That article is a bit overhyped. Several fundamental flaws.
1. The article references 'via Tomshardware' article which ends in....
"... and the first PCIe 5.0 devices should debut this year. Broader availability should come in 2020. ..."
Some announcements may come in 2019 but product? Neither the PCI-e docs , the Tom's article , nor the notebookcheck article have any example of any CPU that was coming with PCI-e v5 in 2019. There are explicit examples of PCI-e 4.0 ( e.g., AMD ) , but there are zero for 5.0. Huawei has a ARM server chip with PCI-e 4 announced last month. IBM has had Power 9 announced and shipping since 2017.
2020 isn't a "short term" timeline to getting to another Mac Pro.
2. The first CPU onto PCI-e 4 was IBM Power 9 (2017). Roadmaps from around 2017 era had them on PCI-e 5 in 2019 with Power 10. .... but then Global Foundries flaked on 7nm (IBM was shooting for a custom high density '10nm' ). GF stopped at 14nm. IBM has had to shift to Samsung for 7nm and
the roadmaps now say 2020+ for Power 10 arrival.
CPU (and to lessor extent GPU) development hiccups can kick out the time lines.
3. It will be shorter than the PCI-e v1 -> v2 -> v3 pace that had updates at 4 year intervals. But that doesn't mean it is going to get skipped in a significant way. In part, it is going to be short because IBM has been the
only implementer for almost two years. So since the others have been hitting the snooze button for over a year .... yeah once they get to 4.0 they will probably transition to 5.0 not hitting the snooze button for extra two years. It will certainly be shorter than the 3.0 -> 4.0 transition if use that as a 'baseline'. It will probably be shorter than 4 years. but shorter than two. Probably not. Even two is a stretch.
there are parts and techniques short of fully implementing 5.0 that can be 'borrowed' from 5.0 and used to implement 4.0 ( which more than several are going to do because it is simpler. ).
That the gap between 4.0 -> 5.0 is only two years but that may not be a broad spectrum deployment path. Look at Ethernet. 10GbE BaseT passed in 2006. [ and Ethernet speed standards increasingly came over that decade. ] A Mac with 10GbE standard didn't arrive until 2017 ( 10 years later) . if PCI-e 5.0 cards remain extraordinarily expensive then 5.0 won't deploy into the Mac Pro (let alone general Mac) zone for a long while. (e.g., if the primary drivers are > 100Gbps I/O cards ). Nvidia pushing NVLink and AMD now moving to push Infinity Fabric (for GPGPU card to card backplane) will also dampen deployment to the mainstream. [ Proprietary backplanes puts more money in those GPU vendors pockets so they aren't in hurry dump them. ]
When 3.0 came PCH chipsets held onto 2.0 for a long time. It was more cost effective to implement and that is where the bulk of the I/O device inertia was.
In part, v 4.0 didn't come quicker to the broader market because to a large extent v3.0 was fast enough for an extremely large set of solutions. There are data center folks moaning at the 'delay' but the broad workstation market grossly impeded because didn't have PCI-e v4.0 for almost two years. Almost nobody was 'having a cow' over that delay.
Indeed, a part of the Toms article that notebookcheck glosses over.
"... PCI-SIG expects the two standards to co-exist in the market for some time, with PCIe 5.0 used primarily for high-performance devices that crave the ultimate in throughput, like GPUs for AI workloads, and networking applications. That means that many of the leading PCIe 5.0 devices will land in data center, networking, and HPC environments, while less-intense applications, like desktop PCs, are fine with the PCIe 4.0 interface. ..."
4. In the context of a Mac Pro ( being on target in this forum). Intel isn't jumping to 5.0 this year. AMD isn't either ( in fact sticking with the same socket has a decent chance of meaning have issues getting to 5.0 and need another whole set of motherboards just to get to 4.0. ).
Pragmatically need a CPUs, motherboards, socket (both sides) , and cards/plug-ins to get to the new solutions. Getting all those typically disparate groups to synch up logistically wise takes time.
If want to engage in Apple ARM A-series fantasies for Mac Pro , it isn't even particularly a full user of 3.0 . There is zero momentum there for 4.0 let alone 5.0
![117612--3-von-5-[1].jpg](https://forums.macrumors.com/data/attachments/736/736398-6941f2c0b8258afd71ece37ddaf9f641.jpg "117612--3-von-5-[1].jpg")
![35-0[1].png](https://forums.macrumors.com/data/attachments/736/736400-0da77978d5173b771a1cbf1aa30b3846.jpg "35-0[1].png")
![4092855[1].jpg](https://forums.macrumors.com/data/attachments/736/736401-dca6d1f77e4cf1a6a54c5b6eda849d0b.jpg "4092855[1].jpg")
