News/Info Core i7-10700K processor review by Guru3D

[emailx45]

Core i7-10700K processor review
Pack it up, pack it in, let us begin.
Guru3D - 07/Jul/2020

Spoiler: Content - 1st part

[SHOWTOGROUPS=4,20]
An overdue review is the 8-core Core i7-10700K processor from Intel. The 10th generation Intel Core desktop processors previously known as CML aka Comet Lake has been received with mixed feelings. Will the Core i7-10700K be a more attractive offer? It has been a long wait, but finally, we can share with you features, metrics, and benchmarks of the new 10th generation Intel Core desktop processors, previously known as CML, aka Comet Lake with the second-best flagship model Core i7-10700K, has eight cores and can reach a turbo frequency of up to 5.0 GHz, and 5.1 GHz with proper cooling.

What a ride the past 12 months already have been. We've seen many-core plethora releases from AMD followed by Intel Cascade Lake-X series HEDT processors. With the more recent introduction of Ryzen 3000 and the announcement of Threadripper 3000 processors the processor market and channel once again are turned upside-down (with a multitude of processors to choose from). And then everybody was waiting for that mainstream desktop processor release from Intel codenamed Comet Lake-S. Most of us expected this desktop series far earlier back in Q4 2019 already, but then COVID-19 struck the world, slowing any and all technology release down.

i7-10700K	8C / 16T	3.8 GHz	5.0 GHz	5.1 GHz	4.7 GHz	125W	$374

A few weeks ago intel released many forms and products SKUs starting with the unlocked K series, the flagship desktop Core i9 10900K processor has a very significant 10 cores, which for intel is a pretty staggering number as their primary processor business for the desktop consumer market had been releasing and refreshing quad-core processors for many years. That slowly changed thanks to pressure from AMD, and with the more recent Coffee lake series Intel has stepped up their game, and we have seen some nice up-to 8-core parts in the series. The reality is simple, you cannot point your finger Intel for that long quad-core strategy either, as there was too little competition - hence they had no rush and have been relaxed all the way. However, they've been perhaps a bit too comfortable in that position. And with AMD a shift in that paradigm changed everything, rapidly. All of a sudden there were 6 and 8 cores processors, nice and fast. But two to three years ago another dynamic changed as Intel got (more) hit by another problem, security weaknesses. With the industry and the world in general changing, everything online is attacked in some way or form. Yes, processor vulnerabilities started to stack up for Intel, statistically significantly more than AMD. And albeit the vast majority of these vulnerabilities have been patched in software or firmware, it did have a tangible effect on performance, as well as Intel's long-lasting reputation. Current processors starting at Coffee and Comet Lake have been hardware hardened for the bigger part of Spectre and Meltdown like varieties as well as other vulnerabilities. To date, this, however, remains a sore spot and point for Intel. This entire paragraph was the perfect storm for the competition, in which they thrived.

The 10th generation into Core desktop processors running up-to 10-cores. It requires a new platform as the processor socket has changed towards 1200 pins, aka LGA1200. On April 30th of this year, 2020, Intel announced its tenth generation Core processors codenamed “Comet Lake-S”. Among the main features we have HyperThreading through the entire line of Core products, so that's from Core i3 to Core i9, up to 10 cores and 20 threads, and up to 5.3GHz for a single-core boost. As mentioned, these processors require a new motherboard, the 400 Series motherboards with an LGA 1200 socket. The new motherboards this round should last for Comet Lake-S (this gen) and Rocket Lake-S (Next-gen). Much like Generation 9 (Coffee lake-S), Comet Lake-S and the following next-gen Rocket Lake-S are based on Intel's 14nm process and you could safely say are based on the underlying microarchitecture that is Skylake, but over time refined and advanced and scaled upwards.

That means the flagship desktop processor is the Core i9-10900K with ten cores + twenty threads hyperthreading (SMT), where a single core can go up to 5.3 GHz. A Core i7-10700K with eight cores and sixteen threads would lock in at of 5.1 GHz on a single core. There are also an unlocked Core i5 processors, the best being the Core i5 10600K which we offer as a review in another article. The six cores and twelve threads will run a maximum of 4.8 GHz. You will notice the flagship processors to have a Turbo Boost frequency listing a 5.3 GHz thanks to what is called Intel Thermal Velocity Boost technology. If the processor is kept cold enough, it can run a turbo bin at that frequency.

On the next page, you can look closely at the processor prices and compare these back and forth, as Intel as roughly slashed prices in half. Most Comet Lake-S processors in the high-end segment are tagged with a 125W TDP. Priced at 374 USD, for this review, we'll be checking out that 8-core model Core i7-10700K in this specific review.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 2nd part

[SHOWTOGROUPS=4,20]
10th Generation Core series processors (CML)
Intel has been expanding with an up-to 10-core and thus 20-threads processor in the mainstream desktop segment (thus not HEDT). Intel markets this series as its Core processor series whereas HEDT is Core-X. More recently it has been adding the denominator the Core i9 series into that desktop lineup. Comet lake-S based processors need to be paired with the 400 series chipset for motherboards, as it has a new processor socket, LGA1200. The top of the line processor is the Core i9-10900K with 10 cores and 20 processing threads (SMT). It's Base / Turbo frequency is 3.70 / 5.10 GHz (TB2) / 5.2 (TB3) GHz and 5.3 GHz with what is called a velocity boost. It has 20 MB of L3 cache, 16 PCI-Express 3.0 lines on the CPU, DDR4 @ 2933 MHz Dual-Channel memory configuration support, and a 125W TDP.

