Finally the Skylake-X series is complete. After hastily announcing the 18-core Core i9 last May during Computex 2017, we now have Intel's new 16- and 18-core processors on-hand four months later.
We don't doubt that Intel had planned to release Skylake-X all along, but did they plan to offer something higher than a 12-core part before catching wind of AMD's Threadripper?
It also seems unlikely that they meant to announce the X299 platform in late May and release it in June, made evident by the rushed motherboard development and the mess that ensued.
We received the first Intel Core i9 part -- the $1,000 10-core 7900X -- in late June along with a few Core i7 models featuring 8, 6 and even 4 cores, but it wasn't until August that we learned the official specifications for Intel's new $1,200 12-core, $1,400 14-core, $1,700 16 core and a $2000 18-core processors, the last two of which we'll be reviewing today.
Before covering the Core i9-7980XE and 7960X in detail, it's worth remembering that during all the chaos of Intel's latest desktop platform release, AMD launched its Ryzen Threadripper series on August 10 including the 1950X and 1920X, the latter being a $1,000 16-core CPU that shamed Intel's then flagship 7900X.
As we are about to see, there was more to come from Intel but at the time we questioned if the $1,200 12-core 7920X could even beat the 1950X and suspected that it would probably take the $1,400 14-core 7940X to match AMD's 16-core CPU. While we don't have the 12-core or 14-core models, Intel has served up the ultra-expensive 16-core and 18-core chips which should give us an idea about where the 12 and 14-core versions will sit, and of course we're always keen to see what the flagship parts have to offer.
To quickly recap, the Core i9-7980XE packs 18 cores and with its support for Hyper Threading it touts an insane 36-threads. Each cores has its own 1MB L2 cache and thus there is a total of 18MB L2 cache. Meanwhile, there is 24.75MB worth of L3 cache and due to the restructuring of the cache for the Skylake-X architecture the 7980XE only has slightly more L3 than the previous generation 6950X.
The 7980XE operates at a base frequency of 2.6GHz with a Turbo Boost 2.0 frequency of 4.2GHz and a Turbo Boost Max 3.0 frequency of 4.4GHz. The 16-core 7960X features a base clock speed of 2.8GHz with the same Turbo Boost frequencies. It has a slightly smaller 22MB L3 cache while there is a total of 16MB for the L2. Both CPUs offer the full 44 PCIe lanes, quad-channel memory support and a 165 watt TDP rating.
Both CPUs use the LGA2066 socket and are supported by existing X299 motherboards. For testing I'm using the Gigabyte Aorus X299 Gaming 9 with 32GB of G.Skill TridentZ RGB DDR4 memory clocked at 1600MHz, so DDR4-3200, using CL14 timings. The GTX 1080 Ti was used for all the testing while we've thrown Vega 64 Liquid Cooled in for a few game tests. The focus though will of course be on the productivity testing, these are 18 and 16-core CPUs after all.
Ryzen Threadripper System Specs
Skylake-X System Specs
Ryzen 7 System Specs
Memory and Application Performance
First up let’s check out the memory bandwidth performance. These DDR4 quad-channel memory controllers look to be good for around 62-64GB/s of memory bandwidth when using 3200 memory. Please note all configurations were tested with the same DDR4-3200 CL4 memory.
Cinebench R15's multi-threaded test gives us a good idea of how the productivity tests are going to look if all cores and threads can be fully utilized. Here we see that the 7960X is just 4% faster than the the 1950X for the multithreaded workload, both are 16-core parts of course. The the 7980XE is 10% faster than the 1950X with an impressive score of 3317pts, though it does cost twice as much at $2,000 -- we'll certainly discuss pricing more at the end of the review.
When it comes to single-thread performance, the higher clocked Skylake-X parts are able to best Threadripper.
