OcuLink Expansion Bay Module

[Aurelien_D]

@nrp

We release documentation and keep interfaces open because we believe in openness, but threads like this explain why many companies don’t.

The problem here is quite simple : the lack of communication. If you’d take the time to say publicly something like ‘Hey Josh, we’re waiting our supplier to give the answer you’re looking for, so… bear with us’ the whole community would be like “Wow ! Framework is amazing !”.

But from the community stand point, only thread/project like this are ‘speaking’. And what do we see ? “Still no message from Framework” and a year and a half of 500+ messages long thread.
What I don’t understand is why so much muteness ? So far, your willingness does not deserve such reaction from the community but the fact is, since there is no big project out (especially around gaming possibilities) of the creators yet, you have to keep us informed to be trusted !

[Josh_Cook]

@Mark_Balliet

@Josh_Cook if I get the M.2 adapter working, is there any data I can get for you that might be helpful?

Nope.

[Gmanny]

I agree, and I think you have a lot of understanding throughout the community that you’re doing a novel and hard thing with not many people.

And many of your customers are invested in the idea of repairability and open hardware, which is also perceived as having a certain level of camaraderie and frankness to the community.

IMO, this doesn’t mean you have to disclose plans ahead of when they’re ready or be really fast at anything. It just maybe means having the authoritative answer quicker, even if it’s something like “We’re investigating this, it’ll take a long time”, repeated every N months if necessary in the case of FL16 thermal issues. Well, if Josh really didn’t receive any reply to his two prior emails, the same reply would work there as well, IMO.

Instead, it’s radio silence, and then when people become unsettled it’s passive aggressive replies like

@Destroya

I didn’t know that Framework was not supporting the creators anymore, where did they come from? is it because Josh Cook said that he didn’t receive a reply to one of his emails? I’m confused.

Which I can also understand from what seems like a sole employee monitoring the forum.

To me, it seems like either the company underestimated the need for people doing different flavors of outside communication (technical, promotional, support), which would need fixing. Or there’s a critical shortage of resources and the communication is not first priority, which I understand you wouldn’t want to disclose in any way.

---

Edit: Maybe, a simpler thing: the novel hardware thing you’re doing would also benefit from a novel communication and support strategy.

I do understand that everything is money, though, so it’s very easy for me to say “do more”, while coming up with guidelines, directing many people to a different mindset and then also finding enough resources for those extra actions is a hard and slow task. As such, I am not pushing here for anything. I am a happy customer, I just feel with this I can help the company become even better.

[MJ1]

Well, put.
Some people understand that even projects that might seem fairly simple in concept can still take a lot of time. Especially when you have other things going on, and you are working with others who have their own things going on.

But a lot of people aren’t so familiar with the issues of projects like these and the time it can take. The basic concept seems simple enough, that it just feels to them like it should have been finished and shipped months ago.

It’s not Framework’s fault, or Josh’s, or anyone else’s. Just misunderstanding of what’s involved. If some more comments from Framework staff would be possible, just that issues are still being worked on, that would probably reduce the misunderstanding of some. Of course, with limited resources, that may just not be possible. And we in the community might need to remind others that things are likely still moving in the background.

[TheTRUEAsian]

@Gmanny

Instead, it’s radio silence, and then when people become unsettled it’s passive aggressive replies like

You’ve misunderstood Des’ response. Des was genuinely confused by Josh’s ping and the subsequent responses as she was not involved prior to this.

@Destroya

I’m not sure why I’m being tagged. Were you expecting an email from me? I don’t see anything in my inbox.

[Mark_Balliet]

As a software developer myself, nothing moves at the pace the end user expects it to, and internally resources are always stretched to their limits. So for me the delays seem reasonable, and expected (though I suck at being patient) but for a lot of people unfamiliar to how the “magic” works behind the scenes, nothing seems to be happening.

It sounds like the disconnect is repaired at the moment, and hopefully information will be more quickly accessible in the future. Keep in mind Framework is a small dedicated company, the projects that bring money in, aka new laptops, new devices, upgrades, etc will always take priority, they can’t just throw infinite people at all the problems that exist. But Framework here is in a unique position, delivering a product that is time consuming to build, and also derives added value from the hard work the community puts into the ecosystem. I hope that in the future, communications can be re-examined so that adequate resources can be dedicated to monitoring creator requests for information to build the community side of the equation, which I think will feed back into Framework’s sustainability and market appeal.

[ATi]

@nrp

Each iteration takes time, since for both Josh and for our engineering team

@MJ1

But a lot of people aren’t so familiar with the issues of projects like these and the time it can take. The basic concept seems simple enough, that it just feels to them like it should have been finished and shipped months ago.

Sounds like a good topic for a Framework Blog article or video to me:
“What’s involved in product development and why does it take (so) much time?”

@Josh_Cook

@Destroya could you ask internally about what is going on with my emails, since it’s seriously not a good look to be “supporting development” yet not responding to an email for 45 days with a relatively simple yes/no question

@nrp

In the last round of emails, Josh asked a question around PCIe that required us to go upstream and check with suppliers who in turn didn’t respond to us. We then did some further research and provided a response based on our best guess of some possibly underdocumented functionality on the CPU.

