News Gigabyte's new motherboard supports massive 128-lb GPUs — Reinforced slots still can't fix GPU PCB cracking

This isn't an entirely new feature actually. "Jeff Butts" also brought up this issue in his recent article:

https://www.tomshardware.com/pc-com...et-series-numbers-to-run-in-parallel-to-intel

In fact, Benchlife notes that the “B650E Aorus Pro X USB4 motherboard matches the specifications of most AMD X870 motherboards.” It also features a reinforced PCIe slot supporting up to a maximum weight of 58 kg. I can’t for the life of me figure out why you’d need your PCIe slot to withstand so much weight, but there you have it.

This Article: "While the PCIe UD Slot X feature isn't novel, this is the first time we've been given a specific metric of the maximum weight it can withstand"

FWIW, some of the Gigabyte's previous INTEL motherboard also had this feature, the "PCIe UD Slot X" PCIe 5.0 x16 slot, with 10X strength for the graphics card.

Z790 AORUS XTREME X, and also the PRO X model. Few more boards are there, but those are Chinese/APAC-exclusive models though, with some previous Intel chipsets as well.

https://www.gigabyte.com/Motherboard/Z790-AORUS-XTREME-X#kf

j3hTu6O.png
 
Last edited:
  • Like
Reactions: PEnns

Notton

Prominent
Dec 29, 2023
523
451
760
The slot isn't the problem. Reinforcing it doesn't solve anything.
The problem is the GPU is too heavy.

Ideally, the back plate or heatsink should have holes so it can be bolted directly on the case.
Similar to how an engine block is mounted to a frame/chasis.
Or, how a mobo gets mounted to the case.
 

bit_user

Polypheme
Ambassador
My hunch is that people are misinterpreting their claim (perhaps partly their fault, for misrepresenting).

If you consider the problem of an OEM like Alienware shipping a PC with a dGPU, they ship it fully-assembled and need to be sure that even if the delivery driver drops the box or it falls off of a tall stack, that the shock won't damage the motherboard. So, you build in some assumptions about what kind of shock it can handle, like 20 G with a GPU of up to 6.4 pounds. Then, by the magic of Newtonian physics, a cool trick you can play is to say that it can support a static GPU weight of 128 pounds. The two forces are equivalent.

Anyway, this information tells OEMs like Alienware that they must either use better packaging or lighter GPUs, so they don't exceed that force for packages dropped from less than whatever height the shipper will guarantee. Further, they can theoretically use packaging making it evident if the package was subject to excessive force, during shipment, or maybe just embed a little IoT accelerometer.
 
  • Like
Reactions: PEnns

bit_user

Polypheme
Ambassador
Ideally, the back plate or heatsink should have holes so it can be bolted directly on the case.
This used to be a solved problem... I'm having trouble finding a good link, but anyone who built PCs in the 90's should remember how there used to be slots on the front side of the card cage to hold the front edge of full-length add-in cards.

I've also seen cases with an over-the-top brace, to support the cards. This obviously places limits on the height of the card and requires the PCB be exposed at that point.
 
  • Like
Reactions: PEnns and taz-nz

TechLurker

Reputable
Feb 6, 2020
184
105
4,760
Given that GPU weight isn't likely to go down on the high-end, esp. if the rumors that the brick-sized, 4.5 slot reference cooler prototype will be reused on NVIDIA's next top-end GPU, and the continuing push for more power/performance, there's merit to adopting some new design elements mainly to spread the weight across more, besides GPU brackets or braces.

To list off some ideas I've read in the past:
  • A new "AGP 2.0" style slot exclusively dedicated to GPUs (at first), except it's still wired as PCIe but designed to handle thicker PCBs and also provide pass-through power to GPUs to cut the dangling weight of external power plugs. This was pushed around a bit back when GPUs were already hitting 2.5 slots and 3 x8 power pins, but never left the conceptual stages. I do remember a number of articles suggesting GPUs should have their own special slot as power draw and cooling requirements increased, and this seemed the most practical. It's also future proof if it can be expanded to become a new PCIe slot standard, being only slightly thicker than a regular reinforced PCIe slot.
  • The next idea was redesigning the shroud +heatsink (+backplate if included) to be load-bearing. Have it and the backplate functionally become 1 piece (when assembled) that relieves the weight off the PCB some and would have the I/O shield/PCI screw in points connected to the shroud instead of the PCB. It would effectively be its own built-in GPU bracket, esp. if the long and narrow ends of the shroud were made of metal. And given that the heatsink is also designed to connect to the shroud, the weight on the PCB is further reduced.
  • The last idea was tweaking GPU plastic shroud designs to include 4+ integrated, adjustable stabilizers along the slot-side that would let it raise/lower some rubber feet to level the GPU to the motherboard, whether it has a plastic shroud or not over the face. The idea being to at least redistribute some of the weight to other points of contact with the mobo. It would increase the thickness of the GPU shroud though, unless some cooling finstack is cut away to allow nestling of the stabilizers while keeping a thinner shroud.
 

alan.campbell99

Honorable
Sep 11, 2017
36
4
10,545
This used to be a solved problem... I'm having trouble finding a good link, but anyone who built PCs in the 90's should remember how there used to be slots on the front side of the card cage to hold the front edge of full-length add-in cards.

