> Gigabyte removes unofficial PCIe 5.0 support from B650 motherboards in latest BIOS update
It seems that the board was actually running in PCIe 5.0 mode, even though it was only supposed to (by the specs/manual) do PCIe 4.0
It was never advertised as supporting 5.0
This was the root of the article in my opinion:
> The issue may have emerged from the fact that PCIe 5.0 was enabled by default, and not clearly marked as experimental, and that may be something AMD did not like to see.
However, it seems to just be speculation. I too would like to know what the real reason was. Was Gigabyte dealing with returns/RMAs due to the feature? Was there pressure from AMD to stop allowing their lesser chipset to access PCIe 5.0?
Would be my guess. The fact that the B650 can even do 5.0 on the GPU lanes (it is able to do 4x 5.0 on NVME) makes me wonder if the B650 chipset is the same silicon as B650E and the speed is just really a configuration setting. Across the board the B650 seems like it is the exact same as a de-rated B650E.
It would surprise me if motherboard vendors who support 4.0 weren't actually compliant with 5.0 from a signaling / integrity perspective for future proofing and increased headroom.
Gigabyte shipped 4.8 million MBs in the first quarter.
Assuming they sell 20 million MBs in a year, how does a $0.01 additional component cause $1 million in losses? By my math, it's 200k per year.
Further, what restricts gigabyte from raising the price of their MB by $0.10? or even $1 to accommodate the various $0.01 components they might need per board? Do you think the market wouldn't bear such a price variance for $200 boards?
All the AM5 chipsets are the same silicon, except A620 which has less capability (might just be fused off though) and the B840 which seems to be a rebranded AM4 chipset. B650, B650E, B850, and X870 are all one Promontory 21 chipset; X670, X670E and X870E are two Promonotory 21 chipsets (one chained behind the other). The B840 seems to be the Promontory 19 silicon, same as a B550 AM4 chipset; which seems weird, but it has more downstream connectivity than the A620 chipset, although the cpu to chipset link runs at pci-e 4.0 for A620 and 3.0 for B550 and presumably B840, so that's a bottleneck if you have high throughput devices behind the chipset.
The GPU lanes on AM5 all come direct from the CPU, as well as at least one x4 m.2 socket, and there's 4 more CPU lanes not earmarked by AMD. Those don't touch the chipset, and any correlation between the chipset branding and the speed of those lanes is a marketing restriction: If you put B650E branding on your board, AMD says the board must support pcie 5.0 for the GPU and first NVMe slot; but the chipset isn't involved in those lanes, it's just a marketing agreement; I'm not sure if AMD disallows using pcie 5.0 on a B650 board for the GPU slot, but capability is a matter of trace design between the slots and the cpu and maybe what the board maker is willing to certify / what the firmware is willing to enable. For B850, AMD says pci-e 4.0 for the GPU and 5.0 for the first nvme.
I'm sure there are some X870 boards that are significantly different from B850 boards, but there's probably a lot of boards where everything is the same other than the product name. Likely the same was true of B650 and B650E. For reliable PCI-e 5.0 operation, you do need to follow proper design rules for the traces, and that results in a more expensive board than following the design rules for PCI-e 4.0 operation. But it can make product sense to sell some of those 5.0 capable boards with a lesser chipset brand anyway.
If their hardware team designed the PCB and the qualification process to the specifications of the B650 chipset, they would have targeted PCIe 4.0 speeds.
Regardless, the chipset has been out for almost 3 years. Kind of late to put the horse back in the barn. Regardless of certification, lots of systems are out there where it's enabled, and those with those systems and a pci-e 5.0 can tell us how it works regardless of what the branding says. Maybe it's flaky, maybe it rarely works. Maybe it just made sense to follow the pci-e 5.0 design rules for the gpu slots, even without the branding; maybe the pci-e 4.0 design rules are good enough at the trace lengths on a typical am5 board. Maybe once you have the PCB designed and built to run the cpu m.2 at 5.0, it doesn't cost anything to follow design rules for the cpu x16 slot as well.
No AM5 chipset can do 5.0 speeds, they are all linked via x4 PCIe 4.0 lanes. What you're thinking about is PCIe 5.0 speeds from CPU to SSDs/GPU. There's no chipset that does PCIe 5.0 speeds, they are all on 4.0.
There is no AM5 chipset with PCIe 5.0 support. They are all on PCIe 4.0. The lanes come from CPU, they go to SSDs/GPU/chipset. All AM5 chipsets are linked to CPU via PCIe 4.0 lanes, none support PCIe 5.0 speeds.
edit: Not sure why the downvote, this is common information and it should be easy to link an AM5 motherboard that has x4 PCIe 5.0 downlink lanes to chipset. But they do not exist.
My guess (having some high-speed design and manufacturing experience) is that the boards were physically designed and qualified around the official spec, which was PCIe 4.0.
Then after some confusion they discovered that the BIOS team had enabled PCIe 5.0 on boards that were never qualified for it. Batch-to-batch variations of the boards could have caused instability because the manufacturing materials and test processes were only targeting PCIe 4.0
We absolutely want to hold companies accountable, but if we lambast them for reasonable actions, where's the motivation for them to do right?