Framework 16 Suggestion regarding x8 link

I would like to strongly suggest a x16 link on the framework 16 laptop back bay modules instead of x8 link. The reason is that while x8 might be good enough for a gaming GPU, but I would like to have the option of 10 gigabit Ethernet on the back, as well as a GPU. x4 + x4 might not be a great GPU experience for many games.

1 Like

10G networking could be run off of one of the Thunderbolt ports for the expansion cards most likely.

1 Like

A module combining 10 gigabit LAN plus a decent GPU sounds very exotic and unlikely to happen. I don’t think you need to worry about that.

7 Likes

A module combining 10 gigabit LAN plus a decent GPU sounds very exotic and unlikely to happen. I don’t think you need to worry about that.

There’s lots of laptops with Ethernet. It’s actually really weird to have a “gaming laptop” without Ethernet. 10G is standard for content creation and it’s better to not have an adapter hanging off the laptop.

Seems like framework plans to only provide a mid-range laptop? I’m not sure how an x4 link would do for gaming as I don’t think it’s possible to do an x7 + x1 link for even gigabit ethernet. (maybe it is? I don’t know pcie that well)

the Ethernet Expansion card is 2.5G and doesn’t exactly hang off the laptop.

3 Likes

PCIe switches exist, so you could run a 8 lane GPU and a 8 lane LAN adapter on the same 8 lane upsteam connection. They obviously would need to share the bandwidth, but it shouldn’t be a huge problem in most scenarios. This would of course raise the price of the module by about $20.

I think connector supports pcie16x, however only 8 lines are connected, as there is no more lines provided by the cpu. Mobile CPU has less PCIE lines than desktop CPU.

The expansion bay connector only supports PCIe x8, as you can see here on the second page:

https://github.com/FrameworkComputer/ExpansionBay/blob/main/PCB%20Reference%20Design/FW_EXPANSION_BAY_PCB_0.pdf

Look for connection names like PEX_RX?P or PEX_TX?N where the question mark resembles the PCIe lane. You’ll see it only goes from 0 to 7 (so 8 different lanes).

3 Likes

Looks like the 13 Gen Intel HX support 16 PCIe G5 and 4 thunderbolt . Should be enough for 16x to the expansion bay?

Even if some processors have limited IO, why limit the infrastructure of the design

Remember that you also have to allocate PCIe lanes to things like the Wifi module and nvme slots for the SSD(s).

I’m sure Framework will have thought this through.

4 Likes

The chipset has PCIe lanes for those usages

1 Like

I know that Framework now officially had Thunderbolt 4 now but with the capabilities of this MoBo has anyone thought of an M.2 -to-Oculink option for a eGPU instead of the slow and sometimes unstable Thunderbolt 4 eGPU option? I think that Framework could with a bit of modification to create a path for a Oculink cable(via a PCI Express M.2-Key Cable adapter) from one of the (2) M.2 ports on of Framework’s (6) module ports would work great.
I mean if you are wanting to Game with a big desktop card then why settle for the devaluing that GPU with a inferior Thunderbolt 4 line?

If you don’t know what OcuLink is then please check it out. This is a good introduction to it vs TB4

This DESTROYS Thunderbolt - Oculink Makes eGPU Feel Like Desktop!

What are your thoughts on this?
Thank you

2 Likes

So am I correct in saying that the Framework 16 supports x8 lanes external PCI output? If so then all someone needs to do is make a custom PCB designed with OcuLink 8i connector. For a far superior eGPU.

Or is OcuLink the connecter being used for the Framework’s eGPU module?

Some please respond.

This is exactly what Framework’s CEO is saying.

2 Likes

Superior apart from not having hot plug or power delivery or any of the other amenities, but performance wise definitely.

1 Like

Yes, this is true but why else have a eGPU and underpower it Just because one is too lazy to do a little power down sequence?

I mean like, we’re all running solid state drives here right? How long does it really take you to restart your computer. Also typically using a external GPU you’re going to be stationary anyway.

I want my Framework 16 to have the possibility to be a graphical powerhouse if I need it to be. I would like to run The more high-end graphic cards if I wish to on par with or at least close to desktop performance. Thunderbolt 4 doesn’t provide that.

Does that make sense?

2 Likes

I want my Framework 16 to have the possibility to be a graphical powerhouse if I need it to be. I would like to run The more high-end graphic cards if I wish to on par with or at least close to desktop performance. Thunderbolt 4 doesn’t provide that.

I don’t think it’s really realistic to expect similar performance as a desktop in a laptop (look at the rediculous power draw from the rtx 4000 series, you’d have to cram all that cooling capacity in the expansion module).
Additionally, remember that this is new. No other laptops have this, so it’s gonna take a while for vendors to start making things for it (if at all, depending on the market share).

They are talking about connecting an external GPU using an adapter in the Expansion Bay module to provide a OcuLink connector.
https://community.frame.work/t/framework-16-suggestion-regarding-x8-link/28300/14

I am personally ready to take quite a performance hit to not have to close all my stuff down every time I switch from stationary to portable but different people use tech differently.

If they could get pci hotplug working, that would be the holy grail though, having to plug in an additional cable for power and peripherals is completely acceptable to me.

In the end I just find the technology neat, not gonna get a 16 as it’s way too big to count as a notebook to me.

2 Likes

This is where we all need to consider the evolving tech or in this case the OcuLink was actually created back in 2015 but was not standardized and used mostly for servers. Its only in the advent of the Chinese (from my understanding) using it in BitCoin mining that the true potential for eGPU’s was realized. The question is still on how to supply a full ATX sized PSU to the RTX 4000 (or any High-end card) beyond 250W. One Dock has the solution and will be selling their One Dock 1.5 eGPU dock by the end of August. Unlike their original One Dock which has to be modded to support anything over 250W PSU the new one will support ANY size GPU and Natively support Desktop ATX PSU’s to accommodate the power hungry GPU’s of the higher-end.
For those of us who do not mind docking a small portable daily driver laptop or tablet that sports a spare PCIe slot connected to an OcuLink Port than this is a dream come true.
I am planning to run my AMD Radeon RX 7900 XTX from the One Dock 1.5 when it comes out.
As you can see, I am very excited :slight_smile:

2 Likes