Why don't cases seperate CPU from GPU for thermal reasons?

[Pets4Ever]

The two main sources of heat are the CPU and GPU. Why are they located in the same compartment? Wouldn't it make more sense to separate them so that case fan curves for the different zones could reflect whatever component was actually in there? I heard that a riser cable for a GPU doesn't impact performance significantly so this should be entirely possible.

I say this as a user who is considering an AIO because I want better airflow for my mobo and GPU, not because I think the CPU needs it.

 

[_roman_]

I run my Fractal design meshify 2 for an important reason without the top cover and the side glass panel. Open air with a cpu tower cooler and a grahpic card. With good quiet components it does not matter much. I was forced to sell my previous graphic cards msi radeon 6800 z trio as the graphic card was most likely around 40dba or higher with 40% rpm speeds.

There are quiet graphic cards with very low noise.

 

[Pets4Ever]

I run my Fractal design meshify 2 for an important reason without the top cover and the side glass panel. Open air with a cpu tower cooler and a grahpic card. With good quiet components, and not trash graphic cards like my previous owned and soon sold msi radeon 6800 z trio, it does not matter much.

There are quiet graphic cards with very low noise.

That computer will stay a bit longer in that state. Thermaltake may sell maybe end of june 2025 the T30 140mm fans finally.
I have a need to cool my DRAM and therefore I need the top cover mounted again. I want to assemble it with new fans. Noctua could not sell me a working product with a correct invoice a few weeks ago, sold by fake noctua shop on amazon.de. working product means in my case a hassle free and proper product and a customer support who is able to read german and answer those questions after 100€ purchase.

Hmm I'd prefer to keep the case closed personally for both aesthetic and acoustic reasons.

 

[LabRat 891]

The two main sources of heat are the CPU and GPU. Why are they located in the same compartment? Wouldn't it make more sense to separate them so that case fan curves for the different zones could reflect whatever component was actually in there? I heard that a riser cable for a GPU doesn't impact performance significantly so this should be entirely possible.

I say this as a user who is considering an AIO because I want better airflow for my mobo and GPU, not because I think the CPU needs it.

As I recall, more than a few chassis' over the years have done this to varying degrees. Ex. Many ITX mini-chassis' do this via a Riser, into a separate airspace.

Unfortunately, by-nature of AGP/PCI/PCIe, the Add-In-Cards *need* to be somewhat close to the Chipset+CPU/SoC. Working around that, requires much more expense (risers, trace complexity, PCB costs) or subverting universal compatibility.
In the case of the most-recent GPUs, Gen5 PCIe does not like being extended via passive riser. Gen5-> + riser, requires re-driving or re-timing of the signals. -Those ICs are not cheap.


I will mention though, that you can 'fix' this w/o a new case. (you might not be able to make it pretty, though). Any 'paperboard' packaging or craft supplies, can be cut and folded into air ducts, guides, and separators. If you have a windowless case (or don't care), the cooling benefits are surprising and immediately tangible.

 

[Pets4Ever]

I searched it up previously and while they do exist, I think they should be the norm rather than a random alternative. If the standard is not good, the standard should be changed.

 

[LabRat 891]

I searched it up previously and while they do exist, I think they should be the norm rather than a random alternative. If the standard is not good, the standard should be changed.

Conceptually, you are 100% correct. However...

BTW, what you're asking *used to be* standard. Every single pre-built desktop PC I've had hands on made before the 20teens, separated the CPU's airflow from the rest of the case, in some manner.
IIRC, it was part of Intel's 'Thermally Advantaged Chassis Design' guidelines.

IMHO, bigger coolers and DIY PC/PC Gaming taking off is what 'killed' the standardized duct(ing).

 

[Pets4Ever]

Conceptually, you are 100% correct. However...
View attachment 402116

BTW, what you're asking *used to be* standard. Every single pre-built desktop PC I've had hands on made before the 20teens, separated the CPU's airflow from the rest of the case, in some manner.
IIRC, it was part of Intel's 'Thermally Advantaged Chassis Design' guidelines.

IMHO, bigger coolers and DIY PC/PC Gaming taking off is what 'killed' the standardized duct(ing).

But if you slap in a cooler that takes up like, 25% of the case by volume, now you're restricting airflow for all the rest of the components, no? It's always located right next to the main exhaust.

 

[unwind-protect]

My watercooled case still has the radiator externally mounted.

 

[Pets4Ever]

My watercooled case still has the radiator externally mounted.

What case? Did you record before and after temps? Thanks in advance

 

[unwind-protect]

What case? Did you record before and after temps? Thanks in advance

I drilled holes for the hoses myself. Some generic black thing from IIRC from CompUSA.

Lemme see whether I can dig up a photo...

 

[freeagent]

You could try to find a case that runs in BTX, that way your GPU is above your CPU. It definitely helps on the CPU side.

