In a prior blog post, I proclaimed proudly that my latest scratch-built audio workstation would handle any sort of processing load my musical ambitions could throw at it.  High-def 24-bit, 96 kHz audio.  Surround sound.  Blah, blah, blah.  That was a foolish statement.

Yes, we stand at the event horizon of the audio processing singularity, but one thing prevents us from crossing over (and becoming elongated spaghetti noodles):  With the latest generation of processors come new applications (and lazy programmers) that gleefully consume every technological resource Moore’s Law tosses at our hungry little mouths.  Consequently, my music projects have been stalling, just in sight of the finish line.  Yeah, I could save some CPU cycles by dropping down to 16-bit or 48 kHz (or both), but in my quest for excellence (read “audio geek pride”), I’m unwilling to do that.  I am also unwilling to upgrade my hardware, given that the state of the art in CPU technology (compared to four years ago, when I first built this machine) promises only a 30-50% processing boost for audio applications.  I know.  “Huh?!  Aren’t today’s processors magnitudes of order more powerful than those of four years ago?”

Here’s the problem for audio professionals:  Most of the performance gains in recent generations of processors relate to multithreading, where the most powerful processors have an abundance of parallel processing “cores.”  However, a single channel of audio demands serial processing.  That is, a mono channel of audio with, for example, three processors applied to it – say, EQ, delay, and reverb – must, by natural law, process each effect in turn because, after all, the reverb has no way of predicting what the delay will do to the audio signal.  It must wait.  Likewise for the delay and EQ.  So, CPUs with high core counts, while generally beneficial, are somewhat crippled in the audio world, except perhaps where your core count surpasses your channel count (with a few to spare for the OS and host application).  To put this in perspective, most commercial mixing sessions have three dozen audio channels, minimum, so until Moore’s Law brings us 64 cores (and applications that can efficiently implement them), the full capabilities of those processors will be unrealized.

This is where we all hold hands and say, “fark.”

A Technological Alternative to Pseudo-Profanity

Without the promise of a worthwhile return on my investment in hardware upgrades, I chose to investigate another option, spoken of only in dark circles:  Overclocking.

Actually, most folks are by now at least familiar with the concept of overclocking though they’ve probably not tried it, myself included.  It’s scary stuff, operating outside the boundaries prescribed by equipment manufacturers and their eye-squinting warranties.  Executed poorly, overclocking can render your mission-critical audio workstation unstable, putting irreplaceable recording sessions and data at risk.  At the extreme, overclocking can destroy expensive computer components.

But then there’s the promise of overclocking done properly – meaningful gains in processing power for the price of a single Oreo™ cookie.  [I haven’t eaten Oreos in years – too many carbs – so that sounds pretty delicious right now.]  I choose the Oreo cookie.

Thankfully, my computer’s hardware is optimized for overclocking (through sheer dumb luck – overclocking wasn’t among my original shopping criteria).  My motherboard, an Asus P9X79 LE, appears to be purpose-built for overclocking, with an overwhelming array of options available in its BIOS and plenty of how-to videos available on YouTube.  And my processor too, an Intel Core i7-4930K @ 3.4 GHz, where “K” means that the core multipliers and other overclocking features are “unlocked.”  Who knew?

So far, my steps have been of the baby variety.  I performed the simplest procedure recommended by ‘experts,’ adjusting the core multipliers from “Auto” to “44.”  That’s it.  It’s supposed to generate 4.4 GHz, but I instead got a boost to around 4.1 GHz.  It’s a subtle increase but an increase nonetheless, and without cost – no perceivable increase in heat yet, even under full load.  [Technically speaking, my core temps are hovering around 30°C while the purported “danger zone” is way up around 85°C.]

But, it turns out, 4.1 GHz is all I needed to get my bloated projects over the finish line.  I can now insert my icing-on-the-cake (read “CPU-hogging”) audio plugins without running out of processing power.  And if I need more, it’s apparent I can push the overclocking further.

But – let’s not forget – a newb overclocker can only get so far adjusting the core multipliers alone.  To reach 5.0 GHz and beyond will require a much greater commitment.  Studying the arcane scripts.  Speaking the forbidden words.  A lamb sacrifice.  As such, I hope to get by with what I’ve got until AMD or Intel produce something worth upgrading to.

There’s a warm glow on the event horizon.  It’s so close.  One day, we’ll get there.  One day.