The Intel Skylake-X Review: Core i9 7900X, i7 7820X and i7 7800X Testedby Ian Cutress on June 19, 2017 9:01 AM EST
Building a PC is an experience worth having. Finding out what works with what and putting it all together is an experience, and the first time always gives a sense of achievement and accomplishment. In the high-end desktop space, even more so: trying not to break your first $500+ CPU can be akin to feeling like a surgeon. Following the HEDT space can be a journey, especially after five years of iterative updates we are seeing something special happening in 2017. Intel, the incumbent, is changing its 2017 strategy in a few ways which means there is more to understand, and in the future, and also means it is set to release 18-core consumer processors. Today is the first part of that release, with the new Skylake-X processors: the Core i9-7900X, the Core i7-7820X and the Core i7-7800X. These throw a few curveballs into the mix which are worth walking into.
Back at Computex, Intel announced its new high-end desktop (HEDT) platform code named Basin Falls. There are three elements to Basin Falls: the new X299, chipset which brings a number of enhancements from the more mainstream consumer line, a set of Skylake-X processors based on the Skylake-SP microarchitecture (and slightly different from the mainstream Skylake-S processors launched in 2015) from 6 cores to 18 cores, and a pair of Kaby Lake-X processors, reusing the Kaby Lake-S silicon launched earlier this year but with a bit more frequency and power consumption.
We covered all three parts in our launch pieces (linked below), but here is a quick recap.
Intel Announces Basin Falls: The New High-End Desktop Platform and X299 Chipset
Intel Announces Skylake-X: Bringing 18-Core HCC Silicon to Consumers for $1999
Intel Announces Kaby Lake-X Processors: High-End Desktop Getting the Latest Microarchitecture
Skylake-X Today: Three CPUs, Up to 10 Cores, Lower Pricing than Broadwell-E
One of the big criticisms from the launch of Broadwell-E, the platform before Skylake-X, was that Intel significantly increased prices over the preceding generation. Nominally consumers were used to the fact that the top chip in the stack was $999 to $1099, and rather than readjust the 6, 8 and 10 core parts to do so, Broadwell-E pushed the 10-core Core i7-6950X to ~$1721, in order to claim parity with the enterprise processors. Users who scoffed at that price will be pleased to hear that the Skylake-X processors launching today are priced more in line with Intel's traditional HEDT chips.
|Skylake-X Processors (Low Core Count Chips)|
|Core i7-7800X||Core i7-7820X||Core i9-7900X|
|Cores / Threads||6/12||8/16||10/20|
|Base Clock||3.5 GHz||3.6 GHz||3.3 GHz|
|Turbo Clock||4.0 GHz||4.3 GHz||4.3 GHz|
|TurboMax Clock||N/A||4.5 GHz||4.5 GHz|
|L3||8.25 MB||11 MB||13.75 MB|
The Core i9-7900X is the 10-core processor and set to have a tray price of $999, which means the shelf price will be around $1049-$1099. The Core i9-7900X is also the only one with a full complement of 44 PCIe lanes from the processor, as Intel is restricting this feature to the higher cost parts of the stack in order to further differentiate it from the cheaper i7-7800 series. This processor is quite aggressive in its specifications – normally the top LCC processor is clocked down, using the extra cores as the main reason for purchase, but here Intel is supporting a maximum turbo frequency of 4.3 GHz, or up to 4.5 GHz with the new-ish favored core mode. The base frequency is lower than the other CPUs to compensate, at 3.3 GHz, but we saw an all-core frequency at 4.0 GHz in our testing. This all comes in a 140W TDP, with support up to DDR4-2666 in quad channel mode at 1 DIMM per channel (DDR4-2400 at 2DPC).
The Core i7-7820X sits below, exchanging two of the CPU cores and some PCIe lanes for a higher base frequency. This is an eight-core part, which reduces the total cache as well, but has a base frequency of 3.6 GHz to compensate. The PCIe lanes are reduced from 44 to 28 as part of Intel’s product segmentation strategy, although Intel notes this is still enough for a single GPU and three PCIe 3.0 x4 devices (such as storage) directly attached to the processor. DRAM support is the same as the Core i9 at DDR4-2666, as is also the TDP at 140W. Coming in at $599 represents a bit of a step from the 10-core down to the 8-core. This price will arguably draw a good number of users who would rather direct that $400 to something else in their system with the only deficit being two cores and a proper x16/x16 layout for dual video cards.