 

[wheresmycar]

There are mini-ITX cases out there that use the 2-compartment layout but I think you're talking more about true thermal isolation. Essentially the CPU and GPU in completely separate chambers, each with their own airflow path. More like dedicated intake and exhaust fans per chamber, or for the GPU side, letting the GPU manage its own intake but with a clear exhaust route (with fan/s) to avoid heat build-up. This is possible with riser cables but no case currently supports it (not that i know of). The trade-off: thats a thick case to begin with - not everyones cup of tea. Then as Labrat points out you're also contending with riser cable limitations, but not impossible, just adds an additional layer of complexity and cost. If a case of this sort was to come into fruition, the case design would also need to account for the riser's clean signal path. No sharp bends, the least amount of interference, short cable, etc. It's the the kind of design that works best for mITX cases which usually end up being pretty expensive. So, pricey SFF motherboards, cases that will no-doubt cost more because of the materials and complexity, bulkier case that takes up more space, more expensive PCIe 5.0 riser cables (looking to the future), etc. All in all, its a costly setup that probably isn’t worth the effort for most people unless CPU and GPU thermals eventually hit a point where they just can’t coexist as a couple in the same apartment. Personally, I can't imagine seeing my CPU and GPU forcibly divorced

Motherboards adding GPU slots on the back is another possibility but again, the design complexity and increased manufacturing cost just won't cut it.

There is a solution to the madness but unfortunately it hasn’t been standardized across different performance tiers or made affordable: hybrid AIO-cooled GPUs!! If these hybrids didn’t come with a $250–$500 premium (in some cases more), we could have easily gotten away with a single compartment layout without worrying about internal heat build-up. We’re starting to see more case options with support for 'side' fan/rad mounts (or even 240mm at the rear). You could easily configure both CPU and GPU radiators as exhaust while maintaining strong intake from the front and bottom. This seems to be the more practical solution for the way everything is currently standardised and also keeps the GPU firmly in place on its compromise-free board slot. Sadly out of reach for most, since hybrid GPUs are still limited to top-tier cards that cost a fortune. Well, thats one way to keep the CPU and GPU happily married. I should have been a marriage counsellor.

 

[Braegnok]

The two main sources of heat are the CPU and GPU. Why are they located in the same compartment? Wouldn't it make more sense to separate them so that case fan curves for the different zones could reflect whatever component was actually in there? I heard that a riser cable for a GPU doesn't impact performance significantly so this should be entirely possible.

I say this as a user who is considering an AIO because I want better airflow for my mobo and GPU, not because I think the CPU needs it.

My Lian Li A4-H2O chassis has the graphics card located in rear compartment they call it the sandwich layout design,.. there is room for a full size card, but I have SFF card in it.

A4-H2O - LIAN LI is a Leading Provider of PC Cases | Computer Cases

A4-H20A TRIBUTE TO SFF The A4-H2O is an 11-liter small form factor case, a collaboration project with DAN Cases. Based on the original DAN A4-SFX case sandwich layout concept, the A4-H2O continues the minimalist and compact form factor with improved hardware compatibility (including AIO water...

lian-li.com

 

[Caring1]

My Meshalicious System has the GPU behind the Motherboard also.

 

[Onasi]

Frankly, while there ARE cases that make provisions for separating CPU and GPU, the real solution is essentially a full on ground up reappraisal and potentially overhaul of the entire ATX ecosystem. Which is unlikely, to put it mildly. But, and I have been banging this drum for a while, original PCI-e AIB card spec was just never meant for 400+ watt triple-fan quadruple-slot monstrosities. Feels like a comprehensive change is needed.

 

[io-waiter]

The adherence to the ATX standard are looser and looser. As mentioned above many cases that need separation have it and the others solve it by airflow.

My Lian Li Dan A3 is slightly "modified" so that it does not exhaust on the back it breathes in troug the back and bottom and exhaust up and out in top on the front part of the case. PSU takes air in from the side and exhaus out top as well. More and more cases are set up in this way so that all components are getting cool air in, in retrospect the BTX standard was a nice improvement tha we would have benefited from.

 

[Kwadratowicz]

I'm trying to tune up My cooling solution, and for me it's stupid that intake is in front where in front new gpu's have this pass through that will shove hot air just to cpu cooler.
I think that cases should use back as air intake, it can lessen the heat load on cpu, and gpu.

 

[BoggledBeagle]

You can easilly move the heat producing part of a CPU cooler in the case by using an AIO, there are also graphic cards available with them.

You can easilly guide the warm air coming out of the GPU away from the CPU cooler and let it consume the cold air from the front fans by using simple folded sheet of paper or plastic.

B

Thread 'Presenting my new invention - RAMGUARD'

May 22, 2025

I am happy to present my new invention, made from a paper carton from a Kakao box. It needs to be from Kakao, Cacao or any other product would not work!

You may also want to install a sheet of some transparent plastic to be less distracting and ugly, but that would not work either. Only paper from kakao boxes works.

The function is simple - with the air in the case moving towards the back of the case, this simple barrier prevents the hot air coming from the GPU to hit RAM sticks and significantly lowers their temperature, probably also lowers the temperature of the VRM heatsinks on the...

• BoggledBeagle
• Replies: 72
• Forum: Cases, Modding & Electronics

• 
• 
• 

 

[RogueSquadron80]

The two main sources of heat are the CPU and GPU. Why are they located in the same compartment? Wouldn't it make more sense to separate them so that case fan curves for the different zones could reflect whatever component was actually in there? I heard that a riser cable for a GPU doesn't impact performance significantly so this should be entirely possible.

