Back in the day, a new CPU, be it a refresh or brand new architecture, users have come to expect an increase in both clock speeds, as well as some form of IPC increase. Today, in the era of what I like to call ‘core wars’ AMD has started going wide and adding more cores and threads to their CPUs while Intel was behind in core count and pricing, but excelled in single threaded performance as well as loads that did not surpass the core/thread count. In an effort to answer the competition on this platform, Intel has released the Core i9-9900K Processor sporting eight cores and 16 threads with base clocks at 3.6 GHz and Boost clocks hitting the magic 5 GHz mark. This is a core count increase from the i7-8700K which had 6c/12t setup using a faster base clock of 3.7 GHz but slower boost clock of 4.7 GHz.
This review will put those CPUs against each other and see who comes out a winner in our testing suite. We will also talk about the value proposition of this CPU versus others that exist in the market including Intel’s own, as well as AMD’s 2700X (8c/16t) processor. Read on to see the details and see how the product shapes up in our testing suite.
Specifications and Features
Intel’s latest line of desktop processors includes several changes which, on paper, should make the power user and overclocker smile. The new CPU brings with it up to eight cores and 16 threads using higher frequencies, integrated hardware security updates for Spectre and Meltdown, as well as using the STIM which should yield better thermal conductivity. The other side of the coin is the processor is more expensive, and it actually uses more power (regardless if the TDP is the same at 95W).
There are a total of three new CPUs that have been launched, the 5 GHz 8c/16t Core i9-9900K (first mainstream processor to carry the i9 designation), an 8c/8t i7-9700K, and the final model a 6c/6t Core i5-9600K (the i7 and i5 do not have Hyperthreading). All three SKUs are of the K variety and use an unlocked multiplier for easy overclocking. In this latest launch, Intel seems to be moving away from Hyperthreading which will slim down the product stack and helps prevent the CPUs performance/price from tripping on each other.
The cache structure has also changed a bit on these CPUs, particularly the L3 cache. In the past, Core i7 processors had 2 MB of L3 cache for each core, the i5 with 1.5 MB, and the i3 chips had either 2 MB or 1.5 MB. For the 9 series CPUs, Intel only allows the full cache allotment on the Core i9 parts with the i7 dropping down to 1.5 MB of L3 per core.
Pricing has also gone up a fair amount with the 9900K asking $488 (tray = 1000) with the price for single units even higher. This is a $100+ increase over the 8700K which was released at $359 which also sold more when not bought in 1000 unit batches. Newegg currently has these on sale for $580 which is quite a bit over the expected MSRP. This price premium is not unusual considering the lack of stock for the flagship CPU. For about what amounts to a ~35% price increase, gives users 33% more cores and threads, as well as a small clock speed bump. The 9700K is priced at $374 and when compared against the 8600K (released at $260) fetches a $114 premium, or 43% increase. Newegg has these currently for $420. The 9600K priced at $262 and looks to be a potential sweet spot for the budget oriented crowd looking for the latest and greatest. With 6c/6t it will power through most any game without a performance hit due to a lack of cores/threads. These can currently be found for $280 at Newegg and the most likely to find in stock.
|Intel 9th Gen Core Product Stack|
|Core i9-9900K||$580||8 / 16||3.6 / 5.0||95W||16 MB||2666||GT2|
|Core i7-9700K||$374||8 / 8||3.6 / 4.9||95W||12 MB||2666||GT2|
|Core i5-9600K||$262||6 / 6||3.7 / 4.6||95W||9 MB||2666||GT2|
The new chips make a slight change on how turbo works with the 9900K now able to boost TWO cores up to 5 GHz. In the past, and on chips lower in the stack, peak boost is typically with one core. The 9700K will boost one core to 4.9 GHz while the 9600K’s maximum boost is also using one core to 4.6 GHz.