The bottom processor is the Core i7-7800X, with six cores, a lower frequency (3.5 GHz base, 4.0 GHz turbo, no favored core support), 28 PCIe lanes, and support only up to DDR4-2400. While categorically ‘the runt’ of the litter, Intel has priced this part at $389, only $50 more than the mainstream consumer quad-core overclocking processors. Along with a slight cost increase for the new X299 motherboards, users looking to move from the mainstream to the bottom rung of the high-end desktop can do so with around $120-$150 extra for CPU+motherboard, and then a bit more if the users wants to use quad channel memory (these processors work in dual channel as well). As the low-end part, traditionally users may point to it as a potential overclocking processor – we unfortunately did not have time to test our sample for overclocking, but I wouldn’t be surprised to see it hit the same frequencies as the other two processors fairly easily.
Competition and Market
In previous years, Intel only had one serious angle of competition for the HEDT platform: itself. As it introduced new generation parts, the older generation could (sometimes) be picked up cheaper, making the older generation more enticing. But as it stands, for the first time in a long time, Intel has three main sources of competition for the X299 platform.
First is as mentioned, Intel’s older platform in Broadwell-E (or even Haswell-E). Broadwell-E, aside from the steep pricing, is now a solid platform, with extensive motherboard support and well-known functionality. Intel has not officially dropped the prices of the older processors, which unless the retailers decide to shift stock wholesale might discourage users going for brand new CPUs, but the ecosystem for BDW-E is mature.
|Comparison: Core i9-7900X vs.
i7-6950X vs i7-5960X
|Features||Intel Core i9-7900X||Intel Core i7-6950X||Intel Core i7-5960X|
|Cores/Threads||10 / 20||10 / 20||8 / 16|
|Base/Turbo||3.3 / 4.3 GHz||3.0 / 3.5 GHz||3.0 / 3.6 GHz|
|PCIe 3.0 Lanes||44||40||40|
|L2 Cache/core||1 MB||256 KB||256 KB|
|L3 Cache||13.75 MB||25 MB||20 MB|
|TDP||140 W||140 W||140 W|
Second is AMD’s recently released Ryzen set of CPUs. As the first new serious entry into the HEDT space for AMD in almost five years, along with a new x86 core, AMD offered similar-ish performance to Broadwell-E in many aspects (within a few percent), but at half the price or better. The Ryzen CPUs still had other limits – dual channel memory, only sixteen PCIe lanes, limited IO from the chipset, and a miasma surrounding the launch with motherboards and memory support; but several months on the ecosystem for Ryzen is fairly solid and still a new and exciting prospect for the HEDT crowd. For $60 less than the price of the Core i7-7800X, a six-core Intel processor, AMD offers the Ryzen 7 1700, which runs in at 8 cores at slightly lower frequency, with cheaper motherboards as well.
|Comparison: Core i7-7800X vs. Ryzen 7 1700|
Ryzen 7 1700
|6 / 12||Cores/Threads||8 / 16|
|3.5 / 4.0 GHz||Base/Turbo||3.0 / 3.7 GHz|
|28||PCIe 3.0 Lanes||16|
|8.25 MB||L3 Cache||16 MB|
|140 W||TDP||65 W|
Both Intel and AMD would argue the merits of the platforms vs. cost, but Intel still this as an angle for competition. Similarly, the 6-core Core i7-7820X at $599 goes up against the 8-core $499 Ryzen 7 1800X.
|Comparison: Core i7-7820X vs. Ryzen 7 1800X|
Ryzen 7 1800X
|8 / 16||Cores/Threads||8 / 16|
|3.6 / 4.3GHz
(4.5 GHz TMax)
|Base/Turbo||3.6 / 4.0 GHz|
|28||PCIe 3.0 Lanes||16|
|11 MB||L3 Cache||16 MB|
|140 W||TDP||95 W|
Third is AMD’s future HEDT (or Super High-End Desktop, SHED) platform. Announced in the run up to Computex, AMD is set to offer a new X399 chipset with ThreadRipper processors up to 16-core and with 60 PCIe lanes for graphics (plus four for the chipset, which is why AMD keeps saying 64). AMD is using two of its consumer Zeppelin silicon dies on the same package to get up to 16-cores, and has a new socket planned due to the increase in PCIe support. Having had Intel moving from 6 to 8 to 10 cores on the HEDT space over the last 5-6 years meant that AMD announcing a 16-core part was a big enough jolt to this market such that Intel has to provide a response. Pitting a 10-core CPU such as the Core i9-7900X against a 16-core AMD CPU means that AMD would win most the high-throughput heavily-threaded benchmarks, despite perhaps being lower frequency and higher power. For the HEDT/SHED crowd, sometimes it’s all about how quick you can compute. AMD is set to launch the X399 platform and ThreadRipper CPUs in the summer, so we’re still light on details until then, but given AMD's disruptive nature, there's a real risk to Intel that AMD will once again undercut Intel’s equivalent 16-core pricing by a large amount.