I say this as a user who is considering an AIO because I want better airflow for my mobo and GPU, not because I think the CPU needs it.

corsair are making a case soon the air5400 which has a triple chamber, one for psu/cables, one for cpu radiator and one for gpu

 

[SL2]

A PIO-style board with a horizontal slot (instead of BTX) solves some problems, but you're gonna need a new case. I've only seen old budget models tho, and there are no other slots, but it's an interesting idea.

You could easily fit an 8-slot graphics card lol.

 

[Shihab]

Not all systems have discrete GPUs.
Not all discrete GPUs are > 150 Watt monstrosities (ditto for CPUs).
Not all discrete GPUs with high power draw are air cooled (ditto for CPUs).

Factor in the added complexity of both design and manufacturing, and assembly by the end-user, and you get a recipe for way too many 1 star reviews.

If complexity is acceptable, then a watercooler loop (even an AIO) for either or both components would be a much more effective solution, imo.

That said, for a well-ventilated case, I don't think having both the CPU and GPU in the same compartment should be much of an issue. If you're letting air retain inside long enough to have the entire volume heat up significantly, your problem isn't the volume...

Edit: Don't some SFF cases do this compartmentalisation? Even if not intentionally.

 

[lilhasselhoffer]

So, the simple answer is physics. The more complicated answer is that if you decide to start chopping the case into sections then you start having to demand specifications from a whole lot of entities to make that happen.

Take for instance an example of this working in the past. Dell used to use these ugly green plastic shrouds, which ducted the exhaust from the standard Intel cooler out the back 120mm fan. Some units I've taken apart even forego either the CPU fan or the 120mm fan in the back, because one fan was enough air flow. The GPU was often then supported on a rail, with it sucking from the bottom and pushing out the back of the PC. It was a solution that worked...but it was also highly tuned as price sensitive rather than performance enhancing. Their 10 cent plastic duct saved them a few dollars on a fan, and basically kept the thing from thermal throttling with the crappy CPU cooler just barely. Somebody could theoretically pull the fan, get a better one, and wind up with a cooler computer...or assuming they didn't bodge the form factor too much they could buy a better cooler and replace the thing to spend a bit more and get better performance and quieter operations.

So...why not do the same thing today? Well, your average air cooler for a CPU is a tower. The peak temperatures for the CPU and GPU are somewhere near 90 degrees Celsius. I know 95 is usually said, but the math is trivially different. Your average inside temperature is about 20 C. 90-20 =70. Q˙=k A(ΔTΔx). That's the equation for conduction heat transfer. Your q is theoretically at stable temperatures, so if you model A as a constant (same shaped flat rectangular surface), your q is the TDP in Watts, and if you use copper the k is 400 Watts/M/K. You can ballpark area. 100 fins, 1 mm thickness, each fin at 40 cm by 80 cm, or a total of 100*(40*80/100/100) = 32 square meters. So if the TDP of the CPU is 150 Watts, 150 = 400*32*(T*0.001), 150/400/32/0.001 = 11.71875. That means that the air enters the CPU at 20 C, moves away at 32 C, and deals with all 150 Watts of heat the CPU makes. In practice the flow of air means that the actual CPU temperature will be lower, because the difference between the CPU and the air itself will constantly create more heat transfer than the heat energy being output by the CPU.


So...the answer to your question is basically that heat transfer is now so much more efficient that they do not have to separate the sections. Yes, it would result in cooler components with infinite airflow...but in practice cutting the areas off from one another will likely increase opposition to airflow...and cause its own issues. Remember that when the change in energy is greater than the system is capable of taking the change in temperature goes rather quickly to zero, and everything heats up relatively quickly. Assuming your heat transfer is tuned...then 20 C versus 32 C is a trivial matter...

Please note all of this is assuming something like ATX. In an ITX or smaller environment the lack of area for the cooler, or the stagnation of airflow that decreases delta T, will cause issues.

 

[TheDeeGee]

I'm trying to tune up My cooling solution, and for me it's stupid that intake is in front where in front new gpu's have this pass through that will shove hot air just to cpu cooler.
I think that cases should use back as air intake, it can lessen the heat load on cpu, and gpu.
View attachment 402141

I tried that in my O11D, and the CPU temps were shite.

My cooler is a D15S.

 

[bonehead123]

I run my Fractal design meshify 2 for an important reason without the top cover and the side glass panel. Open air with a cpu tower cooler and a grahpic card. With good quiet components it does not matter much. I was forced to sell my previous graphic cards msi radeon 6800 z trio as the graphic card was most likely around 40dba or higher with 40% rpm speeds.

There are quiet graphic cards with very low noise.

Well, although your points are valid, the OP's topic does not really seem to be about noise, but rather HEAT, which is the bane of everything electronic....

And as already stated, open-air rigs are NOT everyone's cup of tea, me included

 

[Athlonite]

I tried that in my O11D, and the CPU temps were shite.

My cooler is a D15S.

How many fans do you have in you O11D