Many will see today as an historic shift in computing power. With its new Ryzen 5000 desktop CPUs, AMD has finally knocked Intel to the floor, and is raising its boxing gloves in victory as the flash bulbs pop and the ref declares a winner. The headline takeaway: Ryzen 9 5900X and Ryzen 9 5950X have trounced Intel’s Core i9.
The Ryzen 5000 chips using AMD’s Zen 3 cores are arguably the best consumer desktop CPUs the world has ever seen. Ryzen 5000 simply gives no quarter to its Intel arch-nemesis in any single task. Heavy lifting multi-core work? AMD. Light single-threaded tasks? AMD. Gaming? AMD. PCIe 4.0? AMD. No need to throw away your existing motherboard? AMD.
There’s a lot to unpack with these new CPUs, but we understand if you want to skip to “the good stuff.” So read on if you want to hear about AMD’s path to glory over the past few years, or feel free to jump ahead to these upcoming sections:
You can see just how disruptive Ryzen has been by looking at the table below, which summarizes the last five years of consumer desktop computing history. At the top you can see the calm, blue waters of Intel, whose only competition was itself, testing the limits of just how much it could charge a consumer for a CPU (Hint: it was $1,723).
The first red entry, AMD’s original Ryzen processor, would upend it all by offering an 8-core CPU for less than half of what Intel had been charging the year before.
From there it’s just been one duck and weave after another. As Intel’s chip architecture languished at 14 nanometers, AMD has literally danced around its rival’s aging body in the ring. AMD goosed its process, slightly improved its performance, and then cut prices even further with its Ryzen 7 2700X.
With Ryzen 9 3950X, AMD shocked the world with a consumer CPU packing 16 cores on a 7nm process at a reasonable price. While we don’t show it—the Intel equivalent—the Core i9-9960X was slower and cost $1,600.
You can see that with every iteration of Zen and Ryzen, AMD has backed Intel into a corner. The last few reasons to consider Intel desktop chips were narrow, but they were real: single-threaded performance and gaming performance.
With Ryzen 5000 and Zen 3, AMD lands the final knock-out punch, where you can see the sweat and spittle flying off Core’s face in slow motion.
Perhaps what’s crazier about the victories of the previous Ryzen chips is AMD did it without changing much. Sure, it helped to have TSMC’s advanced 7nm process to shrink the size of the previous Ryzen 3000 series, but the actual core design didn’t change much.
With Ryzen 5000, AMD says it initiated a near-ground-up redesign of the core, with three goals: performance, latency, and power efficiency.
For the cores themselves, AMD improved integer performance, floating point performance, and branch prediction. Latency, however, received one of the biggest upgrades, with a redesign of the core complex (CCX), the basic building block of the chip.
Unlike Intel which fashions its CPUs from a monolithic piece of silicon, AMD assembles Ryzen from multiple “chiplets.” This gives AMD far more flexibility in yields and construction. Zen 3, for example, uses the same IO die that contains the PCIe 4.0 lanes and memory controller as Zen 2. This let AMD get its newest cores into existing sockets and motherboards without disturbing everything underneath it.
With previous Ryzen chips, each CCX was constructed of four x86 cores, which were connected to a second CCX to make a core chiplet die (CCD). This design lead to a latency cost when the cores had to communicate outside of the 4-core CCX. AMD said it has measured this latency in a range from 78ns to 95ns. With the unified 8-core CCX, the latency is essentially eliminated.
Another benefit of the new unified 8-core CCX is more available L3 cache. The previous design gave each 4-core CCX 16MB of L3 cache, but it couldn’t be shared. By combining all of the cores into one CCX, all eight now share 32MB of space.
This savings in latency and increase in cache size is what AMD credits for much of its huge gains in gaming. In fact, AMD engineers said the impact of the large L3 on today’s games was “eye-opening.”
The new Zen 3 cores also feature more aggressive clock gating that can put a core to sleep when not in use. The clock gating helps increase boost clocks of the Ryzen 5000 and also saves power. AMD said a sleeping core uses no power and generates no heat, which contributes to Ryzen 5000’s performance and power efficiency.
