• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

What is the purpose of not covering the back of GPU die

Joined
Aug 7, 2011
Messages
79 (0.02/day)
System Name BUILD
Processor INTEL® CORE™ i7-13700K
Motherboard MSI MPG Z790 CARBON WIFI
Cooling NOCTUA NH-D15 chromax.black
Memory 32GB G.SKILL 5600 MHz
Video Card(s) MSI RTX 4070 Ti GAMING X TRIO
Storage 2TB SAMSUNG 990 PRO
Display(s) GIGABYTE M27Q 1440p 170Hz IPS
Case PHANTEKS ECLIPSE P500A DRGB
Power Supply ORSAIR RM1000x SHIFT
Mouse COOLER MASTER MM711
Keyboard CORSAIR K70 RGB TKL OPX
A simple question...

Why almost every GPU has the back of the GPU die itself not covered ?
In other words, the backplate covers the entire card, but almost never covers the back where the GPU die is, and a lot of capacitors and resistors remain exposed.

What is the purpose of not covering that part ?

Almost every Custom Cards, every model, every brand NVIDIA/AMD are doing this. Also reference card like the new AMD 6000 series...

sdasda.png
 
Assuming "budget" is unconstrained , yes one could mill out , say a part that would fit snuggly .
 
Lol. Artistic choice? FE models from Nvidia do.
As does my RTX 2060:
maxresdefault.jpg


RX 5700 XT Evoke from MSI also:
9107_707_msi-radeon-rx-5700-xt-evoke-oc-custom-navi-goes-gold.jpg
 
My thinking is:
  • There are components on the back that should not be covered
  • The GPU produces a lot of heat which seeps into the back of the PCB as well. That heat can more effectively escape if that part of the PCB is not covered.
Alternatively, there are solutions like my Asus Rog Strix RX 5700 XT where the back plate actually covers the GPU's back side and is connected by a slab of thermal pad. But I suppose not every PCB design allows this, considering the components in the way.

Edit: Link
 
Last edited:
I think mostly they don’t wanna get labeled incompetent designers by tech reviewers. Heat trap and all that shenanigan.
 
Mine has also few holes for some reason.

1607850416450.png
 
I think mostly they don’t wanna get labeled incompetent designers by tech reviewers. Heat trap and all that shenanigan.

Not one of them actually have proved that have they ??? ( I haven't seen proof at least)
Who's to say that's not just bad case air flow too ?
 
In the OP's example, the GPU's cooler/heatsink bracket might be in the way if the backplate didn't have cutouts for it.

In the other examples, I can't discern the presence of a cooler bracket to cause any issues with the backplate.
 
The reason is mostly mechanical, really. If the manufacturer want to have the backplate to be as close to the board as possible, then the backplate would be at the same level as the bracket. If manufacturers decided that the additional Z-height is acceptable, they do sometimes just cover over it entirely.

As for why the bracket itself is not solid, my guess are 1). bulk caps used for the GPU decoupling is usually one of the tallest component on the backside of the board. It's sensible to just make a cutout for it instead of making a taller bracket to accommodate it. 2). brackets are usually mass produced for the entire GPU line, so whether a design uses thin or thick bulk caps, all get the same bracket. Custom brackets are quite rare. 3). manufacturers sometimes need access to the backside of the GPU for post production testing. Measuring directly at the point of load is the only way to accurately measure power delivery performance.

Personally I don't think thermal is high in the consideration. GPU are predominantly top-side cooled, and while non-negligible amount of heat is transferred to PCB, it would be spread quickly by the PCB power planes. In comparison, most VRM MOSFETs are bottom-side/PCB cooled, but we rarely see cutouts behind them. Heck, to this day I still confused why no one even put thermal compound behind them to the backplate...
 
The reason is mostly mechanical, really. If the manufacturer want to have the backplate to be as close to the board as possible, then the backplate would be at the same level as the bracket. If manufacturers decided that the additional Z-height is acceptable, they do sometimes just cover over it entirely.

As for why the bracket itself is not solid, my guess are 1). bulk caps used for the GPU decoupling is usually one of the tallest component on the backside of the board. It's sensible to just make a cutout for it instead of making a taller bracket to accommodate it. 2). brackets are usually mass produced for the entire GPU line, so whether a design uses thin or thick bulk caps, all get the same bracket. Custom brackets are quite rare. 3). manufacturers sometimes need access to the backside of the GPU for post production testing. Measuring directly at the point of load is the only way to accurately measure power delivery performance.