• TB2 - Intel Turbo Boost Technology 2.0: Dynamically increases the processor frequency up to 4.6 GHz when applications demand more performance. Speed when you need it, energy efficiency when you don’t.
• TB3 - Intel Turbo Boost Max Technology 3.0: Identifies the four best-performing cores to provide increased single- and dual-core performance up to 5.2 GHz, quad-core performance up to 5.1 GHz.

Velocity Boost
Intel has a fairly new classification for its highest boosting frequency, this is called 'Thermal Velocity Boost'. This is an extra boost of 100 MHz, on both the single and all-core turbo, however, it is only active under certain conditions. The conditions, as you can tell by its name, are the temperature of the processor, which needs to remain below 70 degrees Celsius. Once the processor gets warmer than value, the processors that support Velocity Boost (currently the four models in the 10900 SKU range). This means that once the Core i9 10900K passes that heat signature, the single-core boost is limited to 5.2 GHz or even 5.1 GHz without Turbo Boost Max 3.0. This technology was introduced with Cascade Lake-X and is available towards mainstream desktop to ensure single-threaded tasks run on the fastest cores. Pretty much end-users with liquid cooling will benefit from this the most.

Architecture changes
The Comet lake series processors have genuinely been based upon the Skylake architecture as a building block that you know from pretty much all socket 1151 parts, but then scaled upwards and from there on you have seen tweaks and fabrication process updates. So that means Intel is still at its 14nm node. Opposed to expensive architecture changes in caches and so on, Intel has been using the architecture as a modular design, adding cores. Intel applies a familiar L2 cache of 256 kB and L1 is at 64kB in total (32kB data + 32kB instruction caches. However, relative to that the L3 cache is 2 MB per core. With these processors, you will now see three Turbo modes dubbed 2.0, 3.0, and explained in more detail in a lower paragraph, Velocity boost. Comet Lake-S processors offer an increase in clock frequencies and thus performance, as well as addressing security fixes in hardware, better protection against side-channel attacks. This series has 44 PCIe Gen 3.0 lanes accumulated in total (16 from the CPU and 24 available from the chipset and 4 used as DMI 3.0 interlink inbetween the CPU and chipset). You'll notice Wi-Fi 6 and 2.5GbE support for these new processors, don't get confused as that is not native, the controllers simply are made compatible, motherboard manufacturers can opt to add-in the appropriate modules as a feature.

Comet Lake Die, count and spot them, that's 10 cores.



10th Gen Intel Core Desktop Specifications and Prices
It is a bit much to list the pricing of all the models, but we'll cherry-pick some SKUs, the rest you can observe from the slides. The Core i9-10900KF will cost $ 472,-, the i7-10700KF will cost $349,- and the i5-10600KF will cost $232,- We highlight the KF series these as they are the ones that have unlocked multiplier and do not have integrated graphics, similar to the conditions of what AMD offers with Ryzen. The Core i3-10300 goes for $143,- the i3-10100 for $122, these two will be competing with the Ryzen 3 3100 at $99 and Ryzen 3 3300X at $120. Below we created an overview of what is released specifically in the Core range, as there are also Pentiums and Celerons released. Models with a K in the suffix are unlocked, and with an F as a suffix means it has no (active) integrated GPU.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 3rd part

[SHOWTOGROUPS=4,20]

CPU	Cores / Threads	Base Clock	Single Core Turbo	Max Turbo 3.0	All Core Turbo	Thermal Velocity Boost	TDP	USD
i9-10900K	10C / 20T	3.7 GHz	5.1 GHz	5.2 GHz	4.8 GHz	5.3 / 4.9 GHz	125W	$488
i9-10900KF	10C / 20T	3.7 GHz	5.1 GHz	5.2 GHz	4.8 GHz	5.3 / 4.9 GHz	125W	$472
i9-10900	10C / 20T	2.8 GHz	5.0 GHz	5.1 GHz	4.5 GHz	5.2 / 4.6 GHz	65W	$439
i9-10900F	10C / 20T	2.8 GHz	5.0 GHz	5.1 GHz	4.5 GHz	5.2 / 4.6 GHz	65W	$422
i7-10700K	8C / 16T	3.8 GHz	5.0 GHz	5.1 GHz	4.7 GHz	-	125W	$374
i7-10700KF	8C / 16T	3.8 GHz	5.0 GHz	5.1 GHz	4.7 GHz	-	125W	$349
i7-10700	8C / 16T	2.9 GHz	4.7 GHz	4.8 GHz	4.6 GHz	-	65W	$323
i7-10700F	8C / 16T	2.9 GHz	4.7 GHz	4.8 GHz	4.6 GHz	-	65W	$298
i5-10600K	6C / 12T	4.1 GHz	4.8 GHz	-	4.5 GHz	-	125W	$262
i5-10600KF	6C / 12T	4.1 GHz	4.8 GHz	-	4.5 GHz	-	125W	$237
i5-10600	6C / 12T	3.3 GHz	4.8 GHz	-	4.4 GHz	-	65W	$213
i5-10500	6C / 12T	3.1 GHz	4.5 GHz	-	4.2 GHz	-	65W	$192
i5-10400	6C / 12T	2.9 GHz	4.3 GHz	-	4.0 GHz	-	65W	$182
i5-10400F	6C / 12T	2.9 GHz	4.3 GHz	-	4.0 GHz	-	65W	$157
i3-10320	4C / 8T	3.8 GHz	4.6 GHz	-	4.4 GHz	-	65W	$154
i3-10300	4C / 8T	3.7 GHz	4.4 GHz	-	4.2 GHz	-	65W	$143
i3-10100	4C / 8T	3.6 GHz	4.3 GHz	-	4.1 GHz	-	65W	$122

The Series T
So many suffixes, A "K" means an unlocked processor. A "U" means the chip is designed for laptops and mobile devices, as "U" chips are Intel's "ultra-low-power" models. Where "F" means the chip has no built-in graphics processor the "T" means the chips are designed to use less power while also having less performance than the standard chips without any letters. For the 10th gen Core series Intel will also be releasing a T series, these have a low TDP in the 35 Watt range.

