Intel Core i9-9900K 9th Gen CPU Review: Fastest Gaming Processor Ever

Hardware-Based Security Fixes, Architecture & Test Setup

Much like the Cascade and Whiskey Lake processors we recently covered, Intel's Coffee Lake refresh comes with hardware-based mitigations for the Meltdown and L1TF (Foreshadow) vulnerabilities. Current Spectre and Meltdown mitigations, which Intel delivers via software and microcode patches, can reduce performance by up to 10% on newer CPUs, with older hardware suffering even larger losses. The new mitigations, baked directly into the silicon, should reduce or even eliminate the performance impact for a few vulnerabilities.

Vulnerability
Coffee Lake Refresh/Whiskey Lake Mitigation
Cascade Lake Mitigation
Variant 1 (Spectre)
Operating System
Operating System/VMM
Variant 2 (Spectre)
Microcode + Operating System
In-Silicon + Operating System/VMM
Variant 3 (Meltdown)
In-Silicon
In-Silicon
Variant 3a
Microcode + Operating System
Firmware
Variant 4
Microcode + Operating System
Microcode + Operating System/VMM
L1TF (Foreshadow)
In-Silicon
In-Silicon

As we're learning, it may take several processor generations before the fixes for all vulnerabilities are applied at a silicon level. Intel's ninth-gen CPUs do still need a combination of microcode and operating system patches. But at least Meltdown and L1TF Foreshadow are patched fully in hardware.

Architecture

Although we don't have much new information about Coffee Lake refresh architectural changes, David Schoor at WikiChip says the new chips use a familiar ring bus, an internal high-speed pathway connecting the cores and cache.

This stands in contrast to the mesh architecture (deep dive here) that Intel uses on its high-end desktop models, including the eight-core Core i7-7820X. We've found that the mesh architecture has a negative impact on some desktop applications, including games, which Intel acknowledges. The mesh architecture was designed to increase scalability as it expands to higher core counts in the Xeon family.

Overclocking, Rounds One and Two

We tapped Corsair's H115i v2 to test our Core i9-9900K sample in the U.S. lab. This liquid cooler afforded enough headroom to sustain a 5.0 GHz overclock with a 1.33V Vcore and a Load Line Calibration 4 setting. It kept the chip at a steady 85°C during extended non-AVX stress tests. Folding in AVX instructions did, unfortunately, overwhelm the all-in-one. To reign in the thermal output, we set the AVX offset to -2, meaning the chip ran at 4.8 GHz during AVX-optimized workloads and 5.0 GHz in the absence of AVX instructions. We maintained a temperature of 95°C during three hours of Prime95 using those settings.

To model real-world settings attainable by enthusiasts with closed-loop liquid coolers, we applied the -2 AVX offset for our 5.0 GHz overclock in the gaming, office and productivity, and rendering tests.

We did not use an offset for the workstation graphics, compute, power consumption, and temperatures sections.

MEG Z390 Godlike

We're using MSI's MEG Z390 Godlike as our test platform for all Intel processors. This pricey board retails for $600, but has the power delivery subsystem to support aggressive overclocking.

The MSI MEG Z390 Godlike sits at the top of MSI's motherboard hierarchy. It has a decked-out 18-phase power delivery subsystem that's designed to squeeze every drop of performance out of Intel's new processors. It also comes with a few nifty accessories like an M.2 PCIe riser card and an HDMI streaming card.

Comparison Products

Loading...

Test System & Configuration
Hardware
Germany

Intel LGA 1151 (Z390)

Intel Core i9-9900K, i7-9700K, i5-9600K, i7-8700K, i5-8600K, i5-8400
MSI MEG Z390 Godlike
2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667 & DDR4-3466

AMD Socket AM4 (400-Series)
AMD Ryzen 7 and Ryzen 5
MSI X470 Gaming M7 AC
2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667, DDR4-3466

Intel LGA 2066
Intel Core i7
MSI X299 Gaming Pro Carbon AC
4x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2666

All Systems

GeForce GTX 1080 Founders Edition (Gaming)
Nvidia Quadro P6000 (Workstation)
1x 1TB Toshiba OCZ RD400 (M.2, System SSD)
4x 1TB Crucial MX300 (Storage, Images)
be quiet! Dark Power Pro 11, 850W
Windows 10 Pro (All Updates)

U.S.

