Intel’s Core i9-7900X CPU reviewed, part one

Intel’s Core X CPUs are here, and we’re kicking off this new era with the highest-end chip in the lineup so far: the Core i9-7900X. As it traditionally does for its high-end desktop platform, the company is repurposing silicon from its upcoming Skylake Xeons to serve as Skylake-X chips. That means some unusually large changes are in store for us enthusiasts as Skylake makes its transition from mainstream desktops to the data center.

The Core X family of CPUs needs a new socket, LGA 2066, and a new platform, called X299. We’ve already covered the Core X lineup and the X299 platform, as well as the entry-level Kaby Lake-X CPUs for that platform, in a dedicated article. The short take is that X299 is an evolutionary step forward from the X99 platform. It keeps that platform’s quad-channel memory support (out of a total of six on at least some Skylake-X dies) and pairs it with a chipset powered by a lot of the same DNA present in the Z270 platform. If you need to brush up on Core X before reading on, feel free.

Now, back to Skylake-X. The fundamental pipeline of this chip isn’t much different from the various Skylake and Kaby Lake desktop parts that we’ve known and loved for almost two years now. We never did a deep dive into the Skylake architecture, but compared to Haswell and Broadwell, the basic Skylake integer pipeline is wider and can have more going on at once. To prevent this wider engine from burning power on execution of the wrong instructions, Skylake also features a better branch predictor than Haswell and Broadwell, according to Intel.

The basic Skylake core as seen in the Core i7-6700K and friends. Source: Intel

That’s a gross oversimplification, of course, but it’s generally how Intel has improved its chips over the past few years. Let’s have a look at the big differences between mainstream and server Skylake CPUs now.

AVX-512 unhinges its jaw

The first big change in Skylake-X is support for the AVX-512 instruction set. These new instructions add important new capabilities to Intel’s SIMD implementation, including scatter-gather support, dedicated state and mask registers, and much, much more. To support this generation of AVX, the chip’s vector data registers are now twice as wide, and there are twice as many of them. These wider registers are fed with more load and store bandwidth. Skylake-X can now handle two 64-byte loads and one 64-byte store per cycle, compared to a single 64B load and a single 32B store per cycle in mainstream Skylake.

On top of its wider and more numerous registers, the Skylake-X core also has a dedicated AVX-512 fused multiply-add unit (FMA) on top of the pair of 256-bit-wide AVX FMA units in Skylake-S. This unit can only handle AVX-512 FMAs, and it resides on port five of the Skylake-X unified scheduler. Since the pair of 256-bit FMAs can execute a single AVX-512 FMA in parallel alongside the dedicated AVX-512 FMA unit, throughput for that common instruction is effectively doubled in the best case compared to mainstream Skylake.

The AVX register structure in Skylake-X CPUs. Source: Intel

While those performance improvements may sound impressive, real-world performance of AVX-512 has caveats. Firing up those monster SIMD units requires large amounts of power (and therefore produces more heat), so Skylake-X CPUs might be forced to clock below Intel’s specified Turbo speeds (or not Turbo at all) when executing AVX-512 instructions. That clock-speed tradeoff might result in lower-than-expected performance from AVX-512 code, and Intel says developers will need to be able to amortize the expected performance gain of their AVX-512 applications over time versus the clock-speed drop their code might incur. Mixed workloads with a small proportion of AVX-512 instructions in the overall mix are apparently not an ideal case for speedups from that SIMD hardware, either.  

Model Base

clock

(GHz)

Turbo

clock

(GHz)

Turbo

Boost

Max 3.0

clock

(GHz)

Cores/

threads

L3

cache

PCIe

3.0

lanes

Memory

support

TDP Socket

Price

(1K

units)

i9-7980XE 18/36 LGA 2066 $1999
i9-7960X 16/32 $1699
i9-7940X 14/28 $1399
i9-7920X 12/24 $1199
i9-7900X 3.3 4.3 4.5 10/20 13.75MB 44 Quad-channel

DDR4-2666

140W $999
i7-7820X 3.6 4.3 4.5 8/16 11MB 28 $599
i7-7800X 3.5 4.0 NA 6/12 8.25MB Quad-channel

DDR4-2400

$389
i7-7740X 4.3 4.5 NA 4/8 8MB 16 Dual-channel

DDR4-2666

112W $339
i5-7640X 4.0 4.2 NA 4/4 6MB $242

What’s more, not every Core X chip in the lineup will enjoy the same boost in SIMD performance from AVX-512. Only the Core i9 series of CPUs will ship with the dedicated AVX-512 FMA. The Core i7-7800X and Core i7-7820X will still have the wider registers for AVX-512, but they’ll only execute instructions using the pair of 256-bit AVX units common to all Skylake chips. This exercise in segmentation might surprise people expecting a uniform performance increase from AVX-512 across all the CPUs that support it. (The Kaby Lake-X Core i5-7640X and Core i7-7740X won’t support AVX-512 at all.)

 Because of those caveats, we may be waiting a while for mainstream desktop applications that can really take advantage of all the extra parallelism on offer from these new instructions. Scientific-computing, deep-learning, and financial-services folks will probably be drooling for AVX-512, but regular Joes and Janes probably won’t see any major speedups until companies recompile their software (at the very least). That assumes AVX-512 is coming to mainstream Intel CPUs, as well.

 

Bigger L2 caches for better performance

Skylake-X also has a much different cache allocation per core compared to its mainstream counterparts. Instead of the relatively small 256kB L2 cache (or mid-level cache, in Intel parlance) in Skylake-S and Broadwell-E, each Skylake-X core enjoys a whopping 1MB of private L2. In support of AVX-512, the bandwidth between the L1 data cache and the L2 cache has been increased to 128 bytes per cycle. On top of that size increase, Intel quadrupled the associativity of the cache from four ways in  to 16 ways in Skylake-X. Intel says the move to a larger private cache lets programmers keep usefully large data structures close to the core, and the result is higher performance. Pretty cut and dry.

Intel says it undertook this change because it felt its older architectures placed too much emphasis on data sharing through the L3 caches. In turn, Skylake-X’s architects reduced the shared last-level cache allocation to as much as 1.375MB per core, compared to as much as 2.5MB per core for Broadwell-E chips. This last-level cache isn’t inclusive of the L2 caches, and it serves as a victim cache for the L2. The tradeoff for this rebalancing of cache allocations is a higher L3 cache access latency, according to Intel.

Getting meshy

Finally, Intel is abandoning the ring topology it’s used to connect CPU cores in its many-core CPUs for several generations. In place of its ring, the company is introducing a new (or at least new outside of the Knights Landing accelerator) mesh interconnect topolgy that promises several improvements. First off, Intel says its mesh interconnect delivers lower latency and higher bandwidth than the ring bus, all while operating at a lower frequency and voltage. Those last two characteristics are important, because they should result in less power consumption from the interconnect portion of the chip as it scales up.

Intel also says that the mesh design also allows it to include units like I/O, inter-socket interconnects, and memory controllers in a modular, scalable way as core counts increase. The company claims the distribution of these elements across the chip using the mesh minimizes undesirable “hot spots” of activity that might ultimately constrain cores’ access to those critical resources, limiting performance.

Mesh architecture conceptual representation. Red lines represent horizontal and vertical wires connecting CPU components; green squares represent switches at intersections. Source: Intel

The mesh design should also offer a boon for applications that need to do a lot of inter-core communication. The last-level cache in Skylake-X is distributed across each core, and thanks to the more uniform access characteristics of the mesh, Intel claims that application developers no longer have to worry about non-uniform latencies when accessing data in those caches. Cores should also enjoy more uniform access characteristics when accessing the die’s I/O and memory controller, as well.

Multiple ring buses, as seen in high-core-count Haswell Xeons.

Previously, the shared L3 caches on a chip might have resided on different rings, requiring cores to communicate across the buffered switches formerly used to join discrete rings on the die. These switches added latency on top of that incurred by traversing the ring bus in the first place—something that Intel gave customers the opportunity to avoid in past chips with a “cluster-on-die” mode that turned each ring into something resembling a NUMA domain of its own. The mesh topology in Skylake-X should make headaches from the non-uniform distribution and access latencies of resources among rings a thing of the past.

As for characteristics of the Skylake-X silicon itself, Intel honchos clammed up when we asked about die size and transistor count. The company believes that disclosing this information will lead to unfounded conclusions from its competitors about the quality of their chip designs and process technologies compared to Intel’s. Only the paranoid survive, we suppose.

 

The Core i9-7900X and LGA 2066 in the flesh
We’ve already touched on the Core X CPU family and the X299 platform in depth, but it’s good to get an up-close look at the Core i9-7900X and Asus Prime X299-Deluxe motherboard that Intel has provided us for testing. 

Clockwise from upper left: Core i7-5960X, Core i7-6950X, Core i9-7900X, Core i7-7700K, Ryzen 5 1700X

Outwardly, little has changed that would help us identify the Core i9-7900X at first glance. The eagle-eyed will note more rounded corners on the edges of the integrated heat spreader, but that’s about it.

From left to right: Core i7-5960X, Core i7-6950X, Core i9-7900X

Flipping these chips over reveals a dense forest of surface-mounted components, but it’s otherwise hard to notice the extra 55 lands on the Core i9-7900X. If you want to count, we’ll wait.