TDP and PL States
Comet Lake processors are allowed to run load clock frequency values considerably higher and longer than 9th gen Coffee Lake. The TDP that is specified/listed is now 125 Watt for the more core parts with 65 Watts for the more energy-friendly models. However, the story does not end there though as these values are based on what is called the P1 state. For a 9900K that value sits at 95W, with the Comet Lake S thus 125W, of course, it is two more cores so nobody will wonder about that. However, there is also a PL2 state where a higher power limit that can be applied for a relatively short time. If you allow me to take that 9900K again as an example, it had a PL2 allowance of 1.25 times PL1, = 119W, for a duration of 28 seconds. Here's what's so different and where Intel yields most of its performance for Comet lake, if I take the 10900K as an example, PL2 now is 2 times PL1, so 250W. That PL2 state now also is longer, 56 seconds. And therein reaps the benefit of additional performance, substantially, but at the cost of energy and heat.

CPU	P1 State	PL2 State	Duration
Core i9 10900K	125 Watt	250 Watt	56 seconds
Core i7 10700K	125 Watt	229 Watt	56 seconds
Core i5 10600K	125 Watt	182 Watt	56 seconds
Core i9 9900K	95 Watt	118 Watt	28 seconds
Core i9-9900KS	127 Watt	159 Watt	28 seconds

In short, this means the overall performance of Comet Lake can be a lot higher compared to Coffee Lake (9th gen processors) and older processors, but that also will have an effect on power consumption and inevitably, heat levels. That new and close to a minute PL2 state duration, however, is significant. With such significant power states, we do fail to see the significance and reliability of a listed TDP value these days if you allow double the TDP for almost a minute these days.



Overclocking related features
As in all generations, Intel brings new developments for overclocking in this new tenth generation. The ones highlighted by the company are Hyper-Threading per core, overclock of the PEG and DMI buses, and better control of the voltage-frequency curve. They will also renew their XTU utility with more functions and a better graphical interface.

Another improvement is a thinner die, improving the thermal transfer from the cores to the IHS, which should help to reach better temperatures or at least get heat faster away from the processor.

Z490 and what is the deal with that listed PCI Express 4.0 support?
While the new Z490 motherboards mechanically can handle PCI Express 4.0, the reality is that Comet lake-S is a PCI Express 3.0 compatible product series. Z490 as a platform, however, will also support the following series, Rocket Lake-S, and that processor will move towards What about PCI Express 4.0 The following is unconfirmed but, Comet Lake-S actually was designed to support PCI Express 4.0. However an issue prevents PCIe 4.0 to become usable, so the platform will support PCI-E 3.0 instead. The reason behind this is the presence of Jitter in the Z490 chipset signal, that causes instability in the data signal and that means that specifications required for the PCI-Express 4.0 certification are simply not met. PCI-E 4.0 is working at very high frequencies, and that requires that the signal be completely clean. This information remains speculation and hearsay though. That also means Comet Lake is using an x4 DMI PCIe Gen 3 interlink (the link between the processor and chipset). Intel informed us specifications of the three consumer chipsets: Z490, H470 and B460. A cheaper H410 chipset will follow at a later date. The Z490 chipset largely resembles its predecessor in terms of its feature set.

	Z490	H470	B460	Z390
OC / Tweaking	Yes	-	-	Yes
PCIe 3.0 lanes CPU	16	16	16	16
PCIe 3.0 lanes chipset	24	20	16	24
SLI / Crossfire	Yes	-	-	Yes
SATA3	6	6	6	6
USB 3.1	6	4	0	6
USB 3.0	10	8	8	10
USB 2.0	14	14	12	14
AX WiFi 6	Yes, by CNVi	Yes, by CNVi	No, controller required	-
2.5 Gigabit LAN	Yes, by PHY	Yes, by PHY	Yes, by PHY	-

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 4th part

[SHOWTOGROUPS=4,20]
Product Showcase
On this page you can see the Core i7 10700K. Socket LGA1200 with that number meaning the number of pins means quite a bit of pin contact points right there. We use Z490 motherboards with the latest motherboard BIOS revisions. For any Comet Lake-S generation processor, you will thus need a new motherboard based on the new socket.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 6th part

[SHOWTOGROUPS=4,20]
CPU-Z Screenshots
And here we have CPU-Z screenshots of the processor(s) seated, armed, ready, and waiting for some load, let's have a peek. So, that's all looking alright. If interested, you can download CPU-Z here. The cache sizes look rather familiar.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 7th

[SHOWTOGROUPS=4,20]
Power Consumption
We show energy consumption based on the entire PC (motherboard / processor / graphics card / memory / SSD). This number depends and will vary per motherboard (added ICs / controllers / wifi / Bluetooth) and PSU (efficiency). Keep in mind that we measure the ENTIRE PC, not just the processor's power consumption. Your average PC can differ from our numbers if you add optical drives, HDDs, soundcards etc.