Intel LGA 1151 (Z390)

Intel Core i9-9900K, i7-9700K, i5-9600K, i7-8700K, i5-8600K, i5-8400
MSI MEG Z390 Godlike
2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667 & DDR4-3466

Intel LGA 2066
Intel Core i9-7820X
MSI X299 Gaming Pro Carbon AC
4x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2666, DDR4-3200

AMD Socket AM4 (400-Series)
AMD Ryzen 7 2700X, Ryzen 5 2600X
MSI X470 Gaming M7 AC
2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933

All Systems

EVGA GeForce GTX 1080 FE
1TB Samsung PM863
SilverStone ST1500-TI, 1500W
Windows 10 Pro (All Updates)

Cooling
Germany

AMD Wraith Ripper
Alphacool Ice Block XPX
Enermax LiqTech 240 TR4
Thermal Grizzly Kryonaut

U.S.
Wraith Ripper
Corsair H115i
Enermax Liqtech 240 TR4 II

Power Consumption MeasurementContact-free DC Measurement at PCIe Slot (Using a Riser Card)
Contact-free DC Measurement at External Auxiliary Power Supply Cable
Direct Voltage Measurement at Power Supply
2x Rohde & Schwarz HMO 3054, 500 MHz Digital Multi-Channel Oscilloscope with Storage Function
4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100 kHz, DC)
4x Rohde & Schwarz HZ355 (10:1 Probes, 500 MHz)
1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement1x Optris PI640 80 Hz Infrared Camera + PI Connect
Real-Time Infrared Monitoring and Recording
Acoustic MeasurementNTI Audio M2211 (with Calibration File, Low Cut at 50Hz)
Steinberg UR12 (with Phantom Power for Microphones)
Creative X7, Smaart v.7
Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H)
Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm
Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA)
Graphical Frequency Spectrum of Noise

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy

MORE: All CPUs Content

148 comments
    Your comment
  • dlim389
    "The better the cooling, the better power consumption"? What?! This doesn't make sense at all. You meant the better the cooling, the lower the temperature.
  • siman0
    "Redefine" at double the price of a 2700x it needs to do more than that. The price diffrence is more than enough to go up another GPU tier ie 1070ti to a 1080ti. Id rather have more pcie bandwidth and the ability to upgrade my cpu till 2020. The only way Id recommend a upgrade to something close is if you already have the motherboard. But even then Id say the 9700k.
  • s1mon7
    Wait, so it performs within a spitting distance of the 2700x with DOUBLE the power consumption and price? Holy smokes, I thought Intel will be able to easily take on AMD after they launch their 8-cores. I have to say that these results were very surprising to me, since I believed in this being the ace up Intel's sleeve. This is really interesting, and a big win for AMD. The 9900K goes through twice the power just to squeeze out that extra clock speed edge to outperform the 2700x by a mere 10%, at double the price, mind you.

    Intel clearly needs 10nm and a new architecture to go back into the game. As is, I struggle to think of any reason to buy the 9900k.
  • TCA_ChinChin
    Interesting chip from Intel, but if you are looking at CPUs from a performance per dollar point of view, its rather disappointing. The real disappointment is the 9700k which is more expensive than the 8700k but performs basically the same on average.
  • Adrian Ocampo
    As a gamer, why would I buy a 9700K when an 8700K trades blows with the 9900K in both gaming and productivity. It just doesn't make sense. Its like its just competing with its own product as this point. I would be better off buying an 8700K deliding it, put liquid metal and overclock to 5.0Ghz for a much lower price than a 9900K.
  • gfaiii
    Guys just as a heads up you should say second generation 1151 socket, these are NOT compatible with 200 series boards that have 1151 sockets (without modding)
  • sonichedgehog360
    Ladies and gentlemen, Intel’s FX 9000 series.

    (By the way, I totally saw this coming what with the crazy clock speeds they were pushing.)
  • Johnpombrio
    I was surprised on how well the i9-9900K did on stock clocks. I may not even bother with overclocking as it does well even without it and I may be able to use air cooling that way. Compared to my Kaby Lake i7-7700K, it definitely is a big step up. I already have the ASUS Strix Z380 mobo, 32GB Corsair Dominator Platinum memory kit and have preordered the CPU.
  • sstanic
    how is this an editor's choice is beyond me. but not beyond marketing people, is it?
  • redgarl
    For the money, you can buy a motherboard, a CPU and a 1080 GTX for the same price as the 9900k with it's motherboard.

    Also, you tested this system on a 600$ motherboard... 600$ and a prenium cooling solution.

    This system is above the 2000$ threshold compared to an AMD one barely hitting the 1000$.
  • lperreault21
    This is a terrible cpu. How much did Intel pay you to give it a 4.5/5. Yet anothrt reason to not trust Tom's with anything
  • jimmysmitty
    2809234 said:
    Wait, so it performs within a spitting distance of the 2700x with DOUBLE the power consumption and price? Holy smokes, I thought Intel will be able to easily take on AMD after they launch their 8-cores. I have to say that these results were very surprising to me, since I believed in this being the ace up Intel's sleeve. This is really interesting, and a big win for AMD. The 9900K goes through twice the power just to squeeze out that extra clock speed edge to outperform the 2700x by a mere 10%, at double the price, mind you. Intel clearly needs 10nm and a new architecture to go back into the game. As is, I struggle to think of any reason to buy the 9900k.


    Torture loop power numbers are hard to use as a real definition of power draw as most no one maxes any CPU out 100% 24x7. Add in the clock speed difference and thats makes it look worse than most people will veer experience.

