Core Scaling and Gaming Performance — How Many Cores Do You Need?
Selecting the best processor for your new gaming PC build is a tough choice but can be critical to getting the optimal performance and gaming experiences for years to come. Recently, the discussion around gaming CPUs has expanded, and I wanted to share some interesting data that shapes how Intel builds products and why I believe our technology provides the best gaming experience on the planet.
At Intel, we believe that the best processor for any task is the one that balances all vectors of performance including core count, frequency, IPC, and more for optimal performance. Our engineers and product planners are always optimizing designs based on what’s possible. Sometimes we expand cores to do more for creators, other times we optimize within a current core count envelope; we constantly look for new ways to add value to users’ experiences.
One of my repeated themes during the Intel Real World Performance Events (of which I wrote about the IFA iteration on this blog) is that it “isn’t an accident” that we excel in these areas of performance. Gaming requires a different type of performance than CPU-based ray traced rendering, just as an example, and optimizing your product stack for the workloads that matter and are utilized in each segment just makes sense.
For gaming, 8-cores is the optimal spot for performance scaling in modern PC gaming, and at 8-cores we get the best frequency scaling out of our Coffee Lake products and 14nm process technology today. And clock speed is what feeds the hungry primary threads of game engines today!
To back up this point with data, I wanted to showcase a couple of sets of benchmarks. First up is some testing on the Core i9–9900K, configured in 4-core, 6-core, and 8-core mode, with a fixed frequency of 4.0 GHz and HyperThreading disabled. The fixed frequency and HT disabled state helps to remove differences that might result from varied Turbo frequencies at different core residencies.
As you can see from this data, core scaling does exist in modern PC gaming. Moving from a 4-core to a 6-core configuration results in a modest 7% gain in Shadow of War but up to 33% in Assassin’s Creed Origins.
Interestingly, of the six games tested in this pre-existing data set, only three scale from 6C to 8C. We measured a total gain of 50% in Assassin’s Creed Origins and over 30% for both Ashes of the Singularity: Escalation and Hitman 2. The other three titles, including Far Cry 5, Shadow of War, and Rise of the Tomb Raider, do not scale with the move to 8-cores.
To demonstrate game performance scaling beyond 8-cores, I turned to some testing that our evaluation labs had handy based on the Core i9–9980XE 18-core CPU. Again, the clock speed is fixed at 4.0 GHz and HT was disabled.
This data set happens to be a bit larger (as the testing was done for a different project) but it is a superset of the previous data from the 9900K, so it provides for an interesting extrapolation.
The first thing I see with this set is that going from 8-cores to 12-cores results in less than 10% scaling in all games, and in that 12-game set, 9 of them improve by 5% or less.
What about that 12-core to 16-core jump? Only one game in our results today sees an improvement (Ashes of the Singularity once again) while the other 11 games not only do not scale up in performance, THREE of them actually regress slightly, offering lower performance than a 12-core configuration.
The data backs up what Intel has been saying since its release: The Core i9–9900K is the world’s best gaming processor and the new Core i9–9900KS will be even faster. Adding core count just because you can, without a corresponding increase in sustained frequency and architectural design decisions necessary to feed these cores (like low latency memory systems), doesn’t result in better performance. The software engines that power games across the PC ecosystem scale best with frequency and IPC, and Intel plans to lead in this space for years to come.
Legal Notice
Performance results are based on testing as of October 15, 2019 and may not reflect all publicly available security updates. No product or component can be absolutely secure.
- 9900K Test Configuration — CPU: Core i9–9900K, MB: Asus Prime Z390-A, Bios: 1005, Storage: 2TB Intel 660p SSD, Graphics: Nvidia GeForce RTX 2080 Ti Founders, Driver: 26.21.14.3602, Memory: G.Skill Ripjaws V 16GB DDR4–2666 MHz, OS: Windows 10 Pro 64-bit RS6
- 9980XE Test Configuration — CPU: Core i9–9980XE, MB: Asus Prime X299 Deluxe, Bios: 1902, Storage: 2TB Intel 660p SSD, Graphics: Nvidia GeForce RTX 2080 Ti Founders, Driver: 26.21.14.3602, Memory: G.Skill Trident Z RGB 64GB DDR4–2666 MHz, OS: Windows 10 Pro 64-bit RS6
- All games tested with integrated benchmarks and shown resolution and settings. Exceptions: PUBG and CS:GO tested with 5 minute gameplay scenarios.
For more complete information about performance and benchmark results, visit www.intel.com/benchmarks.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.
