Table of contents
- Intel Core i9-11900K Review
- 11th-Gen Intel Core Desktop CPU Architecture
- What’s the deal with 14nm?
- Test Setup and Methodology – Core i9-11900K Review
- Thermal Behavior and Power Consumption- Core i9-11900K Review
- Overclocking the Core i9-11900K Review?
- CPU Test
- Gaming Performance (720p)
- Gaming Peformance: Esports
- Gaming Performance: Cont’d
- Value & Conclusion
Intel Core i9-11900K Review
Today marks the official release of the Intel’s Rocket Lake 11th-generation Core CPU but this was already preempted by many reviews out today of early retail units of the Core i7-11700K. We also have a review of the Core i7-11700KF which can read here. In this review, we’ll focus on the Core i9-11900K and I’ll say it outright, this isn’t the saving grace for Intel and the blue giant knows that.
That said, we’ll skip a lot of the technical details and we’ll dive straight to gaming performance. That’s the ultimate question that many Intel users such as myself would have. So let’s get straight to the point, in this review we’ll be taking a closer look at the gaming performance of the Intel Core i9-11900K and see how it stacks up head to head against the Intel Core i9-10900K, Intel Core i7-10700K and in this review we’ll also compare it with the AMD Ryzen 9 5950X and Ryzen 7 5800X.
11th-Gen Intel Core Desktop CPU Architecture
Its not easy being an Intel fan the past year and many have been clamoring for Intel’s promises for a new architecture and a refined process but we’re sort of getting that with the 11th-gen Intel CPU… sort of. Built on the 10nm Sunny Cove architecture, due to production difficulties, constraints, hurdles, etc. on their 10nm node, Intel has decided to backport and build Rocket Lake-S for desktop on 14nm. The Sunny Cove cores aren’t the only ones getting a backport as the integrated graphics for Rocket lake is also getting the 14nm treatment from 10nm on the Xe-LP IGP, as seen on Tiger Lake. That said, this marks another addition to Intel’s 14nm history perpetuating the meme of the 14nm+++++++++++++++^n. Intel’s new CEO Pat Gelsinger is confident that they can overcome this soon as they have the capacity and capability to do so.
What’s the deal with 14nm?
Intel’s tick-tock cycle has been on a standstill for a while and Skylake saw it becoming one of the most mature desktop CPU if the the only one that lasted as long as it did when it saw Kaby Lake, Coffee Lake, Coffee Lake Refresh and Comet Lake. Intel’s story regarding its 10nm process is known but details are not very public and but Intel has recently come out and stated that efforts to bring 10nm for socketed desktop CPUs has seen them on the negative end of the development process. As Ice Lake already existed, Intel decided to end Skylake’s prolonged stay with Rocket Lake. This birthed the decision to backport Ice Lake to 14nm together with Xe-LP.
The issue here is that with modern desktop CPUs being marketed with >4Ghz clocks, both AMD and Intel have put it upon themselves that they can make architectures that can operate with such clock speeds despite also being having lower power draw. AMD managed to get over that hurdles thanks in part to third-parties capable of producing at a smaller process with TSMC handling their manufacturing. As a company that generally produces things in-house, Intel’s choice to keep it within themselves to create their solution to 10nm has been largely a bust and has cost them a growing chunk of the desktop market, particularly the gaming segment.
Test Setup and Methodology – Core i9-11900K Review
|Intel Core i9-11900K||Intel Core i7-11700KF||Intel Core i9-10900K||Intel Core i7-10700K||AMD Ryzen 9 5950X||AMD Ryzen 7 5800X|
|uArch||Cypress Cove||Cypress Cove||Comet Lake||Comet lake||Zen3||Zen3|
|Base Frequency||3.5 Ghz||3.6 Ghz||3.7 Ghz||3.8 Ghz||3.4 Ghz||3.8 Ghz|
|Turbo Frequency||5.3 Ghz||5 Ghz||5.3 Ghz||5.1 Ghz||4.9 Ghz||4.7 Ghz|
|PCIe||Gen4 (20 lanes)||Gen4 (20 lanes)||Gen3 (16 lanes)||Gen3 (16 lanes)||Gen4 (24 lanes)||Gen4 (24 lanes)|
This review will focus primarily on gaming performance but to touch on CPU performance, we’ll have some reviews as well. The motherboard used for the 11th-gen CPUs is the ASUS ROG Maximus XIII Hero and our RAM of choice is the G.Skill Trident Z Royal DDR4-4000 C17 32GB (16GBx2) memory kit. This setup makes up our previously drafted AMD test bench for our GPU and we will use it here in hopes of achieving the same expectations from the current performance king amongst the mainstream CPUs which is the AMD Ryzen 9 5950X.
