AMD Ryzen 7 2700X Overclocked to an Insane 6GHz Across All Cores

AMD Ryzen 7 2700X Overclocked to an Insane 6GHz Across All Cores

Every time AMD or Intel launch a new CPU, you can bet some group of people somewhere are planning a liquid nitrogen party. With AMD’s Ryzen 7 2700X, that took the form of an overclocking competition hosted by Asus at its headquarters, with competitors like Der8auer fighting to hit the insane 6GHz level.

Spoiler alert: They managed to do it, courtesy of some aggressive LN2 (liquid nitrogen) overclocking, a pile of processors to play with, and a complete lack of cold bugs in the Ryzen 7 2700X. For those of you who aren’t aware, certain CPUs (and sometimes entire CPU families) are known to respond poorly to LN2 or liquid helium precisely because they don’t operate correctly below a certain temperature range. Everyone knows that running too hot can damage a chip, but not every CPU can handle -50C or below. Liquid nitrogen evaporates at approximately -196C, while liquid helium (which was not used at this event) pushes temperatures down to -270C.

One very interesting chart leapt out from Der8auer’s video and I think it’s worth discussing further. The clip above is the entire video; this link drops you off at the section in question.

Six years ago, I wrote a story called “Physics, Ivy Bridge, and the Slow Death of Overclocking.” Back then, we could already see how each successive generation of CPUs was struggling to match the overclocking potential of the previous. Today, the trends are even more stark. At 1.0v and 3.75GHz, the 2700X is a relatively svelte chip, drawing just 73W of power with a Cinebench 15 score of 1696. At 4.3GHz all-core boost, the Ryzen 7 2700X is pulling 185W and turning in a Cinebench score of 1937. Performance went up 1.14x, while power consumption went up 2.53x.

Back in the old days, when you couldn’t hit the overclock you wanted, dumping more voltage into the CPU often solved the problem. While this is still true today, the amount of wiggle room you have at any given moment is much smaller. Consider the end of the graph, where Der8auer records four different results at the same 4.3GHz clock speed. I’m not sure why he did this, and the video doesn’t explain it, but I’m glad he did — it’s an excellent example of why raising CPU voltage doesn’t produce results the way it used to.

At 1.425v and 4.3GHz, the 2700X draws 185W. At 1.5v and 4.3GHz, the 2700X draws 219W. A 1.05x increase in CPU voltage drove a 1.18x increase in power consumption. This graph illustrates precisely why the argument that Intel is somehow choosing to hold back CPU clock speeds to line its own pockets is so wrong. In the old days, you could rely on a voltage tap to hit higher clocks. Today, that voltage tap could easily make the problem much worse by contributing to hot spot formation on the die. The reason top-end CPUs don’t have the overclocking headroom they used to enjoy is because both Intel and AMD have used that margin for themselves, to enable them to continue launching high-end products. Hitting 6GHz on the Ryzen 7 2700X required as much as 1.85v — and that’s a voltage I used to pencil mod into a Duron to enable a 133MHz native FSB boot back in 2001.

Back in 1999, Intel’s Mendocino Celeron was beloved for its ability to hit 450MHz on air from a base clock of 300MHz, for a 1.5x overclock. Today, AMD’s Ryzen 7 2700X (and, to be fair, all of Intel’s CPUs as well) need LN2 to go from an all-core 4GHz to 6GHz. That’s not an indictment of AMD or Intel’s CPU engineers — it’s an inevitable consequence of the materials we use to build microprocessors and the limits of those same materials.

Der8auer was unable to break the Core i7-7820X’s record in Cinebench, but they did manage to beat the world Geekbench 3 record with a Ryzen 7 2700X at 5.8GHz winning over a Core i7-7820X at 6.2GHz. AMD might not have beaten Intel in Cinebench, but it can claim some new records that it didn’t previously own as a feather in its proverbial cap.