Let me just quickly share some results of overclocking on 5600X and 5950X before I go deeper and share with you my methods of how I chose to do it. In Cinebench R20 single core test we got about 6% more performance from Ryzen 5 and lost about 4% performance on Ryzen 9. This is almost a perfect scaling when compared to the overclock amount.
And in the multi core test we see 11% improvement on Ryzen 5 making it the same performance as the 8 core Intel 10700k while Ryzen 9 gets an incredible 18% bump as it’s no longer throttling down.
These chips are very much temperature limited. At the moment we are using a 240mm liquid All-In-One cooler (Fractal Design Celsius+ S244 Dynamic) with ambient temperature being 27C, so if we had a better cooling system then I feel we could push another 50 to 100MHz. If you are interested in getting this kind of performance, stick around as you may learn a thing or two.
I want to highlight a very important point – any kind of overclocking is not covered by warranty, and for someone who is not experienced in overclocking – there is a very good chance of breaking things. So do not attempt this unless you are fully aware of the risks involved as you may damage your CPU or worse.
There are a few different avenues you can choose for your overclocking. I will briefly go over several of them, starting with Precision Boost Overdrive or PBO for short – it is an extension of AMD’s precision boost which manages clock speed based on different workloads. With PBO you are able to adjust parameters like package power target, thermal design current and electric design current.
After raising these parameters the chip is given more headroom to push its performance, while it is not going to boost past max boost clock, it will allow it to boost more often providing it is staying under other limits such as temperatures. This is almost an effortless tweak and results can be very much hit or miss.
Another method of overclocking is using tools like auto OC. In this example either motherboard or software will automatically tweak the key parameters like voltages to get the most out of the CPU. Generally this provides a small amount of improvement and there is a potential of overriding boost by up to 200MHz with very low effort. Personally, I never really had any good experience with this.
Next method is manual adjustments using Ryzen Master, and here you can control per CCX clock speeds, peak and actual voltages, overclock and configure memory as well as infinity fabric speed. The nice thing about using Ryzen Master is the ability to tweak without restarting the system as you can run some quick tests within the software.
On the other hand if you go a bit too far with the tweaks you may end up crashing your PC and actually not being able to boot, you will need to default BIOS and start from scratch again. It is very beneficial to have motherboards with BIOS reset buttons so you can quickly recover, alternatively most motherboards have a more hands on process to reset the BIOS, so please check the manual.
A more old school type of overclocking is tweaking settings through BIOS. This is my preferred way, but also is probably more time consuming as you need to make changes, boot into the OS, test for stability and performance, check if it is ok – if not, restart, make the changes, test and repeat. With Zen 2 last year, there have been so many days spent optimising, literally pulling out my hair just to finally realise that stock performance is actually good enough.
Another thing to note – when you do manual overclocking, you lose automatic boost. With Zen 2 you have to choose either leaving your CPU at stock to have great single thread performance from boost or improve all core performance by overclocking. Unfortunately all core overclock is normally lower than maximum boost speed so you will certainly make sacrifices.
Zen 3 on the other hand seems to have much higher overclocking headroom. On the smaller 6 core 5600X we managed to overclock it to 4.85GHz which is 250MHz above maximum boost clock across all cores, improving single and multi threaded performance. On the 16 core 5950X we were able to overclock to 4.7GHz which is 200MHz short from the max boost clock. If you remember, the last generation, 3950X could only go as far as 4.3GHz, so this is not bad at all. Do bear in mind that every CPU is different, some will overclock better than others.
So how did we get here? We overclocked through BIOS. I’ll repeat myself again but this is very important – tinkering with overclocking voids your warranty so do it at your own risk. Back to it – first boot into BIOS, here you can turn on DOCP to overclock your RAM straight away. Then scroll down to CPU core ratio and set your multiplier. I would recommend starting at 4.6GHz for the smaller core count CPUs like Ryzen 5 or Ryzen 7 and at 4.5GHz for Ryzen 9.
After it’s done, head down to voltages and set CPU core voltage to manual, then change it to 1.3V to start with. Go to VRM control (in ASUS motherboard it is called External DIGI+ Power control), here under CPU load line calibration set it to level 3. Load line calibration helps to apply more voltage while the CPU is under heavy load which in turn provides better stability, overwise when you do intensive tasks your voltage level will drop and you will likely experience a crash.
A thing to consider – the higher the overclock the more voltage you will need which produces higher temperatures, so it’s paramount to have a good cooling system. Now that the settings are ready, we can save and boot into windows.
