Monday, June 3rd 2019

AMD 300-series and 400-series Motherboards to Lack PCIe Gen 4 with Ryzen 3000

This shouldn't really need to be spelled out, but AMD clarified that you can't have PCI-Express gen 4.0 running an upcoming Ryzen 3000 "Matisse" processor on older socket AM4 motherboards based on AMD 300-series and 400-series chipsets, and that the processor's PCIe root-complex will run at PCI-Express gen 3.0 speeds. AMD's official reason for this is that the older motherboards can't guarantee reliable signaling needed for PCI-Express gen 4.0 and hence the company decided to blanket-disable PCIe gen 4.0 for the older platforms. This statement was put out by Robert Hallock, senior technical marketing head for CPUs and APUs, on Reddit.

Unofficially, though, we believe there are technological barriers standing in the way of PCI-Express gen 4.0 on the older motherboards, the least of which are the lack of external PCIe gen 4.0-certified re-driver/equalizer components, and lane-switching on boards that split one x16 PEG link to two x8 links. There may be other less technical issues such as PCI-SIG certification for the older platforms. Intel faced a similar challenge with its 3rd generation Core "Ivy Bridge" processors, which introduced PCI-Express gen 3.0 to the mainstream desktop platform, and were backwards-compatible with Intel 6-series chipset (eg: Z68 Express). The older 6-series motherboards could only put out PCIe gen 2.0 with the newer processors.
Source: Robert Hallock (Reddit)
Add your own comment

62 Comments on AMD 300-series and 400-series Motherboards to Lack PCIe Gen 4 with Ryzen 3000

#1
jeremyshaw
I venture the primary reason is MB vendors are unwilling to go though the trouble of recertifying their boards for something those boards were never deisgned for. Also helps sales of newer boards.

Since AMD has goodwill and knows to keep their board vendors happy (vs the messy Ryzen 1000 launch), AMD will take the minor PR hit. Win win for everyone, except for the consumer (who will self police, anyways).
Posted on Reply
#2
PLSG08
Still ok as there's not a lot of Devices out there supporting PCiE Gen4. Tbh I was surprised they announced support of Gen4 for the new gen. Until we get more Gen4 ready devices then the old mobos are still good to go
Posted on Reply
#3
Xzibit
Oh, do I remember. "NEW PCI-E 3.0 Bus" boards where 2xSata6 was a luxury but you got 4xSata3 and depending on the board you got asmedia or marvell for additional that were slower if they worked properly
Posted on Reply
#4
TheLostSwede
jeremyshaw, post: 4058467, member: 92377"
I venture the primary reason is MB vendors are unwilling to go though the trouble of recertifying their boards for something those boards were never deisgned for. Also helps sales of newer boards.

Since AMD has goodwill and knows to keep their board vendors happy (vs the messy Ryzen 1000 launch), AMD will take the minor PR hit. Win win for everyone, except for the consumer (who will self police, anyways).
No, that's not it. It turns out older boards can't achieve full PCIe 4.0 speeds and as such, even if they wanted to, they couldn't be certified for PCIe 4.0.
Posted on Reply
#5
jeremyshaw
TheLostSwede, post: 4058471, member: 3382"
No, that's not it. It turns out older boards can't achieve full PCIe 4.0 speeds and as such, even if they wanted to, they couldn't be certified for PCIe 4.0.
"Yes, no, maybe."

Also, Gigabyte already thought otherwise, and will be reversing their decision (as noted by the quote itself, of course).
Posted on Reply
#7
Xzibit
cucker tarlson, post: 4058478, member: 173472"
Intel did it so it's fine,right @btarunr ?
Intel did it differently. They just connected one x16 for the 70 series. Wasn't until the 100 series they connected the chipset.
Posted on Reply
#8
robot zombie
Didn't really expect them to. I never would have thought the older hardware was up to spec for compliance. Nothing nefarious about that IMO. Of course, any mobo vendor might have wanted to be able to offer it... add a little value to their older boards and maybe keep up sales until they're done. Keep in the spirit of things and gather some karma in the market. And they may honestly have thought it would've flown from an engineering perspective. Such is the nature of adapting new tech to old hardware. Win some, lose some.
Posted on Reply
#9
cucker tarlson
Xzibit, post: 4058483, member: 105152"
Intel did it differently. They just connected one x16 for the 70 series. Wasn't until the 100 series they connected the chipset.
That's right they were on dmi 2.0 even with x99.
Wasn't x470 pcie 2.0 too?
Posted on Reply
#10
btarunr
Editor & Senior Moderator
cucker tarlson, post: 4058478, member: 173472"
Intel did it so it's fine,right @btarunr ?
Yes.
Posted on Reply
#11
Xzibit
cucker tarlson, post: 4058485, member: 173472"
That's right they were on dmi 2.0 even with x99.
Wasn't x470 pcie 2.0 too?
Yes. Intels transition lasted 3 gens Ivy, Has, Brd & 2 sockets 1155 & 1150 before completing the full switch on SkyL 1151.
Posted on Reply
#12
cucker tarlson
btarunr, post: 4058486, member: 43587"
Yes.
what are you gonna do,amd needs money more than they do anyway
Posted on Reply
#13
Crackong
Please don't.
We all know how motherboard manufacturers handling things.
If they really put PCI-E 4.0 bios updates to older chipsets, expect your PCI-E devices to be unstable for at least 1 or 2 bios revisions.
Better not doing that.
Posted on Reply
#14
sutyi
jeremyshaw, post: 4058472, member: 92377"
"Yes, no, maybe."

