[Hardware]CPU Benchmark and Hierarchy 2021: Intel and AMD Processors Ranked

[Revo]

 

 

Our CPU benchmarks performance hierarchy ranks all the current and previous generation processors, including all of the Best CPUs for Gaming, based on performance. Your CPU has a huge effect on overall performance and, to many, is a computer's most important component. CPU benchmarks help us suss out the differences, but when it comes time to buying a CPU for your desktop, you'll find a dizzying collection of model numbers and specs from both Intel and AMD. We've listed our the best CPUs for gaming and best processors for work in other articles, but if you want to know how each chip stacks up against all the others and how we come to our decisions, this CPU benchmark hierarchy is for you. We also recently added a new article covering the CPU benchmark hierarchy for Cyberpunk 2077, which is a dedicated piece to show performance scaling in the most hotly-anticipated game in recent history.           

We recently added the Ryzen 5 5600X, Ryzen 9 5950X, 5900X, and Ryzen 7 5800X to our rankings, along with all of the previous-gen Zen 3, Zen+, and Zen 1 versions of those chips. These new processors represent a watershed moment for AMD: The company has swept our benchmark hierarchy and taken the lead in every metric, including our gaming, single-threaded, and multi-threaded categories.

We are busily retesting our full arsenal of processors. We will re-rank them with our latest tests and hardware, but for now, we have split the latest results for each category (gaming, single-threaded, and multi-threaded) into two different listings. One is listed as "Post-Zen 3," and the other is listed as "Pre-Zen 3." 

We're using different tests and a newer version of Windows 10 to measure performance in the new rankings, so the two lists aren't directly comparable. This applies doubly-so for the gaming hierarchy, as we are now using an Nvidia GeForce RTX 3090 for gaming tests, which is significantly faster than the old 2080 Ti we used previously. For now, we've only tested 21 processors in the Post-Zen 3 era, but you can use the old table of results to get a rough sense of how the newly-tested processors compare to the un-retested models.     

We'll transition to a single table for each category once we've reached a tipping point in the number of processors tested. We're working on that endeavor full time.    

We'll explain how we ranked the processors under each table. The game testing ranking is first. We also include an application performance metric in our application score tables, which we've split up into single- and multi-core measurements (below gaming table). Make sure to check the Post-Zen 3 rankings for all three categories. 

The most powerful chip gets a 100, and all others are scored relative to it. If you want our recommendations for specific price bands, please check out our Best CPUs for gaming page. 
We previously only listed 1080p game rankings, but we've now added 1440p testing into the mix. As such, we have two rankings for each chip, and the chart is aligned sequentially based on the 1080p game results. However, the 1440p listings aren't listed in sequential order due to unfortunate limitations with our tables. Pay attention to the 1440p rankings: Some faster chips at 1440p may be listed below slower chips simply because of the 1080p results. 

We measured performance for the 1080p games from a geometric mean of Borderlands 3, Far Cry 5, Hitman 2, Project CARS 3, Red Dead Redemption 2, Shadow of the Tomb Raider, and The Division 2. 

We measured performance for the 1440p games from a geometric mean of Borderlands 3, Far Cry 5, Flight Simulator 2020, Project CARS 3, Red Dead Redemption 2, Shadow of the Tomb Raider, and The Division 2.  

As you can see, AMD's Ryzen 9 5950X and Ryzen 9 5900X take a healthy lead over Intel's flagship Core i9-10900K in both 1080p and 1440p results, and as you'll see in our Ryzen 9 5950X and 5900X review, overclocking can't even the score for Intel. 

The stock Ryzen processors beat the overclocked Intel processors, which is quite the feat, and Ryzen is even faster after overclocking, too. Also, check out those performance deltas between the previous-gen Ryzen processors compared to the 5000 series. That's impressive.

Most folks overlook the incredible power efficiency of the Zen 3 processors, but that equates to a faster, cooler, and quieter system that doesn't require super-expensive cooling solutions. Take note of the TDP divide in our charts - it's surprising. Check out our review for more in-depth power testing. 
*indicates an APU tested with a discrete GPU. Note: These types of processors are geared for performance with integrated graphics - please see individual reviews for those performance rankings.

This is our Pre-Zen 3 rankings, which you can use to gauge relative rankings compared to the table above this one. We've ranked all the consumer Intel 10th, 9th, 8th, and 7th Gen processors, along with AMD's Ryzen and Threadripper chips from all four generations. We determined the order in the first table based on a geometric mean of gaming performance in four titles: Grand Theft Auto V, Hitman 2, Final Fantasy XV, and World of Tanks. 

We used the Nvidia GTX 2080 Ti GPU for our tests at HD resolution. We have also included several newer processors that we haven't tested yet, but we've assigned them basic positions in our hierarchy based on their capabilities. This list will soon be absorbed into the Post-Zen 3 rankings as we test more chips.