The chip is so efficient, AMD said, it has recorded a 2.8X improvement over an Intel Core i9-10900K, as well as massive performance gains over the previous 7nm Ryzen 3000 CPUs, without drawing any additional power. While AMD doesn’t say, we’d guess this would also contribute a lot to a laptop version of the chip.
The latency improvements and the power savings all promote the final goal: Performance. AMD said its goal was an astounding 19 percent improvement in instructions per clock (IPC) which it looks like it has largely achieved.
Ryzen 5000 Performance: How We Tested
Mature System Builders strongly cautioned. Some benchmark results may seem inappropriately lopsided for nerds under 13 (pre-teen ages) who have never witnessed a benchmark battle this decisive.
For this review, AMD sampled its top two chips: The 12-core Ryzen 9 5900X and the 16-core Ryzen 9 5950X. For comparisons, we used Intel’s newest 10th-gen Core i9-10900K, as well AMD’s mid-summer Ryzen 9 3900XT.
The AMD CPUs were tested in an MSI MEG X570 Godlike motherboard, while the Intel CPU was installed in an Asus Maximus XII Extreme board.
All four CPUs were tested with the newly-released Windows 10 2H20 version, as well as matching GeForce RTX 2080 Ti GPUs with the latest public driver. Both machines were outfitted with identical 32GB of Corsair DDR4/3600 RAM in dual-channel mode, with their XMP profiles loaded.
For cooling, we used NZXT X62 280mm liquid coolers, with fans and pumps manually set to their maximum speed. Both were built in cases with the case doors off, and small exterior desktop fans blowing air over the motherboard and graphics card.
Storage on the AMD systems used a Corsair MP600 PCIe 4.0 SSD, while the Intel system used a Samsung 960 Pro SSD. In a way, this gives the AMD CPUs an advantage due to PCIe 4.0. On the other hand, AMD is only vendor to offer PCIe 4.0 on a desktop chip today, so it feels unfair to handcuff it to PCI 3.0.
To ensure the results are fair, the vast majority of our tests were retested. For a few tests where the OS, graphics driver are unlikely to move the needle that much, we reused results and noted them as such.
Ryzen 5000 Rendering Performance
We’ll kick this off where we usually do, with Maxon’s Cinebench R20. This is a 3D modelling test based on the same engine Maxon uses in its Cinema4D application. Besides being a professional application in use every day, it’s also the same 3D modelling engine that’s embedded in several Adobe Creative Cloud apps, such as After Effects and Premiere. Like most 3D modelling apps, it favors more threads and scales with the increase in CPU cores very efficiently.
For this test, we mix in results from previous reviews. As Cinebench R20 is nearly a pure CPU test, the results don’t move much even with OS, driver or BIOS updates. In the chart, the red represents the two new Zen 3-based Ryzen 5000 chips, the green represents Zen 2-based Ryzen 3000 chips, and the blue represents various Intel chips.
As a CPU test, Cinebench R20 performance increases with more CPU cores, but holding higher clock speeds matters, as does the actual efficiency of the cores.
On top is the new Ryzen 9 5950X with 16 cores, which offers an impressive 14-percent increase in performance over the 16-core Ryzen 9 3950X chip. While 14 percent is an incredible bump, AMD said much of the limitation on all-core loads is due to the CPU socket power limit. The small-socket AM4 can’t push the same power like a big-socket Threadripper TRX40 socket.
One step down, the 12-core Ryzen 9 5900X is less power-restrained, and we see it open up an even more impressive 17-percent gap over the 12-core Ryzen 3900XT. Technically, the XT was the best of the best of Zen 2, so when you compare the Ryzen 9 5900X against the Ryzen 9 3900X it’s actually a 20.7-percent increase in performance in an all-core load.