That outward similarity might lead one to believe that LGA 2011 and LGA 2066 chips are interchangeable among X99 and X299 motherboards, and the LGA 2066 socket doesn’t help. The dimensions of the socket are the same as those of LGA 2011, but chips for that socket absolutely will not work with LGA 2066. Don’t make an expensive mistake by eyeballing it. The only things that builders can carry over from LGA 2011 systems are their DDR4 kits and cooling hardware.

Intel sent us home with Asus’ ultra-ritzy Prime X299-Deluxe motherboard to host the Core i9-7900X. This board boasts everything one might want out of a high-end platform: four USB 3.1 Gen 2 ports, built-in 802.11ac and 802.11ad Wi-Fi, tasteful RGB LEDs, and a bevy of PCIe x16 slots. Asus also offers two M.2 slots, one of which allows gumstick SSDs to stand up vertically for better cooling. Even if both M.2 slots are occupied, the horizontal M.2 SSD can still enjoy plenty of heat-dissipation potential thanks to an integrated heatsink for the slot and chipset. This mobo even comes with a Thunderbolt 3 card and an add-on fan control board with several extra headers.

Our testing methods

As always, we did our best to collect clean test numbers. We ran each of our benchmarks at least three times, and we’ve reported the median result. Our test systems were configured like so:

Processor Ryzen 7 1800X
Motherboard Gigabyte Aorus AX370-Gaming 5
Chipset AMD X370
Memory size 16 GB (2 DIMMs)
Memory type G.Skill Trident Z DDR4 SDRAM
Memory speed 3866 MT/s (rated)

3200 MT/s (actual)

Memory timings 15-15-15-35 1T
System drive Intel 750 Series 400GB NVMe SSD

 

Processor Intel Core i7-5960X Intel Core i7-6950X Intel Core i9-7900X

(DDR4-2666)

Intel Core i9-7900X

(DDR4-3200)

Intel Core i7-7700K
Motherboard Gigabyte GA-X99-Designare EX Asus Prime X299-Deluxe Gigabyte Aorus GA-Z270X-Gaming 8
Chipset Intel X99 X299 Z270
Memory size 32GB
64GB
32GB
64GB
16GB
Memory type G.Skill Trident Z

DDR4 SDRAM

G.Skill Trident Z

DDR4 SDRAM

G.Crucial Ballistix Elite

DDR4 SDRAM

G.Skill Trident Z

DDR4 SDRAM

G.Skill Trident Z

DDR4 SDRAM

Memory speed 3600 MT/s (rated)

2666 MT/s (actual)

3200 MT/s
2666 MT/s 3200 MT/s 3866 MT/s (rated)

3200 MT/s (actual)

Memory timings 15-15-15-35 2T
16-18-18-38 2T
16-17-17-37 2T 16-18-18-38 2T 15-15-15-35 2T
System drive Corsair Neutron XT 480GB SATA SSD Samsung 960 EVO 500GB NVMe SSD Samsung 850 Pro 512GB SATA SSD Samsung 960 EVO 500GB NVMe SSD

They all shared the same common elements:

Storage 2x Corsair Neutron XT 480GB SSD

1x Kingston HyperX 480GB SSD

Discrete graphics Nvidia GeForce GTX 1080 Ti Founders Edition
Graphics driver version GeForce 382.33
OS Windows 10 Pro with Creators Update
Power supply Seasonic Prime Titanium 1000W

Thanks to Intel, Corsair, Kingston, Asus, Gigabyte, Cooler Master, G.Skill, and AMD for helping us to outfit our test rigs with some of the finest hardware available.

Some further notes on our testing methods:

  • The test systems’ Windows desktops were set at a resolution of 3840×2160 in 32-bit color. Vertical refresh sync (vsync) was disabled in the graphics driver control panel.

  • For our Ryzen systems, we used the AMD Ryzen Balanced power plan included with the company’s most recent chipset drivers. We left our Intel systems on Windows’ default Balanced power plan.

The tests and methods we employ are usually publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

 

Memory subsystem performance

Since Skylake-X has such a drastically different cache allocation versus its predecessors, we simply had to run it through SiSoft’s bandwidth and latency tests. We’ll start with the Sandra utility’s cache and memory bandwidth test.

To examine the effect of the increased L2 cache size in an intuitive way, we started with the single-threaded test. To interpret these results, remember that the block size identifies which level of cache the test is exercising. From 2KB to 32KB on all of these CPUs, we’re in the L1 cache. Past that point, we’re generally in L2 for most of these CPUs out to 256KB for Haswell, Broadwell, and Kaby Lake, 512KB for Zen, and 1MB for Skylake-X. Finally, we spill over into L3. 

The most interesting result in the chart above unsurprisingly ocurs between the 512KB mark and 1MB, where we’re testing most of these chips’ L2 caches. Thanks to their 1MB L2s, each Skylake-X core can sustain much higher bandwidths at the 1MB block size than the competition. Save for the Zen core, which is still in its L2 at 512KB, all of the older Intel chips are already into their L3 caches.

Sandra’s multithreaded cache and memory bandwidth test shows the combined bandwidth from every core on each chip. In turn, we have to account for the combined size of each cache level on each chip when mapping block sizes to the cache that we’re in at a given size. That’s because the same source block is being split among multiple cores. If that sounds complicated, it is, and there are a lot of moving parts to account for here between core counts and clock speeds. Still, this measurement is a good way of showing just how much data these chips can move around in total.

Because Sandra supports AVX-512 instructions, and because Skylake-X doubles the L1 cache load bandwidth available per core, we get much more throughput compared to older Intel CPUs at most block sizes. Since the test is still hitting the i9-7900X’s L1 cache from 64KB out to 256KB, for example, the increased load bandwidth for that cache lets the 7900X turn in incredible throughput increases across the chip. That trend continues as the test moves into the 7900X’s L2 cache, where it can move roughly 50% more data compared to the Core i7-6950X, itself no slouch. All told, the i9-7900X sets a high new bar for cache throughput—not an easy task in this company.

Next, let’s look at some tests of main memory bandwidth from the popular AIDA64 utility.

Interesting. Though the i9-7900X takes a commanding lead over the i7-6950X in AIDA64’s read tests with both DDR4-2666 and DDR4-3200 RAM, the Skylake-X chip falls slightly behind Broadwell-E in write bandwidth, and it needs DDR4-3200 to match the i7-6950X in the copy test. None of these chips are slouches, to be sure, but the results for the i9-7900X aren’t quite as eye-popping as they were in Sandra’s cache tests. Time to talk latency.

Sandra also offers a detailed benchmark for cache and memory latencies. To reduce the effect of prefetching on this result, we use the “in-page random” access pattern. Like the single-threaded cache bandwidth test above, this test isn’t multithreaded, so it’s easy to keep track of what cache is being measured at each block size. 

As Intel suggested it would, the L2-L3 rebalancing on Skylake-X offers more bandwidth from its L2 caches at the same latency as before, at the cost of slightly higher latencies in the L3. Seems like a fine tradeoff to us, given the apparent benefits of the larger L2.

Compared to past Intel architectures, Skylake-X lags in main-memory access latencies. That increase may be offset somewhat by the potentially much higher hit rate in the L2 cache, though.

Some quick synthetic math tests

AIDA64 offers a useful set of built-in directed benchmarks for assessing the performance of the various subsystems of a CPU. The PhotoWorxx benchmark uses AVX2 on compatible CPUs, while the FPU Julia and Mandel tests use AVX2 with FMA.

 

PhotoWorxx doesn’t seem to benefit from the improvements in the Core i9-7900X, but the Skylake-X CPU enjoys a commanding lead in our tests of floating-point prowess. At least in these synthetics, the i9-7900X seems to justify its more-than-twice-the-price tag of the Ryzen 1800X. Let’s see if this strong opening translates into real-world performance now.

 

Javascript performance

The usefulness of Javascript benchmarks for comparing browser performance may be on the wane, but these collections of tests are still a fine way of demonstrating the real-world single-threaded performance differences among CPUs. These tests may be an ideal stage for Skylake-X’s improved Turbo Boost Max 3.0 implementation.

Turbo Boost Max 3.0 doesn’t go all the way toward eliminating the single-threaded performance gap we tend to see between mainstream and high-end Intel chips using the same architecture, but the feature does let the i9-7900X get most of the way there—and in Kraken, the Skylake-X chip does catch the i7-7700K. That’s good news for folks who like a responsive machine in both lightly-threaded and all-out workloads.

Compiling code in GCC

Our resident code monkey, Bruno Ferreira, helped us put together this code-compiling test. Qtbench records the time needed to compile the Qt SDK using the GCC compilers. The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.

Bang. The i9-7900X neatly shaves 20% off the compile time of our previous champion, the Core i7-6950X, and it does it for hundreds of dollars less.

7-Zip file compression

Put the i9-7900X to work zipping and unzipping archives, and it delivers a small-to-moderate boost over its Broadwell predecessor. Memory speeds don’t seem to make the least bit of difference to the 7900X’s performance, though.

