Intel i9-11900K and i5-11600K Review: A Stopgap Until 10nm

Less than a year after the launch of Intel’s Comet Lake CPUs, we’re back with something new. Well, mostly new. The Intel i9-11900K and i5-11600K we have here today are part of Intel’s Rocket Lake CPU line-up and bring with it some significant changes over its predecessors. These improvements are covered in-depth in our specifications and features section below.

The flagship CPU is an 8-core, 16-thread CPU with a maximum boost speed of 5.3 GHz that Intel claims will deliver the best gaming experience in the desktop market. This time around, the i9-11900K will cost about $40.00 more than the 10900K with an MSRP of $539.00. We have included the full list of SKUs with specifications and pricing below as well.

Today we are pitting AMD against Intel, and we’ll see who comes out a winner in our testing suite. The showdown will include the i9-11900K (8-core @ $613) and the i5-11600K (6-core @ $269) from Intel up against the Ryzen 7 5800X (8-core @ $499), and the Ryzen 5 5600X (6-core @ $299) in an effort to align core count and pricing.

Specifications and Features

Intel’s latest line of desktop processors is a combination of new and older technologies designed to take advantage of the benefits both offer. Rocket Lake still uses the trusted 14++ nm process node, and with it, the higher frequencies we saw in Comet Lake maintaining the 5.3 GHz Thermal Velocity Boost (TVB) and up to 5.1 GHz all-core boost with the all-new Adaptive Boost Technology (ABT).

The new Rocket Lake CPUs use Intel’s Sunny Cove microarchitecture variant found in the 10 nm Ice Lake mobile processors called Cypress Cove. With it, we have a new micro arch and features such as PCIe 4.0, enhanced Intel Xe graphics, and an IPC uplift of up to 19%, according to Intel’s internal testing.

PCIe Gen 4.0

With the Rocket Lake series of processors, Intel is finally introducing PCIe Gen 4.0 to its lineup. Generation 4.0 effectively doubles the available bandwidth for connected peripherals to 16 GT/s translating to nearly 32 GB/s with an x16 interface. AMD was the first to release PCIe 4.0 devices to the mainstream with its Ryzen 3000 series CPUs paired with a 500 series motherboard along with the Radeon RX 5000 series GPUs in July of 2019. We all expected Intel to follow suit, but 10nm proved to be a roadblock for its mainstream CPUs, and delays followed.

So here we are today, nearly two years later, and Intel has caught up offering 20 PCIe 4.0 lanes with the new Rocket Lake CPUs. These 20 lanes are divided into a full PCIe 4.0 x16 for add-in cards and an x4 link for storage. As of yet, PCIe Gen 4.0 hasn’t offered much in the way of performance gains for compatible PCIe 4.0 graphics cards, but storage, on the other hand, does see some benefit from the higher transfer rates. We also have 24 PCIe Gen 3.0 lanes from the PCH for additional storage and peripherals.

Z590 and Socket 1200

This leads us to the motherboard support. The new Rocket Lake CPUs continue with socket 1200 and introduce a new 500 series chipset. The good news is 11th gen CPUs are backward compatible with the 400 series motherboards, including PCIe 4.0 support if the board has it. We also have forward compatibility with 10th generation Comet Lake CPUs in the 500 series motherboards, making upgrading a bit easier. However, this won’t give you access to PCIe 4.0 since that is dependent on using an 11th Gen CPU.

Aside from PCIe 4.0, there are some other changes worth noting. In particular, Intel has doubled the DMI 3.0 link from DMI x4 to a DMI x8 link. The DMI link connects the chipset to the CPU directly, very similar to PCIe lanes. This should mean that two PCIe Gen 3.0 x4 storage devices can be attached and run at full speed in addition to the PCIe Gen 4.0 x4 drive on the PCIe link to the CPU. Undoubtedly, SATA port sharing is inevitable, and some will be disabled in this situation.

We’ll also see native support for USB 3.2 Gen 2×2 (20 Gbps) through the chipset and Thunderbolt 4 (40 Gbps) for USB connectivity using Intel’s Maple Ridge controller. Thunderbolt 4 doesn’t offer any additional data transfer bandwidth but does improve video output now with resolutions up to 8K, dual 4K monitor support, and improved security with Intel VT-d DMA protection.