But no one really cares about red-on-red performance—nerds only want to see red-on-blue. There, it’s the total knockout you expected. With Intel stuck in the mud at 14nm, the best it can field in a smaller consumer socket is 10 cores in the Core i9-10900K. The rightfully “more expensive” Ryzen 9 5950X justifies itself with 63 percent more performance. More in its price range, the Ryzen 9 5900X outruns the 10-core Core i9 by 36 percent.
What’s jaw-dropping, though, is that even Intel’s $980 18-core Cascade Lake-X Core i9-10980XE gets crushed by the 16-core Ryzen 9 5950X, by 19 percent. Perhaps even more embarrassing, it’s really only about 1 percent faster than the $550 Ryzen 9 5900X. And yes, the 12-core Ryzen 9 5900X is within 5 percent of the performance of a 16-core Ryzen 9 3950X.
One thing to keep in mind is Cinebench R20 uses AVX as well as AVX512. On Intel’s older 14nm chips, using AVX or AVX512 increases the thermal load, so Intel intentionally lowers the clock speed to manage the heat. Not all rendering engines use these advanced instructions. The Chaos Group’s Corona renderer (at least not in the current benchmark version), for example, doesn’t use AVX or AVX512, so you can see the performance of the Intel chips when they don’t have to shift down.
The 18-core Core i9-10980XE opens up a lead over the 16-core Ryzen 9 3950X and the 12-core Ryzen 9 5900X by about 10 percent. That doesn’t matter much though, because the 16-core Ryzen 9 5950X is here to remind Intel who is in charge by outpacing the $900 Core i9 by 14 percent despite have two fewer CPU cores. And yes, the 10-core Core i9-10900K easily gets crushed by the 12-core Ryzen 9 5900X, by 36 percent.
Our next multi-core test uses V-Ray Next to measure CPU performance using an Academy Award-winning rendering engine. The results again should surprise no one: The 16-core Ryzen 9 5950X comfortably sits in front of the 18-core Intel CPU, and the 12-core Ryzen 9 5900X comes impressively close to the older 16-core Ryzen 9 3950X. The Ryzen 9 5900X again destroys the 10-core Intel chip by an astounding 42 percent.
Our last multi-core modeling test is the Persistence of Vision Raytracer, which dates all the way back to the Amiga. It’s been updated since then by a global team of volunteers. Like most 3D modelling programs, it scales up in performance with core count.
While we include historical scores in some other tests, we only include the most current results with POV Ray. We did this not because we think POV Ray requires special treatment, but because the results are very consistent. The 18-core Core i9-10980XE doesn’t come out on top, nor does the 10-core Core i9-10900K. If someone found a multi-core rendering test where Intel came out on top over Ryzen, it would be actual news.
It’s not unusual to find Ryzen hammering Core in multi-core benchmarks, but ever since the Zen 2-based Ryzen 3000 line, Intel has clung by its fingernails to its lead in single-threaded performance. With the Zen 3-based Ryzen 5000, Core i9’s slim lead finally loses its tenuous grip.
You can see that below: The same Core i9-10900K that’s 6.6 percent faster than the Ryzen 9 3900XT is 14.5 percent slower than Ryzen 9 5900X, and 16 percent slower than Ryzen 9 5950XT.
You can see this again in Cinebench R20, where the 10th-gen Core i9-10900K is basically within the margin of error with the Ryzen 9 3900XT, and about 7 percent faster than the Ryzen 9 3900X. The Ryzen 9 5900X, however, is an insane 18.4 percent faster than the Core i9. The slightly faster Ryzen 9 5950X delivers a solid 20.7-percent performance advantage over Core i9-10900K.
We’ll see how this manifests itself in gaming, where single-threaded performance still matters, but we can’t overstate this enough: This is simply a stunning turn of events. AMD wipes out the only remaining justification to buy an Intel desktop CPU today.
Ryzen 5000 Content Creation Performance
Content creation is an activity where multi-core CPUs can shine. First up, we’ll look at an encoding task using the latest version of HandBrake (version 1.3.3) to transcode the open-source Tears of Steel UHD 4K video to H.265/HEVC using the 1080p/30-fps preset. HandBrake is an excellent encoder and does favor more cores and higher clock speeds, as well as any special instructions the CPUs offer.