|Intel Core 9 Series Turbo Boost (Non-AVX)|
|i5-9600K||3.7 GHz||4.6 GHz||4.5 GHz||4.4 GHz||4.4 GHz||4.3 GHz||4.3 GHz|
|i7-9700K||3.6 GHz||4.9 GHz||4.8 GHz||4.7 GHz||4.7 GHz||4.6 GHz||4.6 GHz||4.6 GHz||4.6 GHz|
|i9-9900K||3.6 GHz||5.0 GHz||5.0 GHz||4.8 GHz||4.8 GHz||4.7 GHz||4.7 GHz||4.7 GHz||4.7 GHz|
|i7-8700K||3.7 GHz||4.7 GHz||4.6 GHz||4.5 GHz||4.4 GHz||4.4 GHz||4.3 GHz|
Outside of what is mentioned above, the processor is still built on the dated, yet effective, 14 nm process node which with this iteration of the 9th Generation Coffee Lake refresh is dubbed “14nm Class” by Intel. The 14nm process has been around since 2014 introduced with 5th generation Broadwell-based CPUs, refined 2016 with 7th generation Kaby Lake (14nm+), and then with 8th generation Coffee Lake S/U/H and Whiskey Lake-U CPUs debuting in 2017. It’s a bit long in the tooth, but overall it is keeping Intel in the game while they develop new chips on a smaller process.
With new CPUs comes a new line of motherboards from the board partners based on the latest Z390 chipset. The Z390 chipset was released in mid-October bringing with it native USB 3.1 (10 Gbps) ports as well, integrated CNVi based Wi-Fi, all integrated on the chipset. The Z390 chipset is “made for” the new 9-Series CPUs accepting them natively. On the other hand, Z370 chipset based motherboards will require a BIOS update in order to work with these new CPUs. We published an article going over some of the features of 50 new Z390 motherboards that is worth a read.
Meet the i9-9900K
Before things get started, below are images of the 9900K in CPUz at its stock settings with XMP enabled.
And finally, a picture of the packaging as it was sent to us, and the CPU itself alongside an i7-8700K. Note in the last image, the PCB used is thicker than the 8700K.
Test Setup and Results
Here we take a slightly different approach to CPU testing with ours based on a lot of Hwbot.org benchmarks since that is what we are known for, overclocking and benchmarking. We use real-world testing as well with Cinebench, x265, POV-Ray, and 7Zip in order to give readers a good idea of the general performance of the product tested.
|Test System Components|
|Motherboard||MSI MPG Z390 Gaming Edge AC|
|CPU||Intel i9 9900K and Intel i7 8700K (stock)|
|CPU Cooler||EVGA CLC 240|
|Memory||2×8 GB G.Skill Trident Z 3200 MHz CL15-15-15-35|
|SSD||Toshiba OCZ TR200 480 GB (OS + Applications)|
|Power Supply||EVGA 750W G3|
|Video Card||NVIDIA RTX 2080 (411.63 drivers)|
Thanks go out to EVGA for providing the CLC 240 CPU Cooler and 750 W G3 Power Supply to cool and power the system, G.Skill for the Trident Z DRAM, and Toshiba OCZ for the 480 GB TR200 SSD storage running the OS, benchmarks, and games. With our partners helping out, we are able to build matching test systems to mitigate many differences found between using different hardware. This allows for multiple reviewers in different locations to use the same test system and compare results without additional variables.
- AIDA64 Engineer CPU, FPU, and Memory Tests
- Cinebench R11.5 and R15
- HWBot x265 1080p Benchmark
- SuperPi 1M/32M
- WPrime 32M/1024M
- Intel XTU
All CPU tests were run at their default settings unless otherwise noted.
We have updated our gaming tests and dropped down to two games for CPU reviews. In many cases, even at 1080p, the difference between CPUs isn’t that much and the two titles we use covers both a CPU heavy title in AOTSe as well as a GPU bound title in SOTR. All game tests were run at 1920×1080 with all CPUs at default settings unless otherwise noted. Please see our testing procedures for details on in-game settings.
- Shadow of the Tomb Raider
- Ashes of the Singularity: Escalation
AIDA64 CPU, FPU, and Memory Tests
The memory channels and configuration hasn’t changed on these refresh CPUs, so bandwidth results should be similar to the 8700K or 8086K at default settings. We can see the raw data below the chart the 8700K and 9900K are quite similar when using the same sticks of RAM. The 8086K was using faster RAM speeds results in its notably higher results. We have seen some pretty high memory speeds achieved with this CPU and Z390 based boards, in particular on ITX board(s) due to their shorter trace paths, so if pushing memory is preferred, a 9-series CPU and Z390 based board can help in that endeavor.