Also, Gigabyte already thought otherwise, and will be reversing their decision (as noted by the quote itself, of course).
Higher-end X470 might be able to handle at least the main x16 slot and an M.2 x4 connection due to better tracing and PCB. Everything else just does not have traces and shielding that is required to make it happen, no point in enabling it when the rest of the boards was never designed to handle it in the first place. I'm just happy that I can drop in a 3600X or 3700X in place of my R5 1600 and maybe upgrade mobo later to a B550 if need be.
Posted on Reply
#15
Melvis
Doesnt worry me in the least and im sure 99% of people that are currently on AM4 will feel the same. Not going to really matter until we are all running RTX 3080 Ti's lol
Posted on Reply
#16
laszlo
even gen4 speed is twice the gen 3 the real gain is not so high for an average user so what is the loss?
Posted on Reply
#17
Patriot
I guess enough x470 boards couldn't hit spec to keep them from making it limited to a chipset x470/x370

There were a couple demoed at ces in jan running pcie 4 on their x470 boards.
Posted on Reply
#18
tigger
I'm the only one
Are there actually any PCI-E4 devices out there yet?
Posted on Reply
#19
TheLostSwede
sutyi, post: 4058490, member: 112688"
Higher-end X470 might be able to handle at least the main x16 slot and an M.2 x4 connection due to better tracing and PCB. Everything else just does not have traces and shielding that is required to make it happen, no point in enabling it when the rest of the boards was never designed to handle it in the first place. I'm just happy that I can drop in a 3600X or 3700X in place of my R5 1600 and maybe upgrade mobo later to a B550 if need be.
It's not just about the "traces", that's about one quarter of the equation.
Cheaper boards are often only four layers, whereas higher-end boards are six or even eight layers. This allows for much beter signal integrity as well.
Then you have the type of PCB design that's employed. Apparently PCIe 4.0 based boards use improved materials and you can find some details about the various options available here (it's a tad old, but still kind of relevant to the topic) https://www.lboro.ac.uk/microsites/research/iemrc/documents/EventsDocuments/2012 conference/presentations/A Morgan Low Loss High speed laminates - What is really needed.pdf
Then there's the "eye" diagram, you'll also find that in the PDF linked above. This is a standard test for signal quality and if the eye part isn't even or open enough, then your implementation doesn't meet the signal quality required. I would guess this is a big issue on older PCB designs that didn't take into consideration the possibility of a higher speed signal being passed through the board traces when these PCBs were designed. No-one tends to design these things for a possible future standard, so it's just done to a satisfactory level for the current technology available.

tigger, post: 4058501, member: 24505"
Are there actually any PCI-E4 devices out there yet?
There will be SSDs by the time the boards launch, but I guess you didn't read any of the Computex coverage on this site?
Posted on Reply
#20
Metroid
Melvis, post: 4058491, member: 50520"
Doesnt worry me in the least and im sure 99% of people that are currently on AM4 will feel the same. Not going to really matter until we are all running RTX 3080 Ti's lol
pcie4.0 is only important at moment for servers that need that bandwidth for many use cases, researchers need that, gaming in general gpu wise is 2 to 3 generations to maximize it, so 3nm gpus or so. The only benefit we have at moment is, ssd's will double their sequential speed, is that important? no for 99.99% of people, yes for very very small group of people that rely on that, so other than that I see as a waste of money. I might end up getting a cheap b450 or x470, power consumption is 4.8 watts, x570 11 watts, 2.2 times more power than that and reason it has a fan on it.

I will have to wait reviews x470 x x570 and if there is a speed difference between them, memory wise which tends to contribute a lot on the final ipc.
Posted on Reply
#21
biffzinker
tigger, post: 4058501, member: 24505"
Are there actually any PCI-E4 devices out there yet?
NVMe SSDs have been announced to go along with your new X570 motherboard purchase. From AMD/Nvidia no, nothing yet.
Posted on Reply
#22
tigger
I'm the only one
TheLostSwede, post: 4058502, member: 3382"
I guess you didn't read any of the Computex coverage on this site?
Skimmed some of the bits that got my attention, but there was so much of if, so not really. If i had I would not have posted the question.
Posted on Reply
#23
biffzinker
tigger, post: 4058513, member: 24505"
Skimmed some of the bits that got my attention, but there was so much of if, so not really. If i had I would not have posted the question.
I'm guilty of skimming through, so don't feel bad.
Posted on Reply
#24
Flaky
PCIe riser manufacturers had no problem with "maybe/partially/reduced performance" kind of working ¯\_(ツ)_/¯

It is motherboard manufacturers responsibility to verify whether the physical layer they provide complies to signaling requirements or not.
Such step by AMD shows that they do not believe that MB makes can competently validate their own products, i.e. would cut corners or just enable 4.0 with no checking at all.
Posted on Reply
#25
TheLostSwede
tigger, post: 4058513, member: 24505"
Skimmed some of the bits that got my attention, but there was so much of if, so not really. If i had I would not have posted the question.
At least that means the TPU team did a good job of covering the show...
They still have more coming, as I bumped into them at a few manufacturers that haven't been covered yet.

Flaky, post: 4058521, member: 173700"
PCIe riser manufacturers had no problem with "maybe/partially/reduced performance" kind of working ¯\_(ツ)_/¯

It is motherboard manufacturers responsibility to verify whether the physical layer they provide complies to signaling requirements or not.
Such step by AMD shows that they do not believe that MB makes can competently validate their own products, i.e. would cut corners or just enable 4.0 with no checking at all.
I'm also sure a lot of those products weren't certified...
There's a difference between a cheap accessory and something that has to be certified or you can't sell it.

You might be right there. Big difference between the tier one guys and the rest...
Posted on Reply
Add your own comment