Comment by paulmd
3 years ago
Try asking the same enthusiasts who are adamant that switching to USB-C is no big deal and legacy cruft needs to be eliminated what they think about switching to a new PCIe power connector.
Forget the 12VHPWR problems, we had a push for a new connector last gen too and it was universally reviled for some reason. People do not want a new connector, let alone a new voltage rail to help move that power a little more safely, they just wanna keep doing 8-pin forever even when it swells to 4+ connectors and starts dictating PCB layouts etc.
If people want to be evenhanded and honest about seriously looking at legacy standards, it's time to have a chat about ATX and PCIe add-in-card form factorand their relevance in the 21st century. It's time for a new connector and new voltage rails and new form factors for GPUs in general, that standard dates back to the original AT PC (ATX is literally "AT eXtended") and a PC no longer looks anything like it did in 1980.
The "add-in-card" is now pulling more than the actual CPU and the growth of the coolers makes supporting them physically difficult, and the chaos in the market has made solutions impossible because there is no standard for physical dimensions or connector plug-points etc. This is textbook, it's the exact same situation as pre-USB connector standards, everyone is doing their own sizes and layouts and it's chaos. Why should we not have government intervention to impose some common standards and get everybody on the same page so solutions can be engineered and we move forward? Why should a GPU not be a 6x4x12 inch box (or a set of other standard sizes) that slides as a module into the case with high-current connectors in a standardized location? No more GPU sag, no more 4x8pin, just one XT90 jag that contacts as you slide the module into its receptacle on a rail. Sounds like heaven to me.
EU should look at giving a standards body a mandate to move forward past that and we can work towards mandatory adoption of the new standard and outlawing the sale of legacy ATX within 5-10 years. That sounds extreme but it's what we did for USB-C, right? But it's different when it's your sacred cow and you have to pay for new hardware...
> The contrast between "force Apple to stop adhering to legacy standards, USB-C is the future!!!" and "gubmint hands off muh 8-pin aux connectors!!!" is interesting.
Both of these are the same stance to me, "existing industry standards appear to suffice, why foist something different on the market?"
I don't think governments are behind a lot of these changes so that last bit is probably exaggeration but seems misplaced; it's Nvidia, Apple, etc choosing these connectors for their products.
> existing industry standards appear to suffice
reminder: lightning came first, because the industry couldn't get its shit together and agree on a successor to micro-b, despite acknowledgement of the shortcomings.
lightning is the pcie 8-pin in this comparison: you may not like the technical limitations it imposes, but, it works well enough that you can build products around it. So why impose a change to The Current Thing and do this whole big changeover if it's already working well enough?
could it be better? sure. And you could do better than pcie 8-pin too. But that's the hardware that's in people's hands right now.
> I don't think governments are behind a lot of these changes
sure they are, what do you think this article is literally about? The EU handed a technological monopoly via legislative fiat to the USB-IF here. You don't have to follow it, but you can't sell your goods in the EU if you don't.
Why could we not require DIY parts sold in the EU to adhere to a new "pcie module" standard and outlaw sale of legacy form-factors outside enterprise/specialty use-cases (like engineered HPC systems and other enterprise, non-consumer products)? Most of those guys are on their own mezzanine standards anyway for the important stuff. Just declare "we're gonna make the change" and give companies a few years to implement it and get hardware into consumer hands, and then flip the switch and ban sale of the legacy products. It's not a big deal to update if you actually get buy-in from companies, but you can't have half the industry go one way and half the industry go another, because then it's a mess, and in the meantime if you have half the companies digging in their heels nothing changes. Just like USB connectors.
The really ironic thing is that if PCIe could make the same transition as USB - towards 48V power delivery - it would all be a lot easier and safer. USB is literally the model for moving forward here lol, and if you upgrade your PSU you could even keep your legacy connector (48v lets you deliver 4x the power at the same current). But people aren't interested in moving forward for the sake of moving forward - the interest in ditching legacy standards vanishes as soon as you bring up a standard their brand favors.
It's just an extension of the brand wars. It's not about USB-C vs Lightning, you could simply mandate adoption of the power delivery standards and usb 3.0 speeds (ipad pro already supports usb-3.0 over lightning) if you wanted. It's about Android vs Apple and people are only interested in ditching legacy standards when it can be tied to their brand war. People absolutely don't care about the connector, and it's not about what came first (which was Lightning). They care about The Other Guy getting a thumb in their eye and having the EU legally outlaw the Apple standard. The thrill of having the EU legislatively confirm that Android Is Better.
And ATX vs future-power is turning into a proxy war for AMD vs NVIDIA too. It's mildly disgusting, like sure I'd love to move forward on eliminating legacy standards cruft, but if you want to talk sacred cows then ATX/pcie-add-in-card are #1 and #2 on the list. People aren't raising the "legacy standard" thing in good faith at all, it's just brand warriorism and the instant you ask them to change their crufty legacy standard the answer is no.
Brands are now competing and marketing to customers on the basis of refusing to adopt the new standard despite its official adoption into ATX - that's so obviously problematic that it invites the same kind of governmental action as the EU took with USB. This is how you're supposed to do it according to the USB model and industry-consensus model that the EU favors. But oh, this is my brand and my sacred cow, you can't make me change a cable, that's not fair! And literally it's the most mild possible change, ATX working group is bending over backwards to maintain PSU compatibility and it's causing problems with heating/connection quality/damage. Really we need to move up to 48V like USB did, but enthusiasts are going to pitch an even bigger hissy fit over that.
> lightning is the pcie 8-pin in this comparison: you may not like the technical limitations it imposes, but, it works well enough that you can build products around it. So why impose a change to The Current Thing and do this whole big changeover if it's already working well enough?
That would be valid if everyone was using it.
If all phones had lightning I would be much less inclined to make them change it.
> But that's the hardware that's in people's hands right now.
The hardware that's in people's hands right now is overwhelmingly USB C. That's why the motivation exists to get Apple to be compatible.
> 48V power delivery - it would all be a lot easier and safer
I don't know, that's getting pretty high. Do you then convert directly from 48v to 1.1v or do you need to add a second conversion stage onto GPUs?
Edit: Though given this post, the new PCIe connector seems like an overall good idea, and it's just some adapters that need to be recalled. https://www.igorslab.de/en/adapter-of-the-gray-analyzed-nvid...
Proprietary anything is bad. Apple is bad because it uses proprietary hardware. There is no "Apple standard." There isn't a "brand war" either. USB isn't a "brand," it's an industry consortium that designs connectors that anyone can use. Apple is bad and it should feel bad.
>> existing industry standards appear to suffice
>reminder: lightning came first
Lightning was never a standard.
4 replies →
> legally outlaw the Apple standard.
Apple "standards" are DRM proprietary stuff, I would not dare to call it a standard.
3 replies →
Consideration of cost and product volume are big here.
Not nearly so many people are affected by PSU and motherboard specs as they are phone connectors.