AMD Ryzen 9 3900X and Ryzen 7 3700X Review: Kings of Productivity
It’s finally time to review AMD’s new 3rd-gen Ryzen processors. On hand today we have the Ryzen 9 3900X and Ryzen 7 3700X, with more content to come in the next few days. AMD decided to release and lift the review embargo on both Zen 2 and Radeon Navi at the same time, but with much testing to be done we’re going to delay our Navi coverage to focus on these Ryzen processors. It’s interesting AMD didn’t choose to dominate the news cycle for a few weeks, so for now we have a very thorough CPU review and tomorrow we’ll have our full Radeon RX 5700 review live.
The Ryzen 9 3900X is a 12-core, 24-thread processor with a massive 64MB L3 cache. It runs at a base frequency of 3.8 GHz with a boost frequency of 4.6 GHz. It costs $500, placing it in direct competition with the Core i9-9900K. Then the Ryzen 7 3700X costs $330 and AMD suggests it’s taking on the more expensive 9700K. It’s an 8-core, 16-thread CPU with a 32MB L3 cache and clocks ranging from 3.6 GHz to 4.4 GHz.
Both 3rd-gen Ryzen CPUs are bundled with the Wraith Prism RGB cooler and we’ll be using this for the majority of our testing. This means Intel will have a slight performance advantage, but keep in mind that the cost of the cooler will be factored into our value analysis for each processor. We’ll also provide some 3900X and 3700X performance figures using an all-in-one liquid cooler.
The MSI X570 Creation motherboard was used for testing the 3900X and 3700X, while the Asus ROG Crosshair VII Hero was used to test the 1st and 2nd-gen Ryzen chips. All have been tested with DDR4-3200 CL14 memory, but there will also be some memory scaling benchmarks included in the review as AMD recommends DDR4-3600 CL16 memory for best results… we’ll look into that.
The 8th and 9th-gen Intel Core processors were benchmarked on the Gigabyte Z390 Aorus Ultra, using the same DDR4-3200 CL14 memory, but they were cooled using the Corsair Hydro H115i RGB Platinum 280mm liquid cooler. Do note the Intel CPUs are not TDP restricted as that’s not the out of the box experience, so we are showing the absolute best case scenario for out of the box performance. Finally, our graphics card of choice was the MSI Trio GeForce RTX 2080 Ti.
There’s no better place to start than Cinebench R20’s multi-threaded benchmark, and boy does the Ryzen 9 3900X looks mighty. Scoring an incredible 7086 points makes it 24% faster than the Threadripper 2920X. Moreover, it decimated the Core i9-9900K by a 45% margin.
The Ryzen 7 3700X was equally impressive. The new 8-core part matched the 9900K and that meant it was 22% faster than the 2700X and 30% faster than the more expensive 9700K. These new 3rd-gen Ryzen parts are already looking like kings of productivity.
The R9 3900X matched the single core performance of the 9900K and was a full 19% faster than the 2700X. The 3700X also did well hitting 500 points and that placed it roughly on par with the 9700K, so this is a massive performance improvement for Ryzen.
Here we have some memory bandwidth figures and this is sustained read/write performance. As we said earlier all CPUs here are using DDR4-3200 CL14 memory, but interestingly 3rd-gen Ryzen is a little down on 1st and 2nd-gen Ryzen, but there is a good reason for this.
AMD’s made a compromise as client workloads do very little writing, so rather than use this space to improve something that isn’t required, they’ve invested the silicon real estate in more beneficial ways to achieve tangible performance gains. Whereas the Core Complex Die to IO Die link for reading memory is 32 bytes wide, it’s only 16 bytes wide for writing, and this significantly reduces write performance which impacts the SiSoftware copy test.
The changes AMD made with the Zen 2 architecture have had a profound impact on WinRAR performance, for example. The 3700X is a staggering 84% faster than the 2700X, allowing it to comfortably beat the 8700K and almost match the Core i9-7900X. The 3900X also put on a show, though it was only 15% faster than the 3700X and couldn’t quite match the 9900K. Overall a remarkable performance uplift over the 2nd-gen parts.
Moving on to 7-zip, first we have the compression test where traditionally Ryzen hasn’t done that well. For example, you can see how the 2700X is 14% slower than the 9900K. However, now we see the 3700X brushing aside the 9900K by a 15% margin, basically matching the 10-core 7900X. Then we see the 3900X beating the 2920X by an 11% margin and the 9900K by a whopping 45% margin. King of productivity right there.
Then when it comes to decompression work, 3rd-gen Ryzen still enjoys a handy performance advantage. Here the 3700X was 11% faster than the 9900K and a whopping 52% faster than the 9700K. Meanwhile the 3900X was a little over 60% faster than the 9900K and even offered a 17% performance uplift over the Threadripper 2920X.
