Recently, Intel launched a new high-end platform, the X299, that uses the new LGA2066 socket. The new CPUs for this platform are codenamed Kaby Lake-X and Skylake-X, and so far the highest-end model is the Core i9-7900X, which has 10 physical cores, 20 threads, 3.3 GHz base clock, 4.5 GHz Turbo clock, and supports quad-channel DDR4 memory. Let’s test it and see how fast is it.
The X299 platform comes to replace the previous HEDT (high-end desktop) platform, X99. The main feature of this platform is, besides the support for high-end CPUs, the memory access in four channels.
For this platform, Intel launched not only high-end CPUs, with six or more cores, but also two four-core CPUs, one Core i7 and one Core i5. These two processors are based on the “Kaby Lake X” architecture, while the other ones are based on “Skylake-X” architecture. Besides having less cores, these two basic moldels have only 16 PCI Express 3.0 lanes, and access memory with only two channels.
The table below shows the CPUs launched (so far) for this new platform. Intel already annouced a few more models to be launched briefly.

Model Base Clock Turbo Boost Clock Cores/Threads L3 Cache PCI Express lanes TDP Memory channels MSRP
Core i9-7900X 3.3 GHz 4.3 GHz 10/20 13.75 MiB 44 140 W 4 USD 999
Core i7-7820X 3.6 GHz 4.3 GHz 8/16 11 MiB 28 140 W 4 USD 599
Core i7-7800X 3.5 GHz 4.0 GHz 6/12 8.25 MiB 28 140 W 4 USD 389
Core i7-7740X 4.3 GHz 4.5 GHz 4/8 8 MiB 16 112 W 2 USD 339
Core i5-7640X 4.0 GHz 4.2 GHz 4/4 8 MiB 16 112 W 2 USD 242

The Core i9-7900X CPU also supports a 4.5 GHz Turbo Boost 3.0 Clock when running on up to two cores.
Figure 1 shows the Core i9-7900X CPU.

Core i9-7900XFigure 1: the Core i9-7900X

In Figure 2 we have the underside of the CPU.

Core i9-7900XFigure 2: underside of the Core i9-7900X

For our benchmarks, we compared the Core i9-7900X to its predecessor, the Core i7-6950X (read the review here), which has the same number of cores. We also included the highest-end CPU from AMD we have in our lab, the Ryzen 7 1700X (review here). Please mind that we included this AMD CPU as a curiosity, since the Core i9-7900X costs about three times more than the Ryzen 7 1700X, so they are not direct competitors. AMD is about to launch a new series called Ryzen Threadripper, that will be direct competitors to the tested processor.
We used a GeForce GTX 1080 video card on all tests.
Let’s compare the main specs of the reviewed CPUs in the next page.
[nextpage title=”The Reviewed CPUs”]
In the tables below, we compare the main features of the CPUs included in our review.

CPU Cores HT/SMT IGP Internal Clock Turbo Clock Core Tech. TDP Socket Price
Core i9-7900X 10 Yes No 3.3 GHz 4.3 GHz Skylake-X 14 nm 140 W LGA2066 USD 1050
Core i7-6950X 10 Yes No 3.0 GHz 3.5 GHz Broadwell-E 14 nm 140 W LGA2011-v3 USD 1650
Ryzen 7 1700X 8 Yes No 3.4 GHz 3.8 GHz Sumit Ridge 14 nm 95 W AM4 USD 400

Below you can see the memory configuration for each CPU.

