AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force

Manufacturer: AMD AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force

Ever since receiving the details about AMD Ryzen 3000 CPUs at Computex on May 27 we have been eagerly awaiting the launch of these new Zen 2 CPUs, and at last the wait for 7nm is over – on 7/7 of course.

So will these new processors give Intel’s 9th-gen desktop parts some serious competition at the top of the desktop CPU market? There is only one way to find out, and it involves benchmarking.

Our own Josh Walrath is working on a dive into Zen 2 architecture as you read this, but we have performance from our Ryzen 7 3700X and Ryzen 9 3900X CPUs to share with you right now.

Let’s dive in!

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  1

Specifications

While we must wait until September for the massive 16-core Ryzen 9 3950X announced at E3, the rest of the Zen 2 lineup is available now.

AMD Ryzen 3000 Series Desktop Processors
Model Cores / Threads Base / Boost Unlocked GameCache Memory TDP Price Launch Date
Ryzen 9 3950X 16 / 32 3.5 Ghz / 4.7 GHz Yes 72MB DDR4 3200 105W $749 September
Ryzen 9 3900X 12 / 24 3.8 Ghz / 4.6 GHz Yes 70MB DDR4 3200 105W $499 7/7/2019
Ryzen 7 3800X 8 / 16 3.9 GHz / 4.5 GHz Yes 36MB DDR4 3200 105W $399 7/7/2019
Ryzen 7 3700X 8 / 16 3.6 GHz / 4.4 GHz Yes 36MB DDR4 3200 65W $329 7/7/2019
Ryzen 5 3600X 6 / 12 3.8 GHz / 4.4 GHz Yes 36MB DDR4 3200 95W $249 7/7/2019
Ryzen 5 3600 6 / 12 3.6 GHz / 4.2 GHz Yes 36MB DDR4 3200 65W $199 7/7/2019

CPU Benchmarks

While by no means a comprehensive list, we did put our two Ryzen samples through a few benchmarks to see how performance – single-threaded in particular – has improved with Zen 2. As you will see there have been some significant gains from Zen+ and the Ryzen 2000 series. But is it enough to best Intel?

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  2
PC Perspective Test Platforms
Motherboard AMD: GIGABYTE X470 AORUS GAMING 7 WiFi
GIGABYTE Z390 AORUS PRO WiFi
Memory AMD: G.Skill Flare X 16GB (8GBx2) DDR4-3200 @ 3200 MT/s XMP
Intel: Corsair Dominator Platinum 16GB (8GBx2) DDR4-3200 @ 3200 MT/s XMP
Storage Corsair Neutron Series XTi 480GB
Power Supply CORSAIR RM1000x 1000W
Operating System Windows 10 64-bit (Version 1903)
GPU Drivers NVIDIA: 430.86

Cinebench R20
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  3
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  4

The results in this particular rendering benchmark are very good for AMD. It’s still pretty remarkable to think of 12 fast cores in a desktop CPU for $499, and this is an example of how optimized software can take advantage of the Ryzen 9 3900X, though the Ryzen 7 3700X result might be more impressive considering its $329 price and proximity to the Intel Core i9-9900K here.

Geekbench 4

Moving on to the overall results from the latest version of Geekbench for Windows (4.3.4) we see a different picture from this group, with Intel taking the single-core advantage. AMD’s 12-core Ryzen 9 3900X outclasses the group in multi-threaded performance, and it will take Intel’s HEDT lineup to compete in this benchmark.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  5
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  6
7-Zip

Using the built-in CLI benchmark to measure both compression and decompression performance we now take a look at 7-Zip.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  7

As you can see the Ryzen 3000 CPUs claim the top two spots here, with the Ryzen 7 3700X edging out the Core i9-9900K. Impressive! But is any of this affected by the use of CL14 memory with the AMD CPUs and CL16 with Intel? Yes, to some degree, as this benchmark is certainly affected by memory performance.

x264 Benchmark

To get a feel for transcoding performance we ran the x264 benchmark at its default settings, with the averaged results of its four iterations (two passes each) below.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  8
POV-Ray

Quite an old test by modern standards, this freely available ray tracing benchmark can be run in single or all-core modes for another breakdown of these CPUs. As you can see Intel occupies the first four positions in the single-core test, but the new Ryzen CPUs are on top in the multi-core tests.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  9
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  10

Game Benchmarks

It turns out that benchmarking games to test CPU performance can be a tricky thing. The strategy of using 1080p to force a CPU-bound state with a powerful GPU is not new, and here we used an NVIDIA RTX 2080 FE to stress the processors on test.