We use Adobe Premiere on a daily basis, currently with a TR 2950X. Here we see the 3900X is 8% faster than the 2920X, so the 3950X will beat the 2950X when it’s released in a few months’ time, we can’t even imagine what the 3rd-gen Threadripper series will deliver.
The R9 3900X was 22% faster than the 9900K and even the 3700X edged out the 9900K, making it 25% faster than the 9700K.
Making these results more impressive is that Premiere is a very pro-Intel piece of software. AMD’s shown the 3900X to be a little over 50% faster than the 9900K in DaVinci Resolve, and yes, we know we need to get a DaVinci project sorted and start adding it to our benchmarks, promise we’ll get that done soon.
Next up we have V-ray 1.0.8 (older version), but we’ve also tested the latest version and we’ll check those results in a moment. Using the older build the 3900X took 48 seconds to complete the workload while the 3700X took 68 seconds. This meant the 3700X was slightly slower than the 9900K but much faster than the 9700K. Meanwhile the 3900X beat everything including the 2920X.
We see similar margins in the newer version, though here the Core i9-7900X manages to edge ahead of the 2920X. Despite that the 3900X easily conquered all while the 3700X wasn’t much slower than the 9900K.
Similar margins are also seen when testing with Corona, the 3900X was 30% faster than the 9900K, while the 3700X was only 10% slower. The 3700X was also 12% faster than the 2700X, a decent uplift.
The last application we’re going to look at is Blender. Once again we see similar margins between the tested processors, though this time the 3700X is much closer to the 9900K than it is to the 2700X. While running the Blender Open Data benchmark we also measured entire system power draw, so let’s check those results before getting into the gaming benchmarks.
Well, would you look at that. The R7 3700X consumed less power than the previous-gen 2600 and even less than the Core i7-8700K. In fact it was comparable to the Ryzen 5 2600 and 1600 along with the old quad-core 7700K.
Just as remarkable is the 3900X which is comparable to the 2700X and 7900X, making it worlds more efficient than the 9900K and 2920X.
Put in other words, the R9 3900X was 41% faster than the 9900K in Blender and yet it reduced total system consumption by 8%. This is an stellar result for AMD and on that high note lets cautiously move into the gaming benchmarks.
Testing with Assassin’s Creed: Odyssey, 3rd-gen Ryzen is ~10% faster than the 2700X which is good, but not good enough to beat the 9900K, at least when looking at the average frame rate. Despite similar frame time performance, the 9900K was 4% faster on average at 1080p with an RTX 2080 Ti. It’s really close, so let’s see what happens at 1440p.
Oddly we don’t see a coming together of the results at 1440p, but rather the 9900K is now 6% faster than the 3900X and even more bizarre is that the 2920X is now matching the 3900X, though it wasn’t far behind at 1080p. The 9700K was also 6% faster than the 3700X. Not a huge margin but we were hoping we’d see less of a gap at 1440p.
Battlefield V sees 3rd-gen Ryzen parts trailing Intel’s Core i9s again. Both Ryzen CPUs were 11% faster than the 2700X which is a decent gain, but both were down on the 9900K’s average and 1% low performance. The 9900K was 8% faster on average and 13% faster when looking at the 1% low result.
At 1440p the 9900K was just 4% faster than the 3900X which is a close call.
Here’s our biggest deficit yet, this time the 9900K was 17% faster than the 3900X when comparing the average frame rate and 20% faster for the 1% low result. The 3700X did better relative to the 9700K, but again gaming looks solid on Intel’s side.
Bumping the resolution up to 1440p reduces the margins significantly. Now the 9900K was only 8% faster for the average frame rate and this is certainly a more realistic resolution for the RTX 2080 Ti.
We were hoping these 3rd-gen Ryzen processors would be a little more punchy in The Division 2. Of course, gameplay was silky smooth, but comparatively the 9900K was 11% faster than the 3900X though with similar 1% low performance.
As we move to 1440p we’re entirely GPU bound with these higher-end CPUs and this showed with the 9900K, 9700K, 3700X and 3900X all delivering identical performance.
Far Cry New Dawn has always been troublesome for Ryzen processors. Here we see the 2700X allowing for just 98 fps on average and that made the 9900K 26% faster. The new 3rd-gen Ryzen parts don’t completely solve this issue but they do push things in the right direction. The 9900K is now just 10% faster than both the 3900X and 3700X.
Then at 1440p the margin is reduced further and now it’s just the 1% low results that noticeably favor the Intel 8-core processors.
The deficit seen in Far Cry New Dawn was a little disappointing and the same can be said when discussing the World War Z results. Actually this is by far the most disappointing result we’ve seen, and that’s because the 3rd-gen Ryzen processors are no faster than the 2700X. This means the 9900K was 18% faster than the 3900X and te 9700K was 21% faster than the 3700X.