|Comparison: Core i9-7900X vs.
i7-6950X vs i7-5960X
|Features||AMD Ryzen 7 1800X||Intel Core i9-7900X||AMD ThreadRipper|
|Cores/Threads||8 / 16||10 / 20||16 / 32|
|Base/Turbo||3.6 / 4.0 GHz||3.3 / 4.3 GHz||?|
|PCIe 3.0 Lanes||16||44||60|
|L2 Cache/core||512 KB||1 MB||512 KB|
|L3 Cache||16 MB||13.75 MB||32 MB|
|TDP||95W||140 W||>140W ?|
What Is This Review
This review comes in two big meaty chunks to sink your teeth into. The first part is discussing the new Skylake-X processors, from silicon to design and covering some of the microarchitecture features, such as AVX-512-F support and cache structure. As mentioned, Skylake-X has some significantly different functionality to the Skylake-S core, which has an impact on how software should be written to take advantage of the new features.
The second part is our testing and results. We were lucky enough to source all three Skylake-X processors for this review, and have been running some regression testing of the older processors on our new 2017 testing suite. There have been some hiccups along the way though, and we’ll point them out as we go.
An extra morsel to run after is our IPC testing. We spend some time to run tests on Skylake-S and Skylake-X to see which benchmarks benefit from the new microarchitecture design, and if it really does mean anything to consumers at this stage.
It should be noted that Intel did not provide all the CPUs for this review. As with standard Intel policy, only the top CPU came direct from Intel. All CPUs in this review are classified as engineering samples for partner testing and not for resale, although they are identical to retail CPUs in performance.
Post Your CommentPlease log in or sign up to comment.
View All Comments
mat9v - Tuesday, June 20, 2017 - linkTo play it safe, invest in the Core i9-7900X today.
To play it safe and get a big GPU, save $400 and invest in the Core i7-7820X today.
Then the conclusion should have been - wait for fixed platform. I'm not even suggesting choosing Ryzen as it performs slower but encouraging buying flawed (for now) platform?
mat9v - Tuesday, June 20, 2017 - linkPlease then correct tables on 1st page comparing Ryzen and 7820X and 7800X to state that Intel has 24 lines as they leave 24 for PCIEx slots and 4 is reserved for DMI 3.0
If you strip Ryzen lines to only show those available for PCIEx do so for Intel too.
Ryan Smith - Wednesday, June 21, 2017 - linkThe tables are correct. The i7 7800 series have 28 PCIe lanes from the CPU for general use, and another 4 DMI lanes for the chipset.
PeterCordes - Tuesday, June 20, 2017 - linkNice article, thanks for the details on the microarchitectural changes, especially to execution units and cache. This explains memory bandwidth vs. working-set size results I observed a couple months ago on Google Compute Engine's Skylake-Xeon VMs with ~55MB of L3: The L2-L3 transition was well beyond 256kB. I had assumed Intel wouldn't use a different L3 cache design for SKX vs. SKL, but large L2 doesn't make much sense with an inclusive L3 of 2 or 2.5MB per core.
Anyway, some corrections for page3: The allocation queue (IDQ) is in Skylake-S is always 64 uops, with or without HT. For example, I looked at the `lsd.uops` performance counter in a loop with 97 uops on my i7-6700k. For 97 billion counts of uops_issued.any, I got exactly 0 counts of lsd.uops, with the system otherwise idle. (And I looked at cpu_clk_unhalted.one_thread_active to make sure it was really operating in non-HT mode the majority of the time it was executing.) Also, IIRC, Intel's optimization manual explicitly states that the IDQ is always 64 entries in Skylake.
The scheduler (aka RS or Reservation Station) is 97 unfused-domain uops in Skylake, up from 60 in Haswell. The 180int / 168fp numbers you give are the int / fp register-file sizes. They are sized more like the ROB (224 fused-domain uops, up from 192 in Haswell), not the scheduler, since like the ROB, they have to hold onto values until retirement, not just until execution. See also http://blog.stuffedcow.net/2013/05/measuring-rob-c... for when the PRF size vs. the ROB is the limit on the out-of-order window. See also http://www.realworldtech.com/haswell-cpu/6/ for a nice block diagram of the whole pipeline.
SKL-S DIVPS *latency* is 11 cycles, not 3. The *throughput* is one per 3 cycles for 128-bit vectors, or one per 5 cycles for 256b vectors, according to Agner Fog's table. I forget if I've tested that myself. So are you saying that SKL-SP has one per 5 cycle throughput for 128-bit vectors? What's the throughput for 256b and 512b vectors?