Before we move onto the more serious benchmarks I just wanted to see how these extreme CPUs stand up in PCMark 10 which looks at office type workloads (typically not core-heavy). To my surprise, Threadripper scored well in this test and did considerably better than the Skylake-X CPUs. The 7980XE looked particularly weak here. Of course none of these CPUs are slow for these more general office type tasks so it's somewhat irrelevant.
Excel provides us an office type benchmark that can utilize many threads, especially when running the extreme Monte Carlo simulation. Here the Threadripper 1950X really impressed taking just 1.66 seconds. However the 16-core and 18-core Skylake-X parts go one better as they dip down into the 1.4 second range allowing the 18-core part to complete the workload 14% faster.
For those of you unaware, VeraCrypt is an open-source utility used for on-the-fly encryption and features optimized implementations of cryptographic hash functions and ciphers which boost performance on modern CPUs. It also supports parallelized encryption for multi-core systems as well hardware-accelerated AES to further improve performance.
In short, it takes full advantage of the many cores these CPUs have on offer and we see that here. That said, the Threadripper 1950X provided the best results and even the 7980XE couldn't beat it -- a disappointing result for Intel here.
Next up we have non-encrypted compression and decompression performance using 7-Zip. Whereas Hyper-Threading sees a similar level of efficiency when compressing and decompressing, SMT is significantly more efficiency for decompression work.
For decompressing, the Threadripper 1950X is again able to best the 7980XE, though it was much slower for the compression test. If you do a lot of compression work then it looks like Intel high core count CPUs are king here.
Encoding and Rendering Performance
Handbrake is a popular application for encoding video and we've used it to convert a 4K H.264 video to 1080p using H.265 and recorded the average frame rate. Here the 7960XE managed 22.3 fps while the 7960X was actually 8% faster. The reason for this is down to utilization, HandBrake really only utilized about 70% of the 18-core CPU and because the 7960XE has a lower operating frequency the 7960X actually came out on top.
The Premiere Pro CC results interest me the most as I spend a lot of time rendering these 4K videos on my Core i7-6950X editing machine. I do plan to upgrade to Threadripper soon but I have to say the 7980XE and 7960X export times are enticing. That said, it will be interesting to see how many can justify spending 70% more on the 7960X to shave around 13% off the encode time.
It's also worth noting that once again the 16-core 7960X was faster than the 7980XE in this application due to Premiere not being able to fully utilize the 18-core CPU.
Moving on to the rendering tests we have Blender and first we're running the Ryzen Graphic 27 test. Here the 7980XE and 7960X both delivered similar results, taking about 12 seconds each and this meant at best they were 10% faster than the Threadripper 1950X.
The Gooseberry workload takes some serious firepower to complete in around half an hour and we see this as the 1950X turns in a time of almost 32 minutes, though the 7980XE and 7960X were mighty impressive here taking around 23 minutes. Again the 16-core 7960X outpaced the 18-core 7980XE.
The 7960X was a whopping 27% faster than the 1950X for this extreme Blender workload so that was impressive to see.
Corona comes as a standalone benchmark. It renders a fixed scene six times and we take the time it takes to complete the task. This application loves threads so the more the merrier.
Here the 7960X was 14% faster than the 1950X while the 7980XE was 20% faster and this is one of the better results we've seen for Intel. Still a 20% increase in performance for a 100% increase in price might be a tough sell.
POV-Ray is another ray-tracer, it's been around for many years and we're using the official benchmark. Here the 7980XE was just 11% faster than the 1950X while the 7960X was 7% faster. So while Intel new 16-core and 18-core parts are certainly quicker, it's not by much.
Time for a few quick gaming benchmarks before we look at power consumption and temperatures. Here we're testing Battlefield 1 using the DX12 API with the ultra quality preset enabled. All the games have been tested using the GTX 1080 Ti and Vega 64 GPUs, and we have the GeForce results up first.
Here we see that Threadripper doesn't look particularly impressive but it has to be said with over 100 fps at all times the game was incredibly smooth. Please note I'm not bothering with the gaming mode, resetting your workstation to play some games isn't practical or realistic so we're only testing the distributed memory mode.
