Intel Core i5-10600K vs Ryzen 7 3800X
If you want to build a PC there will always be the question: Which CPU should I take? Which CPU offers the best performance and, for budget conscious users, which offers the best price-performance ratio?
You don’t want to spend 30% more for only 5% increase in performance.
In this review we tested the Intel i5 10600K vs the AMD Ryzen 7 3800X.
We explain all features and especially performance benchmarks for gaming and applications used in workstations, i.e. graphic design applications etc.
In the last section of the articel, down below, you can find our final verdict and recommendation, the winner of the battle Intel Core i5-10600K vs Ryzen 7 3800X.
Test Results: Intel Core i5-10600K vs Ryzen 7 3800X
Ranking First: AMD Ryzen 7 3800X
- Excellent value compared to Intel competitors
- It is fast and its single-core performance beats many of Intel’s similar processors
- It is an excellent choice for gaming, and also for productivity
- Bundled cooler
- No integrated graphics
Best performing CPU: AMD Ryzen 7 3800X
At the start of the third Ryzen generation, we put two first models to the test: the Ryzen 9 3900X and the Ryzen 7 3700X.
With eight to twelve cores, these models attracted attention and put Intel under a lot of pressure, especially in terms of multi-core performance.
Thanks to numerous improvements within the Zen 2 architecture, the new CPUs also caught up well in games.
In the weeks that followed, further Techtestreport tests of the Ryzen 5 3600X and Ryzen 5 3600 followed, some of which proved to be real price-performance hits with six cores.
The AMD Ryzen 5 3400G is somewhat out of line.
As an APU, it not only provides an integrated Vega graphics unit, but is also based on the Zen+ architecture despite its new name, which is why the improvements over its predecessor were very limited.
With the AMD Ryzen 7 3800X we are closing another gap in our test database shortly before the launch of the new 16-core flagship for the AM4 socket and the X570 chipset.
The AMD Ryzen 7 3800X, like the Ryzen 7 3700X we’ve already tested, offers eight cores with 16 threads, but further increases the frequencies, which makes the TDP on paper increase from 65 W to 105 W.
In return, the base clock rate increases from 3.6 to 3.9 GHz, whereas the boost frequencies increase only slightly from 4.4 to 4.5 GHz.
The 32 MB L3 cache and the 4 MB L2 cache are of course still available.
The AMD Ryzen 9 3900X, the next larger model, offers four more cores, thus a little more cache and finally clock rates of 3.8 to 4.6 GHz with a TDP of 105 W as well.
While AMD has announced a recommended retail price of 429 Dollar for its Ryzen 7 3800X, the street price is now far below that.
At Geizhals.at, the eight-core is listed for less than 390 Dollar.
A Ryzen 9 3900 and unfortunately also the Ryzen 5 3500X will not be available for end customers for the time being – they are reserved for the OEM market or the Asian region.
Measurements: Power consumption
Under load, the AMD Ryzen 7 3800X allows itself just over 130 W, which is a good 30 W or almost 30% more than the Ryzen 7 3700X, the next smaller model.
This is more in line with the older Ryzen flagship models, such as the Ryzen 7 1700X or Ryzen 7 2700X.
Intel’s Core i7-9700K proves to be the counterpart with 131.6 W.
A similar picture emerges when looking at all components of the test system.
Then the test computer with AMD Ryzen 7 3700X draws just under 200 W from the power outlet and sorts itself between the Ryzen 7 1700X and Ryzen 7 2700X, whereby the Intel counterpart scores slightly better.
The increase from 65 to 105 W TDP is thus clearly measurable.
Verdict: Best performing CPU: AMD Ryzen 7 3800X
The AMD Ryzen 7 3800X has a heavy stance or very strong opponents in its own ranks.
Although it closes the gap between the Ryzen 7 3700X and the Ryzen 9 3900X in terms of performance, it doesn’t do so very efficiently.
Compared to the next smaller model, the performance increases only in the single-digit percentage range, while the hunger for electricity increases by a good 30%.
The efficiency advantage of the smaller models and the Zen 2 architecture evaporates.