Even at the more GPU limited 1440p, the 9900K was able to beat the 3900X by a 9% margin.
Moving on we have some Rage 2 results and here the 3900X and 3700X perform very well, offering slightly better 1% low performance than the 2700X which puts them on par with the 8700K. The 9900K and 9700K do offer slightly better 1% low performance, around 7% better. Not a big difference though and as you might expect the gaming experience was identical.
Then at 1440p we’re entirely GPU bound, so from the 3700X up we were able to get the most out of the RTX 2080 Ti.
Hitman 2 is another one of those titles that’s never been that friendly with AMD processors and we see that it’s still an issue for the 3rd-gen parts. Here the 9900K beat the 3900X by a 15% margin, though that’s nothing like the 27% margin it beats the 2700X by, so there’s that.
The margins at 1440p are reduced and now the 9900K is just 5% faster than the 3900X. The 3700X was still 9% faster than the 2700X, so we see recurrently how 3rd-gen Ryzen series is a good step forward when it comes to gaming.
Total War: Three Kingdoms delivers a solid result for AMD. The 3700X was able to improve on the 1% low performance of the 2700X by an 8% margin, making it just 1% slower than the 9900K and 5% slower than the 9700K.
Then at 1440p we’re once again almost entirely GPU bound, leaving just a frame or two at the most in it.
DDR4-3200 vs. DDR4-3600
DDR4-3600 is the fastest spec memory AMD recommends using with 3rd-gen Ryzen as higher clocked memory will actually reduce performance, at least when clocked higher than 3733 as this changes the Infinity Fabric to a 2:1 mode rather than 1:1. Basically 2:1 sees the Infinity Fabric clocked at a quarter of the memory speed, while 1:1 is half.
Since AMD recommends DDR4-3600 for optimal performance and provided us with a CL16 kit, we tested it just to make sure we’re not hampering performance by using the CL14 3200 kit.
The good news is we’re not, here you’ll see virtually identical performance in Corona, WinRAR, Far Cry New Dawn, Assassin’s Creed Odyssey and World War Z from either RAM kit.
Due to time constraints we didn’t go further in testing memory. This is something we’ll follow up on at a later time but for now we can see key performance metrics are not greatly altered by the RAM choice or by not maximizing Ryzen’s capabilities on this regard.
X570 vs. B450 Chipset
We also took a quick look at chipset performance, comparing B450 to X570, using the MSI B450 Gaming Pro Carbon AC and MSI X570 Creation. We didn’t include any X470 testing due to time constraints but considering we know B450 and X470 boards deliver the same performance you can fill in the blanks pretty easily.
Looking at the Cinebench scores there appears to be a marginal difference between the two platforms, but we’re talking a 1% difference within the margin or error.
Enabling PBO + the AutoOC feature in Ryzen Master we see similar performance on the X570 and B450 boards again. As claimed by AMD the PBO feature of the 3rd-gen Ryzen processors can be enabled on all motherboards that support these new processors.
A quick look at gaming performance with World War Z we see no major performance difference on both boards. At a later time we’ll check out how this looks for B350 and X370 motherboards as well.
Time was of the essence to get you all this testing on time, but it seems pretty clear that like the 1st and 2nd-gen Ryzen CPUs, there’s virtually no overclocking headroom with 3rd-gen Ryzen. This isn’t surprising as both AMD and Intel are locked into a fierce battle which sees both extract the most amount of performance they can from these high-end CPUs.
At best we were able to boost the 3900X multi-core score by 4% while the 3700X saw a 6% increase. However, as was the case with 2nd-gen Ryzen, the simple multiplier OC method that we used for the 4.3 GHz all core, is not the method you want to use.
Here we see when looking at the single core results, it does reduce performance, leaving PBO overclocking as the most effective method overall. Still we feel most won’t even bother overclocking just to extract 4-6% more performance and from what we’re hearing from fellow publications, 4.3 GHz seems to be one of the better overclocks.
On an even more disappointing note, we somehow managed to end the life of our 3900X sample at this stage of the review. We don’t recall exactly what settings were applied, but we know we hadn’t manually adjusted voltages yet. We believe after testing the 4.3 GHz overclock with auto voltage, we increased the LLC to see what impact that had on temperatures and during our first CB20 pass the system crashed and reset, and never booted up again.
The CPU now gets stuck at code 07 after microcode. We tried running the chip on a different X570 board among other tests before sadly declaring it dead. AMD says no other reviews had managed to kill their 3900X so we’re just special or unlucky, your pick. AMD has sent us a replacement but in the meantime we were unable to test the 3900X on B450 boards, or include it in the IPC test with a few cores disabled in each chiplet.
