Cranking up 16 cores to new heights of performance (and watts)

For as long as I can remember, I've had love of all things tech, spurred on, in part, by a love of gaming. I began working on computers owned by immediate family members and relatives when I was around 10 years old. I've always sought to learn as much as possible about anything PC, leading to a well-rounded grasp on all things tech today. In my role at PCMag, I greatly enjoy the opportunity to share what I know.

The 16-core Ryzen 9 7950X delivers exceptional performance that blows previous AMD consumer chips out of the water and even outpaces Intel's top competing "Alder Lake" efforts.

If you follow the field of PC processors, and the recent history of AMD and its bigger competitor Intel, you know that most of the time, semiconductor change is gradual. On rare occasions, though, we see performance improve by leaps and bounds. For AMD and its new Ryzen 7000 series of processors, this is one of those times.

No question about it: The outgoing Ryzen 5000 series of processors, headed by the Ryzen 9 5950X, was a very capable, competitive set of options if you were buying or building a PC that could leverage lots of cores. But Ryzen 7000 takes that much further, with significantly better performance in both single-threaded and multi-threaded workloads.

We’re in the process of testing and reviewing the initial line of Ryzen 7000 chips released so far, and we decided to start at the top with AMD’s flagship 16-core entry. At this writing, the $699 Ryzen 9 7950X is the fastest CPU that you can buy without jumping into server-grade hardware or AMD's Threadripper line. Of course, Intel’s 13th Gen Core (“Raptor Lake”) line looms soon; we’ll have to see what the 2022 Core i9 flagship from Team Blue can do with its Performance and Efficient cores. In the meantime, though, AMD will enjoy a strong claim to the CPU-performance crown that’s hard to argue against. The Raptors have their work cut out for them.

The Ryzen 9 7950X is based on AMD’s new architecture, Zen 4. Like with the previous Zen 3, AMD is continuing to use a chiplet-based design in this new architecture, which gives it a similar high-level look to all of the prior Zen processor architectures. Still, some changes in the core architecture and a move to a new 5nm manufacturing process have enabled AMD to boost performance considerably. In general, the trend for Zen 4 is to be like Zen 3, but bigger and more robust. Zen 4 has a beefier front end than the outgoing Zen 3 architecture that includes more powerful branch-prediction hardware that is able to make two branch-per-cycle predictions.

This improvement helps to reduce pipeline stalls and leads to better overall throughput on the CPU cores. AMD also increased the size of its L1 B2B cache by 50% compared to Zen 3. The size of the L2 B2B cache was also increased by an undisclosed amount, and the Op cache was expanded by 68% compared to Zen 3, too. The Op cache’s throughput was also increased to allow for an additional three macro ops to be able to be processed through it per cycle.

On the load/store end of the chip, we also see larger caches, with the load queue being 22% larger and the L2 DTLB cache increased by 50%. The core’s execution engines have 25% larger instruction retire queues, as well, with larger register files for integer and floating-point operations. The total L2 cache size was increased in capacity from 512K to 1,024K per core, and in general, the buffers throughout the chip were increased to further help keep the cores fed with data.

A few new instructions were added to Zen 4, as well, most important among them being AVX-512 related. These aren’t going to be useful to everyone, but if you are doing heavy compute-intensive workloads, this could benefit you greatly if AVX-512 is supported by the software you are using.

Cumulatively, AMD claims that these architectural changes enable roughly 13% better IPC throughput on Zen 4 processors. This, added to sizable increases in clock speed, are estimated by AMD to yield an average 29% performance increase in single-threaded operations. All Zen 4 Ryzen 7000 CPU cores are crafted on a 5nm TSMC process, but not all components on the processor are made with the same process.

AMD has been adding integrated graphics processors (IGPs) to a select subset of its desktop processors for more than a decade now; with the Ryzen chips, those models were traditionally daubed with a "G" on the end, like the Ryzen 7 5700G. But starting with the Ryzen 7000 series, this will become a standard feature. The graphics portion of Ryzen 7000-series processors is derived from AMD’s RDNA 2 graphics technology. These chips are built on a 6nm process from TSMC and melded into the chiplet processor alongside the CPU cores.