VeraCrypt disk encryption

VeraCrypt is a continuation of the handy TrueCrypt project. This disk-encryption utility lets us test throughput for both the hardware-accelerated AES algorithm and the implemented-in-software Twofish cipher.

The AES half of our VeraCrypt testing does appear to be memory-bound for once, so DDR4-3200 allows for a whopping 2.4 GB/s higher throughput. Twofish doesn’t enjoy any such speedup from faster memory, although the i9-7900X still handily pulls away from the i7-6950X.

 

Cinebench

The Cinebench benchmark is powered by Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. The test runs with a single thread and then with as many threads as possible.

In Cinebench’s single-threaded test, the i9-7900X once again proves its Turbo Boost mettle against the Core i7-7700K. Core for core, the Skylake-X chip also hands in a roughly 16% improvement over the Core i7-6950X. Pretty darn good considering the low-single-digit performance improvements we’ve come to expect from Intel’s generation-to-generation advances.

Blender

Blender is a widely-used open-source 3D modeling and rendering application. The app can take advantage of AVX2 instructions on compatible CPUs. We chose the “bmw27” test file from Blender’s selection of benchmark scenes to put our CPUs through their paces.

The i9-7900X carves up our test file in fine form with a 12.5% decrease in render time over its predecessor. The i7-6950X is probably looking at its original suggested price tag and sweating a bit by now. Once again, faster memory makes no difference in performance for the i9-7900X.

Handbrake video transcoding

Handbrake is a popular video-transcoding app that recently hit version 1.0. To see how it performs on these chips, we’re switching things up from past reviews. Here, we converted a roughly two-minute 4K source file from an iPhone 6S into a 1920×1080, 30 FPS MKV using the HEVC algorithm implemented in the x265 open-source encoder. We otherwise left the preset at its default settings.

Noticing a pattern yet? The i9-7900X completes this job in 14% less time than the i7-6950X needs, and its performance doesn’t vary with memory speed.

 

picCOLOR

It’s been a while since we tested CPUs with picCOLOR, but we now have the latest version of this image-analysis tool in our hands courtesy of Dr. Reinert H.G. Mueller of the FIBUS research institute. This isn’t Photoshop; picCOLOR’s image analysis capabilities can be used for scientific applications like particle flow analysis. In its current form, picCOLOR supports AVX2 instructions, multi-core CPUs, and simultaneous multithreading, so it’s an ideal match for the CPUs on our bench. Check out FIBUS’ page for more information about the institute’s work and picCOLOR.

picCOLOR offers an interesting shake-up in our results for once. Although the i9-7900X still totally outruns every other chip in our stable, this time it’s trailed by the Ryzen 7 1800X instead of the Core i7-6950X. Faster memory still doesn’t do squat for the 7900X, though.

STARS Euler3D

Euler3D tackles the difficult problem of simulating fluid dynamics. It tends to be very memory-bandwidth intensive. You can read more about it right here. We configured Euler3D to use every thread available from each of our CPUs.

Interesting. Euler3D seems to lean on the memory subsystem in ways that are particularly amenable to high performance with the i7-6950X and not so much so for the i9-7900X. Not every workload benefits from Intel’s rebalanced cache hierarchy with Skylake-X, it seems.

Digital audio workstation performance

DAWBench is a popular addition to our CPU test suite, and we’re now working directly with the creator of DAWBench, Vin Curigliano, to refine our testing methods. As part of that collaborative effort, Vin provided us with a beta version of the DAWBench DSP 2017 benchmark. We’re leaving DAWBench’s virtual instrument test on the bench this time around, however, since these high-powered Intel CPUs tend to max out the benchmark and an updated version of the test isn’t quite ready yet.

DAWBench DSP 2017 relies on the freely-available Shattered Glass Audio SGA1566 VST plugin. We used the 64-bit version of this VST in our testing. DAWBench DSP lets us enable instances of this plugin until the session becomes unresponsive. We used Reaper as our host DAW for the test, and we monitored the project using a Focusrite Scarlett 2i2 interface with the company’s latest USB ASIO drivers. We set a 96 KHz sampling rate and used two ASIO buffer depths: a punishing 64 and a slightly-less-punishing 128.

As we’ve come to expect, the i9-7900X delivers a modest performance improvement in this test versus the Core i7-6950X. Strangely, the 7900X-and-DDR4-3200 pairing actually performs worse than its DDR4-2666-equipped configuration, though. We’ll have to monitor this test and see whether it’s a behavior that changes as the X299 platform matures.

 

A quick look at power consumption and energy efficiency

Skylake-X’s performance improvements wouldn’t be worth much if they came with a corresponding decrease in energy efficiency. We can get a rough idea of whether the Core i9-7900X is as efficient as it is fast by monitoring our test system’s power consumption at the wall with our trusty Watts Up power meter and estimating the total amount of energy it needs to complete a task. Our observations have shown us that Blender consumes about the same amount of power at every stage of the bmw27 benchmark we test with, so it’s an ideal guinea pig for this kind of calculation. First, though, let’s check idle and peak load power consumption numbers.

At idle, the X299 platform paired with the Core i9-7900X sips only a bit more power than the Core i7-7700K. Under load, however, the 7900X system consumes the most power at peak load by a wide margin. Of course, peak power draw only tells part of the efficiency story.

To really get a sense of how efficient the Core i9-7900X is, we need to take the task energy consumed over the course of our Blender benchmark into account. Not only does the Core i9-7900X cut 94 seconds off the Ryzen 7 1800X’s bmw27 render time, it does it while expending only just a bit more power to do so. That’s impressive performance per watt.

 

Conclusions

It’s time once more to sum up our results using our famous scatter plots. To spit out this final index, we take the geometric mean of each chip’s results in our real-world productivity tests, then plot that number against retail pricing gleaned from Newegg. Where Newegg prices aren’t available, we use a chip’s original suggested price.

Our value scatter tells the entire story of the Core i9-7900X: if your workload scales to many threads, this chip is generally the one to run it on. The server version of Skylake delivers an unusually large performance boost for a modern Intel CPU revision in many tasks. Core for core and thread for thread, the already-beastly Core i7-6950X can sometimes lag the 7900X in the range of 10% to 20%. All that oomph comes for a jaw-dropping $724 less than the 6950X’s initial suggested price, too. Competition is a wonderful thing.

In a milestone for Intel’s high-end desktop platform, the Core i9-7900X mostly ends the tradeoff between single-threaded swiftness and multi-threaded grunt typical of some older Intel high-end desktop chips. For lightly-threaded workloads, the i9-7900X’s improved Turbo Boost Max 3.0 behavior lets it trail our single-thread-favorite Core i7-7700K by only a few percentage points at most. In typical desktop use, then, the i9-7900X and its TBM 3.0-enabled brethren should feel about as snappy as their mainstream desktop cousins. I need to get the i9-7900X paired up with a GeForce GTX 1080 Ti soon to see whether that single-threaded performance translates to similar gameplay smoothness.

I’ll also need to explore Skylake-X overclocking in depth soon. Thanks to immature firmware and monitoring utilities, I didn’t feel comfortable pushing my 7900X too hard at this point. That said, I see a lot of promise for overclocking this chip. My test system made it to the Windows desktop without a hiccup at an astounding 4.7 GHz on all cores, and thermal limits seem as though they’ll be the primary obstacle to fully exploiting that speed. I’ll explore Skylake-X’s overclocking potential more once the X299 platform has had a bit more time in the oven (and once I’ve picked up some seriously beefy cooling hardware in the meantime).

We’ve always been loath to recommend the top-end CPU in Intel’s high-end desktop family (and yes, that is this chip for the moment). Despite the Ryzen-inspired price reshuffling that’s coming with Core X, the i9-7900X still isn’t a great value. The star of the Core X lineup may actually be the Core i7-7820X, whose eight cores and 16 threads have clocks similar to those of the i9-7900X. You may lose a couple of cores in the bargain, but even so, the i7-7820X should perform better than a Ryzen 7 1800X for not that much more money. We hope to play with one of these more attainable Skylake-X CPUs soon.

Of course, the performance of the Core i9-7900X is beyond question: it’s the fastest single-socket CPU we’ve ever tested. The X299 platform may need a little polish yet to let Core X chips really shine, but the performance bar the i9-7900X is already setting promises an exciting standoff this summer as AMD prepares its Ryzen Threadripper CPUs for launch. If you need as many cores and threads as possible from your desktop, times have never been more exciting. Stay tuned as we see whether the i9-7900X has got game.

0 0 vote
Article Rating
167 Comments
Newest
Oldest Most Voted
Inline Feedbacks
View all comments
BIF
BIF
2 years ago

I still want to see useful F@H performance for CPUs and GPUs.

*** !HINT! !HINT! ***

Just sayin’.

kamikaziechameleon
kamikaziechameleon
3 years ago

finally more than stinking 6 cores for a consumer desktop. My cell phone had 8 core a number of years past. Intel has been reluctant to broaden their consumer CPU thread count. But its happening and its glorious.

Klimax
Klimax
3 years ago

Quick note: x264 in recent builds should support AVX-512.

psuedonymous
psuedonymous
3 years ago

With the single-thread performance being barely behind the flagship 115x part, this is the first time I’ve considered moving to the HEDT platform. Previously the loss of SP performance has not been worth the few cases where extra cores would have been worth it.