CPU L2 Cache L3 Cache Memory Support Memory Channels
Core i9-7900X 10 x 1 MiB 13.75 MiB Até DDR4-2667 4
Core i7-6950X 10 x 256 kiB 25 MiB Até DDR4-2400 4
Ryzen 7 1700X 8 x 512 kiB 2 x 8 MiB Up to DDR4-2667 2

[nextpage title=”How We Tested”]During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions, the only variable devicand was the CPU being tested, besides the motherboard and memory, which had to be replaced to match the different CPUs.
Hardware Configuration

Operating System Configuration

  • Windows 10 Home 64-bit
  • NTFS
  • Video resolution: 1920 x 1080 60 Hz

Driver Versions

  • NVIDIA driver version: 384.94

Software Used

Error Margin
We adopted a 3% error margin. Thus, differences below 3% cannot be considered relevant. In other words, products with a performance difference below 3% should be considered as having similar performance.
[nextpage title=”PCMark 10 and 3DMark”]

PCMark 10

PCMark 10 is a benchmarking software that uses real-world applications to measure the computer performance. We ran the “Extended” test, which included applications opening, web browsing, writing, gaming, photo editing, video chat, video conversion, and rendering. Let’s see the results.
Core i9-7900X
On the 3DMark 10 Extended benchmark, the Core i9-7900X was 4% faster than the Core i7-6950X.

3DMark

3DMark is a program with a set of several 3D benchmarks. Time Spy runs a DirecX12 simulation; Fire Strike runs a “heavy” DirectX 11 simulation, and Sky Diver also measures DirectX 11 performance, but is aimed on average computers.

Core i9-7900X

On Time Spy, the Core i9-7900X was 4% faster than the Core i7-6950X.

Core i9-7900X

On the Fire Strike benchmark, the Core i9-7900X performed similarly to the Core i7-6950X.

Core i9-7900X

On the Sky Diver benchmark, the Core i9-7900X was 4% faster than the Core i7-6950X.

[nextpage title=”Performance in programs”]

Cinebench R15

Cinebench R15 is based on the Cinema 4D software. It is very useful to measure the performance gain obtained by the presence of several processing cores while rendering heavy 3D images. Rendering is an area where a bigger number of cores helps a lot, because usually this kind of software recognize several processors (Cinebench R15, for example, can use up to 256 processing cores).
We ran the CPU benchmark, which renders a complex image using all the processing cores (real and virtual) to speed up the process. The result is given as a score.
Core i9-7900X
On Cinebench R15 CPU benchmark, the Core i9-7900X was 22% faster than the Core i7-6950X.

Blender

Blender is a image and movie redering software that uses all the threads of the CPU. We used the program to render a heavy image of a project named Gooseberry Benchmark. The graph below shows the time the CPU used to finish the image, so the less, the better.
Core i9-7900X
On Blender, the Core i9-7900X was 24% faster than the Core i7-6950X.

CPU-Z

On its current version, the well-known hardware identification software CPU-Z comes with a benchmarking tool, which measures CPU performance for one core and for all available cores.
Core i9-7900X
On the single thread benchmark, the Core i9-7900X was 10% faster than the Core i7-6950X.
Core i9-7900X
On the multiple thread benchmark, the Core i9-7900X was similar to the Core i7-6950X.

Handbrake

Media Espresso is a video conversion program that uses the graphics processing unit of the video engine to speed up the conversion process. We converted a 1 GiB, 1920x1080i, 23,738 kbps, .mov video file to a smaller 320×200, H.264, .MP4 file for viewing on a smartphone. The results below are given in seconds, so the lower the better.

Core i9-7900X

Here the Core i9-7900X was 36% faster than the Core i7-6950X.

WinRAR

Another task where the CPU is very demanded is on file compacting. We ran a test compacting a folder with 8 GiB on 6.813 files to a file, using WinRAR 4.2. The graph below shows the time taken on each test.
Core i9-7900X
On WinRAR, the Core i9-7900X was 21% slower than the Core i7-6950X.
[nextpage title=”Gaming Performance”]

Deus Ex: Mankind Divided

Deus Ex: Mankind Divided is an action RPG with FPS elements, launched in August 2016, that uses the Dawn engine, being compatible with DirectX 12. We tested it using the benchmark included in the game, with DirectX 12 enabled, Full HD, and graphic options as “high”.
The results below are expressed in frames per second.
Core i9-7900X
On this game, the Core i9-7900X was similar to the Core i7-6950X.