Memory Overclocking

After many years Intel is finally allowing memory overclocking on more than its enthusiast class chipset. They now extend memory overclocking to the H570 and B560 chipsets for all Rocket Lake CPUs. We also see a jump in the base support for memory from DDR4 2933 to DDR4 3200. At a glance, this can be a bit misleading as Intel has also introduced a gear ratio for their memory controller, much like AMD Ryzen CPUs. The memory controller now runs at 1:1, which is gear one but can also be switched to gear two, a 1:2 ratio meaning the memory controller will operate at half the memory frequency. The only CPUs that officially support DDR4 3200 in gear one are the i9-11900K and KF SKUs; all others support DDR4 3200 in gear two and DDR4 2933 in gear one.

That being said, we tested the i5-11600K with DDR4 3600 in a 1:1 ratio, and it ran flawlessly. Having this gear ratio will allow the CPU to run some outrageous memory speeds as indicated by the motherboard specifications. The ASUS Maximus XIII we have for testing lists speeds up to DDR4 5333; checking the QVL list, we find maximum speeds listed at 4800 MHz. Regardless, the cost associated with these super high-speed memory kits like this Corsair kit at 5000 MHz for $1400 leads us to believe this is for the extreme overclockers and setting world records.

AVX-512 and Deep Learning Boost with VNNI Support

With Rocket Lake CPUs, Intel is bringing AVX-512 (Advanced Vector Extensions 512) to the mainstream desktop. AVX (128-bit) first appeared in Sandy Bridge, then a few years later, we had AVX2 (256-bit with FMA) show up in Haswell. AVX-512 expands to 512-bit support. The whole idea here is to complete more work in every CPU cycle. More work creates a higher load and more heat at lower speeds; it also requires more power. This is why, since the inclusion of AVX2, we have had the option to set an offset for the CPU in the BIOS if it encounters these types of instructions. The CPU then downclocks to avoid instability or overheating. We also have the ability in BIOS to disable AVX-512 altogether, which is great since most mainstream software today isn’t coded for AVX-512.

Where AVX-512 and VNNI come into play is in Deep Learning and AI-type workloads by accelerating the computations. VNNI (Vector Neural Network Instructions) is part of the AVX-512 instruction set architecture that combines three instructions into one and further accelerates the computation. This is very dependent on the workload and software being used. Below is a slide from Intel’s Server division that’ll give you a visual. For additional information, visit the Intel Deep Learning Boost site.

Adaptive Boost Technology

We’re all familiar (or not) with Intel’s collection of turbo boost technologies that accelerate the CPU frequency depending on factors such as power use and temperature. A quick rundown is an all-core boost, Turbo Boost 2.0, Turbo Boost Max 3.0, and Thermal Velocity Boost. We now have a new boost bin on the block, and just like Thermal Velocity Boost, Adaptive Boost only works with Core i9 K and KF-11th Gen CPUs and is dependent on your cooling solution motherboard power delivery. In effect, this new boost is akin to a 300 MHz overclock above the 4.8 GHz all-core boost allowing the i9-11900K to run consistently at 5.1 GHz while maintaining its 5.3 GHz TVB and downclocking attributes when encountering heavier workloads such as AVX2 and AVX-512. ABT is a one-click BIOS option that negates the need to manually over clock the i9-11900K/KF CPUs.

The Rocket Lake CPU lineup is quite extensive but limited to desktop CPUs only. Based on the Cypress Cove architecture, the Core i5 to Core i9 CPUs bring new features. The Core i-3 and Pentium CPUs, on the other hand, are a refresh still based on Comet Lake with a slight speed boost and five at the end of the name instead of a zero. You can see the charts below for the full list of SKUs with their respective operating speeds, TDP, and MSRP.

Today we’ll be testing the Core i9-11900K 8-core, 16-thread, and the Core i5 11600K 6-core, 12-thread CPUs specifically. Both of the CPUs have unlocked multipliers and a 125 W TDP making them ideal for enthusiasts. Just add cooling and a motherboard to support them. Intel allows up to 250 W for optimal turbo power operation, making cooling very important for maximum speed.