Of the following game benchmarks not everything is what you might expect, and a few games were omitted after some anomalies were found when creating the charts for this review.

In the days to follow the selection of games presented here may be updated, but for now we can at least take a look at four recent titles, beginning with Far Cry 5.

Far Cry 5
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  11
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  12
Metro Exodus

After the rather poor showing from AMD CPUs in the first test, would Metro Exodus be a different story? Slightly, it turns out, but Intel still has an edge with these first two games.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  13
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  14
Shadow of the Tomb Raider

This next title was run in DX12 mode and cranked up to the “highest” preset. As you will see things tighten up considerably with this game.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  15
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  16

Shadow of the Tomb Raider appears to be more of a GPU bound game even at 1920×1080, possibly thanks to those ultra-high detail settings. For the last test (for now, anyway) we will turn to something a little less strenuous, even at the ultra detail/ultra memory preset settings.

Civilization VI: Gathering Storm
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  17
AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  18

The new Ryzen CPUs fare pretty well in the Civilization VI Gathering Storm benchmark, but even a 0.5 ms frame time disparity translates into 3.4 FPS lead for the Core i9-9900K over the Ryzen 9 3900X here. Intel’s Core i7-9700K is on top here, and it has proven to be a competitive option for gaming – though it may need to drop a bit in price with the Ryzen 7 3700X offering solid performance (and 16 threads) for less money.

Power, Pricing, and Conclusion

Power Consumption

While idle numbers in the chart to follow can be safely ignored as a true baseline for these CPUs as the “High Performance” power plan was enabled for all testing (and AMD’s “Ryzen High Performance” plan for these new Ryzen CPUs), we can at least get a feel for max power draw under load.

These numbers were captured using a watts up! meter at the wall, with the RTX 2080 FE card installed. Loads were generated using the all-core Cinebench R20 benchmark.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  19

Under load the Ryzen 7 3700X system pulled a total of 132W from the wall (remember this is not actual power consumption as the efficiency of the PSU is a factor), while the 12-core Ryzen 9 3900X managed to consume less power than the 8-core Intel Core i9-9900K.

Pricing

The Ryzen 7 3700X provides a tremendous value at $329 for a part that builds very nicely on the success of the Ryzen 7 2700X. Performance is up considerably over its Zen+ predecessor, and at just 65W it is quite efficient as well. This might just be the sweet spot of the new lineup – though we have yet to test the entire offering.

With an incredible value proposition for core/thread counts previously reserved for the more expensive Threadripper or Intel’s HEDT offerings, the $499 Ryzen 9 3900X is an exciting product for anyone looking to get the most from their multi-threaded dollar. Later this year AMD will release the $749 Ryzen 9 3950X, but the 3900X will remain a compelling value even at its launch price.

Conclusion

Zen 2 is finally here, and these new Ryzen processors fulfill much of the promise of the next generation of AMD CPUs on 7nm. So far we have tested the Ryzen 7 3700X and Ryzen 9 3900X, and both make a compelling case if you’re looking to upgrade or build up a powerful – and versatile – new system.

AMD Ryzen 7 3700X and Ryzen 9 3900X Review: Disruptive Force - Processors  20

Editor’s note: A look at AMD’s new X570 platform is underway as we encountered some early issues which were eventually resolved, but not in time to complete our testing for this review. Have no fear we will give you our X570 (and PCIe Gen4) impressions soon!

Those expecting AMD to thunder past Intel in gaming performance might come away disappointed with the few benchmarks shown here, but Zen 2 Ryzen – while still not up to Intel’s level in our game testing thus far – is still an improvement over the previous generation of Ryzen CPUs in this regard (and if you game at higher resolutions than 1080p you are going to be more GPU bound anyway).

In short, these new Ryzen processors continue along the disruptive path AMD has followed since early 2017, and if you crave more cores for your money, and aren’t after the last word in gaming performance (Intel appears to retain its advantage after our early testing), it’s an easy choice to make right now.

