Contribute
Register

Gigabyte Z490 Vision D (Thunderbolt 3) + i5-10400 + AMD RX 580

Any links to where you found this info ?
The Vision D page as CaseySJ linked above. Again I am trying to figure the extent of the PCIe4 support with my question. Therefore I have no solid document to link to. I will probably post in the gigabyte forum to get clarification. There is no clear documentation stating what portion of PCIe4 is implemented by putting an 11 Gen CPU in. It would be nice for a hypervisor situation to allow two GPUs to function in full speed in two different VMs. PCIe4 x8 is essentially PCIe3 at x16. Windows and MacOS running at full gpu speed simultaneously, Albeit splitting the cpu cores in half. But still it would be a nice environment to run for some.
 
Interesting point.
  • Comet Lake CPU has 16 PCIe 3.0 lanes
  • Rocket Lake CPU has 20 PCIe 4.0 lanes (so 4 more lanes)
  • When we install Rocket Lake CPU, the M2A_CPU slot must get routed through the CPU because the Z590 chipset only provides PCIe 3.0 lanes
For absolute verification, we can always check the Gigabyte forum or submit a question to their online support desk.
I will write to Gigabyte to clarify. This curiosity all stemmed from reading this document on MSI’s website.

 
I will write to Gigabyte to clarify. This curiosity all stemmed from reading this document on MSI’s website.

Nice find! Probably no need to contact Gigabyte because all such motherboards will behave the same way.
 
Good morning @CaseySJ
Sorry to bother you, I would need to contact you but I am unable to send you a PM. Maybe this function is available only for admin/Mod. However I didn't see anything in the forum rules.
Thanks
 
My concern is whether the M.2 slot reroutes to "add" 4 additional PCIe4 CPU Lanes instead of going through the chipsets 24 lanes. I assume that becoming a PCIe4 enabled motherboard, the M2A_CPU lanes do in fact switch to CPU Lanes.
So you hope that metal traces magically migrate within the PCB when changing the CPU…
I'm sorry to pour cold water on your parade, but that's unlikely to happen.

Currently, if you look at page 7 of the Z490 Vision D manual the M2A_CPU slot is not actually part of the 16 PCIe CPU Lanes; it goes through Chipsets 24 Lanes.
So it runs at PCIe 3.0 speed from the chipset no matter which CPU is used.

If you look at the manual for the Z590 Vision D on page 7, you might see what I am getting at in comparison with the Z490.
CPU lanes, at PCIe 4.0 speed, when using 11th gen. CPU and not available at all when using 10th gen., as shown in the specifications for the Z590 Vision D:
1644744134572.png


The specifications for the Z490 Vision D do not promise any PCIe 4.0 at all.

I will write to Gigabyte to clarify. This curiosity all stemmed from reading this document on MSI’s website.

To do this, MSI would need a double set of traces in the PCB, one coming from the CPU and one from the PCH, going into a PCI switch chip which then feeds the M.2 slot. I suppose that's possible, but it adds to the costs and the design would have been impossible to validate before PCIe 4.0-capable CPUs were released.
Taking MSI Z490 Unify as example, I found this block diagram at page 27 of the manual:
1644745211190.png

No switch is shown, but a plain M.2 slot from the PCH, sharing lanes with a PCIe x4 slot (in x16 physical slot).
The specifications make no mention of PCIe 4.0—and the table doesn't bother to distinguish between the physical PCIe slots, of which two are electrically x16/x0 or x8/x8 from the CPU and the third one is x4 from the PCH. Lazy listing, and sloppy drawing.
1644745860718.png


The specifications and manual may be wrong, and the blog post, right. But I rather suspect the opposite: The block diagram is right and the blog post is a pipe dream from a marketing guy.
 
...
The specifications and manual may be wrong, and the blog post, right. But I rather suspect the opposite: The block diagram is right and the blog post is a pipe dream from a marketing guy.
The blog post is correct. The specifications are also correct for 10th gen, but not for 11th gen because the specification were published at a time when 11th gen was not available and they did not want to add confusion.

Even the Z490 Vision D product page (not spec page) shows PCIe 4 support for 11th gen.

The MSI blog post contains this table that clearly shows PCIe 4.0 x4 support on M2_1 M.2 slot with 11th gen. Again, the blog post is correct.

Look at the Remarks for first row: "When 11th Gen Core Processor detected, M2_1 will be connected to PCU".
Screen Shot 2022-02-13 at 2.28.44 AM.png
 
Heads up, I disabled SIP and InstallAssistant.pkg finally ran successfully!
 
So you hope that metal traces magically migrate within the PCB when changing the CPU…
I'm sorry to pour cold water on your parade, but that's unlikely to happen.

I am not hoping, I am asking if someone here knew if the PCIe4 features of the motherboard are enabled when adding a Gen11 CPU. There is a reason they called it M2A_CPU. It's not a far fetched expectation. Motherboard manufacturers implemented PCIe4, yes with reinforced traces, as per the information page CaseySJ linked to above. The M2A_CPU is handled by a switch which I assume can be rerouted to CPU if the lanes for the PCIe4 are accessible from the Gen11 Processor. I am waiting to hear back from gigabyte. Manufacturers were not sure at the time how well the PCIe4 would be implemented as they waited for Intel to finish Gen11, is what I have read.
 
I concede that the M2A_CPU designation is strange, but as shown in the block diagram it comes from the PCH, from the PCH only, and it has to come from the PCH to be able to use SATA lanes.
You may hope that there's some undisclosed trickery to make the best of Rocket Lake CPUs—and I'd be happy to be proven wrong on documentation reading.

Personal experience with undisclosed features: The Asus P10S WS motherboard has a 5-pin Thunderbolt header (TBT), which is not mentioned anywhere in the manual. I could not get the motherboard to recognise any Asus or Gigabyte Thunderbolt AIC connected to it, and no known trick could get Thunderbolt drivers to load under Windows or a Thunderbolt section to appear in the BIOS. It looks like Asus engineers tried and failed to implement Thunderbolt, and that marketing and documentation tried to hide any intent to implement Thunderbolt, although the board is manufactured with a neatly labelled header.
 
I concede that the M2A_CPU designation is strange, but as shown in the block diagram it comes from the PCH, from the PCH only, and it has to come from the PCH to be able to use SATA lanes.
You may hope that there's some undisclosed trickery to make the best of Rocket Lake CPUs
Yup. Exactly why I was seeing if anyone here knew. I would expect they would update the motherboard manual to disclose if the PCIe4 switch over actually enabled full PCIe4 features. I want to triple confirm before even thinking of spending the money. It would be a really nice perk to extend the life of the system. That MSI document had me thinking it might actually work. Of course doing this will break iGPU support as 11gen uses Intel Iris XE. Perhaps another question is, does macOS work properly with PCIe4?? I would think this would be hardware level and the only thing macOS requires is compatibility with the z490 chipset. Hence it should?
 
Back
Top