A few months ago, engi­neers from AMD set a world record for proces­sor speed. Using liq­uid nitro­gen and liq­uid helium cool­ing, they clocked an FX-​8150 proces­sor up to 8.429GHz for just barely long enough to qual­ify for the record.

This is an amaz­ing feat. It also shows the chal­lenges fac­ing chip design­ers who want to push the lim­its of how fast chips can go.

For some rea­son — per­haps I’m feel­ing overly nos­tal­gic this week — this story reminded me of the Com­modore 64. For those who don’t remem­ber this dar­ling of the 8-​bit per­sonal com­puter era, the C64 ran at a pal­try 1MHz, or 1/​8429 of the speed those AMD engi­neers achieved. But the clock speed wasn’t just a lim­i­ta­tion of the 6510 proces­sor inside.

Indeed, the proces­sor actu­ally ran at “approx­i­mately 1MHz”. In many early machines, the crys­tal oscil­la­tors for gen­er­at­ing the video sig­nal also were respon­si­ble for the sys­tem bus. Steve Woz­niak started the prac­tice of using a 14.31818 MHz oscil­la­tor in Amer­i­can Apple II’s, which is 4x the 3.579545 MHz fre­quency needed to make an NTSC color TV sig­nal. Divide it by 14, and you get the CPU clock speed of early Apple II’s and the Com­modore 64: 1.023 MHz.

In both the Apple II and the Com­modore 64, the video chip and the CPU share the same mem­ory, because RAM wasn’t exactly cheap back then. The RAM essen­tially runs at 2MHz, with the CPU and video tak­ing turns access­ing it. Now, the 4116 DRAM chips in the Com­modore 64 are designed for a 375ns cycle, or roughly 2.6MHz. So you’d need to replace the RAM with chips designed for faster clock speeds.

But then there’s also the mat­ter of keep­ing every­thing in sync. In its built-​in C64 com­pat­i­bil­ity mode, the Com­modore 128 can actu­ally run its CPU at 2MHz. But enabling this fea­ture dis­ables the dis­play. A faster CPU can’t share the RAM with the video chip. And since the CPU clock is derived from the color video sig­nal, the fastest you can clock the CPU is 14.31818 MHz unless you turn things around and drive the col­or­burst clock from the CPU clock, instead of the other way around.

What a mess! Maybe the big, excit­ing thing about the 8 GHz proces­sor exper­i­ment is the proof that we’ve decou­pled all the clocks inside our com­put­ers, that we’ve made crys­tal oscil­la­tors so cheap we can throw a bunch inside one com­puter. We can use dan­ger­ous liq­uids to clock our proces­sors to insane speeds yet still have video and access RAM!