Power consumption measurements will differ per PC and setup. Your attached components use power but your motherboard can also have additional ICs installed like an audio controller, 3rd party chips, network controllers, extra SATA controllers, extra USB controllers, and so on. These parts all consume power, so these results are a subjective indication. Next, to that, we stress all CPU cores 100% and thus show peak power consumption.

Above can see that once we average out the cores over system power draw per core, we're looking at total system power + CPU cores stressed / the number of cores.



Temperatures
The reason we do not table up temperature results is that we'd need to apply the same cooling over and over on all platforms. Also, coolers (RPM) react differently to TDP and variables like BIOS on all motherboards let alone brands. The Core i7 10700K peaks towards 70 Degrees C under full load on the processor package, this temperature is measured during a looped wPRIME 1024M runs. Overall not bad at all. We are using an LCS cooling by Corsair. BTW notice the Power state PL2 and PL2 behavior as the plot shows it perfectly, the processor goes full throttle for a while and then drops down in frequency a bit lowering temperatures.

The fastest core

In light of the recent Turbo clock frequency discussion, we want to show you the fastest core frequency on a single thread. The processor reaches its advertised 5100 MHz on two cores, that's the Thermal Velocity Boost). You can see an all-core clock of 47x. And with any threads in between you can reach 5.0 GHz.

​

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 8th part

[SHOWTOGROUPS=4,20]
Hardware & Software Used
We now begin the benchmark portion of this article, but first let me show you our test system plus the software we used.

Mainboard

Z490 (ASUS Maximum XII Extreme)

Processor

Intel Core i7-10700K

Graphics Cards

GeForce RTX 2080 Ti - Review

Memory

2x8 GB DDR4 3600 CL16 MHz G.Skill Trident Z RGB

Power Supply Unit

RM850X Corsair - Review

Monitor

4K UHD Monitor at 3840 x 2160 - Review

OS related Software

Windows 10 (latest patches & updates installed)
DirectX 9/10/11/12 latest End-User Runtime (Download)
Nvidia GeForce driver Latest WHQ (Download)

Software benchmark suite

• 7-Zip 19.00
• SpecWorkstation 3
• IndigoBench 4.0
• Blender 2.81a
• Google Chrome / Mozilla Kraken / Jetstream
• Aida
• 3DMark Time Spy
• Handbrake
• CineBench 15
• CineBench 20
• POV-Ray 3.7
• V-Ray Next
• Corona Benchmark
• WPrime
• CPU-Z benchmark
• FryRender
• Shadow of the Tomb Raider
• Strange Brigade
• Codemasters Formula 1
• Far Cry 5

A Word About "FPS"

What are we looking for in gaming, performance wise? First off, obviously Guru3D tends to think that all games should be played at the best image quality (IQ) possible. There's a dilemma though, IQ often interferes with the performance of a graphics card. We measure this in FPS, the number of frames a graphics card can render per second, the higher it is the more fluently your game will display itself.

A game's frames per second (FPS) is a measured average of a series of tests. That test is often a time demo, a recorded part of the game which is a 1:1 representation of the actual game and its gameplay experience. After forcing the same image quality settings; this time-demo is then used for all graphics cards so that the actual measuring is as objective as can be.

Frames per second	Gameplay
<30 FPS	very limited gameplay
30-40 FPS	average yet very playable
40-60 FPS	good gameplay
>60 FPS	best possible gameplay

• So if a graphics card barely manages less than 30 FPS, then the game is not very playable, we want to avoid that at all cost.
• With 30 FPS up-to roughly 40 FPS you'll be very able to play the game with perhaps a tiny stutter at certain graphically intensive parts. Overall a very enjoyable experience. Match the best possible resolution to this result and you'll have the best possible rendering quality versus resolution, hey you want both of them to be as high as possible.
• When a graphics card is doing 60 FPS on average or higher then you can rest assured that the game will likely play extremely smoothly at every point in the game, turn on every possible in-game IQ setting.
• Over 100 FPS? You either have a MONSTER graphics card or a very old game.

Monitor Setup

Before playing games, setting up your monitor's contrast & brightness levels is a very important thing to do. I realized recently that a lot of you guys have set up your monitor improperly. How do we know this? Because we receive a couple of emails every now and then telling us that a reader can't distinguish between the benchmark charts (colors) in our reviews. We realized, if that happens, your monitor is not properly set up.

​

Looking at the above graph, what are you looking for?

• Top bar - This simple test pattern is evenly spaced from 0 to 255 brightness levels, with no profile embedded. If your monitor is correctly set up, you should be able to distinguish each step, and each step should be visually distinct from its neighbors by the same amount. Also, the dark-end step differences should be about the same as the light-end step differences. Finally, the first step should be completely black.
• The three lower blocks - The far left box is a black box within the middle a little box a tint lower than black. The middle box is a lined square with a central grey square. The far right white box has a smaller "grey" box that should barely be visible.

You should be able to distinguish all small differences, only then you monitor is set up properly contrast and saturation wise.