@Gmanny

It just maybe means having the authoritative answer quicker, even if it’s something like “We’re investigating this, it’ll take a long time”, repeated every N months if necessary in the case of FL16 thermal issues. Well, if Josh really didn’t receive any reply to his two prior emails, the same reply would work there as well, IMO.

Instead, it’s radio silence, and then when people become unsettled it’s passive aggressive replies like

I think it’s dangerous to make any assumptions from the bits and pieces presented in this forum thread regarding this issue, without knowing the whole picture.
Having been employed in IT-Support myself, I can tell you from my own experience, that support tickets tend to get out of sight, whenever the involvement of a 3rd party becomes necessary, which doesn’t reply timely, because the employees usually can’t proceed said issue on their own and get busy with other matters at hand. In such a case the best course of action is imho to ask for a status report every 2-3 weeks proactively. In case there’s no response, other channels like a PM can be justified, but a thread is definately not the right place for direct communication.

As I haven’t heard of any similar issues yet, I think we can treat this matter as a special case.
I plead in favor of refraining from further criticism to stop wasting time from the parties involved for arguments, that would be better invested in advancing the development of the project itself.

[Ethan_Kiers]

I did some googling earlier and it looks like it is really hard to turn 2x 4i into 1x 8i. There seem to be lots of 8i to 2x 4i, but not the other way around. The reason I saw on one forum was that combining two separate pcie connections together required some more advanced timing and switching chips that was not worth the cose or bot easy to develop. Im not an expert and im roighly paraphrasing what I remember from google so I could be totally off.

[Josh_Cook]

@Ethan_Kiers

I did some googling earlier and it looks like it is really hard to turn 2x 4i into 1x 8i. There seem to be lots of 8i to 2x 4i, but not the other way around. The reason I saw on one forum was that combining two separate pcie connections together required some more advanced timing and switching chips that was not worth the cose or bot easy to develop. Im not an expert and im roighly paraphrasing what I remember from google so I could be totally off.

You can simply just say that the PCIe link is 8x and not bifurcate down to 4+4 and it will work fine, assuming you have the m.2 adapters in the correct order, since both halves of the link come from the same chip

[Mark_Balliet]

I ordered the 4i adapter, but now I’m wondering if it might be better to have the 8i adapter, would that be able to be split into 2 4i’s and have an external M.2 drive AND the GPU, while allowing more flexibility for getting an 8i GPU later on?

[MJ1]

@Mark_Balliet

now I’m wondering if it might be better to have the 8i adapter, would that be able to be split into 2 4i’s and have an external M.2 drive AND the GPU, while allowing more flexibility for getting an 8i GPU later on?

Looks like no?

See here:

@Josh_Cook

but with the 8x link, it would normally be splittable, but due to FW16 limitations, it will only act as a 8x link, that is not splittable.

[b473A]

I think what Josh meant is if you have the m.2 + 4i card, but add oculink to the m.2, you could tell your OS it’s actually 8 lanes and the CPU will send data like it’s a PCI-E 4.0 x8.

For what it’s worth, running current-gen graphics cards show little performance loss when run at PCI-E 4.0 x4 compared to x8. As in less than 5% (https://www.reddit.com/r/eGPU/comments/1cpl435/egpu_over_oculink_8i/). I’m happy to let Cunningham’s law do its thing if anyone has better data.

[Philipp_Lamprecht]

Your reply is very much focused on the hard, important work that is being done, but not so much on the importance of perception of this hard work. The post shows that it was about a lack of communication (as I referred to in my previous post and others mentioned now too).

The mission of Framework is sensational, as is the work being done. The fact that this forum exists and that you yourself take time to respond is sensational too. I think Framework has the potential to be a truly unique company.

My point would only remain to please invest a bit more in communication. This could have been exactly a response such as “we are working on it, but it will take some time as we wait for others”. It is difficult for members like Josh to get it right if they do not know what is happening.

PS: And this is also just to add that I am writing this merely as a happy customer, so this is just meant as a suggestion towards one of my favourite companies in tech.

[Xiao_Jin_Ling]

Out of curiosty, has anyone thought of what they would do with the 7700S after hooking up their FW16 to an eGPU?

[Gmanny]

Gaming in trips

[TheLPeink]

I would love to be able to buy an interposer → PCIe adapter to use it as a desktop GPU. There are devboards that do this. I would buy one of these in a heartbeat when the next dGPU comes out. Please Framework allow us to buy these (or see the blueprints so we can order them!)

[Philipp_Lamprecht]

I didn’t buy it, but would certainly buy a RTX 5080 module if it comes out ;).

[Gu_tally]

me too, but i really don’t know the problem of my egpu’s dismal performance, I don’t know if it is just the CPU bottleneck, or the TB3 issue, or the AMD GPU issue (I am using 7900xtx now). Maybe the oculink will tell me what is the problem. Becuase I know for a GPU that is powerful like this, it is going to use all of the pcie 3.0x4, or even 4.0x4, because when I did this with m.2 to pcie adaptor egpu setup with a 6800xt, it already can eat up almost all of the bandwidth of 4.0x4. For even beefier GPU, I don’t know what the result will be but I assume 4.0x4 will have at least some impact to the performance.
But on the other hand, the 7940hs is a really old CPU by today’s standard…