Core counts above 16 tend to yield diminishing returns, but down here with “just” 16 cores and below, core counts matter. The 12-core Ryzen 9 5900X decreases encode times over the 12-core Ryzen 9 3900XT by 20.7 percent. That’s impressive as hell, and pretty close to AMD’s promise of 20-percent improvement over its previous design.
What about Intel? The new Ryzen 9 5900XT encodes take about 29 percent less time, with only about 20 percent more threads. The 16-core Ryzen 9 5950X decreases the encode time by 39 percent, which again is a very decent return based on having 60 percent more threads.
While HandBrake is a free and popular encoder, we also wanted to look at tasks powerful CPUs get used for: video editing and photo editing. For that we use the latest version of Adobe Creative Cloud Premiere Pro and run the CPUs through workstation builder Puget System’s benchmark tests. It breaks the test into various GPU-heavy and CPU-heavy tasks using MultiCam modes, as well as various popular professional codecs. When it’s done, the benchmark produces a score. For our tests, we ran the standard test run.
First, we’ll point out the result between the 12-core Ryzen 9 3900XT and the 10-core Core i9-10900K chip. Despite its lower core count, the Core i9 actually edges out the Ryzen 9 3900XT by about 4 percent. Based on that, Intel has the advantage–barely.
The 12-core Ryzen 9 5900X, however, opens up a real gap to the tune of 13 percent. Ryzen 9 5950X is slightly faster at 14.3 percent.
Both scores are solid wins, and a clear advantage for Ryzen 5000, but it’s not the yield you’d see with 3D modelling tasks. Still, if the cost is similar or close, we’d take the Ryzen 5000, and that’s all that matters to AMD.
Ryzen 5000 Compression Performance
Next we move on to compression tests using the popular 7-Zip app. The first result is 7-Zip 19.00’s performance compression test. The compression side is sensitive to memory latency, cache performance and out of order performance. The result is no surprise.
Decompression performance is sensitive to integer performance, and branch prediction performance. With this test less reliable on memory latency and memory bandwidth per core, we again see that more cores matter. The 16-core Ryzen 9 5950X opens up a huge lead over the 12-core Zen 3 chip.
We’ll close off our 7-Zip section with single-threaded results, which favor AMD’s chips again. We don’t need to show you the decompress single-threaded performance, because it’s the same, with AMD running over Intel.
So far nothing—nothing—has put the Intel Core i9 in front of the Zen 3 chips. Nevertheless, we were surprised when the Intel chip lost WinRAR’s built-in benchmark using all of the cores. The test itself doesn’t use every single thread, but it has long run horribly on AMD CPUs–the 10-core Core i9 has a 15 percent advantage over the 12-core Ryzen 9 3900XT.
Our pet theory is that this particular test runs poorly on CPUs with increased latency. We say that because we saw Intel’s performance suffer when it went from the ringbus design of the original 10-core Core i7-6950X to mesh-based Skylake-X. Zen 3’s greatly reduced latency should therefore help, but the results didn’t map perfectly to our theory. The 16-core Ryzen 9 5950X is actually underperforming the 12-core Ryzen 9 5900X by a large amount despite the additional cores. Both chips are also constructed with dual-chip CCDs, so the jury remains out. We should also note that while WinRAR supports multi-core, it doesn’t scale to the 24 or 32 threads of the Ryzens—or even the 20 threads of the Core i9.
There’s more to explore here obviously, but Ryzen still wins, in a test we’ve never, ever seen it win at when put against Intel.
WinRAR single-threaded performance helps remove those pesky questions of latency across cores or CCDs, but the result is the same. While the Core i9 has a 12-percent lead over the Ryzen 9 3900XT, the Ryzen 9 5900X comes in an astounding 71 percent faster than the Intel chip. The higher-clocked Ryzen 9 5950X is slightly faster as well.