For most Ryzen 7000-series processors, the graphics processor integrated into them will consist of just two compute units with 64 shaders each, adding up to a total of 128 stream processors. The IGP also has eight TMUs and four ROPs, and it was clocked at 2,200MHz on both our Ryzen 7 7700X and Ryzen 9 7950X samples, which suggests that it may be clocked at this speed on the other Ryzen 7000-series processors so far, too. Our test motherboard, Asrock’s X670E Taichi, wouldn’t allow us to set any more than 512MB of RAM for the IGP.

Given its specs, it’s clear that this IGP is really meant to be used only as a basic display adapter. It’ll work without issue for video playback, office work, and web browsing, but it is not meant for gaming. 

Later on, it’s likely we could see AMD offer Ryzen 7000 processors without IGPs, and some with much stronger IGPs. Maybe the "G" will make a comeback. But so far, all the ones we’ve seen have had only this utilitarian graphics chip.

We’ve run through the key points of AMD’s new AM5 platform in earlier articles on the first flagship motherboards and the chips and platform proper, but let’s quickly summarize the supporting hardware behind Ryzen 7000 here.

AM5 is AMD’s first change to a new socket since its first-gen Zen processors were released. Momentous here is that AMD has abandoned pins on the bottom of the CPUs now, opting for an LGA design like Intel's Core chips have used for ages, relegating the pins to the AM5 motherboard socket. You'll also note a distinct, industrial-looking change to the thermal interface module (TIM) design in the photos here.

In addition to a new socket, AM5 also supports DDR5 memory, as well as PCI Express 5.0 for ultra-high-speed NVMe SSDs and next-gen graphics cards. AM5 and PCIe 5.0 aren’t really being utilized to their fullest yet, but this is one rare occasion when we can testify that the platform looks future-proof, at least for a while.

Four chipsets have been announced for AM5 so far: X670E, X670, B650E, and B650. (The “E” is for “Extreme.”) The E models are supposed to ship with enhanced overclocking features and there are a few other minor changes between the E and non-E models (having to do with PCIe 5.0 support), but nothing else major.

Note, also, that AMD claims all AM4 CPU coolers are compatible with the AM5 socket without the need of any alternative mounting hardware. If you invested in a high-end cooler on an earlier AM4 system, that's good news, because you'll need it here with this Ryzen 9. (We'll talk more about cooling the Ryzen 9 in a bit.)

The shining star of the new Ryzen 7000 series is the AMD Ryzen 9 7950X. It ships with 16 CPU cores, with SMT support that enables each core to handle two processing threads, for a total of 32 simultaneous threads. 

The processor has a high base clock of 4.7GHz and can turbo-boost as high as 5.7GHz depending on the workload. Each core has 1MB of L2 cache for a total of 16MB of L2 across the chip, and there is a shared 64MB pool of L3 cache that all cores can access.

The chip has a relatively high thermal design power (TDP) rating of 170 watts, which soars past the 105 watts of its predecessor, the Ryzen 9 5950X. The chip supposedly can handle a maximum of 230 watts from the socket, though we would be hesitant to try pushing the chip this hard as we aren’t in the game of extreme overclocking. The chip also has a max supported temperature of 95 degrees C, which we found it already colliding with during our tests at stock speeds.

No major change was made to how the core regulates its clock speeds under load, with Ryzen 7000 continuing to use AMD’s Precision Boost 2 algorithm.

AMD set a maximum officially supported DDR5 RAM speed of 5,200MHz, though higher is possible using AMD’s EXPO tech or by manually overclocking. (Hit the preceding link for a primer on AMD EXPO, which allows for easy setting of memory parameters.) In total, the processor’s CCD measures 70mm square with 6.5 billion transistors. The CCD sits atop an I/O die that measures 122mm square and consists of 3.4 billion transistors that are used to interconnect the various components.

We tested the AMD Ryzen 9 7950X on our new AM5 test bed. It consists of an Asrock X670E Taichi(Opens in a new window) motherboard, a Cooler Master MasterLiquid PL240 Flux(Opens in a new window) 240mm water cooler, a G.Skill 2x16GB DDR5 RAM kit, and a SilverStone DA850(Opens in a new window) power supply. The whole works is installed in a Praxis Wetbench open-air testbench case.

An Nvidia GeForce RTX 3080 Founders Edition card was used for all game benchmarks. It was also used as the display adapter for all tests except those testing the IGP specifically.