[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 9th part

[SHOWTOGROUPS=4,20]
Challenges of testing a processor
Each year it is getting increasingly complicated to accurately test processor performance, the motherboard manufacturers are in control of a lot of settings. Intel, on the other hand, does not release reference review motherboards anymore, thus motherboards manufacturers will do tons of tweaks 'at default' in the BIOS just to stay ahead of the competition and tho spread their lineup relative for performance from mainstream towards premium. Settings can include: All cores frequency locks, clock tweaks, and enhancements in power states and duration. For a motherboard review, this is all fine, however, for a processor review, we want to test as close as possible to reference performance. To bypass this challenge, we opted the ASUS ROG Maximus XII Extreme for this reference tests, as the BIOS offers a setting that enables reference Intel Turbo ratio settings, see below:

You expect you will see lots of reviews with higher performance numbers as many media will load up BIOS defaults and start testing without checking. We thus test at reference performance (or as close as possible) with settings as referenced by Intel. Before anyone asks, yes ASUS in the premium models also offers an all-core enhancement, this is set manually to off as well. For motherboard reviews, we, of course, revert to motherboard defaults as it is well within their right to segment and optimize their lineup performance-wise.

Processor performance: CineBench 15
CB15 supports systems with up to 256 threads. The performance of processors and graphics cards is, as usual, determined on the basis of 3D scenes. A selection of test results allows a rough classification of the benefit of your own system. For the CPU test is a scene with around 280,000 polygons used, while the GPU test based on OpenGL comes with about a million polygons, high-resolution textures, and various effects. The results will be issued in final points (CPU) and fps (GPU). According to the developers, the software has been "extensively developed to exploit the performance of new hardware as possible." The results are unsurprisingly not comparable with those from earlier versions. You'll notice we still need to add a number of processors, all in due time. You'll notice the single-core perf paints a completely different picture here.

Instructions per cycle (IPC)
This IPC test will build up and get updated over time. We lock all processor cores at 3500 MHz. That way you can see the architecture performance of the processor clocked at exactly the same frequency. This is a single thread measurement. For many people, this is the holy grail of CPU measurements in terms of how fast an architecture per core really is. I, however, tend to say there's more to it than that and that would be higher frequency allowances, caches, and memory latency defining that per-core performance.

Processor performance: CineBench 20
Maxon released their Cinebench R20 benchmark, more capable of dealing with the heavily threaded processors. You need a PC with at least 4 GB of memory and SSE3 instruction set support. Maxon states Cinebench R20 is now using four times the memory and eight times the CPU computational power compared to Cinebench R15.

Processor Performance - Blender 2.83
With processors getting more threads and cores almost exponentially each year, we have now added Blender, v2.8x. We fire off two heavy intensive renders, BMW27 and Classroom. The software likes many cores and threads. It's a relatively new benchmark suite asset, so over time results will build up.

Content Creation POV-Ray v3.7
POV-Ray is 3.7 introduced; with support for symmetric multiprocessing (SMP) to allow the renderer to take advantage of multiple processors. Shortly before the embargo of this review, a new build was supplied, opening up 64-core processor support. And that certainly shows.

Performance - SpecWorkstation 3
Effectively there are than 30 workloads containing nearly 140 tests in the SPECworkstation 3 benchmark exercise CPU, graphics, I/O and memory bandwidth. The workloads are divided by application categories that include media and entertainment (3D animation, rendering), product development (CAD/CAM/CAE), life sciences (medical, molecular), financial services, energy (oil and gas), general operations, and GPU compute. We have introduced this title to the test-suite to create hefty industry-related workloads on many-core systems, as it turns out the software scales perfectly up-to 32 processor cores.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 10th part

[SHOWTOGROUPS=4,20]
CPU-Z Benchmark
CPU-Z offers a pretty nice performance measurement to test RAW CPU performance, performance is measured both as single core and multi-threaded core performance. It offers a fast and easy manner in which to quickly view single threaded and multi-threaded SMT performance. Give it a try yourself, it is easy to use. Here are some numbers, we'll build up and update more results over time.

Benchmarks: De/Compression - 7-Zip Multi-threaded
In this segment of the benchmarks, we'll walk through compression software applications. We'll use 7-ZIP and look at both compression and de-compression performance. 7-Zip is a multi-threading archiver, especially in de-compression that shows exceptionally well.

Benchmarks: Google Chrome - Kraken and Jetstream Browser
We added this benchmark as it is an underestimated thing, it's what we do all the time on the PC hence we are introducing a proper browser performance test. The Mozilla Kraken-browser benchmark applies multiple complex calculations based on javascript. This will bring you a good overview of browser performance. We test with the latest iteration of Google Chrome. JetStream is a JavaScript benchmark suite focused on the most advanced web applications. Bigger scores are better. For Jetstream, it is clear that it likes non-SMT (hyperthreaded) processors.

Content creation: FryRender
FryRender is a benchmarking framework for everyone, not just for 3D users; anyone out there, from hardware integrators or hardware reviewers to die-hard gamers. Since its conception, FryRender has been designed with the aim of being the most muscled engine in its category. As a result, and after several years of intense development, FryRender's core doesn't let a single CPU cycle be wasted. Its routines have been written to be cache efficient, and to take the maximum advantage possible of the new multi-threading capabilities present in modern CPU architectures. Being a highly-optimized and extremely math-intensive application (mostly in floating-point) which makes a very efficient use of the system's cache, we think that FryRender is the near perfect tool for measuring how much 'brute computational power' a computer is able to deliver.

• FryRender utilizes a maximum of 32 threads

If you are wondering what is up with the latest FryBench results, well I mentioned already that it will be hard to find software that can actually support more than 32 threads. FryBench supports precisely that, 32 threads.