|AIDA64 Cache and Memory Benchmark – Raw Data|
|Intel Core i9 9900K Defaults||47163||46232||43269||45.5|
|Intel Core i7 8700K Defaults||46988||46231||42230||44.0|
|AMD Ryzen 2700X Defaults||49063||47287||45180||66.5|
In the AIDA64 CPU tests, we are able to see what the higher core count and clock speeds bring to the table. Nearly across the board we see the 9900K beating out the other Intel CPUs tested. The AMD Ryzen 2700X shined bright in the AES and Hash testing leading the pack.
|AIDA64 CPU Tests – Raw Data|
|Intel Core i9 9900K Defaults||101379||23297||843||43137||10854|
|Intel Core i7 8700K Defaults||72487||25265||573||29420||7402|
|AMD Ryzen 2700X Defaults||94678||25718||742||71525||24429|
Moving on to the AIDA64 FPU tests, the 9900K performed well in the Julia and Mandel tests easily beating both the 8700K and 2700X. The VP8 testing shows similar performance out of the Intel with the AMD offering almost 10% faster here. SinJulia results show the 9900K easily out working the 8700K but still behind the AMD Ryzen 2700X.
|AIDA64 FPU Tests – Raw Data|
|Intel Core i9 9900K Defaults||6830||80539||43140||11411|
|Intel Core i7 8700K Defaults||6765||54933||29561||7788|
|AMD Ryzen 2700X Defaults||8151||41789||21831||13775|
Real World Tests
Moving on to the real world testing, the 9900K shows a clean sweep across these five benchmarks. It bests the 8700K by 29-40% and is 17-37% faster than the 2700X. Overall a good showing for the CPU.
|Cinebench R11.5/R15, POVRay, x265 (HWBot), 7Zip – Raw Data|
|Intel Core i9 9900K Defaults||22.3||2054||4350||67.3||69831|
|Intel Core i7 8700K Defaults||15.3||1417||2991||47.2||40632|
|AMD Ryzen 2700X Defaults||19.3||1826||3729||45.9||47344|
Pi and Prime Based Tests
Next up are the Pi and Prime number based tests along with Intel’s own stress testing and benchmarking utility, XTU. In this set of testing, the 9900K sits on top of the heap due to its faster clock speeds. All that is, except for Intel XTU. At this moment we are not entirely sure why this result was so and are looking into it.
|SuperPi and wPrime Benchmarks – Raw Data|
|Motherboard||Spi 1M||SPi 32M||WPrime 32M||WPrime 1024M||Intel XTU|
|Intel Core i9 9900K Defaults||7.407||410.995||2.563||71.825||3254|
|Intel Core i7 8700K Defaults||7.865||429.363||3.467||99.57||2368|
|AMD Ryzen 2700X Defaults||9.75||543.338||3.219||84.188||N/A|
For gaming, we are using Ashes of the Singularity: Escalation in DX12. The game leans pretty heavily on the CPU for all it has to do in the game, while Shadow of the Tomb Raider uses a fair amount of CPU, but a more typically GPU bound title. As we can see from the graph below, there was little difference in either title at 1080p.
Power Consumption and Temperatures
Our power consumption results are interesting in that the 9900K, though faster and with 33% more cores, runs at the same TDP as the 8700K of 95W. Intel defines TDP as a value of the cooling required to maintain its base frequency. TDP does not take into account turbo speeds and as one may imagine, the all core turbo value is notably above the 95W TDP.
Intel uses a set of variables called Power levels: PL1, PL2, and PL3. PL1 is the cooling limit (TDP), PL2 is sustained power delivery (Turbo), and PL3 is the power delivery limit. PL2 is the maximum sustainable power the CPU can handle until thermal issues. Intel has set the value of PL2 to 210W though board manufacturers can set their own. Really, if you want to get the most out of this CPU, you are going to have to bring the cooling, even with the improved thermal interface material.
We did not get too granular here and used system power use tests, but we can see the 9900K in stock form is using a fair bit more power than the 8700K. When pushing it using the latest Prime 95 with AVX/FMA3 instructions, the system peaked at 322W at the wall.
Temperatures at stock speeds running these tests were just about our thermal limit. Using a test that is a little less stressful like AIDA64 shows the CPU hit 82 °C at stock using EVGA’s CLC240 CPU cooler. I am imagining hitting around 5 GHz all cores and threads with this cooling when all is said and done. Adding an AVX offset could help push past that value saving a few degrees in those heavy tests.