Grand Theft Auto V

Grand Theft Auto V, or simply GTA V, is an open-world action game released for PCs in April of 2015, using the RAGE engine. In order to measure the performance on this game, we ran the performance test of the game, measuring the framerate with FRAPS. We ran GTA V at Full HD, with all image quality set as “high” and MSAA off.

The results below are expressed in frames per second.

Core i9-7900X

On GTA V, the Core i9-7900X was 9% slower than the Core i7-6950X.

Hitman

Hitman is an action/stealth game, launched in March 2016, that uses a DirectX 12 compatible version of the Glacier 2 engine. To measure performance in this game, we ran the benchmark in it, measuring the framerate with FRAPS. We ran this game with DirectX 12 enabled, with image quality set as “high”.
The results below, in Full HD and 4K, are expressed in frames per second.
Core i9-7900X
On Hitman, the Core i9-7900X had a tie with the Core i7-6950X.

Rise of the Tomb Rider

Rise of the Tomb Rider is an adventure/action game launched in January of 2016, based on Foundation engine. In order to measure the performance using this game, we ran the benchmark included on it, using Full HD resolution and graphics quality set to “medium”.
The results below are expressed in frames per second.

Core i9-7900X

Also on Rise of the Tomb Rider, the Core i9-7900X was 10% slower than the Core i7-6950X.

Tom Clancy’s Rainbow Six Siege

Tom Clancy’s Rainbow Six Siege is a tactical FPS game launched in December of 2015, based on AnvilNext engine. In order to measure the performance using this game, we ran the benchmark included on it, using Full HD resolution and graphics quality set to “high”.
The results below are expressed in frames per second.

Core i9-7900X

On this game, the Core i9-7900X also performed similarly to the Core i7-6950X.

[nextpage title=”Overclocking”]
All the LGA2066 processors have unlocked clock multiplier, which means you can overclocking it just by changing its multiplier, as long as the motherboard has this feature.
We were able to configure the CPU to run stable at 4.7 GHz (100 MHz reference clock and x47 multiplier), with the original voltages. It may be possible to reach higher frequencies if you “play” with the available adjusts, as long as you have a good power supply, motherboard, and cooling system.
It is also good to keep in mind that the overclock capability depends on pure luck, since two CPUs of same model can reach different maximum clocks.
[nextpage title=”Conclusions”]
The Core i9-7900X is the very first Intel CPU to use the Core i9 brand, letting clear that it is a higher-end lineup than the Core i7 models.
If we compare the highest-end CPU from the previous generation, the Core i7-6950X, we see some differences. Besides keeping the same number of cores (and threads), the first thing to draw the attention is the higher clock rate and the better overclocking potential, since the new model reached almost 1 GHz more than the older model.
You can also notice the reduction of the L3 cache, which looks weird until you see that the L2 cache has been multiplied by four, being now 1 MiB per core. This change in the cache topography may be the reason why it was way faster in some applications, but a little slower in some others.
You can also notice the price difference: while the Core i7-6950X costs around USD 1700 since it was launched (it was the highest-end model of its lineup), the Core i9-7900X arrives with the MSRP of USD 1000, which makes sense since Intel already announced new higher-end models to be launched on the next days, obviously costing more.
In order to analyze the performance of the Core i9-7900X, let’s do it one thing at a time. First, is it good for gaming? It is obviously not bad at all, but it is clear that it is no the focus of this CPU. And to be honest, buying a CPU that costs more than USD 400 for gaming is never the better choice.
On the other hand, the Core i9-7900X is focused in tasks that need a big multithread performance, which tipically occours in video processing and encoding, as well as in image and video rendering. In this tasks, this CPU is extremely fast, being superior to its predecessor in Blender, Cinebench and Handbrake, that are software that use all the computational power of every core and thread.
So, if you need a really fast CPU for this kind of work, and the investment will bring more productivity for your work, the Core i9-7900X is an excellent choice.
Now we have to wait the arrival of the new Core i9 models from Intel, with even more cores yet, as well as of the Ryzen Threadripper models from AMD, to see how the high-end CPU scenario will be from now on.