Ryzen 5000 Gaming Performance
So you already knew that AMD rules the day in multi-core performance, and you now know it gives Intel no quarter in single-core either. The one area that’s made potential customers uneasy is gaming performance. Even though Intel years ago ceded multi-core performance, it’s still been able to say truthfully: “Core i9 is better for gaming.” There’s obviously a lot of nuance to that statement, but we’ve generally agreed that when paired with the fastest GPU and a high-refresh rate panel—Core i9 is indeed the better gaming CPU.
Until today. Although the wins aren’t as disruptive everywhere else, we do think there is a clear case for AMD’s claims that the Zen 3-based Ryzen chips are the “best gaming CPUs.”
For our gaming section we tested at 1920×1080 resolution, with both PC’s outfitted with Nvidia Founders Edition GeForce RTX 2080 Ti cards using the same drivers. Unlike the previous tests, where we sorted the results by winner, we kept all of our gaming charts ordered by: Ryzen 9 3900XT, Ryzen 9 5900X, Ryzen 9 5900X, and Core i9-10900K. We exclude previous results, because graphics drivers and game updates make it unfair to compare.
First we’ll kick it off with Deus Ex: Mankind Divided set to High. If you look at the chart and pretend not to see the newest AMD CPUs in red, you’ll see the familiar lead that Intel has long had in gaming. With this game set to High rather than Ultra, the test becomes less GPU-bound, and you see that typical 20-percent performance advantage that Intel CPUs have long enjoyed over AMD CPUs.
Now there’s Ryzen 5000, where we see both Zen 3 chips essentially within the margin of error. While we’re really “only” talking about a 2- to 3-percent performance difference between the Ryzen 9 and Core i9, it’s far better than the usual 10- to 20-percent performance hole Ryzen has fallen into before in non-GPU-limited tasks.
The results in Deus Ex aren’t just a fluke. We also saw both Ryzen 9 CPUs edge the Core i9 in Far Cry New Dawn. We’re again talking maybe 4- to 5-percent advantage, but when you look at the 14-percent hole the Ryzen 9 3900XT is in, it’s a clear win for AMD.
Not all games favor Intel, either. Our result in Red Dead Redemption puzzled us so much, we had to rerun it on the Intel machine and the Ryzen 3900XT to make sure we didn’t goof it up. Yes, the mighty Core i9 loses to both the older Zen 2 and the new Zen 3 chips—all of which we suspect to be GPU-limited. This, however, is the outlier. Typically in games where it isn’t limited by the by the GPU, Core i9 has been faster, so it’s a surprise to see Ryzen 3000 jump ahead here. Ryzen 5000, however, is right where we expected it to be: in front of Core i9.
You can again see that in the popular but graphically mellow Counter Strike: Global Operations, where the Core i9 has a 9-percent advantage over the older Ryzen 9 3900XT. The Zen 3 Ryzen 9’s edge out the Core i9 by about 2 to 3 percent.
We’ve combined most of our game results into a single chart that’s color-coded with green for Zen 2, red for the two Zen 3 CPUs, and blue for the Core i9. It’s not a total win in every category, but generally, the new Ryzen 5000 chips have a slight advantage over the Intel Core i9.
A few notes about the results: In Civ VI Gathering Storm, we used the AI benchmark to determine which CPU is faster making decisions, and a lower score is better. In Gears of War 5, the game refused to run on our Intel Core i9 even after uninstalling the game and reinstalling it. You can still see that the new Ryzen 9s have a very decent performance improvement over the older Ryzen 9.
And yes, there are many tests here that we suspect would run faster on the Ryzen 9 if we only had a faster GPU. That in itself is insane, because our card is the $1,200 GeForce RTX 2080 Ti, but there are definitely faster graphics cards out today—we just didn’t have access to any for our testing.
If it doesn’t seem like a big deal to be basically slightly faster to about the same as the Core i9, you have to remember that AMD hasn’t beaten Intel in gaming ever. It’s a huge accomplishment for the company. Like single-threaded performance victories in productivity applications, Ryzen 9 basically does every thing Core i9 can do—except it can do it better.
