Radiator size, CPU vs GPU

[Maxxtraxx]

Hi all, had a thought swirling around in my mind and wanted to see what other LTT peeps think.

 

Sooo... My setup:

 

i7-4790k has an 88 watt TDP when stock, maybe a 130 watt draw when overclocked?

 

Nvidia Titan X has a 250 watt TDP when stock, maybe a 300 watt draw when overclocked?

 

for the CPU a "good" AIO like my H100i GTX uses a 240mm radiator to dissipate 130 watts.

 

for the GPU a "good" AIO like my EVGA Hybrid uses a 120mm radiator to dissipate 300 watts.

 

Does the anyone else see the connection my mind sees?

 

Are we as PC users going overkill on the CPU? has GPU AIO cooling just not matured yet?

Why do we consider them both good but have seemingly backwards priorities when it comes to cooling our components?

Would you rather... Have a 120mm radiator for the CPU if you could have a 240mm radiator for your GPU? OR does this current setup seem logical to you?

 

Just looking for some thoughts on the subject, Thanks everyone!

[Lotus]

Yup. People go nuts with their CPU coolers. They think having a load temp of 65­°C is somehow better than 80°C (Intel CPUs only). In reality, a Cooler Master 212 Evo is perfectly fine for consumer platforms. I'd only step up to something like an NH-D15 if I were cooling a very high TDP chip like an i7-5820k or an FX-8350.

[The_Stig]

-edit- misunderstood post

[skyress3000]

power draw does not equal heat output, but yeah, GPUs output more heat.

[Benji_w]

Yup. People go nuts with their CPU coolers. They think having a load temp of 65­°C is somehow better than 80°C (Intel CPUs only). In reality, a Cooler Master 212 Evo is perfectly fine for consumer platforms. I'd only step up to something like an NH-D15 if I were cooling a very high TDP chip like an i7-5820k or an FX-8350.

 

Because heat kills components and the cooler you can keep an overclock the better it will be for that part. Also for the 8350 if you keep it under 65c you will remain stable, heat is a massive thing for the FX range.

[Maxxtraxx]

power draw does not equal heat output, but yeah, GPUs output more heat.

 

power draw does not equal heat output? My impression is that more electrical power consumed by the silicon "engine" equals more power lost in the process of performing the needed calculations in the form of heat.

 

Just like a gasoline engine, a higher horsepower/performance level consumes more air and fuel and necessitates more cooling ability to remove the extra excess heat generated by the engine.

[Maxxtraxx]

Sooo...

 

Would anyone rather have a 120mm CPU AIO and a 240mm GPU AIO vs the opposite?

Does that make sense? Thoughts?

[Lotus]

Because heat kills components and the cooler you can keep an overclock the better it will be for that part. Also for the 8350 if you keep it under 65c you will remain stable, heat is a massive thing for the FX range.

There is so much wrong with this. Chip degradation doesn't occur on Intel chips unless you're sustaining 85°C+ 24/7. You can regularly hit 100°C (thermal throttling) with no problems if you don't regularly stay at that range. You can sit at 82°C 24/7 and not damage your CPU. It won't hurt it. Voltages on the other hand can be more problematic and are more likely to cause chip degradation.

 

The FX chips don't have an internal thermometer, but instead rely on a socket thermometer that guesses CPU temps since it's not on the CPU itself. This is why 65°C is the max temps for FX series CPUs, due to the different way temps are measured. 65 on AMD is like 85 on Intel.

 

Sooo...

 

Would anyone rather have a 120mm CPU AIO and a 240mm GPU AIO vs the opposite?

Does that make sense? Thoughts?

 

Yup. That would be what I would prefer. Although, generally the GPU 120mm AIOs have super thick rads, so usually they are not comparable to a typical CPU 120mm AIO.

[EdInk]

Sooo...

 

Would anyone rather have a 120mm CPU AIO and a 240mm GPU AIO vs the opposite?

Does that make sense? Thoughts?

If a 120mm is good enough to potentially dissipate heat for a 300W TDP component then clearly 240mm rad is overkill for a potential 130W TDP component. My theory is pump speed is calibrated differently for example the H100i and H80i as that dictates how quickly heat is removed until the law of diminishing returns applies.

[Benji_w]

There is so much wrong with this. Chip degradation doesn't occur on Intel chips unless you're sustaining 85°C+ 24/7. You can regularly hit 100°C (thermal throttling) with no problems if you don't regularly stay at that range. You can sit at 82°C 24/7 and not damage your CPU. It won't hurt it. Voltages on the other hand can be more problematic and are more likely to cause chip degradation.

