As we all know, the first generation Ryzen CPUs that launched a little over two years ago made a huge splash and signaled AMD’s comeback into the mainstream as a serious contender and not the “budget brand” it had been considered for years.
It disrupted the market and pricing so much that Intel was forced to release their 8000-series processors in the same calendar year as they released their 7000-series parts, increasing core counts on the mainstream up to 6-cores with parts like the 8700K. They then of course followed that up the next year with the refreshed 9000-series, which features up to 8-cores with parts such as the Core i7 9700K and Core i9 9900K. This was a huge change for the company, who previously had remained stagnant releasing only quad-cores on the mainstream platform since 2008. It was a clear sign that Intel had taken the new threat from AMD seriously.
That being said, Intel did still possess a strong single-core lead due to their massive clock speed advantage and high frame rate gaming lead. Ryzen first and second generation CPUs did trail behind the latest Intel 8th and 9th gen processors in gaming, while they did have solid offerings, the top of the line gaming performance crown was still firmly in their competition’s hand.
Enter AMD’s all-new 3rd generation Ryzen CPUs, using the company’s Zen 2 architecture, which packs nearly double the cache (which AMD has aptly named “Game cache”) that aims to change the conversation about gaming, and further their lead in productivity workloads.
Ryzen 9 3900X and Ryzen 7 3700X Specifications
Ryzen 9 3900X
Ryzen 7 3700X
Base Clock (GHz)
Boost Clock (GHz)
Default Memory Speed
Wraith Prism RGB
Wraith Prism RGB
In addition to the two processors we’ll be looking at in this report, AMD has also launched an entire new line-up of Zen 2 based Ryzen 3000-series CPUs as well as a pair of refreshed Zen+ based Ryzen 3000-series APUs.
Zen 2 Architecture Overview
One of the of the biggest changes from Zen 1 and Zen+ architecture is the the all-new ‘chiplet’ die package, which Jim has gone over in detail in some of his videos,but we’ll explain here as well:
AMD has previously used a single chip complete with two “Core Complexes” or CCX’s arranged in configurations of 4 cores, with all data shared between them using its “Infinity Fabric” to move the data between cores. This single chip also included all other primary components of the SoC such as the I/O and memory controller. The Zen 2 design essentially takes those two CCX groups and strips them away from the I/O onto their own separate chiplet that is produced using TSMC’s 7nm manufacturing process. These chiplets can then be arranged onto the CPU package in sets of two to give us a maximum of 16-cores.
Along with the ‘Core’ chiplets, we also have a separate I/O die which is made using the older Global Foundries 12nm process to cut down on costs with something that doesn’t necessarily benefit much from the node shrink. This entire package together is called ‘Matisse’ and it is what makes up entirety of the 3rd-generation Ryzen CPUs with the exception of the APU parts.
The 7nm Advantage
With Zen 2, and 3rd-generation Ryzen CPUs, AMD has finally one up Intel in terms of manufacturing process. While the majority of Intel’s products are still using the company’s aging 14nm process, AMD has delivered Zen 2 on TSMC’s 7nm process. What this means is that for first time in the company’s history it has managed to produce products that are more efficient than their competitors.
In fact, AMD is claiming up to 45% better performance per watt when compared to competing products. The new process also brings with it an increase in maximum frequencies compared to 1st, and 2nd-generation, a fact AMD was not originally expecting when planning for the new node.
In fact, looking at the above chart we can see that compared to original Ryzen, frequencies have increased up to ~12% with Ryzen 3000. This alone should greatly increase performance, but as we’ll see see below it isn’t the only improvement they’ve managed.
With Ryzen 3000, AMD is introducing what they’re calling ‘GameCache’ really, this is just a catchy name for L3 cache, but what makes it really special is the fact that its been doubled from 16MB, to 32MB compared to previous-generation Ryzen parts. This alone manages to reduce memory latency by up to a claimed 33ns, and improve gaming performance by up to 21%, the company says.
Floating Point Unit: AVX-256
Another major improvement for Zen 2 is the redesigned floating point unit which is aimed at improving performance in AVX 256-bit workloads, which is an area previous generation Ryzen processors had difficulties with. This is achieved by doubling the Floating Point and Load Store bandwidth from 128-bits, to 256-bits.
All game benchmarks are taken from a 3-run average either in-game or using the title’s built-in benchmark utility. Performance is recorded using OCAT: the open-source frontend for PresentMon, which itself is an open-source capture and analytics tool which supports modern APIs such as DirectX 12 and Vulkan.
Keep in mind this is the first of many articles diving into the performance of Ryzen 3000 CPUs and the X570 Chipset, and due to time constrants as well as some very bad luck on my part (I went through 3 motherboards that failed during the review process) is review is not as complete as I would like it, which is partially why we won’t be covering PBO or overclocking today. Expect full motherboard reviews, storage analysis and a large gaming head to head with the 3rd generation Ryzen CPUs taking on the 9900K in the coming weeks.