|Intel Core i9-11900K||Intel Core i7-11700KF||Intel Core i9-10900K||Intel Core i7-10700K||AMD Ryzen 9 5950X||AMD Ryzen 7 5800X|
|Motherboard||ASUS ROG MAXIMUS XIII HERO (BIOS 0610)||ASUS ROG MAXIMUS XII EXTREME||ASUS ROG CROSSHAIR VIII FORMULA|
|RAM||G.Skill Trident Z Royal DDR4-4000 C17 32GB (16GBx2) Dual-Channel Memory Kit|
|GPU||ZOTAC RTX 3080 Trinity OC|
|PSU||SeaSonic P1050 1050W + Powenetics Power Measurement System|
|SSD||Patriot Viper VP4100 1TB (CPU Test), GIGABYTE GP-AG42TB 2TB SSD (Gaming Test)|
|Cooling||NZXT X63 280mm AIO|
These tests has been done on the week of March 25th with final testing finishing on March 29th. That is just a few hours away from this launch and we hope that we’re also launching with Z590 reviews. That said, I’ve chosen to focus on the CPUs first as they’re integral to the overall scheme of things.
Back to our test bench, all the systems are tested in open test benches. Ambient temperature is maintain at 28*C and motherboards are kept on stock except for XMP. MCE is disabled. All tests are performed in the same time frame, no score is recycled from previous tests. No CPU is simulated as well.
- Arithmetic Test
- SuperPI 32M(Single-Threaded)
- wPrime (Multi-threaded)
- Rendering Test
- Blender 3D
- Cinebench R20
- Cinebench R23
- VRay 4
- Encoding Test
- x265 1080p
- x265 4K
- Memory Test
- AIDA64 Memory Benchmark
- Content Creation Test
- PugetBenchmark for Photoshop
- PugetBenchmark for Premiere Pro
- Productivity Test
- Microsoft Office Word
- Microsoft Office Excel
- Microsoft Office PowerPoint
- Synthetic Gaming Test
- 3DMark Time Spy
- 3DMark Fire Strike
- Counter-Strike Global Offensive (DirectX 9)
- DOTA 2 (DirectX 11)
- Rainbow Six Siege (DirectX 11)
- The Witcher 3: Wild Hunt (DirectX 11)
- Grand Theft Auto V (DirectX 11)
- Shadow of the Tomb Raider (DirectX 12)
- Call of Duty Warzone (DirectX 12)
- Apex Legends (DirectX 11)
- Valorant (DirectX 11)
- PlayerUnknown’s Battlegrounds (PUBG) (DirectX11)
- Cyberpunk 2077 (DirectX 12)
All games tested on 720p, 1080p, 1440p and 4K
Thermal Behavior and Power Consumption- Core i9-11900K Review
Intel has been criticized about its power consumption many times over and that remains the same for this generation. Despite reporting a 125W TDP, Intel and its motherboard partners can freely ignore the restrictions supposedly set by Intel as it provides a free “boost” of performance at the cost of some extra power. That said, turbo performance has been the primary benchmark of performance for many motherboards in the Z-class for Intel and these boards enforce their own prescription of what an Intel CPU can work with. We’ll touch more about how some motherboard try to get around this further in this generation in our Z590 motherboard reviews.
Going back to Rocket Lake, we’re introduced to AVX-512 with this release. AVX-512 has debuted in some form on some HEDT and the server space but Rocket Lake marks its MSDT debut. AVX in itself is a taxing CPU load so AVX-512 should push that workload further causing heat and power to spike.