Coupled with the jam-packed ASRock X299 ITX board, all I need now is a suitable (low-ish profile) heatsink that can fit between the risers, but beefier than the venerable NH-L9i to handle the extra ~60W of heat.

NTMBK
NTMBK
3 years ago
Reply to  psuedonymous

Given the insane heat that Tom’s Hardware found in their tests, I wouldn’t try to cram the 7900X into a mini-ITX build.

setaG_lliB
setaG_lliB
3 years ago

Yeah. Looks good and all, but considering how well Ryzen does in some of these benchmarks, I just can’t wait to see what Threadripper will do with quad channel memory.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  setaG_lliB

Mostly I think the quad channel memory enables the same level of performance per core. Not much else to say, is there?

ronch
ronch
3 years ago

I just kept looking at how the Ryzen 1800X puts up against all the Intel chips. Still a damn good value in my book as shown by the famous TR graph! Holy cow I wouldn’t think twice about getting it if I were out for a high core count chipolata. I might sound like a shill right now (which I swear by my FX-8350 I’m not) but hey, I’m sure you all agree it’s a great choice for $460.

ronch
ronch
3 years ago

Ok, I put this table together just for comparison:

[url<]http://tinypic.com/r/2zfn7k0/9[/url<] You're welcome. Edit - some blokes here just love downthumbing my posts. Hope doing that makes you rich, guys.

albundy
albundy
3 years ago

the non-synthetic tests say it all. the price/performance ratio makes it a questionable purchase. are the small gains worth the cost?

Klimax
Klimax
3 years ago
Reply to  albundy

Yes. At test level they look small, but throw at it real load and you start looking at distinct advantages like saving hours or days.

chuckula
chuckula
3 years ago

Judging by Kampman’s twitter feed it looks like he’s moved on to Austin for an Epyc press conference tomorrow.

Get some Rudy’s and/or Salt Lick while you’re there!

curtisb
curtisb
3 years ago
Reply to  chuckula

Rudy’s? It’s ok, but head on down to San Antonio for some Texas Pride instead. It’s worth the drive. 🙂

ronch
ronch
3 years ago
Reply to  chuckula

It’s World War 3 as far as Intel and AMD are concerned.

the
the
3 years ago
Reply to  ronch

Unfortunately instead of sending nukes to each other, we’ll be getting them in our processor sockets with how much energy these chips will burn to edge each other out in performance.

Krogoth
Krogoth
3 years ago
Reply to  the

Aside from epenis extreme overclocking benchmarking. I really don’t see this happening. This isn’t the late 1990s and early 2000s anymore.

The PC landscape has completely changed. Enthusiasts want performance without resorting to exotic cooling solutions or super-load fans that belong in 1U/2U in a server room.

i9-7900 is already pushing the envelope as it is and Intel “recommends” using some kind of water-cooling if you want overclock the chip with all cores running.

Fonbu
Fonbu
3 years ago

I might have missed the Memo… But which Xeon cpu is the equivalent to this Core i9-7900X?

Beahmont
Beahmont
3 years ago
Reply to  Fonbu

Are you Google, Amazon, Facebook, or some other multi-billion dollar HPC company? No? Then no one really knows yet.

The Skylake Xeons are to the best of my knowledge only available to the ‘Very Special Customers’ right now. And it’s been that way for over 6 months. That said, those very special customers have been stupidly happy with them and have acquired as many of them as they can.

derFunkenstein
derFunkenstein
3 years ago
Reply to  Beahmont

no reason to be a dick about it

Pancake
Pancake
3 years ago
Reply to  derFunkenstein

Uncalled for. Beahmont didn’t attack Fonbu. A modicum of judicious levity helps us all get through the day.

psuedonymous
psuedonymous
3 years ago
Reply to  Fonbu

None yet. The larger Skylake-SP (Purley platform) chips have been announced, but AFAIK no Xeons have ever been mentioned for Basin Falls.

morphine
morphine
3 years ago

Uuuh, right, you’re welcome! 🙂

chuckula
chuckula
3 years ago

Hey Jeff, quick memory question. The 3200 -speed setup was using 64 GB while the 2666 speed setup only had 32 GB.

Is that a function of having 8GB DIMMS for the slower RAM and 16GB DIMMS for the faster RAM or were you using four sticks in one setup and 8 sticks in the other setup?

Jeff Kampman
Jeff Kampman
3 years ago
Reply to  chuckula

Each setup uses four DIMMs. The 32GB kits use 8GB DIMMs and the 64GB kit uses 16GB DIMMs.

chuckula
chuckula
3 years ago
Reply to  Jeff Kampman

Thanks! You had interesting results with those RAM setups, although not out of line with what I saw elsewhere. All the more reason to wait for firmware teething problems to be worked out.

wingless
wingless
3 years ago

INTEL IS DA BEST! AMD SUX! Now if you could all just go to my GoFundMe so I can build an X299 system that would be swell….

chuckula
chuckula
3 years ago
Reply to  wingless

Oh man, you picked the wrong team for that post.

You should have reversed it if you wanted to get enough money for X299.

derFunkenstein
derFunkenstein
3 years ago
Reply to  chuckula

Intel sucks! AMD rocks! Prove my wrong by paying for an X299 build and I’ll take it for a spin.

the
the
3 years ago

Totally called the mesh topology design [url=https://techreport.com/news/28772/well-be-talking-skylake-with-david-kanter-on-tonight-live-stream?post=928457<]two years ago.[/url<] Also need more Podcast.

UberGerbil
UberGerbil
3 years ago
Reply to  the

Yeah, another Kanter discussion would be aces. (I have no idea why I used that particular idiom)

chuckula
chuckula
3 years ago

I could just imagine AMD’s response if Rory was still there and dressed like [url=http://fudzilla.com/media/k2/items/cache/4cf83e0ebadefbfc492725ef71f1c5fa_XL.jpg<]this[/url<]. [quote<]That's an awful nice HEDT platform you got there Intel. I would be a real shame if there was an... 'accident'. You know, you wouldn't want to wake up one day and find that all its threads had been ripped out. I'm just sayin'[/quote<]

abiprithtr
abiprithtr
3 years ago

Jeff, nice review. A positive, and a negative observation – if I may. +ve: The reviews here at TR have largely thrown out subjectivity, and for the most part read very neutral. -ve: Keywords being “most part”. In page where you look at power, you write ” Not only does the Core i9-7900X cut 94 seconds off the Ryzen 7 1800X’s bmw27 render time, it does it while expending only just a bit more power to do so. That’s impressive performance per watt.” This sentence can just as easily read as follows: Not only does the Core i9-7900X cut [b

jts888
jts888
3 years ago
Reply to  abiprithtr

Yeah, that was probably the weakest paragraph in the review.

I would have considered writing it as:

The Core i9-7900X maintains very competitive task energy efficiency despite its high overall performance brought in part by aggressive clocking: it shaves off almost 32% of the R7-1800X’s bmw27 render time while consuming only roughly 4% more total Joules. The tamer i7-6950X is still the champ in this metric, however.

Unknown-Error
Unknown-Error
3 years ago
Reply to  jts888

Nicely put.

jts888
jts888
3 years ago
Reply to  Unknown-Error

Thanks. These sorts of things can be twisted pretty much any way you want.

Check this out:
Despite Intel’s process advantage, they opted to release their 10 core part with clockings yielding 4% worse energy efficiency in our rendering task than the 7 1800X, which is itself already the least efficient 8 core Ryzen. The 7900X does complete our test in 32% less time than the 1800X, but the huge efficiency regression from the i7-6950X in the name of raw throughput has us wondering just exactly to what degree AMD has Intel panicked.

Unknown-Error
Unknown-Error
3 years ago
Reply to  jts888

Damn. I wish I had your writing skills during my college days.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  jts888

Yeah I think that captures the situation.

lmc5b
lmc5b
3 years ago

I love the task energy analysis, but I think it would help to add a scatter plot with “task energy” in one axis and “time to complete” in the other, as lower power consumption only matters within comparable performance levels.

Jeff Kampman
Jeff Kampman
3 years ago
Reply to  lmc5b

This is a fine idea—I’ll try and implement it soon.

chuckula
chuckula
3 years ago

Hey Kampman, I wanna see a coin-mining benchmark here.

Not because I think these chips are any good at mining, quite the opposite.