    251426 said:
    For the money, you can buy a motherboard, a CPU and a 1080 GTX for the same price as the 9900k with it's motherboard. Also, you tested this system on a 600$ motherboard... 600$ and a prenium cooling solution. This system is above the 2000$ threshold compared to an AMD one barely hitting the 1000$.


    Only if you plan to only get the top end $500 dollar motherboards for the Intel system then cheap out for Ryzen boards. If you compare apples to apples there are equivalent Ryzen motherboards that are the same price and offer similar features apart from the different sockets and chipsets.

    It always amuses me when people compare systems then for one they go with a $150 dollar board thats obviously an inferior product.
  • AgentLozen
    Thanks for the review guys. I agree with your conclusion that the 9900k is in a league of its own. Do you think someone could cool a 9900k system (effectively) with a big Noctua air cooler if you're not overclocking?
  • mgallo848
    I think Intel failed trying to market this CPU to gamers. The price/performance does not justify it at all. In multi-threaded editing applications it looks much more impressive.

    For editing applications yes, for gamers no.
  • Gurg
    This is first CPU that would even remotely justify upgrade from my 5820k running OC @4.2. 77% Time Spy increase vs 52% increase in CPU plus motherboard cost from what I spent on 5820K.

    The cost numbers in this review look far worse by TH pairing this with the MSI Godlike $599 vs the ACE $289 Z390 mb.
  • levijonesm
    "The $263 Core i5-9600K at stock settings regularly beat an overclocked $378 Ryzen 7 2700X in games"

    Please correct the typo on 2700X pricing. Should be $329 MSRP, but currently can be bought for $300 or lower.
  • Brian_R170
    In theory, a i9-9900K with Hyper-Threading disabled in the BIOS should perform slightly better than the i7-9700K in the benchmarks where the 9700K took the lead due to the higher clock speeds and larger cache, right?
  • AgentLozen
    Brian_R170 said:
    In theory, a i9-9900K with Hyper-Threading disabled in the BIOS should perform slightly better than the i7-9700K in the benchmarks where the 9700K took the lead due to the higher clock speeds and larger cache, right?


    That's a good question. I'd like to see a separate article that examines the performance difference between the 9900k and the 9700k with hyper threading turned on and off. It would be a good chance to see how much power overhead hyper threading requires and it could answer what the value proposition of hyper threading really is.

    Is it even worth having hyper threading turned on while you're gaming?
  • delaro
    Not much of a performance difference over an 8700K at least not enough to warrant $499. For that matter, it makes a Ryzen 2700x @ $289 look like an even better deal considering the performance gap from this review isn't that huge. The next version of Ryzen shouldn't have an issue with matching the same performance at a much lower cost.
  • logainofhades
    Price/performance wise, I still think I would till rather have a 2700x.
  • tikal
    I wodner why prices of CPUs on the graphs were not fixed. Irt is impossible to get i5 8400 for that price, at the same time Ryzen CPU have a much lower price now.
    If I spot this right away, I have hard time believing someone it has not been done intentionally.
  • volkgren
    It's not too hot or power hungry that high-end gaming systems can't already handle. It's nowhere near what the FX-9000 series was simply because the i9 is the fastest and I'm sure more stable. All in all it's an excellent CPU at a terrible price. Intel's Core i5 is the best value once again.
  • tamalero
    149725 said:
    Torture loop power numbers are hard to use as a real definition of power draw as most no one maxes any CPU out 100% 24x7. Add in the clock speed difference and thats makes it look worse than most people will veer experience.

    No one maxes any cpu?
    If I had this system for rendering and similar tests, it sure as hell its going to tax the cpu at 100%.
    Your excuse for intel is silly.

    149725 said:
    Only if you plan to only get the top end $500 dollar motherboards for the Intel system then cheap out for Ryzen boards. If you compare apples to apples there are equivalent Ryzen motherboards that are the same price and offer similar features apart from the different sockets and chipsets. It always amuses me when people compare systems then for one they go with a $150 dollar board thats obviously an inferior product.


    Hu, you can get very good motherboards for Ryzen for less than 300 USD. Not need to "cheapen out".
    The HALO effect overprices for intel.
    Plus you need to get a HEFTY cooler for intel's chip.

    For a mod, you sure are going pretty far to defend intel.
  • littleleo
    Did I miss it? I don't recall seeing the 50% better performance that "Principled" Technologies found over the AMD Ryzen 7 2700X. I do see the Ryzen 7 2700 selling for $249.53 and the Ryzen 7 2700x selling for $305, and the Intel i9-9700K selling for $410, and the i9-9900K is $580. I don't see the value of paying $105 or $275 more vs the R7-2700x for the increase in performance shown.

    I can put that money towards a better GPU, though I'm not seeing any value in buying the Nvidia RTX hype train either. So what is going on here? Did Nvidia & Intel see the insane prices some people were paying during the crypto mining craze and think they would cash in by making these ridiculously over price pieces of silicon? News flash Intel & Nvidia we want value for our green stamps and the price to value ratio from Intel & Nvidia has not been so good with their latest releases. No one beating down the door this week looking for these or the RTXs. I'm hoping this isn't another bad launch.