Both the 11900K and 11600K have 2 MB of L3 cache per core, adding up to 16 MB for the i9 and 12 MB of L3 cache for the i5 series. Memory support remains the same for Pentium and i3 CPUs at 2666 MHz. Still, the Core i9, i7, and i5 series now have base memory support of up to 3200 MHz, as was discussed above, with all CPUs supporting 128 GB of memory in a dual-channel configuration.

Intel has also incorporated the new Intel Xe graphics into the Rocket Lake CPUs with claims of a 50% increase in graphics performance.  The new iGPU has 32 execution units, supports up to a single 8K 12-bit HDR or dual 4K 10-bit HDR displays. It also has support for Display Port 1.4a and HDMI 2.0 connectivity. The specifications are listed as Intel UHD Graphics 750 and only available in the Cypress Cove CPUs. The i3 and Pentium CPUs still have the last-gen Intel UHD 630 graphics processor.

Judging by the quick benchmark we ran below, we do have a 50%+ improvement in the graphics portion of Night Ranger from 3DMark.

Pricing for the Core i9-11900K is $539.00, which is slightly higher than the launch price for the 10-core, 20 thread 10900K. The i7-11700K, which is still an 8-core, 16-thread CPU with slightly lower clock speeds, is listed at $399.oo, and the 6-core i5-10600K price remains unchanged at $262. Looking at today’s competition, the 8-core Ryzen 7 5800X lists today for $489.00, and the Ryzen 5 5600X comes in at $299.00. Moving up to the Ryzen 9 5900X, we have 12-cores and 24-threads for $549.00, just $10 more than Intel’s flagship i9-11900K.

All 11th Generation SKUs and details provided by Intel:

Meet the i9-11900K and i5-11600K

Before things get started, below are images of the 10900K and 10600K in CPU-Z at its stock settings with XMP enabled.

Here’s a slideshow of the sample packaging, which differs from retail, some close-ups of the CPUs’ top and bottom, plus a comparison of the i9-10900K and the i9-11900K. Pad orientation is the same, but there are striking differences in the SMD cluster at the CPU center.

This slideshow requires JavaScript.

Test Setup and Results

Here we take a slightly different approach to CPU testing with ours based on many Hwbot.org benchmarks since that is what we are known for, overclocking and benchmarking. We use real-world testing as well with Cinebench, x265, POV-Ray, and 7Zip to give readers a good idea of the general performance of the product tested.

We would also like to thank Intel/ASUS for supplying the ROG Maximus XIII Hero, a Z590 chipset-based motherboard that we used for all testing of the i9-11900K and the i5-11600K. This is one beautiful-looking motherboard in a sleek, matte black design with some silver accents.

Some highlights of the Maximus XIII Hero taken from the product page:

• Robust Power Solution: 14+2 teamed power stages rated for 90 Amps, ProCool II power connectors, MicroFine alloy chokes, and 10K Japanese-made black metallic capacitors
• Optimized Thermal Design: Enlarged VRM heatsinks plus integrated aluminum I/O cover, high-conductivity thermal pad, quad M.2 heatsinks with embedded backplates, and ROG Water-Cooling Zone
• High-performance Networking: Onboard WiFi 6E (802.11ax), dual Intel^® 2.5 Gb Ethernet, and ASUS LANGaurd.
• Fastest Gaming Connectivity: PCIe 4.0, quad M.2, USB 3.2 Gen 2×2 front-panel connector, dual USB Type-C^® ports with Thunderbolt™ 4 USB-C^®
• Industry-leading Gaming Audio: ROG SupremeFX ALC4082 with ESS^® ES9018Q2C DAC for high-fidelity audio
• Unmatched Personalization: ASUS-exclusive Aura Sync RGB lighting, including one RGB header and three addressable Gen 2 RGB headers
• DIY-friendly Design: Pre-mounted I/O shield, BIOS FlashBack™, Q-Code, FlexKey, Q-Connector, SafeSlot, and Graphics Card Holder
• Renowned Software: Bundled 1-year AIDA64 Extreme subscription and intuitive UEFI BIOS dashboard with integrated MemTest86

This slideshow requires JavaScript.

Intel i9-11900K and i5-11600K Performance Testing

** Due to unforeseen shipping issues, the Ryzen 7 5800X hasn’t arrived in time for this launch. When it arrives and testing is completed, the following graphs and comments will be updated to reflect any new information. **
03/31/2021- Graphs have been updated with Ryzen 5800X results and some comments have been modified to include the new data.

