ada lovelace NVIDIA’s 4-slot Founders Edition GeForce RTX 4090 Ti/Ada TITAN graphics card has been pictured

[HenrySalayne]

4 hours ago, leadeater said:

It does need sufficiently large contact area. think about the forces involved of something around 2kg with movement, vibration and shock. It might be fine in isolation but computers get moved and shipped. They don't live in space under zero gravity and even then inertia still applies.

 

That connection point between any such PCB simply cannot be 10mm with only a few screws and not the entire length or good portion of the entire length of the card. The graphics card will snap that PCB very easily at either the screw points or potentially entirely.

 

Think about it logically. Get a thick piece of plastic or wood and a metal block or thick plate. Secure it to the wood using screws of glue starting at 10mm contact area and then waive this around. 100% the metal block aka the weight is coming off one way or another. Repeat until it doesn't, you'll find the contact are between the two materials to prevent this is non-trivial.

Sufficiently large, yes, but that might be much less than you expect. 2 kg are only 20 N. You could hang this easily from a mild steel bolt (S235) with a cross section of 1 mm² - tiny. And that's including a comfortable safety margin (dynamic loads with 5x the 20 N shouldn't be a problem).

 

I know it's sometimes tricky to visualize and it feels counter-intuitive, but maybe this example is clearer:

There is a PCIe port on the motherboard.The port gets mirrored on the GPU side and the two ports are connected with a small slab of PCB. Now if gravity pulls on the GPU, these forces are transferred to the PCB via the PCIe port on the GPU side and the slab of PCB transfers these forces to the PCIe port on the motherboard. Everything is symmetrical, every force experienced by the PCIe port on the GPU side is mirrored at the motherboard's PCIe port and the mechanical system must be in equilibrium.

5 hours ago, leadeater said:

Again in isolation yes, in reality no. No matter what you do the slot connector and solder points will be weight/load bearing and over time those joints will get put under a lot of stress from that and movement and will 100% break. This is why in devices like UPS's connections between boards are either cables or solid metal bars with screws. They do that because it's actually a problem. Solder joints simply are not load bearing and can never be.

 

While you could use a plastic standoff or some other kind of reinforcement like I mentioned that will not remove movement and vibration forces, only compressive. Have a think about it, when is the last time you saw solder joints being used in any kind of load bearing capacity when decent weight is involved.

Following up on the example above, the PCIe slot on the MB can bear the load, so there should be a similar solution for the GPU side.

[leadeater]

1 hour ago, HenrySalayne said:

Sufficiently large, yes, but that might be much less than you expect. 2 kg are only 20 N. You could hang this easily from a mild steel bolt (S235) with a cross section of 1 mm² - tiny. And that's including a comfortable safety margin (dynamic loads with 5x the 20 N shouldn't be a problem).

 

I know it's sometimes tricky to visualize and it feels counter-intuitive, but maybe this example is clearer:

All I'll say is you are WAY to confident in that. No simply based on real life experience if they only mount it with a single row of screws the risk point is the screws and there is more than a low chance of damage.

 

Surely you have seen GPUs entirely ripped out of PCIe slots from shipping and those are vastly more secured than a 10mm-20mm and a single row of screws between such a PCB and the cooler.

1 hour ago, HenrySalayne said:

There is a PCIe port on the motherboard.The port gets mirrored on the GPU side and the two ports are connected with a small slab of PCB. Now if gravity pulls on the GPU, these forces are transferred to the PCB via the PCIe port on the GPU side and the slab of PCB transfers these forces to the PCIe port on the motherboard. Everything is symmetrical, every force experienced by the PCIe port on the GPU side is mirrored at the motherboard's PCIe port and the mechanical system must be in equilibrium.

2kg hanging off a pivot point. You have it all wrong here. The PCB is going to be on the back side of the cooler with the cooler protruding outwards from this board. This cooler is going to be above 1kg and blow 2kg most likely, also likely on the higher end.

 

Everything is absolutely not symmetrical. There is also side to side movement.

 

The risk is the twisting from back to front (up/down) and side to side from the end of the card.

 

Tiny screws getting a lever force will come out if not strong enough and I highly doubt very small screws only supported by 10mm contact width/depth is enough to prevent that lever force. The smaller that contact area the larger the risk.

 

This time actually visualize or better yet do the example I said. Get a large weight by mass and volume and screw it to something flat, similar to what is happening here. Waive it around, keep reducing the surface area used to secure it, watch it fly off. 100% going to happen at some point.

 

Nvidia if they do this like above I'm sure would have calculated it and done reasonable amount of confidence checking but if it comes out like this with a PCIe daughter board attached in this way with what I would consider small contact area then I'd simply never buy it and only consider AIB cards that do not do it that way. It might be fine for month 1, 2, 6, 12 but in the interest of longevity that's a big hell nah from me. Only time will be able to show if they got it right for that kind of mounting and I'm not going to play that risk game.

 

1 hour ago, HenrySalayne said:

Following up on the example above, the PCIe slot on the MB can bear the load, so there should be a similar solution for the GPU side.

True, this is the style I'd personally be way more comfortable with. I see less problems and easier to solve ones.

