AMD Ryzen 9 3950X Review & Benchmarks – The Intel Destroyer

What’s up everyone, Eber here with Hardware Canucks. I have something very exciting here with me, the Ryzen 9 3950X. We have all been waiting for this day, and after a lot of teases from AMD we’re finally allowed to talk about the 3950X’s performance. Now I’ll be honest with you guys, I’m super excited about this processor because it offers a ton of cores and threads, but most importantly because it hits the final nail in the coffin in Intel’s expensive high-end lineup. The 3950X will retail for $750 USD while offering 32 threads on the X570 platform, and that platform is not only more modern but also a lot more affordable compared to the X299. It also offers a lot of features compared to Intel’s mainstream Z390 platform. There’s no time to waste, so let’s get into the performance results, but spoiler alert: this CPU is really, and I mean really, fast. However, is it worth $750 USD / $1000 CAD? Let’s find out.

Specs

I’m going to get into the benchmarks soon, but before that I do want to talk about the processor and what it’s all about just in case if you missed our earlier coverage of it. This CPU has 16 cores and 32 threads, which equals the ThreadRipper 2950X, but it also operates at much higher clock speeds too. Other than the smaller cache size, the
only thing the 3950X really lacks is the ThreadRipper’s quad-channel memory support. Compared to the 3900X it has eight more threads and 100MHz higher Max Boost Clock, but the base frequency is also quite a bit lower too. The TDP is also the same for both chips. The issue for some people might be pricing because the 3950X costs 50% more than the 3900X, so don’t expect the same amount of performance uplift compared to that processor. I think you’re really paying a premium for having the fastest AM4 CPU on the planet. Now I do want to be perfectly clear about this CPU. AMD isn’t targeting this for gamers whatsoever. In fact, it’s really meant for content creators or professionals who want to accelerate their workflow without investing huge sums of money into platforms like X299 or X399.

Synthetic Performance

With that out of the way, I think it’s time to get into the performance results. I do want to mention that we have replaced our testing procedure. In fact, we have eliminated a lot of the synthetic benchmarks and we have replaced them with more real-world use case scenarios. Generally, all of our test systems are the same. They are using 32GB of RAM clocked at DDR4-3600, GeForce GTX 2080 Ti, Corsair MP510 2TB SSD, and Corsair AX860i PSU. All motherboard specific enhancements were disabled too. For cooling we used a Noctua NH-U12S heatsink on all platforms except for X399 and X299 which are cooled with the larger Nocta NH-U14S. Let’s start off with Cinebench R15 and as you can immediately see… WOW. The 3950X blazes ahead of pretty much everything, even the i9-9980XE and the previous generation ThreadRipper. Cinebench R20 conveys the same story too, with the new Ryzen 9 leading everything. Single core benchmarks show that AMD has made huge improvements with their lightly-threaded performance, which is really good news since most professional apps and games don’t fully stress all the CPU cores all the time.

Real-World Performance

That’s it for our synthetic test, and now we’re going to move on to some real-world benchmarks and I think this is where it gets even more interesting. Starting off with Blender, it really does show how good AMD’s new Zen2 architecture stacks up against the super old Skylake X architecture used on the i9-9980XE. Remember this is a 32 thread AMD processor literally tied with an Intel HEDT CPU with 36 threads. Metashape is a new program for our benchmarks. It’s used for taking thousands of digital images and combining them to create 3D data. This workload prefers a combination of high core counts and higher sustained clock speeds, and that’s why we see the i9-9900KS being so competitive. However, once again, the 3950X wins.

In Reality Capture, which is a super CPU-intensive program that is a lot like Metashape since it also takes a ton of image data to compile multi-dimensional images – the 3950X is impossible to beat again. It even puts the ThreadRipper 2950X to shame and that is impressive. The Mozilla Compile test is one of our most CPU intensive since in it we are building a version of Firefox for Windows by processing through more than 10GB of data. Here the 3950X impresses again since its memory bandwidth and cache give it an edge. The i9-9980XE is pretty competitive too, but not for a CPU that costs way more than $1000.

Handbrake is a pretty solid test for multi-core video transcoding and here the ThreadRipper dominates, but look at the 3950X beating the i9-9980XE again. Intel should be pretty worried by now. Premiere Pro is an interesting program since it uses the IGP in Intel’s Coffee Lake CPUs to accelerate rendering. Therefore, the i9-9900K and i9-9900KS end up on top, but the 3950X ends up being the fastest non-accelerated CPUs here. In DaVinci Resolve, the i9-9900K gets another narrow win, but that’s because nearly all of our CPUs are slightly bottlenecked by the graphics card. Just take that into account before you assume a super expensive ThreadRipper build will get you better results.

Gaming Performance

Now on to gaming performance. I’m going to go through these really quickly, but the one thing to keep note of is that ran all of our tests at 1080P, so the gap between the CPUs could narrow if you step things up to 1440P or 4K because the GPU would be more of a bottleneck instead of the CPU. In all of these games, the results are pretty close together but you can see that while the 3950X is pretty competitive, the 3900X is a better and less expensive option if you’re building a gaming focused system. The main reason for that is that the 3900X was able to maintain higher overall clock speeds in most titles, and while it wasn’t by all that much the difference did end up showing up in the charts. Another thing that is interesting is AMD has almost closed the gap with Intel when it comes to gaming performance, even when compared to the i9-9900K series. They are not quite there yet, but seriously you wouldn’t be able to tell the difference with your naked eye.

Temps & Power Consumption

Moving onto temperatures, AMD made an odd claim that the 3950X was optimized for liquid cooling. However, I’m going to say this right now, that’s just marketing BS because you don’t really need an all-in-one liquid cooler to cool this processor. Our Noctua NH-U12S with a single 120mm fan was able to easily keep up with the NZXT Kraken X62 when cooling the 3950X while it was under full load. Therefore, unless you’re looking for more RGB goodness, I would recommend just sticking with a very good air cooler. Switching over to power consumption, you can see that the 3950X’s overall efficiency is a big reason why its temperatures are so good. When the CPU is fully loaded it consumes about the same amount of power as the 3900X, and that’s pretty impressive guys but not surprising since they both have the same TDP.

Conclusion

All right, so to wrap things up, I am genuinely blown away by what AMD has been able to deliver with the Ryzen 3950X. I mean they literally brought ThreadRipper-like performance to the AM4 platform for a fraction of the cost of their higher-end CPUs. I mean $750 is not cheap, but it offers an amazing value for content creators or other power users who need to take advantage of all of those threads to chew through their particular workflow.

I also want to take a few seconds to talk about Intel, because they are in a lot of trouble. I think AMD might have purposely sandbagged their early results by comparing the 3950X to the twelve core i9-9920X while this thing is actually competing against the eighteen core i9-9980XE, a processor that was retailing for a ridiculous $2000 USD just a few months ago. This new AMD processor has basically devalued Intel’s entire high-end desktop lineup, and I’m talking about the upcoming Cascade Lake X here too, not just the current CPUs. The $1000 USD Core i9-10980XE is only supposed to be a few percent faster than the i9-9980XE, but that processor was spanked by AMD’s flagship mainstream CPU and that’s a very big problem. I guess the last thing that needs to be accomplished by AMD is to make sure that this processor is available to buy in stores for $750 USD, because from what we have been hearing there’s going to be very limited quantities of this CPU, and that could potentially lead to price gouging. But otherwise, this is all I’ve got to say to AMD: Bravo! Well done, you’ve earned your amazing success.