AMD Shows Off Zen 4 Overclocking, But Questions Remain

Back in the April AMD made news by saying it was “gonna try to make a big splash with overclocking” with its upcoming Zen 4 CPUs. That would be somewhat of a departure from Zen 3, as it’s not exactly known for its overclocking headroom. Now that Computex has come and gone, we’ve been able to see Zen 4 in action.Despite AMD’s statements, it’s not clear how much the overclocking situation has changed. While Zen 4 clearly allows for overclocking, Zen 3 has never impressed in this regard. Zen 4 may not, either.

As a refresher, at Computex AMD showed a prototype 16C/32T Ryzen 7000-series “Raphael” CPU running Ghostwire: Tokyo. A CPU clock speed monitor was running in the corner, so we could see its clock speeds. Although clocks fluctuated in the low 5GHz range during the demo, the chip did hit a notable peak of 5.5GHz. Dr. Su said this is normal, as it will hit variable clocks depending on the workload. We don’t know what the actual boost clock of the chip is, but it’s clearly higher than the 5950X’s 4.9GHz.

In an interview with PCWorld, AMD’s Robert Hallock confirmed nothing fancy was required to hit those clocks. He said they were using a standard 280mm AIO cooler you can buy online. This is a not-so-subtle reference to the time Intel was caught using a chiller to cool a 28-core desktop Xeon chip. Regardless, he said the CPU wasn’t overclocked, and that “most of the threads” were running at 5.5 GHz. This begs the question: if it can hit 5.5GHz on its own, how high can it go with an overclock?

There’s one additional thing to point out. AMD released info on its upcoming AM5 chipsets (above) and you’ll note it doesn’t list overclocking as a feature offered on B650 boards. AMD has clarified that, saying it will indeed allow overclocking. This means every motherboard in the stack will allow it, so it’s open season when AM5/Zen 4 launch this fall.

But what kind of results can we expect? Color us skeptical, but we’re still not expecting much. For example, the 5.5GHz AMD showed off is the current high water mark for 16 core CPUs. It’s the current single core boost clock of Intel’s binned Core i9-12900KS, after all. If AMD is allowing its 7000-series CPU to get to 5.5GHz on its own, right out of the box, it seems like going even further might be a fool’s errand. As we’ve stated before, if AMD could get it to run at 6GHz without fancy cooling, why limit it to 5.5GHz? Even if it’s rated for a single-core boost of 5.5GHz, and you get it up to 5.7GHz, that’s still less than a four percent single-core overclock.

Over the last five years, AMD has chosen to leave relatively little on the table for manual overclockers, preferring instead to ship CPUs that run quite close to their maximum possible frequencies out of the box. While it may still be possible to manually overclock an AMD Ryzen for performance gain, we’ve had far more luck cranking up clocks on high core count CPUs like the Ryzen Threadripper 3990X, where all-core overclocks of 300-500MHz are possible given sufficient cooling. In this kind of scenario, OCing can still pay dividends over and above stock clock — but Threadripper is a workstation platform and a workstation platform limited to artificially lower clocks at that.

For Zen 4 AMD has cranked up the power requirements by a significant amount, which will also allow it to raise clocks. It’s gone from 105W TDP on the 5950X to 170W TDP, with a maximum socket power of 230W. That’s a huge boost, and will give AMD some added flexibility. Still, it seems like the song will likely remain the same. The lion’s share of the benefits could ultimately come down to overclocking, but not on core clock speeds. Instead it’ll be left to overclocking memory and Infinity Fabric, just as it was on Zen 3. Even Robert Hallock himself has noted that’s where most of the gains have historically come from on AMD’s CPUs. This is seemingly confirmed by reports that AMD is focusing heavily on memory overclocking with Zen 4, via its new EXPO technology.

None of this is meant to be a slight to AMD, because as we’ve said before the world has changed when it comes to overclocking. For both AMD and Intel, the days where they could leave 20-30 percent of a CPU’s clock improvement (or more) on the cutting room floor are long gone. As transistor density increases and node sizes decrease, it’s becoming more difficult to achieve higher clock speeds while keeping thermals in-check. This has been the pattern for some time now, and there’s no reason to think that will suddenly change with Zen 4. Intel and AMD may make some limited carve-outs for overclockers, but we expect both companies to reserve the vast majority of their performance improvements for themselves.

Continue reading

Intel Discontinues Overclocking Warranties as Hobby Continues to Die

Intel will no longer offer its overclocking warranty option, though customers with existing plans will still be able to use them.

Intel Reveals Rocket Lake Price, Positioning, and Overclocking Features

Intel's Rocket Lake is set to blast off and the company has lifted the curtain on the last few bits of data we didn't yet know.

Intel’s Arc GPUs Will Include In-Driver Overclocking, up to 4,096 ALUs

Intel's Xe drivers will include overclocking support natively, with a maximum of 32 Xe cores and 4096 ALUs.

AMD Extends Smart Access Memory Support to RX 5000 GPUs, Adds Auto Overclocking

AMD's latest driver includes some new goodies for RDNA and RDNA2 GPU owners.