Video News

About The Author

Sebastian Peak

Editor-in-Chief at PC Perspective. Writer of computer stuff, vintage PC nerd, and full-time dad. Still in search of the perfect smartphone. In his nonexistent spare time Sebastian's hobbies include hi-fi audio, guitars, and road bikes. Currently investigating time travel.

24 Comments

  1. BackwoodsNC

    For the gigabyte gaming 7 board what bios are you running f40? Some reddit users that I seen couldn’t get ram above 2133. Wasn’t sure what X470 board they were using though. Granted they got parts a few hours early.

    • Sebastian Peak

      F40, yes. And this set of G.Skill Flare X runs at 3200 CL14 without any stability issues. The 3600 set on the other hand… That’s an X570 adventure now.
      (Actually, now that I think of it I didn’t test 3600 on F40 on this board. Now I want to.)

      • BackwoodsNC

        Curious to see if that 3600 kit works without issue. I ordered the 3700x from Newegg, who knows when they will ship it. I have the gaming 7 and 3600mhz Corsair kit.

  2. zgradt

    My last build was a gaming only i5-8500, so I’m sad that you dropped the 8400 from the gaming benchmarks. I already knew it’s multicore performance was pretty mediocre.

    • Sebastian Peak

      I’ll add it back in if I can figure out what went wrong with game testing. No reason an i5-8400 is ‘faster’ than 9600K/8700K/9700K/9900K unless I’ve lost what was left of my mind.

      • zgradt

        From the benchmarks I remember, it could almost catch the 8700 in gaming benchmarks. I just thought it would be funny to see it beating the new Ryzens in Farcry 5.

        Gaming benchmarks are mostly irrelevant to me for CPUs. I still have a 6700K in my LAN machine. I figure it has at least a couple more Quakecons in it

      • zgradt

        I’m also curious how the Ryzen 5’s compare. Mid range CPUs seem best suited to gaming. Best bang for the buck and all. I have a feeling that that crown still belongs to Intel. The i5s are tough to beat, and priced way more competitively than the i7s.

  3. Lorash651

    It would be nice if Civ 6 benchmarks would have turn-time completion, in addition to (or even instead of) FPS. I’d like to see what these little babies can do on those huge, packed-to-the-max maps with a lot of AI players where sometimes you can go do some chores between turns 🙂

  4. PCPerFan

    Why didn’t you use a 2080 Ti? Or test at 720p? You mention that the idea is to avoid GPU bottleneck, yet you do not use the highest-end graphics card you have available and you test at a standard resolution instead of a lower one. This seems like bad testing methodology and you seem to be aware of this fact, but I wonder why?

    P.S. – I hate how you have to login to comment now

    • Sebastian Peak

      I chose 1080 since that is by far the most popular resolution for gaming. 720 makes sense to totally eliminate a bottleneck in theory, but I still had to throw out two sets of game results since they made zero sense even at 1080p. AotS Escalation was ridiculous, for example, as it always is at lower resolutions. CS:GO is the one I would for sure bench at a lower res.
      Sorry about the login thing. It’s not been universally accepted. Things over on the pcper discord are pretty lively these days if you’re interested.

  5. funandjam

    I’ve got a 3900x on a x470 taichi mobo, with a noctua nh-d15 and I’ve seen various single cores boost to 4.65.

    What’s more, is that on my particular cpu is that cores 1 – 6 will reach those speeds at various times, while cores 7 – 12 will get to about 4.5 – 4.57.

    I wonder if this behaviour is the same on other 3900x’s too?

    the below is purely just speculation:

    In order to reach those boost speeds, the chipletts are binned, but those higher quality chiplets are limited in number. Since the 3900x uses two chiplets, have one of them be the better binned that can reach 4.6(or better) and the other be a ‘good-but-not-as-good’ chiplet. And why? so they can have more 3900x’s in order to keep the prices down to $500.

    • Sebastian Peak

      Interesting, and worth looking into. Coupled with AGESA updates there are now a couple of reasons to investigate boost clocks.