Using Powenetics, we monitor the power and how much we’re drawing on specific loads. Our primary reading here is the load temps. Under synthethic gaming loads, the Core i9-11900K doesn’t really get loaded much with the later part of the benchmark. We see temps reaching around the lower 80s but with power draw in 140W range. Now I have to stress the part that this test is performed while enforcing Intel’s CPU limits so there’s no 300W boost happening here pushing our temps in crazy territory. That’s ultimately the realm of motherboards being allowed to do so and by default, Intel’s prescribed behavior is with these limits enforced but for performance to be more pleasing, removing these limits would bring about more power and higher temps.
Taking that into consideration, we will be doing those tests specifically for motherboards. This assures that Intel is also closer to their TDP rating without being aided by board partner tuned boosts. Going back the Core i9-11900K, the results look fair enough and gaming as well as stress loads, produce ok results but to add context to this chart, the CPU drops down to around 4Ghz when running AIDA64. The spike you see on the beginning of the AIDA64 curve is Turbo Velocity Boost in action, it gives our CPU a bump to >150W but eventually drops off either by tau time or temp limit.
These are the peak power draw we recorded during our stress testing sessions. Depending on your workload and configuration, these values may change. These results are only comparable to the setup similar to ours. The workload is AIDA64 stability test (only disk and GPU unchecked) and 3DMark Fire Strike.
Overclocking the Core i9-11900K Review?
Despite being supplied by Intel CPUs for these reviews, only ASUS has provided a form of documentation about these CPUs. I was not able to secure a copy of Intel’s reviewer’s guide for this CPU and had no primer regarding the architecture nor the expectations. ASUS did share a quick OC guide which details some of the new behaviors to be expected from this CPU. Regardless, that was a bit late into testing and initial trials on overclocking was challenging.
I am the kind of reviewer that when overclocking is involved, I would make sure I have adequate cooling. My NZXT 280mm AIO proved insufficient as enabling MCE alone and running ASUS’ AI 51/49 multiplier saw CPU temperatures already surging past 96*C causing our CPU clock to drop to 3Ghz on Cinebench R23. The goal here is to find a stable clock speed and given that I can’t tame temperatures on mine, I did manage to somehow get a stable 5.2Ghz all-core here that drops to 3Ghz after a few seconds without crashing, so I am assuming its just temps holding me back. While many of the things you know about overclocking Intel CPUs can help you with Rocket Lake, in the interest of keeping this review free from potential questionable OC scores, we’ll discuss overclocking further in another article.
Arithmetic, Rendering, Encoding
SuperPI is a long-standing benchmark and is popular in the overclocking scene. As a single-threaded application, it calculates the value of Pi to a very large of digits, specifically 32 million for this benchmark. Gives a good idea how fast in math a single thread of a CPU is. wPrime
wPrime is another math benchmark designed to find prime numbers. This multi-threaded benchmark is designed to use up multiple CPU threads to perform its testing operations.
Blender’s Open Data project allows users to publish their own benchmarks so that the Blender project can create a reference guide for users on various systems and generate expected performance. We use the Barcelona Pavilion benchmark running on the CPUs for this test.
Cinebench R20 and R23 are Maxon Cinema 4D‘s benchmarking component and is available freely for download for everyone. These benchmarks are popular CPU rendering benchmarks and creates a taxing workload as it renders a single 3D image. R23 introduces a stress test mode as well for stability and throttle testing.
POV-Ray is another long-existing benchmark which is a raytracer. With raytracing being more commonly heard right now, the the heavily-taxing nature of raytracing during rendering is still a good benchmark to measure CPU performance and POV-Ray is a good example of that.
Just like CineBench, Corona is another rendering benchmark for a static scene and leverages CPU capabilities.
7-zip benchmark is a compression and decompression benchmark and outputs its results in millions of instructions per second or MIPS. The better the system, the higher the score but is primarily influenced by the CPU and RAM.
A straightforward test: shows us a measurement of memory bandwidth.
CPU Test: Content Creation
PugetBench is a collection of benchmarks designed to gauge the performance of workstations and is made by boutique PC builder, Puget Systems. The benchmark measures various operations of the the applications and scores them. Photoshop benchmark measures plenty of the application’s commonly used function and Premiere Pro is tested on its editing and rendering as well effects application performance. The smoother the experience and faster the render, the better the score. This is a system benchmark and many factors affect each individual benchmark.