[HenrySalayne]

10 minutes ago, leadeater said:

2kg hanging off a pivot point. You have it all wrong here. The PCB is going to be on the back side of the cooler with the cooler protruding outwards from this board. This cooler is going to be above 1kg and blow 2kg most likely, also likely on the higher end.

 

Everything is absolutely not symmetrical. There is also side to side movement.

 

The risk is the twisting from back to front (up/down) and side to side from the end of the card.

Still, the forces on the part of the card closest to the PCIe slot and the forces on the PCIe slot are identical. This micro-system needs to be in equilibrium. If 100 N of force pulling on the card these 100 N will pull on the PCIe slot. The mechanical load on a piece of PCB between the GPU and the PCIe slot does not change if the PCB is longer and spanning the whole depth of the card. You can "cut out" that tiny piece of PCB and it will be under the exact identical stress. These are basic principles of mechanics.

 

14 minutes ago, leadeater said:

Surely you have seen GPUs entirely ripped out of PCIe slots from shipping and those are vastly more secured than a 10mm-20mm and a single row of screws between such a PCB and the cooler.

The forces involved if something is dropped are insane. The device can easily experience 100G, that's 100x the static load. A 20 N GPU would be similar to a 2000 N load. Even a solid steel PCB would probably take some serious damage. There are some things you cannot account for.

But a 2 kg GPU with a PCB spanning the entire depth of the card is not less susceptible to damage than a similar GPU with a 3 cm PCB.

[leadeater]

17 minutes ago, HenrySalayne said:

Still, the forces on the part of the card closest to the PCIe slot and the forces on the PCIe slot are identical. This micro-system needs to be in equilibrium. If 100 N of force pulling on the card these 100 N will pull on the PCIe slot.

Not if the force is on the pivot point between the edge of the PCB and cooler. The cooler can be levering against this point putting more force in to the screws and screw holes than in to the PCIe slot.

 

Hold 2 rulers flat against each other, slide outwards holding in the middle and apply force to the ends of the rulers. As you slide them apart where they join in the middle will try and pull away. Sure there is a flaw here that the distance is increasing but you can compensate by applying the force the same distance away from the middle.

 

You can't seriously tell me that you think there is no difference here. Force over area is the issue.

  

17 minutes ago, HenrySalayne said:

But a 2 kg GPU with a PCB spanning the entire depth of the card is not less susceptible to damage than a similar GPU with a 3 cm PCB.

Yes it is. For what I'm talking about absolutely yes there is less risk. I don't know how you can't see it but it's blindingly obvious to me.

 

Edit:

You're actually telling me that the mounting strength between a board and a weight with 100% contact surface area between them is the same as 50% contact area holding all weights and dimensions equal. This is not the case. I'm confident you do realize this and just aren't understanding the issue I'm talking about and looking at something different.

[HenrySalayne]

3 hours ago, leadeater said:

Not if the force is on the pivot point between the edge of the PCB and cooler. The cooler can be levering against this point putting more force in to the screws and screw holes than in to the PCIe slot.

If you introduce a pivot axis (an axis at which the GPU is somehow fixed and can only rotate) into the system which is considerably closer to the mounting point of the PCB than to the PCIe slot, then yes, this could happen (lever arm) But I don't see how you would get a pivot point with these characteristics. If we ignore the I/O bracket, the pivot point of the system would be the PCIe slot.

(With the I/O bracket it gets a lot more complicated, but generally speaking the pivot axis would be a diagonal from the PCIe lock to the screws on the I/O bracket).

 

And I did not say that one screw on each side is sufficient. You would need to fix the PCB structurally to the frame of the GPU - for example with a metal bar on top you screw down or several screws across the length of the PCIe slot. But since today's GPUs are all-metal assemblies completely encapsulating the PCB, I don't see this as a problem.

 

3 hours ago, leadeater said:

Yes it is. For what I'm talking about absolutely yes there is less risk. I don't know how you can't see it but it's blindingly obvious to me.

Edit:

You're actually telling me that the mounting strength between a board and a weight with 100% contact surface area between them is the same as 50% contact area holding all weights and dimensions equal. This is not the case. I'm confident you do realize this and just aren't understanding the issue I'm talking about and looking at something different.

Just so we are on the same page here:

There is only a small part of the PCB in the PCIe slot (roughly 1 cm?)

I'm saying

->  that an equally large part of the PCB to mount the GPU on is sufficient to absorb all forces (for example another PCIe slot on the GPU side).

and

-> that it doesn't matter if the PCB extends throughout the entire depth of the GPU. The forces on the small part spanning the gap to the PCIe slot are equal.

 

These are two completely unrelated failure modes. "will the part spanning the distance between the GPU and the PCIe slot bear the load?" is a different question than "will the mounting of the PCB to the GPU bear the load?"

 

 

And following on your example:

Screwing something down would be (in an ideal world) a friction mount (clamping). Just the force from the screw pressing on the two pieces is keeping them together and blocking movement in all directions. In this case, contact area doesn't actually matter (the pressure), just the force. But there are pressure limits, before the material starts to flow and screw holes are weakening the structure so they will just rip out if they are too close. But you can - within reasonable limits - reduce the contact area and the mounting strength doesn't change.