Personally I don't think thermal is high in the consideration. GPU are predominantly top-side cooled, and while non-negligible amount of heat is transferred to PCB, it would be spread quickly by the PCB power planes. In comparison, most VRM MOSFETs are bottom-side/PCB cooled, but we rarely see cutouts behind them. Heck, to this day I still confused why no one even put thermal compound behind them to the backplate...
Even though I agree that thermals are usually not a consideration when designing back plates, they might as well be. In the instance of my Asus Rog Strix RX 5700 XT, the back plate is connected to the PCB with thermal pads where the GPU and VRAM chips are, and it gets extremely hot to the touch as a result, indicating that there actually is heat transfer happening towards the back side of the PCB. The thick thermal pads also make the back plate bulge out a little bit in those areas, which scared me when I first saw the card (I thought the card itself was bent). :ohwell: I don't think it helps with heat dissipation, but it might be doing something with the even distribution of heat across a larger area of the PCB. What good or bad that does, I don't know. It might be helping maintaining acceptable GPU thermals - or it might have something to do with my terribly high VRAM temps.
 
95% of backplates are utterly useless for thermals or even counterproductive as they trap heat. Its only cosmetic. Form over function.

Bottom line, we never needed the backside covered for anything. Exposed components plus airflow are in fact optimal. Think naked bikes, there they sell it off as cool ;) Gamurs want to apparently add weight to be cool. Puzzling :) You see it in motherboards too. Ridiculous plastic everywhere, and when premium you get alu... yay.

Gotta love commerce
 
95% of backplates are utterly useless for thermals or even counterproductive as they trap heat. Its only cosmetic. Form over function.

Bottom line, we never needed the backside covered for anything. Exposed components plus airflow are in fact optimal. Think naked bikes, there they sell it off as cool ;) Gamurs want to apparently add weight to be cool. Puzzling :) You see it in motherboards too. Ridiculous plastic everywhere, and when premium you get alu... yay.

Gotta love commerce
There are people in every fan base who like being stupid. Look at budget cars with huge wings, "performance" air filters, chip tuning that does nothing, spaced out and overly cambered wheels that only increase tyre and wheel bearing wear, etc.

Back on topic: a well-constructed back plate can add to the rigidity of a heavy card (not much, though). I just wonder why they never come as a single piece with the PCI slot cover.
 
There are people in every fan base who like being stupid. Look at budget cars with huge wings, "performance" air filters, chip tuning that does nothing, spaced out and overly cambered wheels that only increase tyre and wheel bearing wear, etc.
Your ignorance is showing.
It's clear you aren't a fan of those mods, no one is forcing you to do it, but they are relevant to those that choose to do them.
 
Bottom line the first reason for the backplate is appearance, which like it or not, now usually signifies a quality higher-tiered model.

reason number two is to reduce pcb flex. Other methods can be more effective, but would not look as good (see reason one).

third, some do provide additional cooling. EVGA is especially good about placing thermal pads in the appropriate places on the back, using their metal backplate as a heatsink. This however, would not be needed if not for the obsession with appearance (see reason 1), which traps heat.

Finally, the back of the GPU die may be left open on some models because of the need for thermal dissipation.
 
Last edited:
My Aorus 1080 Ti has a copper section on the back of the card for the GPU. The marketing explanation reads:

"AORUS provides the all-around cooling solution for all key components of the graphics card. We take care not only GPU but also VRAM and MOSFET, to ensure a stable overclock operation and longer life. On the front side, the large copper base plate dissipates the most heat from the GPU and VRAM. On the back side, the copper back plate dissipates the heat from the GPU back side. All other key components are taken care of very well by the WINDFORCE cooling module. AORUS ensures the customer a better cooling solution in many ways."

card2.jpg
 
Hi,
Backplate is often just for r & r protection more than thermal cooling but does heatsink a bit if it has slots/.... in it.
 
They probably are doing that so that the card doesn't get too hot and can disapate heat kinda like there are vents on a laptop to let the heat out.
 
Your ignorance is showing.
It's clear you aren't a fan of those mods, no one is forcing you to do it, but they are relevant to those that choose to do them.
There is nothing ignorant about stating the fact that some mods serve no functional purpose. They're merely a device of aesthetic pleasure for some fans, and there's nothing wrong with it. Problems emerge only when these fans fool themselves (or others) into believing that said mods improve driving characteristics (which they don't).

Same thing with e.g. plastic back plates. They might look cool, but that's it.
 
they are relevant to those that choose to do them.

Imagine if they weren't... bit of a strange statement if you ask me. Personal relevance does not equal actually being useful in a practical sense... I mean, its fine if you want to use a water chiller for the last 100 mhz... but please don't tell me its practical or some sort of necessity when its clearly very inefficient.