I just want it to be publicly known that the i9 SUCKS at mining and to make sure those miners don’t even think about buying them up. It’s kind of like a PSA.

synthtel2
synthtel2
3 years ago
Reply to  chuckula

How many people who would buy an i9 for mining would even be in the readership for such an article?

derFunkenstein
derFunkenstein
3 years ago
Reply to  synthtel2

Isn’t that the point of chuckula’s comment? To prove out that it’s not worth buying for mining?
It’s a joke for crying out loud.

chuckula
chuckula
3 years ago
Reply to  derFunkenstein

THE ENTIRE I9 SERIES IS A JOKE!

derFunkenstein
derFunkenstein
3 years ago
Reply to  chuckula

Me, too. And it hurts. </3

Klimax
Klimax
3 years ago
Reply to  chuckula

I am sorry, but AVX-512 will likely mean that for quite few cryptocoins these chips will be very good. (vector rotation and co)

synthtel2
synthtel2
3 years ago
Reply to  Klimax

Nah. AVX-512 is still weaksauce compared to a GPU in absolute terms and in perf/W terms, and even with the inflation due to *coin, it’s still massively cheaper to hook six Hawaii/P10s to a single board than it is to get a half-dozen 7900Xes up and crunching. There may be some efficiency points in favor of the CPU (RAM access patterns etc) but they’re not even in the same league of brute force per your already-questionably-invested dollar. 64 SP FLOPS/cycle * 10 cores * 3.5 GHz = 2.24 TFLOPS, or ~half of one of those $640 480s. Etherium is a… Read more »

GurtTractor
GurtTractor
3 years ago

Nice review, thanks for being one of the very few reviewers out there testing DAW performance. Is there any chance you could share the project file for that DAWbench DSP 2017 beta benchmark? Or at least point me in the direction of where I could find it? The main website doesn’t seem to have been updated in a while, so the only tests there are there 2014 ones. Quite interesting that you’re working with the creator of those, would love to learn more about this new test and the reasoning behind the choice of plugin. Was the SGA1566 plugin just… Read more »

sophisticles
sophisticles
3 years ago

Not a bad review, some good information, particularly regarding AVX512, disappointing to see Intel doing product segmentation by only enabling a gimped form of AVX512 on the lesser chips. You Handbrake benchmarks are better but you still introduced a nasty bottleneck by resizing your 4k source to 1080p. Most filters, including the ones that Handbrake uses do not scale well, I know it seems like you are increasing the workload by resizing but what you’re doing is bottlenecking the cpu. Here’s a tip, when one is working with video (or audio for that matter) and they have access to a… Read more »

Krogoth
Krogoth
3 years ago
Reply to  sophisticles

Depending on the workload you might end-up becoming I/O bound assuming you are running a SATA SSD not a M.2 NVMe media of some kinda.

DragonDaddyBear
DragonDaddyBear
3 years ago

Aren’t some of these tests very sensitive to memory bandwidth? I’d like to see how the results change. I think it could show the IPC difference more clearly between AMD and Intel.

EDIT: RAGEPRO noticed I didn’t actually ask a question. The question is how much does the performance come from Quad vs Dual channel memory (not likely much, but I think it’s worth investigating).

RAGEPRO
RAGEPRO
3 years ago

Can you elaborate on your question? Every test was run with DDR4-2666 and DDR4-3200.

DragonDaddyBear
DragonDaddyBear
3 years ago
Reply to  RAGEPRO

I was trying so hard to word my comment that I spaced what I was really asking about. Quad vs. Dual channel, I wonder how much of an impact the number of channels had. This is assuming it can be run in such a configuration.

UberGerbil
UberGerbil
3 years ago

I suspect that comparison will show that some of the benchmarks that appear to be unconstrained by memory bandwidth actually are when it is reduced sufficiently, so you’d see more variation in the results from the 2666 vs 3200 tests. But what would that tell you in real world terms? That you should fill all your memory channels? Didn’t you already know that? I guess if TR did a scatter plot of (total system cost) vs (performance), that might get interesting. For a given quantity of memory, you have different costs for two channels of fast memory vs four channels… Read more »

Bauxite
Bauxite
3 years ago

[quote<]What's more, not every Core X chip in the lineup will enjoy the same boost in SIMD performance from AVX-512. Only the Core i9 series of CPUs will ship with the dedicated AVX-512 FMA. The Core i7-7800X and Core i7-7820X will still have the wider registers for AVX-512, but they'll only execute instructions using the pair of 256-bit AVX units common to all Skylake chips.[/quote<] Holy crap, they just can't help tripping over themselves. Also, what is that stupid chip in the top right corner of the cpu used for? All I've seen online so far is something about RFID, which sounds like very bad news for the "owners" of cpus.

Krogoth
Krogoth
3 years ago
Reply to  Bauxite

I suspect that’s only for engineering samples.

It is Intel trying to cut-down and keep track of their stuff to prevent “leaks”.

jts888
jts888
3 years ago
Reply to  Krogoth

What about the allegations that Intel was going to use physical access “upgrade” keys to lock out NVMe RAID functionality on some of their platforms? RFID/NFC crap on package lends well to that sort of thing as well.

curtisb
curtisb
3 years ago
Reply to  jts888

As was said in that article about the RAID enabler keys, those are nothing new on high-end workstation and server systems. RFID and NFC don’t have anything to do with those hardware keys.

chuckula
chuckula
3 years ago
Reply to  Bauxite

Intel gimp the functionality of AVX-512 hardware?

THANK YOU AMD!

[Seriously: Intel’s approach to AVX-512 in the lower-end i9 parts is basically what AMD adopted for old AVX in RyZen but all of the sudden Intel hates AVX-512 in a 7800X that fully supports it while the Threadripper with 16 cores will literally never execute a single AVX-512 opcode. Funny how nobody claimed AMD hated 4 year old AVX2.]

derFunkenstein
derFunkenstein
3 years ago
Reply to  chuckula

Lots of folks griped about effectively-halved AVX throughput on Ryzen. TR’s benchmark suite pretty well established how that was a mistake, too.

NTMBK
NTMBK
3 years ago
Reply to  Bauxite

The vector performance was one massive obvious win over Zen, so of course they had to shoot themselves in the foot. 🙁 AVX-512 adoption will be slow enough as it is, without Intel putting more roadblocks in its way.

I just hope they don’t pull the same shit in the Xeon line!

lycium
lycium
3 years ago
Reply to  NTMBK

It’s a crying shame about AVX-512 because SSE is such a mess. The story of how AVX-512 came to be is a really funny and interesting one, worth reading about (Tom Forsyth).

DrCR
DrCR
3 years ago
Reply to  lycium

Link to the read?

lycium
lycium
3 years ago
Reply to  DrCR

Whoops, sorry for the unintentional red herring, it was Mike Abrash not Tom Forsyth who tells the story: [url<]http://www.drdobbs.com/parallel/a-first-look-at-the-larrabee-new-instruc/216402188[/url<]

synthtel2
synthtel2
3 years ago
Reply to  Bauxite

I’m going to be contrary – I think Intel is doing it right this time. Wrong is what they do with Celerons and Pentiums, which makes it annoying for devs because if we want both AVX’s gains and wide support, we have to have both an SSE and an AVX path. This Zen and SKL-X method means that the latest instruction set gets distributed across the board and devs don’t have to worry about it (or won’t, once it’s been out for as long as Sandy has). It decouples instruction set support from the width of the underlying execution. Having… Read more »

Krogoth
Krogoth
3 years ago

IMO, you should probably wait until Threadripper comes along before jumping onto the i9 bandwagon. There’s a good chance that Intel might do a price cut on i9 line-up if Threadripper leapfrogs i9 at workstation-tier stuff.

I wanna see how Skylake-X i7s fare against the Ryzen 7 line since they are much closer to each other in platform cost.

The loaded power consumption figures make sense since these are “rejected” Xeons and Intel is trying to pull ahead of Ryzen R7 by a good margin and are willing to give-up loaded power consumption for it.

the
the
3 years ago
Reply to  Krogoth

I don’t think we’ll see an immediate price cut on Intel’s part. The exclusive deal with Alienware sort of points toward a limited number of 16 core parts as OEM deals are bread and butter for revenue: they won’t sign an exclusivity deal if they had a fairly large supply to sell. Considering that Ryzen chips are easy enough to come by, I just fathom that Threadripper itself is going to be a low volume part by design rather than an underlaying yield issue.

Beahmont
Beahmont
3 years ago
Reply to  the

You might be surprised. I’ve been hearing some interesting grumblings about AMD’s EPYC and ThreadRipper stuff, as well as HBM and interposer stuff in general. Apparently it is much harder in Real Life(tm) to actually assemble the multi dies onto the packaging in bulk than the basic tech specs would have everyone believe.

chuckula
chuckula
3 years ago
Reply to  Beahmont

One of the best marketing points I’ve seen for EPYC was when they held up the package sans IHS and there was no silicon interposer in there at all. They still might have some packaging challenges, but no needing an interposer to get the job done is a good thing.

the
the
3 years ago
Reply to  Beahmont

Oh, MCM does impact yields but only from a packaging stand point. Still overall higher yields to go with two ~300 mm^2 dies in MCM than a single 600 mm^2 dies.

Beahmont
Beahmont
3 years ago
Reply to  the

And yet, without packaging the chips are not in any usable form. If you’re even just breaking 1 out of ever 10 300mm[super<]2[/super<] chips when you insert them into the MCM package you can very quickly lose your margin advantage over that 600mm[super<]2[/super<] single die total package chip. Slightly bad 600mm[super<]2[/super<] chips can still be used as a lesser bin chip. Broken chips are a complete lose. And I imagine you'd have at least double as many broken MCM packages as chips if you're having any kind of real issues at all. That last sentence is complete speculation on my part. I very well could and am wrong. But it 'makes sense' from my understanding of the assembly process.