For our CPU reviews we recieved AMD’s test kit which features a Ryzen 9 3900X, Ryzen 7 3700X, an MSI Godlike X570, an ASUS X470 Crosshair VIII, 16GB of G.Skill Trident-Z Royale 3600MHz, and a 2TB Corsair NVMe SSD. We also used parts supplied by Western Digital, Crucial, XFX, Gigabyte, Phanteks and NZXT.
For Power draw we use HWInfo64 for sensor readings during our Blender BMW27 benchmark. The draw is our highest number drawn throughout the test.
Here we see what should be expected on processors on a newer node. The 12 core 3900X draws less power than the Intel 9900K which has 4 less cores.
AMD Ryzen 9 3900X and Ryzen 7 3700X Productivity Performance
Cinebench is the benchmarked version of MAXXON’s Cinema 4D software which is used by design studios all around the world making CG scenes used in Hollywood movies and animation.
Here we see in Cinebench R20 the Ryzen 7 3700X coming out 41% faster than the first gen 1800X part, which is a great way to start off the benchmarks. Here we see the architectural gains as well as the clock speed giving a huge uptick in performance. The much cheaper part also barely gets beat out by the 9900K in single and multicore tests.
The 3900X also is over double the performance in multicore compared to the original flagship Ryzen CPU which was released at the same $500 price point that the 3900X currently occupies.
Blender is the free and open source 3D creation suite. It supports the entirety of the 3D pipeline—modeling, rigging, animation, simulation, rendering, compositing and motion tracking, even video editing and game creation.
We use the recently created Blender benchmark on their rendering of a BMW which is one of their shorter tests. Those interested in testing on their own systems can download it here.
Here we can see the 3900X beating the old 1800X flagship by taking almost half the time in this render, and again we see the 9900K and the 3700X trading blows which is a recuring theme in this review. The gap between the 9900K and the 3900X is just under 40% which is to be expected as rendering applications tend to scale very well with higher core and thread counts.
Corona is another rendering application, though this one does ray tracing. Score is listed in pixels per second.
This time the $329 Ryzen 7 3700X just barely edges out the Intel i9 9900K by just about 2%, the gap between the 3700X and 3900X is almost exactly 40%. Finally we see the 3900X is almost twice as fast as the first gen Ryzen 7 flagship counterpart the 1800X.
POV RAY 3.7
In POV RAY we see more of the same, the 9900K and 3700X by an incredibly tight margin, the 3900X about 40% faster than both of them and the 1800X being almost doubled in performance compared to the 3900X
For file compression 7-Zip is open source and is a standard CPU benchmark. For this test we use their built in benchmark at the default settings.
Interestingly enough we see the 3700X losing to the 9900K in decompressing by less than 1% but winning by over 23% in compressing. I’d say the 3700X takes the win here. The 3900X is 32% faster than the 8 core 3700X and trailing way in the back is the first generation 1800X.
Winrar historically has not been the most incredibly optimized application for high core count CPUs though they still are used, since we see the 3900X taking the win here, the big story though is the improvement VS the 1800X which is very much due to the additional cache added. Winrar previously was an application that Intel was heavily favored in due to issues with AMDs Infinity Fabric on their first generation parts. Zen 2’s improved performance here is seriously welcomed.
For Handbrake we use a Samsung promotional clip at 4k of the Barcelona soccer team taken in 4k in the TS format. we render it to the YouTube preset in H.264 at default settings and the H.265 preset with the render speed set to slow.
In H.265 we see the Ryzen 7 1800X fall way behind, this time the Ryzen 9 3900X is 133% faster absolutely demolishing the older CPU showing the improvement in architecture and the perfect scaling between cores. As before the gap between the i9 9900k and the 3700X is a few percent and the 3900X is just under 40% faster than the 9900k.
Dolphin 5.0 Emulator
Dolphin is by far the most popular Gamecube and Wii emulator with one of the most well developed and strong communities on the net. Emulators have been known to be incredibly bottlenecked by single threaded performance. This benchmark runs a Wii program that raytraces a complex 3d scene inside the emulator, this is a very intensive task on one core. Results are given in time.
Those looking to test this benchmark on their own system please check here.
Here we see in a very lightly threaded task the Intel CPU pulls away. Intel does barely win though with the slowest performing Ryzen 3000 CPU here coming in at 7% slower, so by no means is this a blowout.
AMD Ryzen 9 3900X and Ryzen 7 3700X Gaming Performance
Counter Strike: Global Offensive
For CS:GO we use the benchmark in the workshop, and have unlocked the frame rate. The run is 60 seconds long and starts at the beginning of the benchmark and ends roughly once the second set of smoke grenades pop.
in CS:GO we see the once Intel dominated game for esports gamers having the gap closed and while it still does win out overall, not by much. for comparison we can see how much the 1800x trails behind in the average FPS, though all of them average over 240FPS and all of them experienced serious slowdown when the smoke grenades went off.
This benchmark is a big deal, considering Intel’s mindshare in the mind of esports gamers. the performance of Ryzen 3rd gen in esports titles is something we will be diving into in future articles to see exactly how close the two companies have become.