It's really confusing the way you keep saying "AVX unit" or "AVX-512 unit" when I think you mean "512b FMA unit". It sounds like vector-integer, shuffle, and pretty much everything other than FMA will have true 512b execution units. If that's correct, then video codecs like x264/x265 should run the same on LCC vs. HCC silicon (other than differences in mesh interconnect latency), because they're integer-only, not using any vector-FP multiply/add/FMA.
> This should allow programmers to separate control flow from data flow...
SIMD conditional operations without AVX512 are already done branchlessly (I think that's what you mean by separate from control-flow) by masking the input and/or output. e.g. to conditionally add some elements of a vector, AND the input with a vector of all-one or all-zero elements (as produced by CMPPS or PGMPEQD, for example). Adding all-zeros is a no-op (the additive identity).
Mask registers and support for doing it as part of another operation makes it much more efficient, potentially making it a win to vectorize things that otherwise wouldn't be. But it's not a new capability; you can do the same thing with boolean vectors and SSE/AVX VPBLENDVPS.
PeterCordes - Tuesday, June 20, 2017 - linkSpeed Shift / Hardware P-State is not Windows-specific, but this article kind of reads as if it is.
Your article doesn't mention any other OSes, so nothing it says is actually wrong: I'm sure it did require Intel's collaboration with MS to get support into Win10. The bullet-point in the image that says "Collaboration between Intel and Microsoft specifically for W10 + Skylake" may be going too far, though. That definitely implies that it only works on Win10, which is incorrect.
Linux has supported it for a while. "HWP enabled" in your kernel log means the kernel has handed off P-state selection to the hardware. (Since Linux is open-source, Intel contributed most of the code for this through the regular channels, like they do for lots of other drivers.)
dmesg | grep intel_pstate
[ 1.040265] intel_pstate: Intel P-state driver initializing
[ 1.040924] intel_pstate: HWP enabled
The hardware exposes a knob that controls the tradeoff between power and performance, called Energy Performance Preference or EPP. Len Brown@Intel's Linux patch notes give a pretty good description of it (and how it's different from a similar knob for controlling turbo usage in previous uarches), as well as describing how to use it from Linux. https://patchwork.kernel.org/patch/9723427/.
# CPU features related to HWP, on an i7-6700k running Linux 4.11 on bare metal
fgrep -m1 flags /proc/cpuinfo | grep -o 'hwp[_a-z]*'
I find the simplest way to see what speed your cores are running is to just `grep MHz /proc/cpuinfo`. (It does accurately reflect the current situation; Linux finds out what the hardware is actually doing).
IDK about OS X support, but I assume Apple has got it sorted out by now, almost 2 years after SKL launch.
Arbie - Wednesday, June 21, 2017 - linkThere are folks for whom every last compute cycle really matters to their job. They have to buy the technical best. If that's Intel, so be it.
For those dealing more with 'want' than 'need', a lot of this debate misses an important fact. The only reason Intel is suddenly vomiting cores, defecating feature sizes, and pre-announcing more lakes than Wisonsin is... AMD. Despite its chronic financial weakness that company has, incredibly, come from waaaay behind and given us real competition again. In this ultra-high stakes investment game, can they do that twice? Maybe not. And Intel has shown us what to expect if they have no competitor. In this limited-supplier market it's not just about who has the hottest product - it's also about whom we should reward with our money, and about keeping vital players in the game.
I suggest - if you can, buy AMD. They have earned our support and it's in our best interests to do so. I've always gone with Intel but have lately come to see this bigger picture. It motivated me to buy an 1800X and I will also buy Vega.
Rabnor - Wednesday, June 21, 2017 - linkTo play it safe and get a big GPU, save $400 and invest in the Core i7-7820X today.
You have to spend that $400+ on a good motherboard & aio cooler.
Are you sold by Intel, anandtech?
Synviks - Thursday, June 22, 2017 - linkFor some extra comparison: running Cinebench R15 on my 14c 2.7ghz Haswell Xeon, with turbo to 3ghz on all cores, my score is 2010.
Pretty impressive performance gain if they can shave off 4 cores and end up with higher performance.
Pri - Thursday, June 22, 2017 - linkOn the first page you wrote this:
Similarly, the 6-core Core i7-7820X at $599 goes up against the 8-core $499 Ryzen 7 1800X.
The Core i7 7820X was mistakenly written as a 6-core processor when it is in-fact an 8-core processor.
Gigabytes - Thursday, June 22, 2017 - linkOkay, here is what I learned from this article. Gaming performance sucks and you will be able to cook a pizza inside your case. Did I miss anything?
Oh, one thing missing.
Play it SMART and wait to see the Ripper in action before buy your new Intel toaster oven.