How big are GPUs going to get?

[Agall]

28 minutes ago, Alex Atkin UK said:

Looking at that photo, I'd say that the far left is not making the same contact with the pins as the right as its being pulled at an angle.  This is what the bend radius is about.

 

The idea is to keep them all straight so that the resistance of each connection is as close as possible to each other.  Its when one pin has a higher resistance than another that we get the melting as more current goes across the other pins and that one with higher resistance, and that one also gets hotter due to the higher resistance.

Its just a very poor design for sure, but here we are.

 

Also, I'm not saying having options with the connector on the end is a bad idea.  Its just given the low margins on these cards it seems unlikely anyone would do it, instead choosing to stick with the reference design location that is more compatible.

Because the cables are individually sleeved, the length of the cables from the point of the bend are all the same that results in a difference in length for the rest of the cable. The difference in length is made back up at some point in the rest of the run since the cables have an overall equal length.

 

28 minutes ago, Alex Atkin UK said:

Looking at that photo, I'd say that the far left is not making the same contact with the pins as the right as its being pulled at an angle.  This is what the bend radius is about.

 

The idea is to keep them all straight so that the resistance of each connection is as close as possible to each other.  Its when one pin has a higher resistance than another that we get the melting as more current goes across the other pins and that one with higher resistance, and that one also gets hotter due to the higher resistance.

Its just a very poor design for sure, but here we are.

 

Also, I'm not saying having options with the connector on the end is a bad idea.  Its just given the low margins on these cards it seems unlikely anyone would do it, instead choosing to stick with the reference design location that is more compatible.

To really boil down my point:

 

Putting a tight bend on the right can reduce contact between the pin and socket on the outer edge while also putting too much of a bend on the inner edge. When the whole set of cables is anchored with a whole sleeve, this can happen, or when we're talking about the adapter.

 

This is simply not an issue with an individually sleeved cable. You can make the bend have every cable equal distance, where the extra few millimeters can be made up for at some point in the rest of the run.

 

I'll reiterate and use the previously included photo from Seasonic as an example.

 

The bends in these photos are no different, the only variation being the distance between the bend and the connector. For an individually sleeved cable, what is the specific reason why this would matter? 

28 minutes ago, Alex Atkin UK said:

Looking at that photo, I'd say that the far left is not making the same contact with the pins as the right as its being pulled at an angle.  This is what the bend radius is about.

Bend radius for individually sleeved cables isn't the same, being my whole point. There is no association between the left and right for an individually sleeved cable compared to one that has an anchor point and whole sleeving, as I demonstrate above.

[Alex Atkin UK]

43 minutes ago, Agall said:

I'll reiterate and use the previously included photo from Seasonic as an example.

 

The bends in these photos are no different, the only variation being the distance between the bend and the connector. For an individually sleeved cable, what is the specific reason why this would matter? 

Bend radius for individually sleeved cables isn't the same, being my whole point. There is no association between the left and right for an individually sleeved cable compared to one that has an anchor point and whole sleeving, as I demonstrate above.

The problem with the picture above is its not realistic, none of those bends shown should really be an issue (as they aren't factoring in gravity in the diagram) as its not showing the real-world case where the cable is tugging at the pin connections.

 

Yes individually sleeved is better (which is what PSUs generally do), but it doesn't fix the issue that if the bend is too close to the plug it will pull the connections at an angle.

 

Example being I have a Corsair normal 12VHWPR:

 

and also their Premium cable.

 

You can see how the sleeved normal cable is not perfectly straight into the plug, but the more flexible one is.  But still if I reduced the bend radius on the latter it would start to pull the connectors at an angle.  That cable separator seems to exist to avoid you doing that.

[Agall]

5 minutes ago, Alex Atkin UK said:

The problem with the picture above is its not realistic, none of those bends shown should really be an issue (as they aren't factoring in gravity in the diagram) as its not showing the real-world case where the cable is tugging at the pin connections.

 

Yes individually sleeved is better (which is what PSUs generally do), but it doesn't fix the issue that if the bend is too close to the plug it will pull the connections at an angle.

 

Example being I have a Corsair normal 12VHWPR:

 

and also their Premium cable.

 

You can see how the sleeved normal cable is not perfectly straight into the plug, but the more flexible one is.  But still if I reduced the bend radius on the latter it would start to pull the connectors at an angle.

Top example would be the scenario where disassociating the cables would be beneficial in my opinion.

 

So, you're suggesting a significant problem with the distance between the connector and bend is tilt within the actual terminal because of the distance from the bend to the plug?

[Alex Atkin UK]

1 hour ago, Agall said:

Top example would be the scenario where disassociating the cables would be beneficial in my opinion.

 

So, you're suggesting a significant problem with the distance between the connector and bend is tilt within the actual terminal because of the distance from the bend to the plug?

Right, as from what I read some models of the original plugs only had a small section that grips the pin, rather than being ribbed so it makes contact in several places along the length of the pin.

 

Also why it not being fully plugged in caused that one connection point to not go far enough onto the pin to make good contact.

Trouble is, you don't know which style plug you actually have from the early batch.

 

Makes you think this was a manufacturer cutting corners, as this diagram implies it was always supposed to be ribbed.

[maartendc]

I must agree with OP and had this exact thought yesterday.

 

I am running a SFF case, and it is almost impossible to buy a modern GPU now. My case is limited to a 2.5 slot + 270mm length card. That used to cover most midrange to high-end GPU options. Nowadays, you can barely find like a 7700XT that fits these requirements.

 

Technology is improving, making the GPU's faster. But at the same time, manufacturers are also just making bigger and more beefy cards to kind of bruteforce the problem. GPU power usage has also just kept increasing every year.

[Dedayog]

3 hours ago, maartendc said:

I must agree with OP and had this exact thought yesterday.

 

I am running a SFF case, and it is almost impossible to buy a modern GPU now. My case is limited to a 2.5 slot + 270mm length card. That used to cover most midrange to high-end GPU options. Nowadays, you can barely find like a 7700XT that fits these requirements.

 

Technology is improving, making the GPU's faster. But at the same time, manufacturers are also just making bigger and more beefy cards to kind of bruteforce the problem. GPU power usage has also just kept increasing every year.

This is an issue for CPUs as well. We just have the cooling solution separate.

 

The 14900k is brute forcing and needing a 360mm to minimize thermal throttling.

 

Can you imagine buying a 4090 and worrying about thermal throttling because they didn't want to make a beefy cooler?

 

You want more and more frames at higher and higher res, how do you think this is going to happen?

 

As always the consumer is the problem.  Always will be.

 

 

[Riyon133546]

soon all gpus gare going to be water cooled lol