Krogoth
Krogoth
3 years ago
Reply to  the

Intel isn’t too shy on doing it though. They did a massive price cut on Q6600 and Q6700 back in the day in anticipation of the original Phenoms. They thought that Phenoms would at defeat their offerings at their original price points, but alas the Phenom Is didn’t hold-up. I can easily see them doing for 16-core and 18-core i9s though assuming that fully-enabled Threadripper deliver(about ~2x R7 1700X performance in hilariously paralleled workloads). It looks like Skylake-X starts experiencing some scaling issues beyond 8-cores though judging from the difference between i7-7820 and i9-7900 from elsewhere.

w76
w76
3 years ago

More performance and a lower price point. Thank you, AMD. Maybe the “monopolies have to compete with themselves” meme justifying Intel’s dominance will end. Only in some corporatist fantasy land was that ever going to lead to lower consumer prices.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  w76

Looking at that downvote, it seems someone around here likes their monopolies.

blastdoor
blastdoor
3 years ago

Stockholm syndrome (and related ailments in which the exploited side with their exploiters against would-be liberators) runs rampant.

Klimax
Klimax
3 years ago

More like it is down vote for manufacturing nonsensical position and “meme”. Aka making strawman…

blastdoor
blastdoor
3 years ago
Reply to  w76

I agree — thank you AMD!

I just hope that AMD can earn enough to keep competing with Intel.

Kurotetsu
Kurotetsu
3 years ago
Reply to  blastdoor

That’ll depend on whether that “thank you” is genuine and people actually buy AMD products, or if its more backhanded and people turn around and buy Intel now that prices are lower. The former keeps competition going, the latter brings us right back to where we started.

w76
w76
3 years ago
Reply to  blastdoor

Same here, their financial outlook is pretty dark with bonds coming due in coming years if they don’t start making some good profit, especially if interest rates rise for their credit cohort (presumably not ‘investment grade’).

Klimax
Klimax
3 years ago
Reply to  w76

As far as I can tell, nobody held that nonsensical position. You have manufactured a strawman. At least I haven’t sen any comment like that in front page discussions.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  Klimax

I’m not sure anyone in the forum cheers monopolies [i<]too[/i<] much, but in the broader political landscape, there is a certain amount of freedom == no regulations == we all win thinking.

BobbinThreadbare
BobbinThreadbare
3 years ago
Reply to  Klimax

There are pro-monopolist economic theories

Jigar
Jigar
3 years ago

That $450 AMD chip is hanging in there with the big boy ($1000 chip), i can only imagine Threadripper bringing in more performance at cheaper pricing.

chuckula
chuckula
3 years ago
Reply to  Jigar

There are some reviews of the cheaper i9 parts out too.

You’ll note that in many instances (including Handbrake video encoding that AMD itself highlighted when RyZen launched) the $389 i9 7800X with only 6 cores can actually beat the 1800X — and that’s with no overclocking.

[url<]http://www.anandtech.com/show/11550/the-intel-skylakex-review-core-i9-7900x-i7-7820x-and-i7-7800x-tested/13[/url<]

tsk
tsk
3 years ago
Reply to  chuckula

It’s interesting to read about the comparison of syklake-s and skylake-sp in that review, it’s a mixed bag as far as IPC goes, definitely some changes due to the new cache.

the
the
3 years ago
Reply to  tsk

It’ll end as a net positive as more AVX-512 enabled software comes out. That’s the catch though with introducing new ISA extensions: any benefit will be delayed until well after launch.

Voldenuit
Voldenuit
3 years ago
Reply to  the

Considering intel is threatening to sue MS over x86 emulation, I’m hoping somebody comes out with open-sourced SIMD ISAs (like 3DNow! I know, that’s unlikely) that become an industry standard instead of having a litigious monopolist be in charge of vector code.

the
the
3 years ago
Reply to  Voldenuit

Well recent ISA extensions can fall under some patents making emulation difficult. The problem Intel has is that due to their insane market segmentation, software developers have to support the lowest end chips to run software. These chips don’t have the latest and greatest ISA extension. Pretty much everything before AVX is fair game for emulation that would run modern software.

Microsoft is in a better position than most on this topic too as they can can dictate terms of the platform so that developers continue to offer a binary that would work well under emulation.

chuckula
chuckula
3 years ago
Reply to  Voldenuit

Intel has never threatened to sue ARM for going off on its own and doing its own vector instruction set. In fact, those instruction sets already exist, it’s called NEON although it’s nowhere near what AVX can do. Intel is only making noises about trying to make an ARM chip act so much like an x86 CPU that it’s basically being a clone. For all the talk about “monopoly” here I promise you that if you try to go in reverse and do the same thing for ARM they’ll come after you big time when they don’t get their cut… Read more »

Voldenuit
Voldenuit
3 years ago
Reply to  chuckula

And PowerPC had Altivec/VMX back in the day.

Interestingly, Motorola and IBM PowerPC CPUs had completely different logic designs to handling Altivec vector code.

This is what I’m asking for. An open (or at least non-litigious) vector ISA to be industry standard. Chip makers could (and should) patent their own specific implementations in hardware. But the ISA should not be used as a bargaining chip (or hammer) in patent disputes.

utmode
utmode
3 years ago
Reply to  chuckula

but:
[url<]http://www.guru3d.com/articles_pages/intel_core_i9_7900x_processor_review,11.html[/url<]

raddude9
raddude9
3 years ago
Reply to  chuckula
jihadjoe
jihadjoe
3 years ago
Reply to  Jigar

IMO the real competition for that $450 AMD chip is the $600 i7-7820x

raddude9
raddude9
3 years ago
Reply to  jihadjoe
Pancake
Pancake
3 years ago
Reply to  raddude9

Cheaper but less features, much less performance. You gets what you pays for. For the price of a full system the i7-7820X delivers FAR GREATER increase in performance than R7-1800X relative to increase in price. Not speculation. Now proven. That gonna hurt AMD some.

Krogoth
Krogoth
3 years ago
Reply to  Pancake

It not that much though and you are still limited by having 28 PCIe 3.0 lanes on the platform which may limit your expansion options down the road (M.2 NVMe media and Thunderbolt 3).

IMO, the only reason Intel did this was to avoid cannibalizing their i9 sales.

Beahmont
Beahmont
3 years ago
Reply to  Krogoth

And yet you’re being disingenuous enough to almost be out and out lying. If the comparison is Ryzen 1800X to Intel Skylake-X then your complaints about PCIe lanes are BS because Intel has more even at 28 PCIe lanes than any Ryzen part. If the comparison is ThreadRipper, you’re month(s) too early because while ThreadRipper is supposed to have more PCIe lanes, we don’t even know how AMD is doing the cross chip interface for that many lanes yet, let alone the overall performance or cost comparison. For all you know AMD’s ThreadRipper boards may be just as expensive as… Read more »

jihadjoe
jihadjoe
3 years ago
Reply to  Krogoth

There are another 24 PCIe lanes going to the PCH.
And you really want to plug those NVMe drives onto the PCH (instead of the CPU) because it’ll let you boot NVMe RAID.

psuedonymous
psuedonymous
3 years ago
Reply to  Krogoth

[quote<]28 PCIe 3.0 lanes on the platform which may limit your expansion options down the road (M.2 NVMe media and Thunderbolt 3). [/quote<] That's still enough for me to build an ITX system with a 1080Ti and three m.2 drives: one 960, one Optane, and one 2tb 'slow' m.2 once 3D NAND enters wider production and pushes prices down a bit more; all operating at full bandwidth. That's something no other platform can do today, and none on the horizon are likely to do. Reports from ASRock, current king of SFF boards, has been "LOLno, have you seen the size of that socket?!" on any non-ATX Threadripper boards. And with SP performance looking to match, or at least approach, the 7700k in some cases, I'm seriously considering moving to X299 with an i7-7820x. Ryzen's SP performance is not sufficient for me (and Threadripper looks to only be going downhill from there), and just hopping from the 6700k to the 7700k would not be worth the motherboard change.

Sahrin
Sahrin
3 years ago

How much did Intel pay you guys to use a consumer CPU as the only AMD chip in the comparison?

JMccovery
JMccovery
3 years ago
Reply to  Sahrin

What non-consumer AMD CPUs worth comparing exist?

Sahrin
Sahrin
3 years ago
Reply to  JMccovery

I don’t think you understand the point. Toyota doesn’t make a supercar, so does it make sense to compare the McLaren, Model S, GT40, Aventador, etc, to a Camry? No. It’s a stupid, pointless comparison.

derFunkenstein
derFunkenstein
3 years ago
Reply to  Sahrin

oh so TR should have covered up the Ryzen comparison? How stupid are you?

chuckula
chuckula
3 years ago
Reply to  Sahrin

Tell ya what, when The Threadripper launches you can complain about TR throwing in some 7700K benchmarks as a reference point.

You know, compiling and Cinebench numbers that will like 16 cores? You can be sure to savage TR for showing The Threadripper in the same graph as those “consumer” parts.

I’m sure you’ll be even-handed in your critique of TR.