      • funandjam

        I have some updates:

        It appears that bios for some, if not most x470 boards to run xmp profiles are not great. What I mean is, on my x470 taichi, xmp profile for 3200 ran flawlessly with my 2600x. I swapped it out for the 3900x and that xmp profile for 3200 no longer works. It ran run no higher than 2400.
        Turns out that AMD has some bugs to work out with Agesa 1.0.0.3 for x470. I found this out after watching the below video from hardwarecanucks, got curious to see how my 3900x compared and saw that the xmp profile I thought I had enabled, it didn’t take and reverted back to 2133. The system works with ram at 2133 or 2400, so it isn’t like I can’t use it, just can’t turn the xmp profile up to 3200.

        https://www.youtube.com/watch?v=GnR6nCk8BRk

      • funandjam

        Asrock just released bios version 3.50 for the x470 taichi mobo’s which contains agesa 1.0.0.3:

        https://www.asrock.com/MB/AMD/X470%20Taichi/index.asp#BIOS

        looking forward to tonight for updating the bios and see if my RAM can finally be turned up from 2400 to 3200.

        I know it isn’t much performance, but it is still some and every little bit counts, hopefully this one isn’t borked too. Will report back in a day or so if it works or not.

        • Sebastian Peak

          It took starting at 2400 and slowly moving up in increments manually after setting the timings to finally reach 3200 on the ASUS ROG Crosshair VIII Hero board we have here. XMP was a no go, BIOS would eventually fall back and warn of a memory problem.

          • funandjam

            last night updated bios from 3.4 to 3.5, and it’s a mixed bag again. I got the taichi x470 because it has a really beefy vrm setup, little did I know about their feature – “once you upgrade bios you can’t revert back to an earlier version”, otherwise I would rollback.

            anyway, I’m stuck with this mobo for the foreseeable future(unless I brick it), it does work and here is what happened after I updated to 3.5 on my x470 taichi:

            So after further reading, it appears that the issue ins’t necessarily with the RAM, it’s that the bios version 3.40 wouldn’t allow the IF to go past 2400, so setting xmp above that would cause the system to never get past post codes, if it got that far.

            Bios 3.50 fixed that, and now xmp works, even though my ram is rated at 3200, xmp will work up to 3133 and the IF will match it, which is ok with me. between family, job and very limited time for gaming, I simply don’t have the time to mess with manually setting RAM and/or CPU settings,so 3133 is fine, much better than 2400.

            Now on to the not-so-good part of the update.

            bios 3.4:
            chiplett one max boost clocks – 4.6 – 4.65, all cores of chiplett one would at least reach 4.6 at various times.
            chiplett two max boost clocks – I was mistaken about max boost clocks for cores on chiplett two, I remembered it incorrectly when I wrote my first post. They actually only get to 4.35 and occassionally one might get to 4.4, but very rarely.
            max all core boost medium workload – 4.35 – typically what I got when running a game and some tabs in web browser
            max all core boost heavy workload – 4.225 – typically what I got when running a game, streaming or rendering a video and some tabs open in web browser.
            idle clocks on various cores would go down to in the 2000 range, sometimes even lower

            bios 3.5:
            chiplett one max boost clocks – 4.5 on 3, 4.4 on 2, and 4.35 on the last
            chiplett two max boost clocks – appear to be the same as with 3.4
            max all core boost clock medium workload – 4.225 ~ 4.25 doing same things as with bios 3.4
            max all core boost clock heavy workload – 4.05 ~ 4.1
            idle clocks stay in teh low 3000 range and never get lower

            changing windows powerplan to various different plans doesn’t have an effect either. Thinking about uninstalling ryzen software, then reinstalling to see if that changes anything. Otherwise, guess i’ll have to wait for asrock to fix this.

            I know, it’s a first world problem, the system still runs games very smoothly and other apps work just fine but it’s what little ocd i have that’s coming out in me, ha

          • funandjam

            ok, one last update:

            with the Asrock x470 Taichi, once the bios has been updated to a bios version that supports Ryzen 3000, it can not be rolled back to a version that does not support Ryzen 3000 series.
            Since both bios versions 3.40 and 3.50 support Ryzen 3000 series, I was able to roll it back to 3.40 so the max boost clocks and idle clocks would go back to what they are supposed to be when everything is set to default.
            What came as a surprise was that not only are the clocks performing the way they are supposed to (max boost is again 4.6~4.65, idle clocks get below 2.0 and voltage at idel drops to about 1.0 or .9v), but the IF is now matching whatever I set as xmp RAM profile.