This is because the fastest Ryzen 7 is placed in a higher TDP class, but is unable to exploit the potential gained.
Despite the higher TDP, this model also rarely reaches its maximum boost clock and in practice is only a good 100 MHz above the Ryzen 7 3700X, which, according to the data sheet, gets by with significantly more economical 65 W.
Usually only about 4.35 GHz are in it. The Ryzen 9 3900X benefits from four additional cores, but also clocks a bit lower, which reduces the single core performance.
If you then also consider the price, you pay a good 60 Dollar more for the Ryzen 7 3800X, which currently costs around 390 Dollar, and in the end you only get a very small performance increase with a significantly higher power consumption.
New buyers of the Ryzen platform should therefore better choose the smaller model or spend a little more and go for the twelve-core Ryzen 9 3900X.
However, it already costs well over 500 Dollar.
Perhaps there will be some price changes in the next few weeks, when AMD finally launches the Ryzen 9 3950X, throws its new third generation threadripper processors on the market and Intel also adds its Core X models.
Ranking Second: Intel i5 10600K
- Impressive single-threaded performance
- Consistently fast in games
- Now includes Hyper-Threading
- Chipset has upgraded networking
- No native PCIe 4.0 support
Best price-performance ratio – 10600K: Final Verdict
The desktop CPUs show how quickly a market can develop when two participants seriously compete for the favour of the customers:
In 2017, Intel was still offering only four cores (Core i7-7700K) for 400 Dollar, and then six (Core i7-8700K) and eight (Core i9-9900K) cores every year.
With the Core i9-10900K the step to ten cores is made, as AMD already offers twelve cores for under 500 Dollar.
Intel, however, has a problem and that is P1272. This is how the manufacturer internally describes the 14 nm process for the production of CPUs, which has been used for over five years now.
Due to the delay of the 10nm process, Intel is also stuck with the Skylake architecture of 2015, which is why more cores and more clock speed is the only way.
The Core i9-10900K is the last of its kind, but Intel is making the most of it.
In addition to the top model with ten cores, there are also those with eight and six cores, with processors with even less performance to follow in a week’s time.
We tested the Core i9-10900K (10C/20T) for 550 Dollar and the Core i5-10600K (6C/12T) for 300 Dollar.
Both have more cores, respectively threads, than their respective predecessors:
In the i3 class there are 4C/8T instead of 4C/4T, in the i5 segment Intel enables hyperthreading for 6C/12T instead of 6C/6T and in the i7 models there are 8C/16T instead of 8C/8T.
The additional logical cores were previously disabled for product policy reasons, but due to AMD’s Ryzen CPUs like the 3600X (6C/12T), the 3700X (8C/16T) or the 3900X (12C/24T), Intel had to react and chose the obvious option.
However, while AMD has been using the AM4 socket to accommodate all processors for three years now, Intel has switched to a different socket:
The LGA 1200 socket replaces the LGA 1151 v2 socket (with the same cooler bracket), and the manufacturer again cites a changed power supply as the reason.
In fact, the actual socket is prepared for PCIe Gen4.
As usual, the new boards are equipped with revised chipsets:
For high-end boards there is the Z490 and for cheaper boards primarily the B460, furthermore the H470 and the H410.
What they all have in common is that they support PCIe Gen3 and at least the Z490 and H470 natively USB 3.2 Gen2 instead of using an additional controller, but no PCIe Gen4.
The new features include a deeper integration of 2.5 Gbps Ethernet (Foxville) and WiFi6 (AX201 via CNVIO interface).
The 10th Gen CPUs, internally called Comet Lake S, also do not support PCIe Gen4, unlike AMD’s Ryzen 3000.
This capability and 20 instead of 16 lanes will be supported by the 11th Gen aka Rocket Lake S with 14 nm and new Cove microarchitecture, which will also fit into the LGA 1200 socket.
As a result, some boards will have an M.2 NVMe/PCIe SSD slot, which is not usable at all with a Core i9-10900K.