Zeratul
Zeratul
3 years ago
Reply to  Sahrin

Uh… you haven’t heard of the Lexus LFA then I take it.

jihadjoe
jihadjoe
3 years ago
Reply to  Zeratul

MkIV Supra is good enough for me!

tsk
tsk
3 years ago
Reply to  Sahrin

As soon as someone turns to car analogies in tech discussions they’ve lost.

maxxcool
maxxcool
3 years ago
Reply to  Sahrin

Why not, there all cars.

JMccovery
JMccovery
3 years ago
Reply to  Sahrin

So, don’t make a comparison to Ryzen (you know “HEDT performance at mainstream prices), because ThreadRipper isn’t out yet?

That makes zero sense, as #1 it isn’t a pointless comparison [b<]since the highest performing processor AMD currently has is the "HEDT-like" 1800X[/b<], and #2 AMD did compare Ryzen to Broadwell-E.

ronch
ronch
3 years ago
Reply to  Sahrin

On the contrary it makes super duper sense to put the best chip AMD has available [u<]for purchase [/u<] here.

MOSFET
MOSFET
3 years ago
Reply to  Sahrin

I thought Ryzen looked pretty good in this review, actually. Huh.

Kougar
Kougar
3 years ago
Reply to  Sahrin

I understand your point. But your point is inconsequential because AMD hasn’t produced Opterons in half a decade, and more importantly Threadripper hasn’t launched yet which makes the 1800X the best possible chip to compare against.

That you’d ignore these key details just makes you sound like either an industry shill or just another fanboy fanatic.

Wirko
Wirko
3 years ago
Reply to  JMccovery

A 16-core Opteron in liquid antideuterium?

JMccovery
JMccovery
3 years ago
Reply to  Wirko

Hmm… Interesting.

tsk
tsk
3 years ago
Reply to  Sahrin

What other AMD chip should they have included?
The 1800X is the only one to compete with those chips until threadripper arrives.

chuckula
chuckula
3 years ago
Reply to  Sahrin

Dayum, usually it takes at least a dozen posts for the first anti-AMD conspiracy theory to be thrown out.

You win a cookie.

Redocbew
Redocbew
3 years ago
Reply to  chuckula

Save the cookies. I’m sure we can think of a better reward.

chuckula
chuckula
3 years ago
Reply to  Redocbew

Threadripper giveaway to the first Intel fanboy who accuses TR of being in an anti-Intel conspiracy when they publish the Threadripper review!

I’m already writing up my conspiracy theory new, Chucky needs 16 cores!

Redocbew
Redocbew
3 years ago
Reply to  chuckula

By “better” I was thinking of lumps of coal, or maybe something involving the use of a shock collar, but yeah I guess that could work.

the
the
3 years ago
Reply to  chuckula

Can I have a cookie?

chuckula
chuckula
3 years ago
Reply to  the

Yes, but you have to make sure they are turned on in your browser.

tipoo
tipoo
3 years ago
Reply to  Sahrin

♬ You’re so vain you probably think i9 is about you ♬

jensend
jensend
3 years ago
Reply to  tipoo

Man, this is old.

It doesn’t really have much relation to his comment. It’s not a good reply in this case.

It was funny when Jeff said it. It was already old and tired and unfunny the first time somebody aped him. After dozens of encores I am pretty sick of it.

“HEY JEFF MADE A FUNNY. LET’S KEEP REPEATING IT OVER AND OVER TILL THE HEAT DEATH OF THE SUN.”

Redocbew
Redocbew
3 years ago
Reply to  jensend

But they made musical notes! From text! You know this is the Internet, right?

UberGerbil
UberGerbil
3 years ago
Reply to  jensend

The point that is being made has nothing to do with humor. (And if the complaints are going to be repetitive and unimaginative, the responses to them might as well be also. Why waste creativity on trolls?)

jensend
jensend
3 years ago
Reply to  UberGerbil

Why even waste a response on trolls?

If all your response accomplishes is registering your disapproval, a downthumb suffices for the task, without feeding the troll.

maxxcool
maxxcool
3 years ago
Reply to  jensend

Purportedly the Gerbils like it.. /ducks/..

the
the
3 years ago
Reply to  tipoo

DON’T YOU
DON’T YOU
DON’T YOU

RAGEPRO
RAGEPRO
3 years ago
Reply to  the

Someone’s been listening to the NIN version.

Krogoth
Krogoth
3 years ago
Reply to  Sahrin

i7 and i9 Skylake-X chips are “customer-tier” chips bro. They are just Xeons that either have “bad portions” and/or ate too much power for Intel to use for their enterprise market.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  Krogoth

Might be right about that power, I wasn’t real impressed that the task energy was actually higher than AMD. If that carries on to servers, its bad for Intel.

I should point out that AMD’s chip wasn’t exactly clocked for efficiency either.

chuckula
chuckula
3 years ago

Considering the differences in platforms with AM4 being a much lower power platform overall than a full-bore X299 system with twice the RAM channels I’m not that shocked by the differences in power usage, although some tuning could probably cut the 7900X power consumption a bit.

As for servers, it’ll be pretty easy to get the same silicon running more efficiently when it’s not turbo boosting north of 4GHz.

People have complained about Intel only focusing on power efficiency for years, but I guess there’s always something new to complain about.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  chuckula

I think you’ll have to agree that if Intel’s extra hardware (more memory channels, more floating point units) raises its power draw in markets where they need neither the bandwidth nor the floating point, for example a lot of server work… then AMD will actually have a server niche where they can flat out beat Intel. I’m really impressed by what AMD has pulled off here. They left behind what they didn’t need (and couldn’t afford anyway) and went right to the point. Intel’s latest chip here is pretty cool for some things, but its looking like they can no… Read more »

Voldenuit
Voldenuit
3 years ago

[quote<] I think you'll have to agree that if Intel's extra hardware (more memory channels, more floating point units) raises its power draw in markets where they need neither the bandwidth nor the floating point, for example a lot of server work... then AMD will actually have a server niche where they can flat out beat Intel.[/quote<] Doesn't the new [url=https://www.pcper.com/news/Processors/Intel-Skylake-X-and-Skylake-SP-Utilize-Mesh-Architecture-Inter-Chip-Communication<]mesh interconnect[/url<] on the Skylake-X chips use less power than the old ring topology (at least according to intel's claims)? I can see this having a benefit in server applications.

Krogoth
Krogoth
3 years ago
Reply to  Voldenuit

It also allows them to cram more cores onto the monolith packaging too.

Intel is also going multi-die chips in their future roadmap (~5 years down the road). It looks like current yielding issues from massive 14nm chips (Broadwell-E and Skylake-X) has convinced that physics and scale of economics aren’t playing nice.

Anonymous Coward
Anonymous Coward
3 years ago
Reply to  Voldenuit

All the server work I do has next to no need for a fancy interconnect, its separate processes doing their thing on VM slices of big servers, often one core per VM, or two cores. I’ve rented up to 32 cores in one VM, but also there, it was lots of separate processes working with minimal chatter between them. You make a good point that the mesh interconnect could be a step in the right direction for efficiency, but I wonder if the optimal solution for me would be AMD’s famously minimalistic approach to interconnects. Who needs a monolithic die… Read more »

mat9v
mat9v
3 years ago
Reply to  chuckula

I wonder about clocks, 7900X as stated by Intel should be running at 4Ghz all core boost and 6950X was 3.4Ghz if I remember correctly? It seems that IPC for new SkyLakeX arch is not so good in some tasks. Cinebench: 7900X – 2205/4Ghz=551,25 6950X – 1892/3.4Ghz=556,47 Blender: 7900X – 203s for 4Ghz > for 1Ghz would be 812s 6950X – 232s for 3.4Ghz > for 1Ghz would be 788s Handbrake: 7900X – 271s for 4Ghz > for 1Ghz would be 1084s 6950X – 315s for 3.4Ghz > for 1Ghz would be 1071s I do not like IPC of those… Read more »

Sahrin
Sahrin
3 years ago
Reply to  Krogoth

You’re comparing a $1000 CPU to a $500 CPU, and using that as the only example. It doesn’t take a rocket scientist to guess what the result will be.

Sick of Intel fanboys ignoring facts, and accusing everyone else of being an AMD fanboy when they do. Not an AMD fanboy – 1800X is slower than Sky Lake and Kaby Lake. Doesn’t mean you run a consumer-range CPU against an enthusiast-range CPU as the only entrant – especially when AMD has a product coming that does compete with these CPU’s.

Zeratul
Zeratul
3 years ago
Reply to  Sahrin

They compared the best thing AMD had available now so people could see where it stands. Very few people buy the BEST CPU, but anyone looking to build a high end system sees where the current AMD offering compares to high end Intel offerings and can compare on price and efficiently to make a decision. They also threw an Intel 7700K in here.

weaktoss
weaktoss
3 years ago
Reply to  Sahrin

The 1800X is there for the same reason that the 7700K is—to provide context for how the fastest mainstream consumer parts perform against the new HEDT chip. That’s relevant information for potential purchasers, regardless of whether or not you think it is.