I've also seen cases with an over-the-top brace, to support the cards. This obviously places limits on the height of the card and requires the PCB be exposed at that point.
Yes, I remember those. Some cards would have either a fixed or screw-on edge for those slots as I recall
 

vijosef

Proper
Feb 26, 2024
62
74
110
This may actually be harmful.

A flexible slot may permit the GPU to descend until it finds support, whereas a rigid slot might compel a loosely fitted card to fracture.

Understanding the distinction between strength and ductility is fundamental knowledge for any mechanical engineer.
 

bit_user

Polypheme
Ambassador
Or PCIe riser cable
This is a good point. If you mechanically decouple the GPU from the motherboard, you can drastically reduce the potential for damage. For a GPU, the next area of concern would be where it attaches to the back of the case, but that can use high-impact plastic or sheet metal with enough flexibility to absorb some shocks. Obviously, the GPU would need to be supported by something other than the back of the case, but whatever supports the length of it can be made more flexible.
 

Varsaggo

Reputable
Dec 26, 2020
5
1
4,515
This isn't an entirely new feature actually. "Jeff Butts" also brought up this issue in his recent article:

https://www.tomshardware.com/pc-com...et-series-numbers-to-run-in-parallel-to-intel





FWIW, some of the Gigabyte's previous INTEL motherboard also had this feature, the "PCIe UD Slot X" PCIe 5.0 x16 slot, with 10X strength for the graphics card.

Z790 AORUS XTREME X, and also the PRO X model. Few more boards are there, but those are Chinese/APAC-exclusive models though, with some previous Intel chipsets as well.

https://www.gigabyte.com/Motherboard/Z790-AORUS-XTREME-X#kf

j3hTu6O.png
Was just thinking the Asus Prime X299 board I bought last year that was part of the X299 refresh a couple years ago have the same reinforcement on the PCI slots. Lol if you pull the PCI slots hard enough to break it it would snap the whole Mobo in half..
 

Varsaggo

Reputable
Dec 26, 2020
5
1
4,515
My Asus Prime X299 Deluxe 2 refresh has the same reinforcement on the PCI slots this isn't really a new thing that been happening. At least not when talking about ASUS, maybe GiGAbyte boards didn't have this but usually the top tier boards would have features such as the reinforcement...
 

Varsaggo

Reputable
Dec 26, 2020
5
1
4,515
This may actually be harmful.

A flexible slot may permit the GPU to descend until it finds support, whereas a rigid slot might compel a loosely fitted card to fracture.

Understanding the distinction between strength and ductility is fundamental knowledge for any mechanical engineer.
Nah it works pretty well my Asus board came with this and it's a few years old. The slots are reinforced pretty well through the board with steal and once u see it in person you would see what I'm saying. It definitely would save your slot if you had it compared to one that didn't and had enough force applied to it.
 

bit_user

Polypheme
Ambassador
Nah it works pretty well my Asus board came with this and it's a few years old. The slots are reinforced pretty well through the board with steal and once u see it in person you would see what I'm saying. It definitely would save your slot if you had it compared to one that didn't and had enough force applied to it.
There are two sides to this issue. The motherboard side and the GPU side. Multiple GPU makers have had issues with their GPU PCBs cracking, and that's often the more expensive component. A more rigid motherboard only increases the likelihood of that occurring.

So, the answer is not necessarily just reinforcing the motherboard to the point that you could hang a small moon by it.
 

Notton

Prominent
Dec 29, 2023
523
451
760
This used to be a solved problem... I'm having trouble finding a good link, but anyone who built PCs in the 90's should remember how there used to be slots on the front side of the card cage to hold the front edge of full-length add-in cards.

I've also seen cases with an over-the-top brace, to support the cards. This obviously places limits on the height of the card and requires the PCB be exposed at that point.
Some examples I can think of were...
Chieftec/Chembro/bunch of rebranded clone full tower cases had those for extra long PCI-X and ISA cards.

Silverstone TJ08-E/PS07 had a flipped mobo tray and the HDD cage was designed to double up as a GPU support bracket.
Silverstone LD01 ditches the HDD cage for an adjustable GPU shelf.
 
  • Like
Reactions: bit_user

aldaia

Distinguished
Oct 22, 2010
535
23
18,995
The issue here is that the GPU is attached perpendicular to the motherboard producing a leverage effect that exerts strong forces on the slot. There is nothing that prevents the GPU to be physically attached in a parallel plane to the mother board and the issue is gone.

Thats what servers do to preserve the pizza box form factor.

SYS-1029GQ-TVRT.png

Actually, GPU's can even be stacked:
nvidia-mgx-ovx-server.jpg