For the time being, however, CML-S remain current, where Intel has at least developed a chip with special features in favor of high clock rates.
Without soldering it doesn’t work
In purely architectural terms, nothing has changed at Comet Lake S since Skylake S in 2015, because the Skylake technology is almost identical in its literal core.
The frontend and backend remain as they are, even the cache levels have not been touched, only the L3 buffer typically scales with the number of cores.
These are connected by a simple ring bus, Intel does not use a mesh like the CPUs with 18 or more cores.
Compared to the original Skylake design, Intel has made changes over the years to help against security holes:
Comet Lake S has a microcode update against Rogue System Register Read (Spectre V3a) and a hardware fix against Speculative Store Bypass (Spectre V4).
Already in older CPUs there were hardware mitigations against L1 Terminal Fault and Rogue Data Cache Load (Meltdown V3).
Additionally Branch Target Injection (Spectre V2) is addressed by firmware and Bounds Check Bypass (Spectre V1) by operating system patches.
Comet Lake S is based on two chips, one with ten native cores (Stepping Q0) and one with six cores (Stepping G1).
As expected, Intel didn’t comment on die size and transistor count even on demand, but we estimate the larger processor at just under 200 mm².
Intel has flattened the 10 core die from 800 µm to 500 µm in order to optimize the heat transfer (Thinning).
The background is that silicon has a comparatively poor heat coefficient, which is why a thinner ground wafer helps to dissipate the heat faster.
All CPUs based on the 10C-Die are also re-soldered by Intel.
The actual silicon wafer must be elaborately coated with several metals, including nickel and titanium.
The top layer is made of gold, and the same applies to the bottom layer of the metal cover on the processor (heat spreader), which is largely made of copper.
For only between the two gold layers does indium adhere, which serves as solder and heat conducting material.
How much a heatspreader soldered by Intel reduces the temperature compared to paste is difficult to determine.
A low temperature plays an important role for the first time in the Comet Lake S: Intel has integrated the Thermal Velocity Boost known from the mobile segment.
The i9 models clock one multiplier stage higher at below 70 Celsius than when they run hotter.
With the Core i9-10900K, the turbo on all cores increases from 4.8 GHz to 4.9 GHz and the boost on two cores from 5.2 GHz to 5.3 GHz.
Practically, these frequencies are only available when applications are started – as soon as there is a constant load, the clock rates drop.
In general, Intel has once again done strong binning for Comet Lake S, i.e. sorted the corresponding chips by pre-selection.
The initial P1272 processes, 14 nm and 14+ nm, were designed for lower frequencies.
With 14++ nm the transistor gate pitch was stretched, so Intel could use even higher clock rates at the expense of a larger chip area.
However, this does not work permanently and on all cores because the frequency of the processors is limited by the thermal power loss.
PL1, PL2, TAU, EWMA explains
Since the Sandy Bridge generation of 2011 and its Turbo Boost 2.0, Intel also defines power limits that limit or even increase the energy consumption and thus the speed of the CPUs (see PDF on page #16).
Intel emphasizes that the power limits are only guidelines, so motherboard manufacturers and OEM partners are free to change them.
Increasing the targets is therefore not overclocking, as long as the maximum frequencies specified by Intel are not exceeded depending on the number of cores used.
Only when a board is preset in such a way that these clock rates are levered out is it actually overclocking.
Intel puts the Core i9-10900K’s thermal design power (TDP) at 125 watts.
Thus, it corresponds to PL1, the power limit recommended by the CPU manufacturer under continuous load.
In older chips, like the Core i9-9900K, Intel however, names a TDP of 95 watts and lists this value as a recommendation for the PL1.
Besides the PL1, there is also the PL2, which is designed for short-term loads, i.e. burst workloads.
This includes not only starting applications, but also processing steps when editing photos, since here all filters are rarely applied directly one after the other.
In addition to this, the PL2 is bound to a time value, the so-called TAU (Turbo Time Parameter).
Up to now it was common that PL2 corresponds to a factor of 1.25 of PL1 and TAU is set to 28 seconds.
