Intel Core i7-8700K and i5-8400 Review: Coffee Lake Goes With More Cores
A few days ago we gave a preview of what Intel shared with us and some ideas as to what the bigger picture may look like. We also took some flak for a recent review of a processor that had been out for quite some time. At the time I suggested that there was a reason for this, but was unable to say much more. That changes today. Intel is beginning sales of a portion of its Coffee Lake lineup and we’ve had a chance to put the brand new 6-core Core i5-8400 and headliner Core i7-8700K to the test in Ubuntu. We know there are going to be a lot of questions and hopefully this first look at the new lineup will provide the answers you are looking for.
Editor’s Note: The i7-8700K and i5-8400, along with multiple components used for testing, were provided to XDA for review purposes. Please refer to the Test Configuration section for additional details.
Rumor Control: Coffee Lake Edition
Before we get into the details and benchmarks though it’s time to take a cue from movies of old and address some rumors that are lingering about Coffee Lake and the LGA1151 socket that looks the same to its Skylake and Kaby Lake predecessors. To begin, Coffee Lake is not the same pinout as the previous LGA1151 processors. This particular rumor stemmed from multiple outlets reporting that anonymous sources from motherboard partners were stating that they were physically the same processor. Thankfully the pin layout was revealed and showed once and for all that they are indeed incompatible with each other.
Intel's #CoffeeLake pin changes:
Kaby Lake -> Coffee Lake
VSS (Ground): 377 -> 391 (+14)
VCC (Power): 128 -> 146 (+18)
RSVD: 46 ->25 pic.twitter.com/wu8sWLZHWZ
— David Schor (@david_schor) October 2, 2017
The other rumor that quickly arose when Intel announced the Coffee Lake lineup is that this was driven by the success of AMD’s Ryzen processor lineup. And to a certain degree it may have very well been influenced by AMD’s announcements of an 8-core processor since Intel may have had enough time to factor that in. But with the Kaby Lake i7-7700K Intel was once again reaching a point it had seen before with the Pentium 4 lineup: thermal limits that prevented it from going any higher with the status quo. This left them with few options on the table, including the one to add additional cores. And this is something they would have known long in advance of the i7-7700K going to market.
Proponents of this rumor do have additional item in their favor for this: the staggered release that we are seeing with Coffee Lake. In the past Intel usually does a broad release of its mainstream CPU lineup. Coffee Lake breaks with that by its staggered release, opting to go with the unlocked i3/i5/i7 variants today as well as a few locked options. The remainder of the lineup, along with additional chipsets to support it, will not be coming until 2018. Would we have seen this if Ryzen had not performed as well as it did? We won’t know for certain until we see sales announced by both Intel and AMD later this quarter, but my guess is that we would not.
CPU, Motherboard Unboxing & Photos
When we received word that we would be receiving samples from Intel to review, that also meant we needed a motherboard to test with. Thankfully the friends at GIGABYTE have once again assisted by offering us the Z370 AORUS Ultra Gaming to accomplish our testing. Unlike the Z170X GAMING 7 that we tested the i7-7700K with, the Z370 AORUS Ultra Gaming does not include a Thunderbolt 3.0 port on the back. Instead it offers a header that, when used with an expansion card such as the Alpine Ridge, allows for native Thunderbolt support. Another nice feature about this motherboard are the raised m.2 sockets, which has made it far easier to add and/or remove m.2 drives as necessary. The lighting is very similar to the AORUS X370 GAMING 5 that we received as part of the Ryzen review package from AMD and continues to impress.
In addition to the motherboard and CPUs Intel has provided a 32GB Optane drive since the Z370 chipset supports it. However, after checking with Intel it does not appear that it is supported yet in Linux. Since our testing generally involves using an m.2 drive to minimize any potential I/O bottlenecks we will be revisiting this particular item at a later time. There have been attempts to put Optane to use in Linux in cache modes like bcache and we hope to explore that and other potential options in the future.
The processors themselves did not come in retail packaging so there isn’t much to show beyond the CPU itself. Regardless, we do have pictures of all of this and are adding them to the gallery below for your viewing pleasure!
Much of the test configuration here has not changed from the Kaby Lake review. Since the results of that review were obtained within the past 30 days we performed a build test of LineageOS to verify nothing had significantly changed that would prevent us from using those results. After this was verified we began testing of the i7-8700K at stock speeds using the exact same Ubuntu 17.10 environment, accomplishing a full run of the Phoronix Test Suite and LineageOS builds of the Pixel XL.