Performance - Vray NEXT and Indigo 4.0
V-Ray is a stand-alone application to test how fast your hardware renders. The results are displayed in render time mode. The standalone application includes a single GPU scene and a single CPU scene. V-Ray is a computer-generated imagery rendering software application developed by the Bulgarian company Chaos Group. It is a commercial plug-in for third-party 3D computer graphics software applications and is used for visualizations and computer graphics in industries such as media, entertainment, film and video game production, industrial design, product design, and architecture. The software supports multi and mega threading, it is not limited by even 64-threads.

Indigo Renderer is an unbiased, photorealistic GPU and CPU renderer aimed at ultimate image quality, by accurately simulating the physics of light. State of the art rendering performance, materials and cameras models - it's all made simple through an interactive, photographic approach with few abstract settings, letting you concentrate on lighting and composing your imagery.

Corona Ray Tracing
This tool is very easy to use, simply save, extract, and run the downloadable file from their site and you will get started and it will automatically give you results at the end that we can then use to compare performance between CPUs. Workstation grade systems with up to 72 CPU threads can be used in this benchmark, meaning that it was made with heavy threading in mind, making it suitable for testing CPUs with both small and large CPU core counts.

Video Transcoding
Video transcoding is well suited for systems that have more CPU cores. Encoding/transcoding to x.264 format is one of the most intensive tasks a processor can perform. In this test we encode a h.264 DTS 1080P trailer to Matroska x.264 with 5.1 channels AC3. This software is nice for benchmarking the CPU and memory. Compute wise this title also allows you to test features like AVX and OpenCL. We are testing brute processor performance right now though.

• Handbrake runs 16 cores max

The displayed number is the number of frames rendered per second averaged out over the encoding process. The higher the number, the faster the performance is. It's exactly in applications like these where processors with more cores really shine as they are all utilized to the maximum.

WPrime
The tool that overclockers like so much to determine performance and stability. In this test, we max out the number of threads corresponding to the processor and calculate 1024M of prime. This software is a computer program that calculates a set number of square roots using Newton's method for estimating functions, verifying the results by squaring them, then comparing them with the original numbers.

3DMark Time Spy CPU score
3DMark focuses on the two areas most critical to gaming performance: the CPU and the GPU. With the emergence of multi-package and multi-core configurations on both the CPU and GPU side, the performance scale of these areas has widened, and the visual and game-play effects made possible by these configurations are accordingly wide-ranging.

• The default Time Spy CPU test does not scale beyond processors with 10 or more threads.

This makes covering the entire spectrum of 3D gaming a difficult task. It's multi-core and multi-threading aware. The standard Time Spy test is thread limited to 10 CPU cores. If a processor has faster turbo bins on these 10 cores, that's where it'll shine. So in this respect, the processor is still doing remarkably well.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 11th part

[SHOWTOGROUPS=4,20]
Performance System Memory: Memory Read Test
For our memory tests, we have used dual-channel at 3600 MHz at CL16. Typically we make use of 3200 MHz CL14 or 3600 MHz CL16 for tests. We will be transitioning towards 3600 MHz for all platforms though. For this review we make use of G.Skill 3600 MHz CL16 TridentZ RGB DDR4 memory setup in dual-channel.

Performance - Gaming - 1280x720 (HD Ready)
We include game tests at 720p (1280x720 pixels) resolution. The games tested are put through 720p using a GeForce RTX 2080 Ti graphics card with the best settings.

Why measure at a 1280x720 pixels? - We include 720p results based on user/reader base requests, plain and simple. It's here where CPU scaling performance relative towards a GPU is extremely visible. If you think a little more about what you are actually seeing in the charts from 720p up to 1440p, you'll realize that what we are showing is the effect of raw CPU performance, not restricted by any sort of GPU limited situation, and your gameplay is pretty much always limited by your GPU. That is also why we opt to test with the fastest and most expensive graphics card. We realize really that no nobody purchases a PC with an RTX 2080 Ti to game at 720p with this processor, the results, however, are of a theoretical and statistical value. 720p numbers also could interest PC gamers with fast 120 Hz and 144 Hz monitors that want to squeeze out every last frame in performance. 720p tests here serve as a theoretical test. A GeForce RTX 2080 Ti is used to bypass any GPU limitation. The reality with, say, GeForce RTX 2070 or Radeon RX 5700 is simple - the scores would be much closer to each other.

Performance - Gaming - 1920x1080 (Full HD)

Performance - Gaming - 2560x1440 (QHD)

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 12th part

[SHOWTOGROUPS=4,20]

Performance - Games - GeForce RTX 2080 Ti - 3DMark Time Spy
3DMark includes everything you need to benchmark your hardware. Time Spy is way more GPU dependent though. But still a representative title in relation to PC gaming performance.

Performance - Games - Integrated GPU - 3DMark Time Spy
3DMark includes everything you need to benchmark your hardware. Time Spy is way more GPU dependant though. But still a very representative title in relation to PC gaming performance.

If your Comet Lake series processor has an active IGP (Integrated Graphics Processor), it is based on Intel UHD Graphics 630. UHD Graphics 630 is an integrated graphics processor introduced by Intel in 2017 for the performance desktop Coffee Lake microprocessors. The 630 are the mid-range (GT2) IGP offered by Intel with 24 execution units. UHD Graphics 630 are found in 8th, 9th and now 10th generation desktop Core i3, Core i5 and Core i7 processors. The UHD Graphics 630 is simply a rebranding and is otherwise identical to the HD Graphics 630 found in desktop Kaby Lake processors.