CPU Tests

• AIDA64 Engineer CPU, FPU, and Memory Tests
• Cinebench R20 and R15
• HWBot x265 1080p Benchmark
• POVRay
• SuperPi 1M/32M
• WPrime 32M/1024M
• 7Zip

All CPU tests were run at their default settings unless otherwise noted. We have also added results for the 11900K using the new ABT (Adaptive Boost Technology) that Intel has introduced with Rocket Lake. Like TVB, it is only available on the flagship CPUs, namely the i9-1900K and KF series.

Gaming Tests

We have updated our gaming tests and dropped them down to four games for CPU reviews. In many cases, even at 1440p, the difference between CPUs isn’t that much, and the titles we use cover both CPU-heavy titles and GPU-bound titles. All game tests were run at 1920×1080 and 1440×2560 with all CPUs at default settings unless otherwise noted. Please see our testing procedures for details on in-game settings.

• Shadow of the Tomb Raider – DX12, “Highest” preset
• The Division 2 – DX12, Ultra preset, VSync Off
• F1 2020 – DX12, Very High defaults, TAA, and x16 AF, Australia track, do not show FPS counter
• Metro: Exodus – DX12, Ultra defaults
• UL 3DMark Fire Strike (Extreme) – Default settings

AIDA64 CPU, FPU, and Memory Tests

Included below are a couple of shots of the AIDA64 cache and memory benchmark results for both the i9-11900K and i5-11600K. As you can see, the number of cores seems to have a slight impact on memory bandwidth, with the 11900K taking the lead over the 11600K. The new Cypress Cove architecture does improve the memory bandwidth, although it lost a bit of latency compared to the i9-10900K from last year.

In the AIDA64 CPU tests, we can see that the higher core count of the 10900K still gives it a lead in some tests. The Intel i9-11900K does hold its own through most of the testing, even though it has two fewer cores. The surprising win in the PhotoWorx test is the 11600K; presumably, the lower core count allowed it to hold its turbo frequency longer. The i5-11600K puts up a good fight against the 5600X taking three of the five benchmarks as does the 11900k over the 5800X. One thing to note here is that there has been a significant improvement in the way Rocket Lake handles the decryption-based AES test.

Moving on to the AIDA64 FPU tests, the Ryzen 7 5800X puts up a great fight with the Comet Lake-based i9-10900K. Despite having two fewer cores it still manages to beat it out in a couple of the tests. The 11900K did its best but wasn’t up to the task losing a couple to the 6-core 5600X. When we compare the 6-core CPUs, the Ryzen 5 5600X has a definite lead in all but FP64 Ray-Trace, where the results are only a few points apart.

Real World Tests

Moving on to the real-world testing, the Ryzen 5800X and 5600X CPU made a clean sweep across these five benchmarks compared to their Intel counterparts. The 5600X even bested the 11900K in the HWBot x265 rendering benchmark, which is surprising considering that the benchmark scales very well with more cores. These benchmarks are multi-threaded and suit the 10-core, 20-thread 10900K very well and you would think this would give it the edge. Surprisingly the Ryzen 5800X was trading blows with it very seriously with a couple of near ties, one narrow loss, and three victories. This really demonstrates AMD’s strengths when it comes to processing parallel workloads such as video encoding, rendering, and compression.

Pi and Prime-Based Tests

Next up are the Pi and Prime number-based tests. In this set of testing, the 10900K and 11900K make a trade-off. With the IPC improvement over the last generation, the 11900K races ahead of the 10900K. Wprime, on the other hand, tells a different story where IPC loses to core count and the 5800X. Comparing the 6-core CPUs, we can see that the i5-11600Ks 4.9 GHz boost speed gives it a definite advantage in SPi, but when the speeds are evened up in WPrime, its 4.6 GHz all-core boost wasn’t enough, and the 5600X took the lead.

Gaming Results

We’ll let the gaming results speak for themselves. As you’ll see, there isn’t much difference between the CPUs regardless of core-count or boost speeds. At 1080p, the results are within 10 FPS of each other but the Ryzen 5800X manages to come out on top. Moving up to 1440p, we see that the CPU has much less to do with the FPS and the GPU takes most of the load leaving the results fairly even.