The Division 2
In the Division 2, we use the canned benchmark for the first 60 seconds of the benchmark. We test in DX11 due to using a Pascal-based GTX 1080 Ti.
Here we see in our results the Ryzen 3000 parts sit in the middle of the pack at around 5% away from the 9900K though even at the high preset we seem to be somewhat GPU bound throughout a good portion of the run.
Forza Horizon 4
For Forza Horizon 4 we test the entirety of the benchmark in the demo with dynamic resolution/settings off at the high preset.
Forza is where we see the biggest gap between the 1800X and the rest of the lineup. All three of the CPUs put up respectable numbers, though arguably the 3700X and 3900X are a bit smoother with a bit less variance. That being said the Ryzen 3rd gen parts and the 9900K sitting within around 5% of each other in the averages and about 10% in the minimums. AMD takes the lead but by a hair.
Grand Theft Auto V
For Grand Theft Auto V, we test during the last scene of the in-game benchmark, beginning once the scene starts at the railway bridge to the end when the car crashes into the oil tanker.
Here we see something a bit different. As Wendell from Level1techs found back in 2017 GTA tends to have an issue when the frame rate jumps above 170FPS where it causes stuttering or stalling that will cause these lower minimums. What was once thought to be a glitch or some sort of issue with Core i series CPUs is actually the hardware being too fast for the engine. This only seemed to affect the 9900K and 3900X CPUs oddly enough and you can easily tell looking at the .01% lows. Some users have reported the best fix for this is to limit the framerate to 160FPS or below.
With all that being said though we see that the higher end Ryzen 9 3900X achieves parity with the 9900K
For our Hitman 2 test we test the Miami level during a single player run walking through the busy vendor area with many NPCs this tends to be incredibly demanding for the CPU and the lower frame rates here show that.
The gap in performance is about 38% this time in our CPU heavy test on the averages between the two flagship Ryzen processors. Overall though the 3900X and 9900K trade blows with the 3700X trailing by about 10% in the .01% minimums. Not bad considering it is about $150 cheaper at time of writing.
Rainbow Six Siege
For Rainbow Six Siege we run the canned benchmark through the House level, the test is done for the duration of the benchmark.
Yet again even at 1080p on high we seem to be hitting a bit of a GPU bound scenario as the minimums are all somewhat close, The Intel CPU has lower average FPS but higher minimums making this somewhat of a wash.
With all that said The Ryzen 3900X and 3700X quite honestly surpassed my expectations. With all products running at stock (which is how most users actually end up running their system), the 3900X did serious work in our productivity suite, generally beating the similarly priced Intel Core i9 9900K by about a 30-40% margin. The 3700X also did a good job keeping up with the much more expensive i9 9900K while still being around $150 cheaper at time of writing.
As far as gaming performance goes, AMD has done quite a lot to close the gap with Intel here, and their decision to double cache has really paid off.
Looking at the above slide, we can see that overall, the Ryzen 9 3900X and the Core i9 9900K are an even match in gaming, with a less than 1% difference between them. While this is a rather small selection of games, it is still a very impressive achievement for the AMD part.
Moving on to the cheaper Ryzen 7 3700X, and the difference is greater, but still just as impressive as the 65W part manages to deliver just over 95% of the gaming performance of the 9900K at a ~33% reduction in price.
While I didn’t get the chance to properly test overclocking or PBO (Precision Boost Overdrive) the majority of reviews have noted that all-core overclocks don’t seem to exceed ~4.3GHz which may be due to the launch yields not being spectacular. It is difficult to say for sure at this point and we’ll likely know much more about the overclocking capabilities of these parts in the coming weeks as they get in the hands of the general population.
I think the biggest question on everyone’s mind though is “is this worth upgrading from my current system” and the answer really depends on what your workload is and what you’re currently using to accomplish it. Anyone running a older mainstream Intel CPU like the Core i7 7700K and below should seriously consider picking one of these up.
For users on first or second generation Ryzen, the fact that most motherboards will support updating the BIOS to allow a drop in upgrade is incredibly appealing, though someone using the previous generation Ryzen 7 2700X, may not feel the pressure to pick up a new CPU just yet and for those users, we’d likely recommend they wait until prices fall a bit once the 3rd-generation Ryzen CPUs have been in the channel for a few months. That said, if you’re using an Ryzen 7 1700, 1700X, or 1800X; the Ryzen 3700X , and even more so, the Ryzen 3900X will be significant upgrades. We’ve not yet tested the Ryzen 7 3800X or the lower-end Ryzen 5 six-core parts, but given the results we’ve seen so far, they will likely be just as appealing. Still, we’ll follow-up in the next week or so with those reports if we can get our hands on them.
So in closing I personally think the 3900X and 3700X are AMD putting their best foot forward and once again disrupting the market. While there will still be workloads and use cases for the 9900K, it’s not quite the end all be all of mainstream CPUs anymore. And with that I give the Ryzen 9 3900X and the Ryzen 7 3700X our gold Adored awards.