At that point the RD400 experienced yet another lock state that, at the time of writing this has still not reset. As a result we were forced to pull the RD400 and use an alternate m.2 drive instead. We compared build times with the RD400 and its replacement to verify that the drive had not impacted significantly affected performance. After that was verified we continued testing of Coffee Lake using the alternate drive. In each case efforts were made to minimize variables in testing to ensure consistency in our results.
Not all items were not purchased directly by XDA or me, and I state who provided each non-purchased component in the list below.
Shared Platform Configuration (Between AMD & Intel)
- Corsair CX-750M 80 Plus Gold Power Supply
- OCZ 512GB RD400 m.2 NVMe drive (Used for all but Coffee Lake)
- MSI NVIDIA GeForce GTX 1080 Founder’s Edition
- Lian Li PITSTOP PC-T60
- GeIL EVO X DDR4-3200 16GB (provided by AMD)
- VisionTek 120GB 2.5″ SSD x2 (provided by VisionTek)
- Cooler Master MasterLiquid Pro 240
AMD Ryzen (AM4) Testing Platform
Intel Core i7 (LGA 1151) Skylake/Kaby Lake Testing Platform
- GIGABYTE Z170X-Gaming 7 (provided by GIGABYTE)
- Intel Core i7-6700K
- Intel Core i7-7700K (provided by Intel)
Intel Core i7-8700K/i5-8400K (LGA 1151) Coffee Lake Testing Platform
- GIGABYTE Z370 AORUS Ultra Gaming (provided by GIGABYTE)
- Intel Core i5-8400 (provided by Intel)
- Intel Core i7-8700K (provided by Intel)
- Team T-Force Cardea m.2 480GB (provided by Team Group)
- Ubuntu 17.10
- Phoronix Test Suite – current using apt
Given that the i7-7700K and Z170X GAMING 7 had a preset to overclock the i7-7700K at 5 GHz, it’s not a surprise to see a similar overclock preset for the i7-8700K at the same speed. It is extremely noteworthy given that its 6-core predecessors, most recently the i7-5930K and i7-6800K, typically only reached a max overclock speed of 4.3 GHz. This means that Intel has been able to pull off a near 20% increase in clock speed with the same core and thread count of previous generations.
In daily use of the system for a week, the system was overclocked to 5 GHz and stable without any halts or errors noticed on Windows. But a quick look at HWInfo indicated a few notes of concern, and for those who have been following the thermal challenges of the i7-7700K it should not be a surprise that thermals are a very real concern when overclocking. The MasterLiquid Pro 240 easily keeps the processor in the mid-30s Celcius when at idle. As soon as the i7-8700K is placed under a full CPU load it shoots to the mid-80s within seconds, even if the cooling fans are manually set to their full speed.
In some instances we even noticed the throttle sensors being triggered. The times that it was throttled were limited exclusively to when the processor was overclocked and under full CPU load. In addition the throttling was not observed to remain for a long period of time. Based on the stability within Windows we attempted to test within Ubuntu at the same 5 GHz speed. Unfortunately the rapid increase in temperature was too much and came too quickly for Ubuntu to be able to do the same throttling, causing the system to freeze instead. A decrease in the multiplier down to 4.9 GHz corrected the issue and so we used this as our overclocking results for both Phoronix and LineageOS build times.
Long story short, overclocking the Core i7-8700K will need to dissipate the heat as much as possible to overcome the thermal barriers it introduces. A 240mm cooler appears to handle this for the most part, and we are considering a future delid test of this CPU to see if the temperatures keep it under throttling levels. So long as the temperatures are managed the 6-core, 12-threaded headliner really does beat out even its larger LGA2011 predecessors significantly.
Regarding the i5-8400, the multiplier appeared to remain unlocked in BIOS so we attempted to force a higher multiplier despite the processor being locked. BIOS reported a change up to 3.8 GHz, but test builds of LineageOS suggested that no actual change was occurring. As a result we are not including any overclock results for the Core i5-8400. At stock speeds the MasterLiquid Pro 240 easily tames both and manages to keep them under 80 degrees Celcius, even under full load.
Speeds for all previously reviewed processors remained the same as before.
Benchmark Notes: Phoronix Test Suite’s CPU suite offers a plethora of tests and not all will be included in this review. The full test results for Coffee Lake from the Phoronix Test Suite are available on OpenBenchmarking.org. The previous test results from the i7-7700K review are already available here. Benchmarks are color grouped by Ryzen, Skylake/Kaby Lake and then Coffee Lake in a color scheme that more closely follows other graphs on XDA.