As for the 7960X and 7980XE, they do quite well despite trailing the lower core parts but that was pretty much to be expected.
When we drop in Vega 64 Liquid Cooled the frame rates don't change drastically for the quad-core Core i7-7740X compared to what we saw with the GTX 1080 Ti installed. However, the AMD processors perform much better and we're now seeing significantly more competitive performance across the board with Vega.
Moving on we have Ashes of the Singularity: Escalation and again we are testing with the GTX 1080 Ti first. Here the Core i9 processors deliver the best results while Threadripper is comparable to the Core i7-7820X.
However, when we retest with Vega we again see much more competitive results from all the tested CPUs. The 1950X matches the 7820X but they are not much closer to the Core i9 models.
Next up we have Civilization VI and here the Ryzen 7 1800X is competitive with the Core i7-7740X when using the GTX 1080 Ti. Meanwhile, the 1950X matched the R7 1700 and both were much faster than the Intel Skylake-X CPUs, particularly the new 16-core and 18-core parts.
Switching the Vega 64 Liquid Cooled again provides quite different results and this time the AMD CPUs are able to pull much further ahead of the Skylake-X parts and even the 7740X falls behind the Threadripper 1950X.
Finishing up the gaming benchmarks we have F1 2017 and this is the only DirectX 11 titles we're testing. Here the Core i9-7960X and 7980XE look weak though keep in mind they're still pushing over 100 fps at all times. Meanwhile the 1950X offers considerably better minimum frame rate performance than even the 7900X, though it's worth noting that the quad-core 7740X is the real hero here pushing over 200 fps for the most part.
Switching to Vega doesn't change too much here though the average frame rate of the 1950X is now much better than any of the Skylake-X parts and in fact Ryzen is the king of the big core count CPUs here.
Power, Temps, Overclocking
Next up we have power consumption and it's important to measure using software that stresses all cores. I've found Corona works well for providing accurate results, so these load figures are based on the Corona benchmark after a single pass and I’m reporting the maximum logged result. This is total system draw and I'm using a Cabac Power-Mate to measure from the wall draw.
The Threadripper 1950X sucked down 257 watts and that meant total system consumption was 8% lower than that of the Core i9-7900X. Interestingly the 7980XE actually drew less power than the 7960X in this test. Both CPUs delivered similar performance so this suggests the 7960X had to clock more aggressively to achieve that result and this meant greater power draw.
Under full load the 7960X system was consuming around 25% more power than the 1950X while the 7980XE consumed 16% more power.
Due to the limited time we had to test these new CPUs I haven't explored the cooling options all that much. I simply threw them both on our custom liquid cooled X299 test bed and got benchmarking. So I’m yet to see how badly they punish air-coolers and AIO liquid coolers.
At stock clock speeds the 7980XE only pushed temps as high as 65 degrees which is certainly getting up there given that we have a massive 360mm radiator attached to the loop.
Let's see how things look after a little overclocking...
Time to tinker. The 7980XE overclocked all cores to 4.1 GHz quite easily and it might be possible to go further but for now this is where I stopped due to power draw and a few other reasons. The 7960X was happy at 4.3GHz and of course we previously got the 7900X to 4.7GHz. Threadripper has also been thrown in as well, clocked at 4GHz.
Clocked at 4.1GHz the 7980XE produced a multi-threaded score of 3974pts so expect to see people breaking the 4000 barrier with this one. That's a 20% increase over the out of the box performance. The 7960X saw a 17% boost as it hit an impressive 3681 pts.
The Threadripper 1950X only managed to boost its out of the box performance by 13% but even so that allowed for a score of 3425 pts. Overclocked, the 7980XE was 16% faster than the 1950X. So what difference does this make in our extreme Blender workload?