 

The FX chips don't have an internal thermometer, but instead rely on a socket thermometer that guesses CPU temps since it's not on the CPU itself. This is why 65°C is the max temps for FX series CPUs, due to the different way temps are measured. 65 on AMD is like 85 on Intel.

 

 

Yup. That would be what I would prefer. Although, generally the GPU 120mm AIOs have super thick rads, so usually they are not comparable to a typical CPU 120mm AIO.

 

I wouldn't run an intel at 80+ during normal operation and it will eventually degrade it, Voltage and heat decrease the life of components, no matter how you see it, it's a fact. Doesn't matter if it's within safe margins or not.

[m_tieu]

Since people usually only have one CPU but multiple GPUs, it's very hard to find cases that can mount multiple 240mm radiators that close together. Most cases have room in the front and top but compatibility is not ideal. They use 120mm radiator because it's good enough and more cases can support this form factor.

[EdInk]

Hi all, had a thought swirling around in my mind and wanted to see what other LTT peeps think.

has GPU AIO cooling just not matured yet?

Motherboards provide heatsinks for the MOSFET and chipset and have a standard for CPU heatsink installation making it easier to make a waterblock for just one component whereas provision has to be made for a GPU to cool it's VRAM and VRM plus the need to cater for several 3rd party PCB configurations. I guess that's why full waterblocks have matured faster than an AIO solution for GPU's

It pretty much died after the Corsair HG10 and NZXT G10 (I think) ...

[Lotus]

I wouldn't run an intel at 80+ during normal operation and it will eventually degrade it, Voltage and heat decrease the life of components, no matter how you see it, it's a fact. Doesn't matter if it's within safe margins or not.

All Apple users would disagree with you. They routinely use thermal throttling (100°C) in their ultra slim form factor stuff on purpose. You don't see their stuff failing. Data simply doesn't support your alarmist claim.

[EdInk]

Since people usually only have one CPU but multiple GPUs, it's very hard to find cases that can mount multiple 240mm radiators that close together. Most cases have room in the front and top but compatibility is not ideal. They use 120mm radiator because it's good enough and more cases can support this form factor.

Then at least give consumers the option to install one with 240mm just as the CPU AIO market is, offering consumers 360mm AIO's. I think they know there isn't significant performance gains going from 120-240mm on GPU''s which I reckon is down to the pump design and limitations.

[Benji_w]

All Apple users would disagree with you. They routinely use thermal throttling (100°C) in their ultra slim form factor stuff on purpose. You don't see their stuff failing. Data simply doesn't support your alarmist claim.

 

I know people who repair those macs who will disagree with you too. Mac's aren't bulletproof.

[EdInk]

I know people who repair those macs who will disagree with you too. Mac's aren't bulletproof.

You do? You think you can hook me up. I'm after a G5 Mac Pro for a PC mod project.

[Benji_w]

You do? You think you can hook me up. I'm after a G5 Mac Pro for a PC mod project.

 

Not really they're already owned by customers.

[ToastyBear]

Semi conductors conduct and resist using different silicon isotopes. At the quantum level, one isotope can change to another and the rate this happens can be increased with heat. This actually happens all the time for CPUs however, a 1 atom change won't ruin the electrical pathways. It only becomes problematic when we get down low as 6nm where each transistor is only a few dozen atoms wide.

[Megahurt]

Cpus have heat spreaders, gpus don't, that's the biggest difference, they also tend to have a bigger surface area, so the heat transfer is better.

[m_tieu]

Then at least give consumers the option to install one with 240mm just as the CPU AIO market is, offering consumers 360mm AIO's. I think they know there isn't significant performance gains going from 120-240mm on GPU''s which I reckon is down to the pump design and limitations.

There are brackets out there like the one from NZXT or Corsair that allow you to install any compatible CPU AIO that you like to your graphics card. Most people install a 120mm but you could install a 240mm or even 360 if you'd like. The benefit of doing so would be little since temperature isn't the limiting factor of the max overclock anyways, especially if it is being watercooled. 

[paradigm249]

Yup. People go nuts with their CPU coolers. They think having a load temp of 65­°C is somehow better than 80°C (Intel CPUs only). In reality, a Cooler Master 212 Evo is perfectly fine for consumer platforms. I'd only step up to something like an NH-D15 if I were cooling a very high TDP chip like an i7-5820k or an FX-8350.

my 5820k never goes above 60 with the hyper 212. i wouldnt say its very hot