This also means we don’t have temperature data of that CPU for now, just the 3700X. With the Wraith Prism RGB cooler you can get close to extracting maximum performance out of the 3700X, though it does get a little toasty. The cooler is very quiet though, so 87 C overclocked is respectable in our book.
Using a more elaborate and expensive setup like the Corsair Hydro Series H115i the R7 3700X peaked at just 73 degrees after an hour long Blender stress test. Shame we don’t have the 3900X results, but we suspect with the box cooler it runs at around 80 C stock under heavy load. Of course, all the 3900X results were recorded with the box cooler as we noted earlier.
Cost Per Frame
What does 3rd-gen Ryzen offer in terms of value? Looking at present asking prices you are paying quite a premium for these new CPUs when compared to 2nd-gen parts, but that’s hardly surprising.
When compared to the Intel competition, purely for gaming, the 3700X is on par with the 9700K and 8700K in terms of value. The 3900X doesn’t stack up quite as well and here the 9900K is actually quite a bit better in terms of value… or is it? While the 3900X comes with a box cooler, non of Intel K parts do, so let’s add in the cost of a good value high-end air cooler.
When we surveyed a group of readers, most seemed to agree that the be quiet! Dark Rock 4 for $75 was a reasonable pairing for the 9900K/9700K, so we added that to the budget for the Intel K-series processors. Factoring in the cooler means that the 3900X is now better value than the 9900K and the 3700X is now very similar to the Core i5-9600K.
Motherboard costs are about the same, so that won’t play a factor in this value comparison.
As for AMD’s first and second-gen Ryzen processors… they’re really cheap these days. The new R7 3700X costs just 14% more per frame than the 2700X and given the massive power savings and improved single and lightly threaded workload performance, we think it’s well worth the premium. It’ll be really interesting to see how the Ryzen 5 3600 stacks up against the R5 2600 and 1600 — our full review is coming in a few days.
Who Is It For?
Over 50 graphs deep into this review, we feel like there’s a lot more we want to cover, but surely the essentials are all here and now you have a good idea of how the Ryzen 9 3900X and Ryzen 7 3700X compare to current generation Intel CPUs. Also at our disposal is a range of nice looking X570 motherboards and PCIe-gen 4 SSDs that we’ve been checking out over the past week, but we’ll have to go back to review those later. At the same time, if you’re mostly concerned about getting the most out of these new Ryzen processors, neither is required but they do complement the platform at the high-end.
For the most part we feel X570 boards — particularly the more premium models — should only be paired with the 3900X or the upcoming 3950X. For those grabbing the 3700X or the more affordable the Ryzen 5 3600, we’d recommend existing X470 or B450 boards. We’ve heard the concerns regarding X570 board’s pricing, but know you can fully unleash Zen 2 processors on B450 and probably even B350 boards, which we’ll set about testing shortly.
PCIe-gen 4 SSDs are nice but for gaming and general usage they’re not worth the premium. There’s a good chance you won’t notice the difference over existing PCIe-gen 3 models, so keep that in mind. Then for those who justifiably need the extra bandwidth, you’re probably better off waiting for Threadripper as it’ll offer many more PCIe lanes and likely be more useful in that regard.
Going back to our gaming benchmark results, we suspect these will provoke the most debate. Annoyingly AMD lead us to believe the 3900X and 9900K would be neck and neck, trade blows even, but that doesn’t appear to be the case. The end result is good and 3rd-gen Ryzen has generally shown a big step forward for AMD. Moreover under realistic gaming conditions there’s almost no chance you’d be able to tell the difference between the 9900K, 3900X or 3700X, as the difference at 1440p with an RTX 2080 Ti was remarkably small.
The 3900X was 8% slower than the 9900K on average at 1080p, so AMD’s halved the deficit to Intel in gaming. Then as we’ve found before with Ryzen, for almost anything else the 3900X buries the 9900K, while the 3700X delivers comparable performance. When it comes to power consumption the new Ryzen processors are extremely efficient.
If you’re exclusively concerned about gaming performance we’d recommend the 3700X and probably the 6-core 3600 models, which we’ll be reviewing soon. But if you want the absolute fastest gaming CPU then that’s still either Intel’s 9700K or 9900K, even though value for money they aren’t great and there’s no upgrade path.
For those of you who also use their PC for work, content creation, or essentially any productivity task that demands a fast CPU, in those scenarios these new Ryzen CPUs are in a league of their own, making the 9900K a bit of a one-trick pony. Overall the new Ryzen 3000 series delivers and the improvements in power efficiency are nothing short of amazing. Pricing is competitive and if this doesn’t force Intel to adjust theirs, nothing will.
Remember tomorrow we’ll be back to deliver our in-depth Radeon RX 5700 coverage.