            when the 2600x was running, xmp @ 3200 was just set and forget. But it appears that the RAM is a pit finicky with Ryzen 3000(not the best RAM) and while the IF does change with the xmp profile to match, the xmp profile is stable at 2966 and anything higher gives a memory error. It’s very close, and I just don’t have the time to sink into manually adjusting RAM to get better results, so it’s good like it is. If for some reason I find a need to go from 16GB of RAM to higher, i’ll look at getting RAM that is a bit better for compatibility at a higher speed(like 3600Mhz), but for now, clock speeds and RAM/IF issues are gone.

          • funandjam

            this is worth anohter followup:

            Asrock released another bios update for the x470: 3.60 which includes agesa 1.0.0.3 ABB

            I updated and observed the following:

            xmp now works for my RAM at 3200(which it is rated for) and my 3900x and IF matches correctly at this higher RAM speed – so it works correctly

            idle and max boost clock behaviour has changed:
            idles around 3.6 ~3.7 and it appears that AMD is a bit more aggresive on putting the cores back at this new idle clock where as before on 3.4 idle clocks would eventually go back down to around 2 or even lower, but would tend to float around between 3 ~4.
            max boost clocks don’t get any higher than 4.55 where before they woudl get to 4.6~4.65.

            ran cinebench R15 on both multi and single core tests several times on each, scores appear to be about 30 pts higher on multicore scores adn single core maybe 5 pts higher, this is after running about 20 tests on each. One other thing I did notice is that on single core runs, the “gold starred” core(core 3) appears to be utilized more of the time. I mean, it still switches between core 3 and core 1, but now tends to be more consistently on core 3 more fo the time. So far, it appears that RAM running at faster speed is helping perf a bit.

            Idle temps appear to be the same as before.

            idle voltages appear to be around 1.45 or higher, but when under a load will go down to 1.35’ish? thisis a head scratcher, no idea why this happens. The only changes I made after updating bios was to turn PBO on and xmp to 3200, nothing else. I know AMD and mobo vendors are making changes to

            all core clocks when running a heavy load like cinebench still appear to be the same, same as single core clocks under load. Since this is the case when under load ineither scenario, going to leave it on bios 3.6 since now RAM can run at 3200 instead of 2800.

          • funandjam

            Still on bios 3.60, haven’t rollled back yet. out of curiosity, I was able to just turn up xmp from RAM rated at 3200 to 3333, any higher and get memory errors, so yet another pleasant surprise.

  6. BigTed

    I’d be really interested to see memory speed scaling data and the effect on frame times. I’m ready to retire my 4790K which I use mainly for gaming and Permiere.

    I know the 3900X is going to demolish a 9900K at multi-threaded workloads, but will I be making much of a sacrifice if I go for a 3900X for 1440p gaming at 144hz with a 1080ti?

    • Litzner@gmail.com

      From what I have seen at 1440p, you would be giving up very little with either the 3700X or the 3900X vs the 9900K. 1080p, yes you lose a little performance, around 5%, but at 1440p there is very little difference and it could be considered in the margin of error most of the time. Also, the 3900X should last longer then the 9900K just simply due to the extra threads.

      • Godrilla

        Yep especially at 3440x1440p and 4k. It will take years before we have a fast enough Gpu to make a bottleneck of any of these cpus at 1440p and higher resolution. By then maybe you cans swap out that 3700x to a 4700x. On intel side you have the performance advantage now but there is no upgrade path with AMD you have an upgrade path and if you are playing at enthusiast resolutions it will be gpu bound for years to come and you have a upgrade path.

  7. razor512

    One thing that I have yet to see a review of these CPUs do, is bring up overclocking along with overclocked competition. for example. I plan to overclock any CPU that I get, but i have no answer to the question of which will be better, a r7 3800x (virtually no overclocking headroom), or the the 9700k that will overclock well.
    Would spending $45 less on the 9700k and overclocking it allow it to mostly match the 3800x in mulththreaded tasks while furthering its lead when it comes to gaming?

    • Sebastian Peak

      An interesting question.

Leave a reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Latest Podcasts

Archive & Timeline

Previous 12 months
Explore: All The Years!