Maybe you meant to say 'people have hobbies let them do them'... and then I can only agree, but even then, its good to call things what they are. If you don't you lose sight of reality - something marketing and commerce are very keen to inspire.

third, some do provide additional cooling. EVGA is especially good about placing thermal pads

You have got to be kidding me dude. EVGA is utterly shit at backplates and cooler designs in general. Get your facts straight. Sorry, but really.

They simply forget pads and then tell users to unscrew their solution themselves to apply them. AFTER memory started dying left and right. They release lots of designs that lack cooling for VRM and memory altogether because it doesn't exceed spec anyway. This one... actually did though. They simply cheap out as much as possible and overdid themselves here. They even completely axed the cooler in this particular case and rebranded to IceX or whatever (yes even for Pascal!), so now we have butt ugly fans :) Note EVGA is also keen to charge a small premium for some sort of brand recognition. Even on rebranded reference crap. The only thing great about EVGA is the warranty, honestly - but you're paying for it anyway.



Want more? I can provide. They also get it right a lot of the time... but do fail more than most at coolers.
 
Last edited:
Want more? I can provide. They also get it right a lot of the time... but do fail more than most at coolers.

You are complaining about Q&A, not the actual design.
 
You are complaining about Q&A, not the actual design.

No. The 1080 FTW design had a design flaw, and so did the 970 ACX. Did you even read the links? Surely this can't be new...
 
No. The 1080 FTW design had a design flaw, and so did the 970 ACX. Did you even read the links? Surely this can't be new...

Those are old cards. I'm talking about in their modern implementation.

Those designs that were truly flawed were all revisited at some point. I had a fix kit sent to me for the 1080 FTW back in the day, so doubt I need to read more about that one.

That said, they do have Q&A issues. Do I even need to bring up the PCIe slot killing 1080 FTW I had before that one? They ended up leaving me a hefty check after that. Their support is good for sure, but I'd rather it works out of the box.
 
Those are old cards. I'm talking about in their modern implementation.

Those designs that were truly flawed were all revisited at some point. I had a fix kit sent to me for the 1080 FTW back in the day, so doubt I need to read more about that one.

Sure, until they screw it up again, I mean they do apparently need to redesign the same thing every gen, for some reason :) Point was, its not very factual to say EVGA excels at 'placing thermal pads'...
 
Sure, until they screw it up again, I mean they do apparently need to redesign the same thing every gen, for some reason :)

It's been a few gens since a redesign and they aren't alone there (see XFX, etc).

Not excusing it, just pointing out this is not an EVGA exclusive issue. See above edit.
 
Imagine if they weren't... bit of a strange statement if you ask me. Personal relevance does not equal actually being useful in a practical sense... I mean, its fine if you want to use a water chiller for the last 100 mhz... but please don't tell me its practical or some sort of necessity when its clearly very inefficient.

Maybe you meant to say 'people have hobbies let them do them'... and then I can only agree, but even then, its good to call things what they are. If you don't you lose sight of reality - something marketing and commerce are very keen to inspire.



You have got to be kidding me dude. EVGA is utterly shit at backplates and cooler designs in general. Get your facts straight. Sorry, but really.

They simply forget pads and then tell users to unscrew their solution themselves to apply them. AFTER memory started dying left and right. They release lots of designs that lack cooling for VRM and memory altogether because it doesn't exceed spec anyway. This one... actually did though. They simply cheap out as much as possible and overdid themselves here. They even completely axed the cooler in this particular case and rebranded to IceX or whatever (yes even for Pascal!), so now we have butt ugly fans :) Note EVGA is also keen to charge a small premium for some sort of brand recognition. Even on rebranded reference crap. The only thing great about EVGA is the warranty, honestly - but you're paying for it anyway.



Want more? I can provide. They also get it right a lot of the time... but do fail more than most at coolers.
I'm well aware that happened with SSC and FTW 1070 and 1080 models. I was also victim of that with a 1070 FTW. Fact is, the fix was quickly sent out and after that the thing ran exceptionally cool. This was almost everyone's experience. The fact that it was an issue is because they forgot to do what has worked well for their cards' cooling properties. This points to a bad QC issue on release, not a design flaw. You cannot say they DO NOT do something, which is about a regular way of doing things, when it happened once with their backplates. That's called a one-off, an aberration, not a regular way of doing things nor a design flaw.. The 970 ACX design flaw had nothing to do with backplates. It was an off-center heatpipe to GPU chip issue. Please don't mix issues.

Was it lazy QC? Yes it was. Are their backplates used as an assist in effective cooling? yes. But please don't skip my other points which all lead back to point one, which says they are unnecessary in the first place.
 
Last edited:
Back
Top