You may be surprised to find that some less-expensive chips ranked above their upscale siblings. For example, the Core i7-9700K comes out ahead of the Core i9-9900K on this list because it offers better performance in some games due to its lack of hyperthreading, which was enough to help it do better on the overall suite of tests. However, depending on the application, you may see stronger results from the higher-end chip.
We calculate the above single-threaded rankings based on a geometric mean of y-cruncher, Cinebench, POV-Ray, and LAME. The latter consists of two tests: One short duration test and one extended duration test to measure performance once Intel's boost duration limits have been exceeded. 

Single-threaded performance is often tied directly to the responsiveness and snappiness of your PC in any number of daily applications, like loading an operating system or surfing the web. This metric largely depends upon a mixture of instruction per cycle (IPC) throughput (the number of operations the chip can execute in one clock cycle) and frequency, which is the speed at which the transistors switch between on and off states. However, a whole host of other considerations, such as cache, architecture, and interconnects (like rings, meshes, and infinity fabric) impact this measure of per-core performance, so these results do not align perfectly based upon clock frequency. Instead, performance varies with each application and how well it is tuned for the respective architectures.  

With all that said, the delta between Intel's flagship Core i9-10900K and the Ryzen 5000 processors is incredible - at worst, Ryzen 5000 is 10% faster in single-threaded performance.
This is our Pre-Zen 3 single-threaded ranking, which you can use to gauge relative rankings compared to the newer table above this one.

We generated the single-threaded metric via a geometric mean of performance in single-threaded LAME, Cinebench, POV-Ray, and y-cruncher applications, giving you a good idea of performance in everyday desktop PC applications. 
The multi-threaded workload column is based on performance in Cinebench, POV-ray, vray, Blender (four tests - Koro, Barcellona, Classroom, bmw27), y-cruncher, and Handbrake x264 and x265 workloads that represent performance in productivity-focused applications that tend to require more compute horsepower.

Like we see with single-threaded performance metrics, multi-threaded performance, which is a measure of a chip's performance in applications that utilize multiple software threads, varies based upon a whole host of architectural factors. It also depends heavily upon how well the software scales with additional compute cores. As such, these results do not align perfectly based upon core/thread count, though it does serve as a decent litmus of multi-threaded performance. Be aware that architectures, caches, and interconnects profoundly impact these results, as all of these factors impact how well performance scales with additional threads. Performance rarely scales perfectly with the addition of more cores/threads, so the scaling factor of each processor architecture weighs in heavily on the value proposition of going with a higher core count processor for your specific application.  

It's noteworthy that both Ryzen 5000 chips beat out Intel's 18-core 32-thread Core i9-10980XE, and the Ryzen 9 5900X only comes with 12 cores. Again, pay attention to TDP - the Ryzen chips have a 105W ranking while Intel is specced at 165W. 
This is our Pre-Zen 3 single-threaded ranking, which you can use to gauge relative rankings compared to the newer table above this one. The multi-threaded workload column is based on performance in Cinebench, POV-ray, vray, Blender, y-cruncher, and Handbrake x264 and x265 workloads represent performance in productivity-focused applications that tend to require more compute horsepower.

The Threadripper rankings need a bit of explanation. With an industry-leading slathering of core counts, these processors offer the utmost performance in many different threaded workloads, but can suffer in some less-demanding desktop PC-class applications that either don't utilize the cores and threads fully, or aren't optimized for the architecture. That reduces the gap separating the rankings in the mainstream applications we use for this list, but in many cases, you can see much larger deltas with specific applications. You should view individual reviews and our best processors for applications list for a better sense of how those processors fare in professional workloads.  
Legacy Desktop Processor Benchmark Hierarchy
Recognizing that a lot of older platforms are going to be paired with graphics subsystems multiple generations old, we wanted to define the top of our range to encourage balance between host processing and complementary GPUs. At this point, anyone with a Sandy Bridge-based Core i7 would realize a gain from stepping up to Coffee Lake or Kaby Lake, for example. And putting AMD's top FX CPUs next to a handful of Core i7s and those older Core i5s represents an upgrade to their status.

Currently, our hierarchy consists of 13 total tiers. The bottom half of the chart is largely outdated; you'll notice those CPUs dragging down performance in the latest games, regardless of the graphics card installed in your PC. If you own a CPU in that range, an upgrade could really take your experience to another level.

Really, it's the top five tiers or so that remain viable. And in that top half of the chart, an upgrade is typically worthwhile if it's a least a couple of tiers higher. Otherwise, there's just not enough improvement to warrant the expense of a fresh CPU, motherboard and RAM (not to mention the graphics card and storage solution you'd be considering as well). 

Legacy Desktop Processor Hierarchy