Thermal Paste on a Plastic Chip/CPU Package

[Tiberiusisgame]

Are there negative aspects to applying thermal paste to a plastic chip package, like the CPU on a Raspberry Pi 2 which does not have an IHS?

 

I appreciate your answer but before you respond, please consider the following:

 

1. I've researched this issue and have not found anything definitive either for or against it. All the equipment I've worked on use thermal pads on plastic packages, but then manufacturers shy away from thermal paste anyway, likely because it's expensive, messy, dries up, etc.

2. Please don't comment on why I need to cool a plastic package. There are plenty of practical applications for dissipating heat in a chip that isn't metal-backed or doesn't contain an IHS. Consider this an inquiry independent of the specific application.

3. I'm talking about typical thermal paste. Not an epoxy, not paste designed for extreme thermal conditions. The manufacturer's websites don't mention conflicts with a plastic package, but then that's not their target market either. The CPUs in their target market requiring active cooling have either shown an exposed die or have an Integrated Heat Spreader. This is an entirely different, albeit plausible application for thermal paste.

4. I've run heat sinks with thermal paste on plastic packaged CPUs for Pis, routers, switches and game consoles for years, but I've never seen anyone else do this, or more importantly, advise not to do this.

 

Thank you for your consideration!

[Hairless Monkey Boy]

Wouldn't a thermal pad do the job just fine? You don't find anything about it because there is no reason to do it because better solutions exist.

 

Otherwise it should be fine, although unnecessary, to use a generic thermal compound in that way.

[Tiberiusisgame]

6 minutes ago, HairlessMonkeyBoy said:

Wouldn't a thermal pad do the job just fine? You don't find anything about it because there is no reason to do it because better solutions exist.

 

Otherwise it should be fine, although unnecessary, to use a generic thermal compound in that way.

Once upon a time, I would have agreed with the "good enough" sentiment. But then, chips weren't running at GHz speeds and didn't contain throttling and burst technologies to manage heat to within a specific margin (consequently translating heat into performance). They also weren't pushed to their absolute thermal limits because a two-year runtime is considered acceptable, and there were no SoC's, increasing the complexity and heat sensitivity. Consider your GameBoy, if you ever had one. That plastic package chip has none of the technologies I mentioned. It runs at a specific frequency and generates a constant temperature well within the chip's performance envelope. Consequently, my GameBoy still works after 20+ years in service. But my Gameboy doesn't route traffic from 50+devices in real time.

 

In some applications, yes, a thermal pad would work. But it wasn't what I had on-hand. That's not the question.

[Hairless Monkey Boy]

1 minute ago, Tiberiusisgame said:

Once upon a time, I would have agreed with the "good enough" sentiment. But then, chips weren't running at GHz speeds and didn't contain throttling and burst technologies to manage heat to within a specific margin (consequently translating heat into performance). They also weren't pushed to their absolute thermal limits because a two-year runtime is considered acceptable, and there were no SoC's, increasing the complexity and heat sensitivity. Consider your GameBoy, if you ever had one. That plastic package chip has none of the technologies I mentioned. It runs at a specific frequency and generates a constant temperature well within the chip's performance envelope. Consequently, my GameBoy still works after 20+ years in service. But my Gameboy doesn't route traffic from 50+devices in real time.

 

In some applications, yes, a thermal pad would work. But it wasn't what I had on-hand. That's not the question.

Fair enough.