 

1. Which PCIe sockets on the X99 A II connect to the PCIe lanes on the CPU

2. Which of the 40 mb PCIe lanes don't connect to the CPU

3. Which sockets just use PCIe-ish technology without counting as part of the 40 lanes on the mb.

 

Trouble is, the X99 A II manual doesn't have this information, and the PCIe schematics I've found online don't tell me how the lanes relate to the mb layout diagram. Where can I look?

 

Clarification: for any given slot/socket on the mb layout diagram in the manual, I need to know how many formal mb PCIe lanes are used, how many formal CPU PCIe lanes are used, and how many of those mb PCIe lanes connect to CPU PCIe lanes.

You don't need to figure anything out. Go download the manual for your motherboard. In there it will specify which slots run at which speeds under certain conditions, and other things like if the M.2 conflicts with anything else (as in, if you use the M.2, the the Sata Express or U.2 get disabled)

 

Edit: I actually missed you note about it not being in the manual. But I found some info on the product page:
https://www.asus.com/us/Motherboards/X99-A-II/specifications/ Read the sections "Expansion Slots" and also the "Notes" at the bottom of the page.

 

EDIT2: There is also pages 1-7 and 1-8 (pages 23 and 24 of the PDF) in the manual.  http://dlcdnet.asus.com/pub/ASUS/mb/LGA2011/X99-A_II/E11090_X99-A_II_UM_WEB.pdf?_ga=1.93037013.998312872.1469432619

Looking to buy GTX690, other multi-GPU cards, or single-slot graphics cards: 

 

Link to comment
Share on other sites

Link to post
Share on other sites

4 hours ago, brwainer said:

You don't need to figure anything out. Go download the manual for your motherboard. In there it will specify which slots run at which speeds under certain conditions, and other things like if the M.2 conflicts with anything else (as in, if you use the M.2, the the Sata Express or U.2 get disabled)

 

Edit: I actually missed you note about it not being in the manual. But I found some info on the product page:
https://www.asus.com/us/Motherboards/X99-A-II/specifications/ Read the sections "Expansion Slots" and also the "Notes" at the bottom of the page.

 

EDIT2: There is also pages 1-7 and 1-8 (pages 23 and 24 of the PDF) in the manual.  http://dlcdnet.asus.com/pub/ASUS/mb/LGA2011/X99-A_II/E11090_X99-A_II_UM_WEB.pdf?_ga=1.93037013.998312872.1469432619

Yup, 1-7 and 1-8 are the most useful pages in the manual on this.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×