FFTW is a single-threaded benchmark of fast Fourier transform. Intel once again sees notable gains by overclocking due to the additional headroom – but just like Ryzen we see no additional gains with the additional 2 cores.
GZip is a common compression method and so it makes sense to check out the performance here. Just like we have seen in previous reviews Intel continued to do better here due to its single threaded performance, but note that the additional cores didn’t help here either, except that the 6 core i5 now shows single threaded performance matching the Skylake i7-6700K, despite having lower clock speeds.
SciMark 2 (Java) v1.3.0
The SciMark 2 benchmark utilizes Java for arithmetic operations and then provides scoring based on those results. Intel continues seeing its wins in this new version. The trend for single threaded performance though: More cores doesn’t equate to better performance.
John The Ripper
On the cryptography front, John The Ripper offers similar results as the previous two reviews of Ryzen and the Kaby Lake i7-7700K. We see some strange numbers in both the i5-8400 and the i7-8700K at stock speeds, but since the numbers are much closer to what would be expected at the overclocked 4.9 GHz it is worth noting here. In the end Intel finally regains the top spot, despite having two less cores and 4 less threads than the Ryzen 7 1800X. (We also saw abnormal results with the PHP timed build, so we are excluding that for now.)
C-Ray also continued the multi-threaded results and shows that the additional cores and threads of the Coffee Lake i7-8700K have indeed matched against the similar core and thread counts of the Ryzen 5 1600X. But in this particular test core and thread counts do make a difference – and so the Ryzen 7 1800X takes a top spot. We didn’t see this in many other test results and it will be interesting to see how Intel fares once it introduces its own 8 core mainstream processor.
Benchmarks: Build Performance
Build Test: ImageMagick
Build tests were working in AMD’s favor thanks to the additional cores and threads of the Ryzen 7 1800X. But now as Intel has added more cores and threads to its mainstream flagship, the Core i7-8700K wrests the top spot from its rival.
While Google announced yesterday the second generation of the Pixel lineup we’re still unable to build from it. As a result we stay with the Pixel XL for timed build benchmarks. As requested by readers the graph will display both build times with and without caches. But it’s impressive again just how well a new Core i5 holds up in comparison to the previous 4-core, 8-threaded flagships. Intel has the fastest build times of the group but does so with a great clock speed advantage over Ryzen.
Intel Coffee Lake Pricing and Final Thoughts
2017 is a fantastic year for consumers and Intel’s change of its mainstream lineup just shook this up even more. Looking solely from within Intel, the price to get what until today was an i5 just dropped significantly, and the doorway into 2 additional cores at the i5 pricing level means that consumers can now achieve a lot of what was done with a Core i7, even with slower clock speeds, for far less. Users looking to upgrade from an older 6 core enthusiast processor can now do so without having to enter the X299 and X-Series lineup, and if the rumors are true 8 core owners will have their day in 2018. But in order to truly harness the power at these higher core/thread counts as well as clock speeds proper cooling is a must – and while Intel is suggesting that the thermals are correct for the average user, overclockers in particular will want to strongly consider their options to lower temperatures without sacrificing clock speeds.
The pricing provided by Intel though, as shown in the slide above, does concern me. If that pricing is the reseller price for 1,000 units, I expect the consumer will pay no less than this and likely more. We’ll update this once we see pricing comes out and know for certain. I would not be surprised to see pricing higher than years past due to the higher core counts, but how much more will affect how attractive these products become to the consumer.
Considering this in the larger market of AMD Ryzen, consumers will have a choice to make. AMD Ryzen will effectively continue its strategy of offering performance at a lower price point, especially if buyers are savvy enough to look for the gems in the lineup. For example, we have already noted how the Ryzen 7 1700 can often match the 1800X or get very close to it but at significantly less than other options from both AMD and Intel. Motherboard costs will also be attractive since a larger range of products are available for Ryzen, at least until the remainder of Intel’s 8th generation lineup arrives in 2018. The same will apply at the lower and middle ranges of each manufacturer’s lineup.
I cannot say it enough. If readers are in the market for a new PC or upgrading their own, it’s worth a good look at this time to explore your options. Those options are now wide open and offer consumers performance upgrades at every price point, now including Intel’s own lineup. Decisions like this may be hard, but it’s great to see that those decisions are getting harder – all thanks to the processor market getting a lot more competitive than it has been in recent years. Even if decide to hold off, all signs are pointing to this just being the start of that tight competition. Good.