If Threadripper chips were already available, of course those would have been in the review too. Do you expect reviews of new Intel products to be delayed until the corresponding AMD products are out? There is no secret, paid-for, anti-AMD agenda here. Just market realities.

chuckula
chuckula
3 years ago
Reply to  weaktoss

If Wasson wants to give TR a world-exclusive early Threadripper sample for fun & profit then I’m pretty sure Kampman will say YES PLEASE.

coolflame57
coolflame57
3 years ago
Reply to  weaktoss

So this is like comparing Michael Phelps to the fastest swimmer in some coastal village, just so that the villagers can understand just how fast Phelps is, right? The village swimmer is just a valuable point of reference for the villager(bil)s, and until they finish training their next pro swimmer, he is the fastest available.

Krogoth
Krogoth
3 years ago
Reply to  Sahrin

That’s what the current mainstream marketplace has right now and why Intel can justify commanding a MSRP of $999 for i9-7900K. It is now the fastest non-enterprise-tier SKU at content creation creation/workstation workloads.

Threadripper isn’t ready yet. Just relax and wait another a few months.

Waco
Waco
3 years ago
Reply to  Sahrin

Perhaps the “it’s not released” part about Threadripper is news to you? They can’t compare something that doesn’t exist.

Top-end AMD stops at $500 these days, like it or not. There is no better comparison you can make until AMD starts shipping Threadripper.

Unknown-Error
Unknown-Error
3 years ago
Reply to  Sahrin

Without accusing people of taking bribes, why not be happy that there is at least one AMD CPU in the charts. It means AMD now actually has something to show. Just a years ago AMD had pretty much nothing to show. AMD may not be beating Intel in any of these charts but at least now they are IN the Charts.

morphine
morphine
3 years ago
Reply to  Unknown-Error

Oh come on, stop making sense!

southrncomfortjm
southrncomfortjm
3 years ago
Reply to  Unknown-Error

This argument is too well reasoned for the internet. Take it somewhere else please.

dodozoid
dodozoid
3 years ago
Reply to  Unknown-Error

Oh, but they do! In that perf/dollar chart, ryzen sits on pretty prominent spot

Concupiscence
Concupiscence
3 years ago
Reply to  Unknown-Error

Seriously, can you imagine an FX-9590 embarrassing itself here? Or one of the long-neglected Opterons?

Redocbew
Redocbew
3 years ago
Reply to  Concupiscence

Either of those would wreck the scale on the graphs and only help to hide the differences between the parts we actually care about. It’d be ridiculous.

JMccovery
JMccovery
3 years ago
Reply to  Concupiscence

I’ve always wanted to see comparisons of the 1600/X vs 6172/6174/6220 and 1700/1800/X and the 6386SE at similar clocks.

Would be a nice way to show just how far things have progressed.

credible
credible
3 years ago
Reply to  Sahrin

Wow really lolololol, holy cow daft to say the least.

K-L-Waster
K-L-Waster
3 years ago
Reply to  Sahrin

And if they *hadn’t* included an AMD chip, it would be “How much did Intel pay you guys to exclude AMD?” right?

Star Brood
Star Brood
3 years ago
Reply to  Sahrin

Whoa dude, Threadripper hasn’t had it’s NDA lifted for reviewers (or hasn’t been sent to them). 1800X is the biggest, fastest AMD CPU currently on the market.

YellaChicken
YellaChicken
3 years ago
Reply to  Sahrin

How much alcohol did it take to convince yourself you had a valid point?

utmode
utmode
3 years ago

Very good review. Thanks Jeff. I was surprised to see this review in June. I thought I will see it in late July.

chuckula
chuckula
3 years ago

Excellent first part as always. The chip in and of itself is quite strong in a lot of multi-threaded workloads and does a lot to close the traditional gap we’ve seen between the larger core count HEDT parts and the faster but lower core count maintstream parts. Obviously the platform is a bit rough around the edges at this point but there’s certainly potential. Just like with RyZen it’s probably a good idea to let the platform bake a while before ordering. I’m pretty excited that the performance is definitely there even in legacy applications since the first round of… Read more »

Pancake
Pancake
3 years ago
Reply to  chuckula

Ahh… sweet Intel chickens coming home to roost. All predictable and as predicted. The 7900X is too expensive and doesn’t particularly interest me. However, for my workloads I do need quad channel RAM to get up to 128GB so 7800X and 7820X are looking good. Single-threaded performance is important and Intel are way ahead of AMD there. It is unlikely Threadripper will be remotely competitive in this regard although it also has the all-important quad channel RAM and a large number of slow cores isn’t that useful to me. So, in the most important part of the HEDT market Intel… Read more »

Krogoth
Krogoth
3 years ago
Reply to  Pancake

You don’t get HEDT chips for single and dual-threaded workloads though. You get 7600 and 7700K for those kind of workloads.

The workloads where having tons of threads makes sense is where i9-7900 and upcoming Threadripper and other i9 Skylake chips are going for.

Pancake
Pancake
3 years ago
Reply to  Krogoth

Some people like me need TONNES of memory to keep large spatial datasets in memory. I can happily use up to 8 cores. More than that and I need more memory.

Anyway, my point was that a lot of the time I’m just coding, web browsing, stuffing around and fast single-threaded performance makes for a good experience.

Krogoth
Krogoth
3 years ago
Reply to  Pancake

The difference between the platforms at such workloads isn’t that large anymore though. This isn’t the Bulldozer versus Sandy Bridge days. It is more like the difference between Skylake versus Haswell (Current Zen).

The point is that single-threaded doesn’t really matter that much anymore expect for hardcore Dwarf Fortress and MAH 144-160FPS pro-gamer types who don’t care for Threadripper/Skylake-X chips though.

Pancake
Pancake
3 years ago
Reply to  Krogoth

Well, the 7900X as reviewed here slaughters the R7-1800X but for an equal core count (and subject of current discussion) comparison have a look at Anandtech’s review which includes the 7820X. The CPU rendering tests are of most interest to me as they probably are a similar workload to what I would do crunching large spatial datasets (using my own code). The encoding tests are somewhat less interesting to me but in both categories there is, on average, a huge advantage to 7820X. Easily enough to justify the cost increase on a new system purchase. Now, if you were to… Read more »

Krogoth
Krogoth
3 years ago
Reply to  Pancake

Snappiness for single-threaded and dual-threaded loads is primarily bottleneck by I/O throughput. The CPU hasn’t been much of an issue in this regard since we moved onto Core 2 and beyond.

It is a different story for multi-threaded workloads though.

Pancake
Pancake
3 years ago
Reply to  Krogoth

Au contraire. I’m talking about bloated PDFs which take an irritating amount of time to scroll through or redraw. I’m talking about web pages full of horrorshow JavaScript bloatage. I’m talking about the many seemingly trivial but irritatingly janky things that we do everyday.

Like me, posting on here instead of concentrating on my work…

nexxcat
3 years ago
Reply to  Pancake

[quote=”Pancake”<]Au contraire. I'm talking about bloated PDFs which take an irritating amount of time to scroll through or redraw. I'm talking about web pages full of horrorshow JavaScript bloatage. I'm talking about the many seemingly trivial but irritatingly janky things that we do everyday.[/quote<] A lot of that is memory and IO bandwidth. If going back to bloated PDFs is janky for you, then likely core bits of memory have been swapped out. I don't know how good Windows 10 is, but I know Windows Server 2008 R2? 2012? will still aggressively swap out bits of memory that's not being used -- we did some benchmarking at my former place of employment, and the use case there was "flurry of activity, then 1-2 hours of quiet, then markets open." When we simulate that in realtime, the 1-2 hours of quiet was enough to shove memory into swap, and even though we were using enterprise SSDs, swap is still slow, and we'd get substantial, and otherwise unexplainable, slowness as a result. They're probably now waiting on Xeons based on these chips. The sweet spot, for us, were 2-socket servers with maxed cores each when considering rackspace and power.

Pancake
Pancake
3 years ago
Reply to  nexxcat

Wut?? Consider yourself lucky if you’ve never had to open a PDF of a complex map on a single page – bajillions of polygons, lines, fills etc where it takes seconds to redraw a zoom or pan. Nothing to do with I/O bandwidth or paging. Take a look at that Anandtech review where opening a PDF is one of their CPU tests. Any of the Intel CPUs slaughter Ryzen there. Making maps is one of the things I do to present results to clients. I would imagine CAD and graphic design people would also appreciate fast CPUs.

Krogoth
Krogoth
3 years ago
Reply to  Pancake

Crappy code is the real culprit here. You are being held back by [b<]I/O throughput[/b<] and memory bandwidth. Again, the CPU isn't the issue unless you are running something prior to Core 2-era.

synthtel2
synthtel2
3 years ago
Reply to  Krogoth

I’m running a G3258, DDR3-2133 CL9, and a cheesy SSD (Sandisk X300 512GB), and I’ve been manually controlling clocks of late because my mobo, my USB DAC, and pulseaudio are conspiring to ruin everything (long story). For the kind of workload in question here, there is a clear difference between 2.4 and 3.7 GHz. It’s not even subtle, and this thing at 2.4 is still waaay beyond Pentium Ds and A64x2s.

When running non-crappy code that doesn’t deal with large amounts of data 1.0 GHz is fine, but there is a lot of crappy code out there.

167
0
Would love your thoughts, please comment.x
()
x

Pin It on Pinterest

Share This