Next up, we have 3DMark Fire Strike Extreme, a DX11-based test that UL says the graphics are rendered with detail and complexity far beyond other DX11 benchmarks and games using 1920×1080 resolution. As you can see below, the overall and graphics scores were fairly close to one another, but the physics results really show a spread. This test is strictly CPU dependent, and typically thread count is the largest factor. What’s interesting is that even with a lower core-count, the 5800X took the crown in the Physics test. In the end, both Ryzen CPUs came out ahead of the Blue Team.

IPC and HT Efficiency

With Rocket Lake, Intel claims up to a 19% improvement in IPC (instruction per clock) during certain workloads. We have our own range of tests to compare IPC across generations and platforms, which we feel has a real-world approach comprised of commonly used software. All CPUs are set to the same clock speed of 4.0 GHz and performed on one thread without any hyper-threading for these tests. We also include a hyper-threading efficiency test to get an idea of how much “extra” work these virtual cores are performing.

As we can see below, when compared to previous generations, Cypress Cove does offer double-digit IPC gains in all except the 7 ZIP benchmark. We also see they still have some ground to make up between them and AMD, which is still has a lead of up to 10% depending on the benchmark. Luckily the higher clock speeds we see close that gap significantly.

In our SMT and HT efficiency test, we also find some improvements in the Cypress Cove architecture. Aside from POVRay, which lost a few percent, the remainder of the tests gained a few, with 7-Zip showing an 8% improvement in efficiency. Compared to AMD, the Cinebench tests are pretty much even now, but they still have the lead in POVRay and 7-Zip.

Power Consumption and Temperatures

Intel uses a set of variables called Power levels: PL1, PL2, PL3. PL1 has a power limit; in this case, 125 W, PL2 is sustained power delivery (Turbo), and PL3 is the power delivery limit. PL2 is the maximum sustainable power the CPU can handle until thermal issues occur. Intel has set the value of PL2 to 250 W though board manufacturers can set their own. The power levels also have a duration specification.

Under Intel specifications, the CPU will run at PL2  for its specified duration then fall back to PL1, which keeps it within 125 W even if the CPU needs to reduce its speed to do so. ASUS has a setting call MCE (multicore enhancement); this pushes the CPU to sustain the PL2 level of 250 W, and the core speed can vary depending on the load and temperature. The chart below shows the differing boost speeds and power levels for the CPUs we’re testing today.

Running the i9-11900K under Intel’s guidelines, we see moderate power usage and temperatures, but behind the scenes, the CPU runs slower. In this case, during the AIDA64 stress test, the i9-11900K ran at 3.9 GHz. Bump that up to a more stressful load, the FPU test in AIDA64, and the speed drops again, this time to 3.7 GHz. Lastly, 3.6 GHz while running Prime95 small FFTs. Keep in mind that all of these tests can take advantage of the AVX-512 instruction set. With AVX-512 disabled in BIOS, the CPU maintained its 125 W envelope, but we saw increased core speed of 4.4 GHz, 4.2 GHz, and 3.8 GHz, respectively.

Using ASUS’ Multi-core Enhancement (MCE), the CPU now maintains a 250 W envelope and adjusts CPU speed accordingly, but the power usage and temperatures skyrocket. As we can see by the Prime95 small FFT results, the i9-11900K sucks down 377 W of power (at the wall) and reaches a peak of 99°C with the 360 mm EK Predator CLC we use. If you want to get the most out of this CPU, you will have to bring adequate cooling.

The i5-11600K behaved much the same way but did manage to maintain close to its all-core boost speed of 4.6 GHz with 4.5 GHz reported for Prime small FFT. The temperatures and power use were still quite high, reaching 339 W and 93°C, this is over 100 W more than the similar Ryzen 5 5600X.

Overclocking the i9-11900K and i5-11600K

Overclocking and stability always seem to be contentious subjects in enthusiast forums. Many people feel that serious stability testing isn’t always necessary, and tests such as Prime95 small FFTs are labeled a “heat virus” and unnecessary for gaming. While this type of stability may not be necessary for some people, it’s my preferred test when overclocking a system that will run for years without corrupting my Windows installation. That being said, with the inclusion of AVX-512, this is going to limit your outcome unless it’s disabled in BIOS.