With 95 watts PL1, this means about 119 watts for almost half a minute, after which the power consumption of the chip is reduced again.
This works without any problems because the thermal inertia ensures that the processor including the cooler doesn’t overheat in this short time because heat doesn’t spread abruptly.
With the Comet Lake S, however, Intel recommends the PL1, PL2 and TAU to be set much more aggressively:
On the Core i9-10900K, the PL1 is at 125 watts, the PL2 at a solid 250 watts and the TAU is 56 seconds.
However, none of the three mainboards we used was able to keep the PL2 value that long.
The reason for this is by no means a too weak power supply, but rather the EWMA.
This stands for Exponentially Weighted Moving Average.
The EWMA determined from several factors is offset against the TAU and therefore reduces the time period in which the short-term PL2 can be present.
The more frequently high or long loads are applied to the processor, the sooner the EWMA reduces the TAU; much idle time improves it.
In our case, however, the 56 seconds – regardless of the Z490 board – were simply not achievable.
Depending on the board and benchmark, the TAU time and thus the PL2 was exhausted after half a minute, sometimes even after a few seconds.
We therefore made our tests once with the Intel recommendation of 125/250 watts and (up to) 56 seconds and also set the PL1 to 4,096 watts so that the Core i9-10900K isn’t limited by a power limit.
We can choose whether we want to use Intel spec or Unlimited on Asus’ Z490 ROG Maximus XII Hero WiFi during the boot process via F1 and F3.
Gigabytes of Z490 Aorus Master and MSI’s Z490 Ace apply 4,096 watts – which corresponds to roughly 250 watts in real life.
The Core i5-10600K on the other hand stays below 125 watts, only in the y-Cruncher (calculation of Pi with AVX code) or via Prime95 it is above that.
AMD’s 3900X stays faster
We tested all Ryzen CPUs on an Asus Crosshair VIII Hero WiFi (X570, firmware v1302), the Intel models on a Gigabyte Z490 Aorus Master (firmware vM4) and on an Asus Maximus XI Hero (Z390, firmware v1502).
The processors run with 32 GByte DDR4 memory according to the respective AMD/Intel specification, all applications and games as well as Windows 10 v1909 are installed on a Corsair Force MP600 with PCIe Gen4.
Let’s start with the Core i9-10900K: It sets itself apart from the Core i9-9900K by an average of 12 percent thanks to ten cores and a slightly higher clock rate.
In games, it’s just 6 percent, but for multithreaded applications it’s 23 percent.
The unleashed Core i9-10900K without restrictive power limits puts another 11 percent on top under load on all cores, in the best case up to 20 percent.
In games, the Core i9-10900K is faster than AMD’s cheaper Ryzen 9 3900X and in applications when clock speed instead of cores is crucial – no matter whether with 125 watts or with more than 250 watts.
On average, we see the 3900X in front.
Compared to the Core i5-9600K, the Core i5-10600K has an easy game, because it can fall back on twelve instead of six threads and also has higher frequencies.
On average, it’s enough for a gap of 19 percent, whereby we measured 14 percent in games and the same 29 percent in multithreaded applications.
The Core i5-10600K also computes faster than all Ryzen 3000 CPUs in games, although the 3800X and 3900X can still keep up in the important frame times.
In applications, the Core i5-10600K is almost as fast as a Ryzen 5 3600X, which costs significantly less with around 200 Dollar, though.
The price-competitive opponent, the Ryzen 7 3700X, has a significantly higher application performance than the Intel chip, at least when many cores are in demand.
In addition, the Ryzen 7 3700X is even more economical and can be operated with a cheaper or quieter cooling in case of doubt.
We used a Noctua NH-D15S with double ventilation for our tests, making it one of the strongest solutions on the market.
The Core i5-10600K requires around 110 to 115 watts of package power under blender load, so it doesn’t exhaust its PL1 or even PL2.
We only reach 55 degrees after 20 minutes with Blender, so the soldered heatspreader and the thin die are effective.
The Core i9-10900K with 125 watts PL1 is a bit higher with 61 degrees, but we land at 85 degrees with a constant 250 watts.