A key thing to note about our test setup is that we set the RAM on each test bed to the maximum RAM speed listed by the CPU manufacturer. For Intel’s 12th Generation (“Alder Lake”) CPUs, this was 4,800MHz, and for AMD’s Ryzen 7000 series processors, this is 5,200MHz. Technically, this gives AMD a slight advantage, but as the memory controller is an integral part of the CPU and improvements in the memory controller are notable areas of improvement in each CPU generation, we felt it made sense to test this way.

Before we dig into our test results, we need to talk briefly about overclocking with the Ryzen 9 7950X. This processor is rated with a maximum safe operating temperature of 95 degrees C. Even with our 240mm water cooler and the system installed in an open-air test bed, our processor still ended up hitting 95 degrees C during some of our tests. This is likely why AMD does not ship a stock cooler with the Ryzen 9 7950X; if a 240mm water cooler is barely up to the job of keeping this beast below 90 degrees, a stock air cooler certainly wouldn't be.

Anyway, we opted not to experiment with overclocking at present, as the CPU was already pressing up against its maximum safe operating limit, which made us feel uneasy about pushing the power and clocks up even more. We're somewhat more curious about undervolting this processor to possibly achieve better performance, possibly by preventing thermal throttling. As that's somewhat more tricky to do, as performance can also be lowered when overclocking, we opted not to do so for this article. If you plan to overclock one of these chips, we'd recommend at least a 360mm water cooler.

Kicking things off with AIDA64’s cache and memory tests, we can gain some insight into the bandwidth available in each CPU. In spite of the Ryzen 9 7950X using DDR5 RAM clocked at 5,200MHz and the Alder Lake Core i9-12900K using the same DDR5 RAM clocked at 4,800MHz, the Core i9-12900K still showed notably better bandwidth between the CPU and RAM. This clearly indicates an inherent advantage in either Intel’s memory controller or the way in which the RAM was connected to the CPU via the motherboard.

The L1 cache bandwidth results are a little mixed, but the test clearly showed that the Ryzen 9 7950X had significantly better L2 cache and L3 cache bandwidth. By itself, these results don’t mean much, but they can help us to understand the results produced by other tests.

Next, on to our pure CPU tests. In Cinebench R23, we see the Core i9-12900K suffer its first crushing defeat, with the Ryzen 9 7950X laying down the law in the multi-threaded workload. It didn’t quite run circles around it when only one core was used, but, let’s say, it strolled past it. Meanwhile, the previous-gen Ryzen 5000-series CPUs are left far behind.

The results we got from our two Adobe content-creation tests were, like AIDA64, a bit more mixed, with the Ryzen 9 7950X and the Core i9-12900K essentially coming to a draw in Photoshop. The situation changes in Premiere Pro, though, with the Ryzen 9 7950X holding an enormous lead over the competition.

Next was our Handbrake 1.5.1 test, converting a 4K video sample down to 1080p. Once again, the Core i9-12900K and the Ryzen 9 7950X performed relatively similarly, but with the 7950X holding a small lead. In Blender, the Ryzen 9 7950X was considerably faster once again, though, and the same was true in the multi-threaded version of our POV-Ray test. Intel’s Core i9-12900K did manage a solitaire win in the CPU tests in the POV-Ray Single-Threaded test results, but it only achieved this with a small performance lead.

In spite of its commanding CPU test scores, gaming remains a bit of a sore spot for AMD and its Ryzen 9 7950X. That’s not to say it did badly, but with the CPU test scores as guidance, we would have expected it to perform somewhat better in some of these tests. Instead, we noted a lot of ties here, as well as some losses for the red team’s new top-ender.

The 3DMark benchmark, in particular, ran better on the Core i9-12900K than on the Ryzen 9 7950X, and the same is true for our tests with Bioshock: Infinite. Only in these two tests did we really see more than 1 or 2 frames per second (fps) of difference, but it is notable that the Core i9-12900K was almost always 1fps or 2fps faster than the Ryzen 9 7950X in the rest of the tests. Many of these results are statistically insignificant and within the error of margin, but as this repeatedly happened, we have to acknowledge that these games did run at least ever-so-slightly better on the Core i9-12900K.

As we mentioned earlier, the graphics processor inside of the Ryzen 9 7950X is not intended for gaming. But, we are gamers on the side…that’s what we do. (Well, that and make components run in circles.) But the point is if we can game on it, we’re going to try. So we did, and no surprise: It was weak stuff.