As for overclocking the i9-11900K, I chose to go with the new Adaptive Velocity Boost option, and you can see results that have been included with the benchmark tables. This seemed like the best approach to an easy, hassle-free option which saved a lot of time and effort. It also forgoes the need to disable any AVX options or setting offsets to maintain stability. The CPU now runs at 5.1 GHz on all cores when it can and will reduce the core speed when needed to accommodate the load plus; it maintains the 5.3 GHz boost speed for light loads. It really does seem like a win-win option. As you can see in the Fire Strike Extreme result below, we had significant gains in the physics test of nearly 2000 points and an overall score gain of 1000.

The i5-11600K, on the other hand, proved to be a bit difficult. The highest I could push it in the Prime 95 small FFT test was 4.7 GHz; this is with AVX-512 disabled in the BIOS. This slight overclock was pulling 377 W and hitting 95°C, which hardly seems worth it. Using the AIDA64 FPU test, I managed a slightly better result getting up to 4.9 GHz; this was stable enough to run some benchmarks like CB R20 but would not pass the Prime 95 Blend test, which is quite a bit lighter than Small FFTs. In this test, it ran up to the mid-90s and started dropping threads almost immediately. 5.0 GHz was out of the question at this point, aside from some light benchmarks, that is. Overall, I felt that the i5-11600K I received might not have been a good example of average and most likely way below.

Conclusion

The Rocket Lake lineup is an interesting addition to the Intel product stack, no doubt in response to AMD and their continuing gains in the desktop CPU market. With the continual delays they have encountered with their 10 nm node, Intel couldn’t very well sit back and ride out Comet Lake until they get the new node operational. Instead of releasing yet another refresh, they have given us a new architecture with IPC gains and new features.

Leveraging the most they can from the very mature 14 nm++ microarchitecture, we continue to see much higher clock speeds, boosting to 5.3 GHz on a single core as long as the conditions are favorable. However, the all-core boost takes a slight drop from 4.9 GHz now will run up to 4.7 GHz, again under Intel specs, which requires certain conditions, which we covered earlier. To get the most from the i9-11900K, you won’t be running with Intel’s limits, or you’ll be losing performance to remain within the 125 W TDP. Enabling motherboard features such as MCE from ASUS will open up the limits and let the CPU run all cores at 4.7 GHz or 5.1 GHz with ABT enabled. The 11600K doesn’t have the boost options of the flagship CPUs, and as such, the maximum all-core boost is limited to 4.6 GHz. The trade-off here is the heat and power consumption, making high-end cooling a must to get the most you can from these CPUs.

Looking at the i9-11900K from a performance per dollar position, it starts to become a tough sell. With an MSRP of $539.00 but listing at $613 on Newegg plus the need for a premium cooling solution soon puts this CPU’s cost around the $700 range. Comparing that to the Ryzen 7 5800X, which can be found for $489.00 on Newegg. That and the Ryzen 9 5900X will outperform the 11900K in multi-threaded workloads and lists for only $60 less at $549.00. We do have to admit that Intel has come up with a great gaming/lightly threaded CPU here with the 5.3 GHz boost since, in general, the faster CPU typically outperforms in gaming. Still, for most of the titles we tested at 1080p, the 5800X was the leader.

Moving down the product stack to the 6c/12t 11600K, we find a much better value in a gaming CPU for those who don’t need all the cores and threads of the 11900K. With a retail price of $269.00, it’s about $30 cheaper than the Ryzen 5 5600X, so this one is a close call. Gaming results are about the same, and the 5600X outperforms in many of the tests in our suite. Add to that; it also comes with its own cooler, so the cost difference is non-existent.

In the end, the Rocket Lake CPUs stick it to AMD clock speed-wise and single-threaded performance with some very impressive boost speeds reaching up to 5.3 GHz. They also bring with them a new architecture with new features and a much-needed IPC improvement. Still, in most tests, these are the fastest CPUs Intel has put out to date. Whether or not the price is worth it for the gains is up to the buyer, but Intel has put out some solid-performing CPUs for the mainstream platform.

Shawn Jennings – Johan45

Related Articles:

• Intel Core i9-1900K and i5-10600K Review
• Ryzen 9 3900X and Ryzen 7 3700X Review
• Intel Core i9-9900K CPU Review
• AMD Ryzen 5 3600X CPU Review