[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 13th part

[SHOWTOGROUPS=4,20]
Overclocking on air
As always we tried to overclock the processor. It certainly is doable but proper liquid cooling is recommended. We got the processor stable at 5100 MHz all-core with a vcore of ~1.5v This resulted in Cinebench in a score of 2109 points, which is a 4 percent higher score than the 2027 points that the processor reaches on stock settings. The heat production was considerable. This processor does have a soldered heat spreader, which in theory should ensure a much better heat transfer compared to thermal interface material.

As it stands, you can tweak to 5100~5200 MHz under the condition that you have a good LCS cooling unit. Realistically I'd say, leave tweaking for what it is as the defaults of this processors are already really close towards these performance regions.

Performance Summary (relative in %)
Averaged relative performance values in percentages are always a bit dangerous to include, as the accumulation of numbers can have a more positive or negative effect on many variables, like the number of cores an application supports, and some software simply likes faster clocks and fewer threads or favors a certain microarchitecture. The averaged out numbers on this page are thus serving as a relative matched index. While not precise for some target charts (as software differs), it does show an index of what to reasonably expect for your real-world experience relative to the performance. Over time this result set will build up.

For chart scaling purposes, I left out the Threadrippers here as especially the 32 and 64-core procs completely warp the charts, we have to blame AMD for that.

The above (last) chart is a bit of a thesis chart, basically a prognosis of what might become the best gaming processors in the years to come if you take certain variables into account. We make the subjective assumption that games in 2020 and later will make good use of up to roughly 10 CPU cores or threads. If you have a many-core CPU with faster Turbo bins (say, an eight-core processor that can run 4.5 GHz on many of its cores opposed to another 8-core proc that does the same at 4000 Mhz, or an 8-core proc with better IPC), then in this chart you can see what processor would be best for future games. Remember though, you are looking at probable best CPU performance for games once they become properly threaded, not relative game performance.



[/SHOWTOGROUPS]

 

[emailx45]

Spoiler: Content - 14th part - Conclusion

[SHOWTOGROUPS=4,20]
Final words & conclusion
We previously tested the 10600K and 10900K which received mildly positive comments, the 10700K, however, is a little more 'meh?'. Throughout our testing of this processor, we continuously fell in the performance bracket of the previous generation 9900K and 9900KS, and considering it's the same microarchitecture with a few voltages and frequency tweaks that's not surprising. That does, however, give that feel of 'more of the same'. With the Core i7 10700K, you buy an i9 9900K for less money, with a higher spec memory controller and a weirder TDP behaviour. You do need that new Z490 motherboard though. Today's tested product is challenging the AMD Ryzen 3000 series alright, but each has that own advantages and disadvantages. For example, some level of PCIe Gen 4.0 support for Comet Lake-S would have been preferred, even if it was only x4 PCIe Gen4 to support at CPU level one M2 unit at PCIe Gen 4.0. It's a bit of a missed opportunity. The relevancy of PCIe Gen 4.0 M2./ NVMe SSDs however still is debatable at best, but the coming year we'll see that host infrastructure advance with much faster NVMe based SSDs, and that's where its importance will grow. The processor series requires a new motherboard, a year from now Rocket Lake-S will be released, and yeah these will be the only two CPU series that will be supported on this 1200-pin infrastructure. I've stated this in a recent discussion here on the website and forums, but considering that you upgrade a motherboard+CPU likely once every four years, we do not see it as incredibly significant for your purchasing decisions. But for you to upgrade, the processor needs to be just right.

From a processor architecture point of view, little is new in features. We do like the excellent sharp peak performance that the series offers, but we can also tell that Intel had to open up a bag of tricks to be able to accomplish that. Of course, I am referring to the long turbo power states of a PL2 mode that last 57 seconds at high energy consumption. There's little wrong with that as it is not something new, however, what is new is the duration, and power allowance nearly doubled up, so that means this 125 Watt processor is allowed to jump to 229 Watts for almost a minute. That will affect heat, so cool this puppy wisely. TDP as a discussion by itself, do people care about it? Not so much really, but processors jumping towards 229 Watts is a lot, let's be honest there. And if you allow your processor to run to 229/250Watts for a minute, should this not have been called a 250 Watt TDP processor?

Purely speaking from a game performance point of view, Intel has the lead. And you will need to acknowledge that the gaming performance is a notch better compared to the competition when you fire an enthusiast-class GPU at the platform. That all comes back towards the high-frequency Turbos, Intel can apply, running up to 5.0/5.1 GHz (depending on conditions). Is it enough though?, well that is hard to answer question, the competition is robust in threaded and single-threaded performance with applications, it's only gaming combined with 1000 USD graphics cards where they perform a notch-less. You do need to look at platform differences, AMD offers PCIe Gen4.0 throughout the eco-system, whereas with Comet-Lake-S sticks at Gen 3.0, albeit the Z490 motherboards are gen 4.0 ready. Still, that likely is introduced with Rocket Lake-S series desktop processors.