There’s no real need to dig into these results too closely, but just know if you plan to use the IGP in the Ryzen 9 7950X for anything more than basic office work or web browsing, you’ll want to play games that have been around long enough to graduate from high school. Something like the original Bioshock will likely be more playable, but Bioshock: Infinite was not.

To be fair, we did push these games fairly hard, and Bioshock: Infinite was run with maxed-out settings even at 720p. But as that game is pushing 10 years old and was perfectly playable with those settings on the Core i9-12900K’s Iris Xe IGP, we don’t feel this was too extreme. Maybe better RDNA 2-based IGPs will emerge with some future Ryzen 7000-family chips, but this isn't the IGP you're looking for if you want to game.

The Ryzen 9 7950X delivers impressive performance, but it’s also a bit of an energy hog. We tested the power draw of the entire system using a Kill-A-Watt meter and found the Ryzen 9 7950X pulling as much as 565 watts during our Adobe Premiere test. It also consumed more power than any other AMD CPU tested during the Cinebench test.

Intel’s competing CPUs aren’t much better, with the Core i9-12900K pulling significantly more power in Cinebench and only a bit less in Adobe Premiere. Software-based hardware monitors used to gauge the power draw of the CPU itself listed the Ryzen 9 7950X at 192 watts, with the Core i9-12900K notably higher at 262 watts.

The Core i9-12900K also ran exceptionally hot during some tests, hitting as much as 100 degrees C. The Ryzen 9 7950X topped out at 95 degrees C, which isn’t much better. We suspect that these max temp readings occurred due to safety limits in the processors that prevented them climbing higher, which means without the safeties in place it’s hard to say which of these processors would have actually ran the hottest.

The Ryzen 9 7950X is unquestionably an extremely powerful processor. The only question is, if you’re a power user in the market for serious multithreaded muscle, should it be your next processor? If you’re already running a system with an Intel Core i9-12900K, a Core i7-12700K, or something similar, then we’d have to say no. At $699, its price is a bit steep if you’ve already paid for a like-priced CPU a year or two ago. It's faster in many respects, but you won’t get enough of an upgrade to justify it. But if you’re using anything older, be it an Intel 11th Gen (“Rocket Lake”) CPU or one of AMD’s earlier Ryzen processors, the Ryzen 9 7950X is an excellent option for you. Of course, you'll have to factor in the cost of a new AM5 motherboard.

Of course, in a larger sense, you will need to ask yourself if you really need something this fast. And if you don’t, I’d look closer at the Core i7-12700K or possibly AMD’s new Ryzen 7 7700X, for which we should have a review posted shortly after this one. But after running our tests, one thing's clear: Zen 4 is bringing the heat, in every way, and the Ryzen 9 7950X could be a worthwhile upgrade even for someone on a Ryzen 9 5950X system if they need all the multithreaded muscle they can get.

The 16-core Ryzen 9 7950X delivers exceptional performance that blows previous AMD consumer chips out of the water and even outpaces Intel's top competing "Alder Lake" efforts.

Sign up for Lab Report to get the latest reviews and top product advice delivered right to your inbox.

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.

Your subscription has been confirmed. Keep an eye on your inbox!

For as long as I can remember, I've had love of all things tech, spurred on, in part, by a love of gaming. I began working on computers owned by immediate family members and relatives when I was around 10 years old. I've always sought to learn as much as possible about anything PC, leading to a well-rounded grasp on all things tech today. In my role at PCMag, I greatly enjoy the opportunity to share what I know.

I wrote for the well-known tech site Tom's Hardware for three years before I joined PCMag in 2018. In that time, I've reviewed desktops, PC cases, and motherboards as a freelancer, while also producing deals content for the site and its sibling ExtremeTech. Now, as a full-time PCMag staffer, I'm focusing on reviewing processors and graphics cards while dabbling in all other things PC-related.

Read Michael Justin Allen's full bio

PCMag.com is a leading authority on technology, delivering lab-based, independent reviews of the latest products and services. Our expert industry analysis and practical solutions help you make better buying decisions and get more from technology.

© 1996-2022 Ziff Davis. PCMag Digital Group

PCMag, PCMag.com and PC Magazine are among the federally registered trademarks of Ziff Davis and may not be used by third parties without explicit permission. The display of third-party trademarks and trade names on this site does not necessarily indicate any affiliation or the endorsement of PCMag. If you click an affiliate link and buy a product or service, we may be paid a fee by that merchant.