CPU Test: Productivity
Microsoft Office is one of the most ubiquitous software in the world and people take it for granted. Our testing will cover typical performance from file opening, copy images, moving data, and various other tasks. This is done on a very sizable project for each application and the total time to complete the overall operation is presented for Word, Excel and PowerPoint.
CPU Test: Syntethic Gaming
UL Benchmark is a popular way to showcase the theoretical gaming performance of a system with its very uniform testing scenes. They have a global database that pools these data and allows you to gauge if your system is operating as expected or not. While it is GPU-intensive, the total score will also be influenced by your CPU.
Looking good for Intel on synthetic testing but do some actual testing. Before we dive right in, I’ll refer your to one of our recent tests for my GPU testing methodology. You can also see our most recent update on how we measure graphics card performance in this article.
Gaming Performance (720p)
Testing games in 720p is a theoretical test to induce a CPU-limited scenario by pumping frames and giving us a good image of how a CPU would perform once it reaches these numbers. A system incapable of handling 720p numbers would theoretically not be able to attain anything higher with the resolution cranked up. That said, this serves as a purely artificial testing scenario especially for this review’s particular tier.
Gaming Peformance: Esports
Counter-Strike Global Offensive
Rainbow Six Siege
Call of Duty Warzone
PlayerUnknown’s Battlegrounds (PUBG)
Gaming Performance: Cont’d
The Witcher 3: Wild Hunt
Grand Theft Auto V
Shadow of the Tomb Raider
Value & Conclusion
The Core i9-11900K is the replacement for 10-core i9-10900K. With Rocket Lake being built on a new architecture and a 19% IPC bump, a drop to 8-core would’ve made sense IF the Core i9-11900K managed to actually beat its predecessor outright. Coming from a year where computing was in-demand, expectations from upgraders is low so people who may have had a Comet Lake 10900K or coming off a Ryzen 3000 CPU but can’t find a Ryzen 5000 are the least potential targets here and Intel is going for a shotgun assault and just bank it in what works and pray it hits. Unfortunately, as good as the performance is, in vanilla form, its just within the same performance level and we’re not getting any power improvements as well.
Its hard to like a CPU that’s just like the older CPU but isn’t that what Intel has been doing in recent years? The 10900K was actually a good try. A last hurrah for the older 14nm++++ era but Rocket Lake is proving hard to like despite achieving its claims of an IPC improvement, this proves nothing in front of true competition. Intel never meant to compete against he mini-HEDT Ryzen ala Ryzen 9s but it fairs enough to stand a chance against the Ryzen 7 5800X but here’s where things go a little bit harder: the Core i9-11900K could see retail prices of nearly $600 in some regions and the Ryzen 7 5800X is around $500-$600 in my region. Pretty tough but with no forward compatibility promise on Rocket Lake and its mainstream boards still gimped by a corporate lock, its tough to justify a $600 CPU.
Intel Core i9-11900K: worth it for gaming?
I’ve tested the Core i7-11700KF before receiving this CPU so I was really curious how Intel will be differentiating the two. In-game, we’re talking a few percent better in some cases and a few percent less in others. As an 8-core, it has its drawbacks
Intel Core i9-11900K: good upgrade?
From a 10900K? A resounding NO. There’s a reason brands are messaging me hours before embargo lift asking to update the BIOS and that’s because many of them are tuning their boards to deliver Intel’s upper PL2 limit to as best they can and also tune their boards further to do an all-core 5.1Ghz. Regardless, if you do enjoy overclocking, this CPU has its challenges and its fun to try and see how you’re gonna luck out on your CPU, but regular folks who just want a gaming PC, its an ok CPU but definitely not an upgrade against the Core i9-10900K and most definitely comes with some shortcomings, particularly the reduction of cores.
Most CPU launches are pretty cut and dry with the distinction between each tier clearly defined. In the case of Rocket Lake, what’s makes an i9, an i9? In this case, its the boost, I guess?
Save your money and spend it elsewhere.