Once again we arrive at my favorite part of the review, the overclocking. The 9900K did bring with it some excitement from the enthusiast and overclocking community when rumors came out Intel moved back to the more effective STIM. This, in theory, should yield better temperatures and more overclocking headroom assuming all other things remain the same. But due to how the processor works and how much more power it can draw, and therefore heat it will produce, we are back to the cooling used, and silicon lottery as to how high we can push the chip.
Our testing consists of running the AIDA64 stress tests in its default form. While this isn’t the most stressful test we can put it under, our settings do NOT use an AVX offset and we still wanted to overclock to the limits of our cooling while leaving as little meat left on the bone while still being “stable.” We are running the A64 test for 30 minutes as this will get us the maximum heat and be considered at least stable enough to benchmark.
Overall the CPU performed quite well reaching a maximum overclock of 5.1 GHz at 1.30V without an AVX offset. This compares similarly to the 8700K CPUs we have overclocked in the past, but the big difference here is the additional cores and threads (2/4 respectively) the 9900K provides over the 6c/12t 8700K. Additional speed can be gained by using an AVX offset as that will allow for additional temperature headroom.
MSI shipped us their MPG Z390 Gaming Edge AC ($190 @ Newegg and Amazon) motherboard for this review and this mid-range board handled all the overclocked testing without flinching. The board is said to be a 10-phase (5 doubled) unit and even includes an optional 4-pin CPU power lead (not used in this testing). As the name implies, the board has built-in wireless networking handled by an Intel 9462 802.11ac Wi-Fi adapter and supports the latest Bluetooth 5 connectivity. The board has four DRAM slots with support for up to DDR4-4400 with a total capacity of up to 64 GB. Storage is managed by a pair of M.2 PCIe 3.0 x4/SATA slots as well as six SATA ports. The board has three full-length PCIe slots with the top two slots using MSI’s Steel Slot armor. Two-way SLI and 3-way Crossfire setups are supported. The back panel includes two USB 3.1 (10 Gbps) ports, Type-A and Type-C, along with two USB 3.0 ports and two USB 2.0 ports with further expansion through internal headers. The audio is handled by the Realtek ALC1220 HD audio codec.
Overall on the overclocking front, it looks like we have had some improvement from Intel with the 14nm tweaks as well as going with the more effective thermal interface material. Though the potential clock speeds reached are still limited by cooling, we are able to get 8c/16t to 5.14 GHz where with the 8700K tends to lose steam and need more cooling around 5 GHz. Now, that is just with this sample, which seems is pretty average when it comes to the silicon lottery. As always, your mileage may vary.
The 9900K has been met with seemingly equal parts positive and negative press. On one hand, Intel has expanded the mainstream line in order to compete with AMD’s similar (by thread count) offerings on their mainstream lineup. This brought 8 core processors from the more expensive HEDT line down to where the masses tend to purchase. With this came slightly increased clock speeds, boosting to 5 GHz with two cores/threads. A feat only managed by the 8086K limited edition processor (and that was only using one core/thread). Price wise, the comparable AMD CPU, a 2700X blows it out of the water costing about half as much. The price to performance ratio doesn’t work out terribly well here, but, that is the price one has to pay for, in general, the faster CPU (clock speed, single threaded, gaming).
The bottom of the product stack poses quite an interesting value proposition, however. The 6c/6t 9600K can be found for $280. As mentioned above, 6c/6t should be enough for most users to game on and of course other activities without being held back by a lack of cores and threads. This also leaves a fairly significant upgrade path as well. Even here, AMD holds the price to performance and value segment in most cases.
We saw the STIM in full effect with our overclocked testing allowing our sample on our “average” cooling (2×120 MM CLC) to reach 5.14 GHz at 1.3V. This was the limit of our cooling, but not really the CPU. I am certain with a better cooler 5.2 GHz+ is in the cards. Remember, we do not use an offset for our overclocking, but use a slightly less stressful program in AIDA64 to determine relative stability.
In the end, the Coffee Lake refresh CPUs look to stick it to AMD’s clock speed wise and single threaded performance. Some tests show the AMD CPU taking a lead due to where it shines (Hash testing to name one) and its slightly better hyperthreading efficiency. Still, in most tests, this is the fastest CPU Intel has put out to date. Whether or not the increased price of the CPU is worth it for the gains is up to the buyer, but Intel put out a solid performing chip with more cores and threads than they have ever used in the mainstream platform. The trick is trying to find one available online at a reasonable price considering stock on these are hard to come by.
Joe Shields (Earthdog)