From an efficiency point of view, the Core i5-10600K and Core i9-10900K are in a worse position than their AMD counterparts, although they should exceed their nominal 65 watts and 125 watts respectively – and do so. The Ryzen 7 3700X reaches a package power of 88 watts, the Ryzen 9 3900X the full 142 watts.
Best price-performance ratio – 10600K: Final Verdict
Intel sells the Core i9-10900K for about 550 Dollar, the Core i5-10600K is supposed to be available for about 300 Dollar.
However, due to the 14 nm shortage, it remains to be seen in what quantities the chips will eventually make it into the shops and how the prices – up or down – will develop.
It is impressive what Intel has achieved with Comet Lake S from 14-nm technology and the Skylake architecture:
Starting with the Core i7-6700K (test) of 2015, the number of CPU cores increased from four to ten and the clock speed from 4.2 GHz to 5.3 GHz.
However, in the same period AMD has left the bulldozer technology including the 28 nm method behind and developed the Zen1/Zen2 architecture with 14 nm as well as 7 nm, currently the Ryzen 3000.
And the Core i9-10900K and the Core i5-10600K are at least moderately to poorly positioned against them in terms of price:
If the focus is purely on gaming, Intel is unassailable. But as soon as applications play a role, especially multithreading, the two Comet Lake chips lose.
The Core i9-10900K computes slower than the 100 Dollar cheaper Ryzen 9 3900X and the Core i5-10600K loses consistently against the equally expensive Ryzen 7 3700X.
If you use a Core i9-10900K, you should definitely look into the motherboard UEFI at the first start:
Intel’s (non-binding) recommendation of 125 watts is ignored on almost all Z490 boards.
The decore is up to 20 percent faster, but under rendering load it needs 250 watts for the CPU cores!
Although the boards are designed for this, the cooling is not mandatory.
We can select whether we want to use Intel’s defaults or unleashed power limits during the Asus ROG Maximus XII Hero WiFi’s boot process via F1 and F3.
Gigabytes of Z490 Aorus Master and MSI’s Z490 Ace directly apply full steam without manual intervention.
Additionally, the socket AM4 platform is more modern due to PCIe Gen4 and four additional lanes from the processor.
Older boards with X470 and B450 chips can also be upgraded from a Ryzen 1000/2000 to a Ryzen 4000 (Vermeer).
The new LGA 1200 version at Intel with its Z490 and B460 chips only accommodates the Comet-Lake and the future Rocket-Lake generation with 14 nm and better Cove microarchitecture.
Depending on the motherboard, however, PCIe Gen4 can be used only partially or not at all, since the 500 chipsets are connected twice as wide with an x8 instead of an x4 link.
From this point of view, upgrading a current LGA 1200 board with Rocket Lake alias 11th Gen makes little sense.
For those who are not purely focused on gaming, we therefore recommend a current AMD platform for a new purchase.
Final Verdict: Intel Core i5-10600K vs Ryzen 7 3800X
In the end, it was a really close battle between the Intel i5 10600(K) and the Ryzen 7 3800X from AMD.
Performance wise both CPUs are very strong.
All in all, if you want the best gaming performance and also application performance, i.e. for a workstation, without any compromises, you should go with the Ryzen 7 3800X.
If you use your PC mostly as a gaming machine and not primarily as a workstation you will also benefit from the Intel i5 10600(K), as it is better for gaming and less so for use in a workstation.
Also: The Intel i5 10600K offers the best price-performance ratio between the two tested CPUs.
Considering you really get 30% more in performance with the Ryzen 7 3800X compared to the Intel i5 10600(K), but pay only about 20% more for the Ryzen 9 3900X, makes for an easy decision for budget conscious users.
All in all we have to say, that we prefer the Ryzen 7 3800X. It just offers amazing performance for a great price.
There are also rumors that at the end of 2020 AMD will announce the new Ryzen 4000 series.
If this series convinces, it could be a deathblow for Intel, especially when AMD keeps its pricing strategy.
We will keep you updated!