I have been reading about Raspberry Pi overclocking recently. However, I still have some questions. On one of the sites, it stated that the frequency of the GPU " It has an impact on ARM performance since it drives L2 cache.sdram_freq -Frequency of SDRAM in MHz." Is there a ratio of ARM speed to GPU core speeds to prevent it from bottle necking the CPU when the CPU is at 1000 or more MHz? The same site also claimed over-volting can improve performance with out having to increase ARM speeds, is this true? Also, I heard that Turbo overclock setting and the "PI2" setting can corrupt the SD card, does it actually happen? I am using an Pi2. Thanks in advance.

1 Answer 1


WARNING: there be dragons here. The way you've asked your question implies that you know little about overclocking, and how it works. This is not necessarily a bad thing. The Raspberry Pi is a good way to learn about such things, but the process of learning is error prone. A mistake could mess up your setup, or even ruin your RPi. Better this device than your main computer though, right?

To clarify the rest of my answer, I'm going to define a few terms:

  • Overclocking: the process by which we increase the speed at which the CPU cycles. The higher the value, the more instructions the CPU processes in a given time interval (usually 1 second)

  • Overvolting: the increase in voltage that's applied to the CPU.

  • Binning: The process by which chips are culled into specific categories based on their performance after manufacturing. For example, this is often what makes a chip an i5 instead of an i7. Please see the wiki page for more info.

For all currently released models of the RPi, the GPU is in no way a limiting factor of performance. It might struggle if it's given too little RAM for its designated tasks, but I'd recommend against changing it in any significant way.

When you increase the speed of your processor, say from 900 MHz to 1 GHz, you're increasing the amount of work that the processor needs to do. 100 MHz is a relatively small increase, so you probably won't notice a difference in power usage.

As you bump the usage up higher and higher, the processor will consume more and more energy. Overvolting is a way to compensate for that. Some chips can see huge performance increases through overclocking and/or overvolting.

The binning processes means that your chip meets the set minimum, and usually a little over for a margin of error. Some chips will be capable of performance significantly above the set default. How much can really only be determined by experimentation.

I haven't stumbled across anything online that suggests you'll see much performance from messing with the RAM. If I recall correctly, it's currently significantly faster than the CPU is able to take advantage of.

When experimenting, make sure to make small adjustments. I'd start bumping up the speed, and if it seems stable, push it up a little more. If the RPi starts to have issues, you can try adding some voltage, or just slowing it back down. It's a lot of trial and error, but some people find it fun.

Possible Problems:

Several things can go wrong with an overclocked setup. You asked about SD card corruption, so I'll start there, but some others should probably be mentioned.

  • SD card corruption: yes, this one actually happened to me on my RPi 1 B. Usually this is due to the RPi not getting enough power for the higher usage, or the CPU moving faster than the SD card can handle. I know it sounds weird, but it can happen.

  • Heat issues: more work = more power = more heat. Electronics are allowed to get pretty warm safely, but if it gets too warm it will damage the electronics. I think I remember reading that the SOC is safe below 80 degrees C.

  • Power issues: on a broader scale of the SD card corruption, your RPi might just randomly restart, have poor network performance, or behave oddly with USB devices if the overclock causes timing issues.

  • Dragons: I warned you... Overclocking can produce some very strange results. Try not to get eaten.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.