Well not a huge amount. Both Skylake-X parts are now 9% faster while Threadripper was 11% faster. Still the 16-core and 18-core Intel CPUs are blistering fast in this test.
What about power draw you ask? Overclocked, Threadripper pushes total system draw to 414 watts and that's pretty extreme. However, it also pales in comparison to the 7900X at 475 watts, the 7960X pushing total system draw just over 500 watts and finally the 7980XE at 530 watts. Remember this is just CPU load so if you have a few high-end GPUs in your workstation you might to require a ZPM to power it.
Price vs. Performance, Conclusion
As usual we have quite a bit of data to sort through. Of course, it's not just about delivering the best performance and instead most people are interested in bang for their buck.
First up we have Blender, this is a lower is better scenario which is why the numbers at the bottom of the chart work backwards. The faster a CPU the further right it will be and the lower the price the lower it will be situated, so ideally you want to be as far right as possible and as low as possible on the chart.
It's a little confusing so I've added a line indicating the very best value CPUs start from the Ryzen 7 1700 and going up to the 1950X. Anything along this line offers a good balance between price and performance. As you can see the only Intel CPU to touch on this line is the Core i7-7820X.
While slightly faster than the 1950X we see that in terms of value the 7960X and 7980XE are shocking -- who's surprised?
Looking at HandBrake we see that the Intel Core i7-7800X and i7-7820X look to be mixing it up with the AMD CPUs. Following our linear line from the R7 1700 to the 1950X we see that the 7800X and 7820X actually present as better value options when compared to the AMD CPUs. Even the 7960X moves closer to the line in this test though ultimately isn't great in terms of value while the 7890XE is again horrible.
Finally let's take a look at the price vs performance ratio when testing with Premiere Pro CC. Here we again see the AMD processors look to all deliver a similar bang for your buck while the Core i7-7800X and 7820X are mixed in there as well. Looking at our line we see that the Core i7 CPUs, with the exception of the 7740X, do quite well. The new 7960X and 7980XE, however, don't look so great when it comes to value.
The price to performance scatter plots summarize the situation quite well. The new Intel Core i9 CPUs are blistering fast in every test but ultimately weren't that much faster than the Threadripper 1950X. Considering the fastest Core i9s cost anywhere from 70 to 100% more, it's hard to justify what often is a 10% increase in performance.
Overclocked, these Skylake-X parts aren't worlds faster either -- about 15-20% over an overclocked 1950X yet they draw at least 20% more power. This puts Intel's new 16-core and 18-core CPUs in a bit of a jam if you ask me. Now that we've seen how they perform it looks like the conclusion I drew in my Threadripper 1950X and 1920X review is pretty spot on.
I said at a guess it looked like consumers would need to spend at least $1,400 for the i9-7940X to match AMD's 1950X and I seriously doubted the 40% price premium would be worth it. This certainly looks to be the case now that we've seen the 16-core model in action.
Besides its extreme price, I also noted that the X299 platform's lack of ECC memory support is an issue. Whereas Threadripper supports ECC memory, the Skylake-X chips don't and that means anyone who is serious about their workstation won't even consider Intel's high-end desktop platform.
Finally, I said that unless Intel is willing to move on pricing I couldn't see why anyone would invest in the X299 platform. That was my honest opinion and it's just as true with the arrival of these 16-core and 18-core parts from Intel. Unless I simply had the money to burn, I struggle to imagine a scenario where I would spend $1,000 (100%) more on the Core i9-7980XE to gain at best 20% more performance.
That said, there are businesses that could justify the investment and by being the world's most powerful desktop/workstation CPUs, Intel's new chips aren't irrelevant -- at least until AMD admits that Threadripper is nothing more than a disabled EPYC CPU and releases a 24-core processor in retaliation. The company says that won't happen, but our bet is in.
Pros: Fastest overall workstation performance available, especially for compression.
Cons: Terrible value compared to AMD's Threadripper chips. No ECC memory support.