I do my initial testing in Cinebench R20 using multithreaded test and recommend installing a hardware monitoring tool, we use Hardware Info to track temperatures and voltages. If it runs smoothly 3 times in a row then I feel comfortable to say it is “draft” stable and will try to push it further. I go back to BIOS, up the frequency by 25 or 50, leave the rest as is and test again. If during the run temperatures do not reach 90 C but it crashes, then it is likely that the voltage is too low, I would increase it by 1 step which is about 0.125V and try again. If on the other hand you are hitting thermal limits, then bring back frequency and also lower the voltage by 1 step. Tweak like this until you find the best balance. Pro tip – for stability find your sweet spot and turn it down a notch.
We already showed the benchmark results in Cinebench, and those were pretty good. To verify we also ran Blender benchmark, both BMW and Classroom test. Here we found similar results but will focus on the Classroom test as it is longer and has better indication on long term performance. Ryzen 5 has about 7.5% improvement while overclocked and Ryzen 9 has completed the test 13.5% faster.
For stability reasons we had to slightly turn down our overclocks to stay within the thermal limits and let me tell you… these chips are definitely having a workout here. Ryzen 5 is now running in the mid 80s which is 15-20 degrees hotter than stock. On the other hand Ryzen 9 is almost reaching 100C which is 35 or so degrees hotter than stock. Pretty toasty but definitely a serious improvement in performance.
Before we make our full recommendation, let’s compare the results in gaming for both. Starting with CS:GO. Here with maxed out settings at 4k we see Ryzen 5 go from worst, if you can call it that… to being the leading chip with 4.5% improvement on average FPS and 3.5 % improvement on 1 percentiles. However Ryzen 9 has not lost any performance even though it’s maximum clock speed has been reduced.
Next game which is mostly GPU dependent is Total War Three Kingdoms and in both 1080p and 1440p unfortunately we see no change from overclock on either of the chips. A quick observation here – Ryzen 9 with higher core count has better 1 percentile performance while in 1080p overall.
Another GPU focused game is Horizon Zero Dawn. Here at 1080p with Ryzen 5 we see 1% improvement on average FPS and 5% improvement on 1 percentiles. Ryzen 9 even while handicaped is too fast for the RTX3090 which is the bottleneck here.
Bumping up the settings to 1440p we see Ryzen 5 with 2% improvement on average FPS but an impressive 10% improvement on 1 percentiles. Here the overclock really makes a difference and actually brings performance much closer to Ryzen 9. Also interesting to see that due to the overclock Ryzen 9 1 percentiles are improved by 3%.
Lastly we have Shadow of The Tomb Raider and here at 1080p we see overclocked Ryzen 5 has 3% lower average FPS and about 1% lower on 1 percentiles. Ryzen 9 again has 3% improvement on 1 percentiles. In 1440p there is no difference between these CPUs as we are heavily bottlenecked by the GPU.
If you feel like you want to go a step deeper into overclocking then there are a few other tools at your disposal, starting with overclocking the Infinity Fabric. It is best to keep the ratio between Infinity Fabric, memory clock and memory speed 1 to 1 to 1. Memory speed that you see on the RAM is actually double as DDR 4 is doing two instructions per clock thus effective speed is half of what is advertised. With this in mind you might be able to get up to 1900 or in some cases up to 2000MHz speed to further up your performance.
The last common type of performance tweaking is actually undervolting. This is normally done to reduce temperatures to either deal with cooling issues or in situations where you want to make your PC very quiet. This is very common in small form factor PCs where you are limited by space for high capacity cooling and naturally you don’t want the PC to sound like a jet plane taking off. This means you will deliberately be reducing the CPU performance though.
So… overclocking on Zen 3, what are we seeing here? At the low end with Ryzen 5 when we overclock we do gain a bit more performance, but that chip already performs really, really well so unless you need that extra few percent performance improvement or you want to do productivity tasks without buying a higher core count chip, then maybe it makes sense. But if you are purely getting it for gaming then probably leave it as is and enjoy the very cool and quiet system with already good performance.
At the high end for people who require maximum performance, overclocking Ryzen 9 can make sense, and considering its cost – it is likely you can also afford a really good cooling system to handle it. Moreover with the performance this CPU has, the loss of 200MHz boost clock is actually not noticeable, in fact due to all core overclock it seems to perform better in most workloads. So it looks like a win win, now all you need is some good luck with getting that golden chip.
Affiliate disclosure: as an Amazon Associate, we may earn commissions from qualifying purchases from Amazon.