Today is a very exciting one! We are witnessing how the tech underdog is now the performance leader. With Zen 3 AMD has pulled out so many wins that I am not even sure where to start. Let me quickly do a few spoilers.
First – They have improved maximum boost frequency.
Second – They also have improved instructions per cycle.
And third – These chips overclock really well!
Let’s start with covering some of the main changes that lead to performance improvement. AMD claims they have achieved 19% improvement on instructions per cycle over the last generation. You might be thinking – what are the instructions per cycle? Well, when you put two chips side by side at the same speed (in this case AMD used 4GHz), then the new CPU delivers 19% higher results.
To get this, AMD had to carry out architectural re-design in terms of how the calculations are executed within the cores. They have also redesigned the SOC, and more importantly the way cores are sharing cache. In Zen 2, there were two core complexes (CCX) per compute die (CCD) making each CCX up to 4 cores and 8 threads with 16MB of shared L3 cache. In Zen 3, the CCXes have been unified and now consist of a single CCX with up to 8 cores and 16 threads. This essentially doubles the available cache pool which can be either shared by all cores or used by a single core, this is where single core performance sees great improvements.
Also, by reconfiguring the cores this way, AMD claims that it reduced core to core communications latency thus improving efficiency and in turn performance. With all of these changes, AMD was able to improve performance and still maintain existing TDP which resulted in 24% improvement per watt. When comparing it to Intel, AMD claims their CPUs are actually 2.8 times more efficient.
Before we show you the testing results, here is the test bench we used for both AMD and Intel systems. Do note that BIOS used in these tests was AGESA 1.0.8.0, with the latest drivers and Windows updates.
As far as settings are concerned, we have enabled DOCP for RAM, ensured the rest of the settings are at stock without any motherboard specific performance enhancements and set our CPU Fan speeds to full. Our studio ambient temperature was at 26.8 degrees C.
We tested the low end Ryzen 5 5600X and high end Ryzen 9 5950X and… I think you are in for a treat.
Let’s start with the standard Cinebench R20 Multi threaded results, and here we immediately see massive improvements. If we compare Ryzen 5, it scores 16% higher in Zen 3. What’s interesting, it is only 10% behind Intel 10700k which is an 8 core processor and is about 80 to 100 USD more expensive, more so – it does not come with a cooler.
When we look deeper into this benchmark we see nice but not crazy high clock speeds, with Ryzen 5 hitting 4.4GHz and Ryzen 9 hovering around 4GHz. Take note that the 8 core Intel CPU is running at 4.7GHz.
When we check out thermals, we see both of Zen 3 chips at around 60 C together with Intel CPU while older Ryzen 5 is mid 70s. This shows that there has been a great amount of optimisation done. It also indicates that there is a lot of headroom to push these chips higher, more on that later on.
Jumping to single core scores we have Ryzen 5 with 14% improvement over the last gen and Ryzen 9 smashing through that 600 point barrier.
While benchmarking a few times we saw 5950X going above max boost clock. Bear in mind that historically the larger core count CPUs had to sacrifice that single threaded performance, so you could only have either loads of cores or a few high speed cores. With Zen 3 it seems like this is no longer the case.
Next we have blender, and here in BMW and in Classroom tests (providing workload is carried out using GPU) the results are basically identical. These CPUs have enough horsepower to load up the graphics card into being the bottleneck.
And for the last synthetic test we’ve done 7-Zip. Here we see Zen 3 Ryzen 5 with 21% improvement on compression and 18% improvement on decompression over Zen 2. Because of that, it actually beats Intel chip by 12% on compression and 4% on decompression. Considering it has 6 cores against 8, that is an outstanding performance.
So far so good, but what about gaming? When the new RTX cards came out, we tried to push the Ryzen CPUs but 10700k would win every time so up till now it was our test bench CPU. But not anymore. Let’s look at the reasons why.
Starting with something easy for the CPU – Horizon Zero dawn. This game is mostly GPU dependent. At 1080p we still see 11% improvement on average FPS and 21% on 1 percentiles from Zen 2 to Zen 3. Also the new Ryzen 5 is beating the 10700k by 4% on average FPS and 8% on 1 percentiles. 5950X is in a league of its own with 3% higher FPS on average and 20% FPS higher on 1% percentiles against Ryzen 5.
Moving to 1440p we have exactly the same improvement between Zen 2 and Zen 3 on Ryzen 5 chip and it is leading over Intel chip by a few percent. In fact 5950X has the same lead. Basically each one of the chips scales really well in this game between 1080p and 1440p.