Performance & tweaking
We'd rate the new 8-core part as "good" for the results as tested. Temps remain trivial to judge but certainly did not disappoint at processor defaults (depending on the choice of cooling), we are going to recommend a proper LCS cooler. At the OC levels you are looking at 1.35V and higher needed on that CPU core for a decent tweak, however, and that is the same for AMD, tweaking many-core processors is complicated as you are bound to run into core limitations, heat and power consumption. With processor like these, you need to wonder if you want to tweak, as really, the best configuration is already there at defaults. At reasonably low voltages you'll reach an all-core 5000 MHz, LCS got us towards 5.2 GHz (all cores). At that stage, we are closing in at ~350 Watts power consumption ergo my plea to leave the proc at its default configuration. Of course, we received ES samples, so we cannot state anything conclusive on the final retail products in matters of tweaking (yours might run a notch better). The infrastructure that Z490 offers is sound as well as has proven to be reliable and easy to use. Tweaking wise you increase the CPU voltage and multiplier and you are good to go. Another plus for the Intel platform is that over the years they have been able to refine their memory controllers, pop in anything XMP 2.0 and you have a 98% chance it'll work straight out of the box with high-speed memories. However, the effect of speedier frequency memory is far less significant, so opt volume would be my advice. We'd always suggest going with a nice affordable 3200 or a 3600 MHz kit.

Gaming performance
As far as gaming goes, the pure raw wins are mostly for Intel, but everything is relative when it comes to gaming as 98% of the time your actual limitation is the GPU, and not CPU. Ergo, for gaming GPUs matter more than CPUs. You can measure the effect of CPU performance with games, but only when you steer away from that GPU bottleneck, that's why we use a 1000+ USD graphics card. Only then in the lower resolutions you'll see substantial enough differences. But with a Radeon RX 5700 or GeForce RTX 2070, these differences would be much closer towards equal for one another. Hey, everything is relative. Let's have a peek here at the chart below. That's a 1920x1080 (Full HD) gaming chart based on the reference GeForce RTX 2080 Ti.

The vast majority of you guys have a far more GPU limited graphics card. With eight cores, the reality is absolutely and unequivocally is that you can game pretty darn and are future proof. Realistically a 6-core proc offering more value will get you there as well. Currently, I find 8-core processors a sweet spot concerning gaming, 6-cores for value, and anything higher than 8-cores in core count, overkill.

Power consumption
With eight cores you get a 125 Watt TDP processor. But as explained, it all is a little more complicated these days. Intel applies a secondary power stage where the processor is allowed to run twice the TDP value for a pretty long time. That's a holy grail for added threading performance, it does, however, make energy consumption peak with higher values during that long boost. With the system at idle with a GeForce RTX 2080 Ti installed / 16 GB memory / SSD and the Z490 motherboard hovers at roughly 70 Watts. That's just fine, really, but the load values are significant. When we stressed the processor 100% run, we reach approximately 250 Watts with this 8-core processor (for the entire PC).

Conclusion
Honestly, if you already own a processor like a 9900K we see absolutely no reason to upgrade. The performance is pretty much the same with the usual oddities and offsets in mind. Considering the Z490 platform doesn't offer anything new other then WIFI6 and 2.5+ GigE ethernet either, this processor simply put is a hard sell for anyone to upgrade. That doesn't mean it's a poor processor though, contrary it's great, but more of the same that we have seen for a while now. BTW there will be media reviews out there showing an excess of 10% additional performance over a 9900K, likely they did not set up their motherboard BIOS towards reference standards. So realistically that is a 5% margin increase.

Processor conclusions are never easy to write up as there are many factors involved as well as multiple pros and cons for each side and brand. Intel certainly gains the most out of the 14nm fabrication node compared to AMD. At 14nm they can reach 5.1 GHz (under strict conditions like PL2 and thermals). This brings Intel good peak performance which shows in gaming performance as well as boosting multi-threaded performance. The flip-side of that coin is that once the processor returns to its original PL1 state, that threading performance becomes less strong. It is one of the reasons that we recently moved towards test suite software that sometimes runs 10 to 15 minutes or even an hour to test, which, in our belief, paints a more accurate picture. Tagged in there is IPC, the main factor to consider as IPC x your clock frequency is the holy grail of it all. So clock for clock AMD with ZEN2 has better IPC, and Intel better (and longer) high clock frequencies, and that's the tradeoff between the two brands.

Each of the architectures will show advantages and disadvantages. The processor series is sound. Platform wise Z490 in combination with Comet lake-S isn't that different from last-gen and makes this a harder sell to spend money on an upgrade. Imho, Intel needed PCIe Gen 4.0 here today, on this platform to make a substantial difference. The majority of platform vulnerabilities will have been patched at the hardware level where possible, and soft and firmware otherwise. The performance certainly is there. Another trump card for Intel is that they came down on prices. At close to 500 USD the 3900X from AMD is now the competing processor. Gaming wise with only the most expensive GPUs Intel wins, threading, and overall performance-wise, for that money AMD has a more mature platform with the performance that comes along with that in all other segments. But as so often, platform and brand preference will be important for any potential consumer, there's a camp intel and a camp AMD. Tweaking wise there is a tiny bit left for the Core i7 10700K, as explained ~5100 MHz all cores is quite easy and I do foresee some lucky folks reaching a stable 5.3 GHz on all cores. We halted at 5200 MHz which was stable. You also really need to wonder if it is worth it (tweaking and overclocking), you are dealing with extensive power consumption and heat levels. But for kicks and giggles, sure why not.

The Core i7 10700K is a fast processor, make no mistake. In the end, though, seen upgrading from the Core i9 9900K(S) things are more challenging to advise though. But yes, we can certainly recommend Comet Lake-S if you take all the variables we mentioned in mind. Regardless, we'll happily hand out a recommended award for the Core i7 10700K, but for an upgrade to make sense, you need to be a few generations away from your current PC configuration.

Handy related downloads:

• AfterBurner
• AIDA64
• HWiNFO64
• SiSoft Sandra
• 3DMark Time Spy
• CPU-Z
• HWMonitor

[/SHOWTOGROUPS]