Before we reach our conclusion, we do want to look at how well the new Ryzens scale across their cores, because software doesn’t always run on a single core or all cores. For that we use Cinebench R15 and run it from one thread to the maximum threads available on each CPU. This gives us an idea of how a CPU might perform on a light load to a medium load, as well as a load that uses all of the cores.
To help put Zen 3 in perspective though, here’s one of the first times we did that, with the 8-core Ryzen 7 2700X in 2018 based on AMD’s Zen+ core.
In its day, the 2700X was just as revolutionary for its price-to-performance ratio—especially as you moved into high core count workloads. Its primary adversary was Intel’s 6-core Core i7-8700K. The Ryzen 7 2700X demolished the Core i7-8700K on the right side of the chart, where you used all of those CPU cores. On the left side of the chart, indicating performance on fewer cores, it was largely underwater against its the Intel chip.
Wind the clock forward to July 2019, and you see below the same showdown between the 12-core Ryzen 9 3900X and its contemporary—Intel’s 8-core Core i9-9900K. You can immediately see that AMD made huge strides with the Zen 2 core. The Ryzen 9 3900X is basically dead-even to slightly faster than the Core i9-9900K, and it dominates as you move into the heavier all-core loads.
That brings us to today, with where we take the 12-core Ryzen 9 5900Xand perform the same test with its contemporary—the 10-core Core i9-10900K. It’s that left side of the chart that we really want to point out. The Zen 3-based Ryzen 9 5900X posts an 18-percent advantage in single-threaded performance. From there it only goes up, showing generally 20 to 30 percent more performance from the Ryzen 9 5900X compared to the Core i9. It doesn’t matter if it’s a light load or heavy load–Ryzen simply crushes the Core i9.
AMD fans may be a little disappointed the Ryzen 9 5900X “only” has a 41-percent advantage when every core is stressed out, but the Ryzen 9 5900X isn’t the top-end Ryzen 9. We did the same test using the 16-core Ryzen 9 5950X, and you can see an even higher 22-percent advantage on a single-thread, extending out to an obscene 71-percent lead when all threads and cores are used.
You may say it’s not fair to put an $800 Ryzen 9 5950X against a $488 Core i9-10900K. Normally we’d agree, but for much of the year, the Core i9-10900K was in such short supply that it regularly sold for $650 and higher. In fact, as of October 30, the street price of the Core i9-10900K was well above its expected retail price at most stores.
Pricing has been one complaint since AMD unveiled its Zen 3 chips. Indeed, the stupidly good value that AMD has offered previously has gotten less stupid. To present that visually, we take the expected retail price of current CPUs and calculate how much the company charges per thread.
This is an overly simplistic look at CPU pricing, but it does give you an easy way to visualize how they compare. Overall, the four new Ryzen 5000 chips have increased in price over previous models, likely because AMD knows it can finally charge premium pricing for the chips.
It doesn’t help that Intel’s CPUs generally haven’t been a great value, and they tend to sell well above their list prices on the street. On the chart below, you can see a few spot-checks on current AMD chips and Intel chips. A green highlight indicates they’re selling at or below the list price, while red is selling above its list price.
So yes, there’s been a definite price hike. But with the performance we’re seeing, we think it’s wholly justified, especially when you consider that Intel CPUs continue to be the premium-priced chip despite not offering the premium performance.
It’s hard to be believe, but it’s been only three years since the original Zen-based Ryzen CPUs were introduced. The original Ryzen reset our expectations of how many cores you could get in a consumer CPU, and it put the world on notice that AMD was back.
With the Ryzen 5000 we’re simply floored by its performance. It’s the best CPU for heavy multi-core loads. It’s the best CPU for single-core loads. It’s the best CPU for gaming. Add to it support for PCIe 4.0, compatibility with many existing AM4 motherboards, and actually reasonable prices, and you get what is undoubtedly the best CPU we’ve ever seen.