However in 4k we hit the bottleneck with this GPU and all of the chips deliver the same results within a margin of error. If you are a 4k gamer, then your main goal here is to get the fastest GPU possible.
Jumping into a less graphically intensive game – Shadow of The Tomb Raider in 1080p we see an even larger gap between Zen 2 and Zen 3. There is 36% improvement on average FPS and 32% on 1 percentiles. The gap between 10700k and new Ryzen 5 is also larger – we see 20% improvement on average FPS and 16% improvement on 1 percentiles. In this scenario Ryzen 9 5950X is only 5% higher on both average and 1 percentiles and it is because the CPU is actually so fast that even at 1080p we have the RTX 3090 being pushed into a bottleneck.
Bumping the resolution up to 1440p we have both of the new Ryzen chips at the top with GPU being a clear bottleneck. The difference from Zen 2 is not as drastic, staying at around 10% on average FPS and 26% on 1 percentiles. Intel on the other hand is still trailing behind, more specifically on the 1 percentiles.
When jumping to 4k resolution, all of the chips even out. This is expected behaviour so we won’t bore you with any more of 4k results.
Next game is Total War Three Kingdoms and it is just like Horizon Zero Dawn – a very GPU intensive game, so the difference in results is pretty moderate. Starting with 1080p we see 12% average FPS and 27% 1 percentile FPS improvement from Zen 2. About 9% average and 6% on 1 percentile improvement over Intel 10700k and about 20% on 1 percentile improvement between Ryzen 5 and Ryzen 9.
In 1440p the results even out as we are bottlenecked by the GPU.
I think that’s enough information on the graphs. Let me do a quick summary:
In every title we have at our disposal, we find great improvements from Zen 2 to Zen 3 and as we have demonstrated earlier in some cases they are over 30% higher, which is phenomenal!
I was worried when comparing Intel 10700k against new Ryzen 5 CPU as it is clocked higher and also has two extra cores, but once we saw the results it is clear the new Ryzen chip runs laps around it. So where is the catch? Is it temperatures?
When we did our Cinebench test, both new CPUs seemed to maintain rather low temperatures but maybe they didn’t get enough of a workout. We kicked it up a notch and put Prime 95 on for 10 minutes and here we see Zen 3 chips performing rather well staying at 60 and 65 degrees.
When we look at the clock speed, this explains the low temperature. They definitely don’t push it hard – Ryzen 5 staying at around 4GHz which to be fair is still above its 3.7GHz base clock. Ryzen 9 on the other hand is at 3.5GHz which is only 100mhz higher than the base clock.
I have a theory that upon release, the BIOS will be tweaked to allow for slightly better boost clocks when under full load. I did mention overclocking earlier, and this is why I believe there will be tweaks as there is definitely some headroom there.
We were curious, so we did a quick overclock on both of these chips. With Ryzen 5 we were able to push it to 4.85GHz all core (yes, this is 250MHz above max boost clock) and with Ryzen 9 we got 4.7GHz all core which is 200MHz short of max boost clock. These were achieved with 1.35V and load calibration set to level 3. This is only a quick test, let us know if you would like a full video on how to overclock Zen 3 CPUs.
Anyway let’s jump into the results from the Cinebench.
In single core 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 a whopping 18% bump as it is no longer throttling down.
When we look at the temperatures we are certainly pushing these CPUs much closer to the limit, but I think with some setting tweaks and optimisations we can bring this lower. So people who like overclocking their PCs will definitely have a great time here. At the same time I believe AMD can give us more performance via BIOS updates after the launch.
Talking about BIOS – if you are using 500 series motherboards, just bring it up to the latest version and it will be compatible with the new chips. If on the other hand you are using the older 400 series motherboards then you will have to wait until January 2021 for a compatible BIOS update.
To sum things up – what do we think about these chips and are they worth your money?
As far as performance and actually price to performance goes – hell yes! For many people out there Ryzen 5 5600X is going to be a perfect chip for gaming and moderate productivity tasks as it already has great performance and includes a cooler as well as plenty of headroom for overclocking.
Normally recommending Ryzen 9 would mean sacrificing gaming performance, but AMD has turned the tables. The new top of the line Ryzen 9 5950X is actually the fastest at every category – gaming, streaming, productivity and whatever else you throw at it and right now there is nothing on the market to compete with it.
Ryzen 9 is a smashing chip. I truly hope the new Radeon cards are as good as these CPUs. This is going to be a very exciting month.
Affiliate disclosure: as an Amazon Associate, we may earn commissions from qualifying purchases from Amazon.
Comentarios