Intel officially launched the 10th-gen Core Comet Lake H-series CPUs for high-performance laptops recently, but it's not until this week that we can discuss performance. Aimed at high-end productivity and gaming laptops, this review will comprehensively analyze the Core i7-10875H, which is the most interesting chip in the new series.
We'll compare it to previous products like the Core i9-9880H, Core i7-9750H and AMD's new Ryzen 9 4900HS to see how it stacks up for high-performance mobile productivity. Over the next few weeks and months we'll hopefully check out a range of other 10th-gen CPUs and doing all sorts of comparisons, including deeper dives into the new six-core part and how these chips fare for gaming.
This has been an interesting generation for Intel. The first 10th-gen laptop parts launched many months ago in August 2019. We saw a split into the confusing Comet Lake and Ice Lake U series, one on 14nm and one on 10nm. There's no such split for these higher-power 45W chips in the H-series: Intel is only offering Comet Lake, which is their 14nm Skylake derivative. Indeed, five years after Skylake first launched, Intel is still pushing out "new" CPUs with the same base architecture.
H-series parts offering 8 cores are not new, but with the 9th-gen family they were restricted to the Core i9 series which was only seen in very expensive notebooks. The inclusion of a new 8-core SKU in the Core i7 range means the Core i7-10875H brings an 8-core, 16-thread processor to a somewhat more mainstream market. The rest of the Core i7 range offers six cores and twelve threads like with the previous generation. Intel has also reduced the Core i9 range down to just one 8 core chip with higher clocks than the i7 CPU. Finally, the Core i5 range remains as quad cores.
The listed specifications for the Core i7-10875H are very similar to the last-gen Core i9-9880H: we have a base clock of 2.3 GHz, 16 MB of L3 cache and a 45W TDP. The only changes have occurred in the boost specs. Intel is now touting the chip's ability to push up to 5.1 GHz on a single-core (up from 4.8 GHz). However this is only achievable with what Intel calls 'Thermal Velocity Boost'. Essentially this gives the processor 200 MHz of extra frequency when running below 65C, and 100 MHz when running below 85C. If the CPU isn't running nice and cool, the actual maximum turbo frequency tops out at 4.9 GHz.
Thermal Velocity Boost debuted in 9th-gen Core i9 processors including the 9880H, but Intel has tweaked the temperature thresholds and added it to the Core i7 line in this generation.
As for all-core turbo frequencies, Intel lists this as 4.3 GHz for the 10875H, up from 4.1 GHz with the 9880H. In a nutshell, we're getting 200 to 300 MHz clock speed increases under turbo conditions with the 10875H compared to 9880H. However with base clocks and power limits remaining the same, we'd expect long term clock speeds to be similar between processors, which we'll explore a bit later.
Before getting into the benchmarks, let's talk about the test conditions...
In for review we have the Gigabyte Aorus 15G XB, which packs the Core i7-10875H in multiple power configurations alongside Nvidia's new RTX 2070 Super Max-Q. We'll be covering the GPU in a separate review, so make sure you check that out. The Aorus came equipped with a 1080p 240Hz display and 16GB of DDR4-2933, the new maximum RAM speed supported by 10th-gen parts.
This isn't a laptop review but we must mention the new Omron mechanical key switches on this laptop, they're really good. In making thin and light gaming laptops, one of the areas that tend to suffer is the keyboard, a lot of them are really junk, but this is a clicky, surprisingly satisfying switch for such a low profile design. The typing experience is nicer than we expected, so kudos to Gigabyte for choosing this.
Testing laptop hardware in an apples-to-apples configuration is challenging as each system is slightly different. For all the results you'll see today, we've tried to remove as many variables as possible. This means all systems have dual-channel memory, and all are configured to use the Intel default processor power configuration and settings unless otherwise noted. We then average together results from equivalent laptop configurations (not just one model) to provide a generalized look at performance from a given CPU.
Often laptops come a little overclocked out of the box and have utilities that allow you to change various power targets and PL1 or PL2 limits. For testing we're setting each laptop to Intel's default 45W PL1 limit. The Aorus 15G, for example, uses a 52W limit by default, and allows you to choose between five different limits from 38W through to 62W. We have also tested the 62W configuration today. The short burst power limit, PL2, is 80W with all setups.
The reason why we do this is so we can compare processor performance at a given power level. Unlike with desktops, power consumption is crucial in laptops: more power hungry parts require larger coolers and therefore larger laptops. Comparing chips on an equivalent power level allows us to see how they would perform in an equivalent type of design (check out the full list of laptops tested for this review). After all, only comparing an 80W CPU to a 35W CPU in laptops wouldn't make much sense, as you simply couldn't put the 80W chip in a design that only has 35W of cooling capacity.
Another thing people might ask of us is undervolting performance, given it's quite popular with enthusiast laptop buyers. Unfortunately, undervolting is disabled with the 10875H on the Aorus 15G and possibly other models as well. We suspect this is in response to the Plundervolt security vulnerability with Intel CPUs, though we didn't receive an official statement on the matter.
Benchmarks
Let's start the run of benchmarks with everyone's favorite Cinebench R20. This rendering workload benefits nicely from the extra two cores added into the mix: the 8-core 10875H ends up 12 percent faster than the last-gen Core i7-9750H with six cores in the multi-thread test. However in what is perhaps a bit of a surprise, performance doesn't match up with the i9-9880H: this new 8-core part is 6 percent slower than Intel's prior Core i9 processor with the same core count. This suggests that the 10875H isn't clocking quite as high as the 9880H.
The 10875H also gets dominated by AMD's Ryzen 9 4900HS despite consuming more power, falling 30% behind the new Zen 2 competitor in this workload. Even when we raise the power limit up to the maximum Gigabyte supports with the Aorus 15G, at 62W, it still falls 18 percent behind the stunning multi-core performance AMD is providing here.
However the story is different with the single-thread workload. Thanks to higher clock speeds, the 10875H is 8 percent faster than the 9880H in this reasonably long single-thread test, as well as 1% ahead of the 4900HS. Previously, the 9880H could quite comfortably run a single core at ~4.6 GHz long term within the 45W power limit, so raising the frequency cap up to 4.9 GHz which is what we typically see in the single-thread workload, does allow for faster performance than previous-gen CPUs here.
We see very similar performance in Cinebench R15, although the 10875H is a bit slower in the multi-threaded test relative to other CPUs compared to the R20 workload. Compensating for this is a better single-thread score, the 10875H is up to 10% faster than the 4900HS.
Handbrake is another long-term multi-threaded workload and we do see a moderate performance boost for this Core i7 CPU compared to the i7-9750H, pulling 11 percent ahead. However, like with Cinebench, the 10875H can't match the 9880H, and falls behind the Ryzen 9 4900HS, even when pushing up to 62W of power consumption. If you plan on using your laptop for x265 video encoding, an 8-core Ryzen APU would be a better choice given its strong performance lead.
Blender follows the trend we have seen across the last couple of benchmarks: the 10875H is decently faster than the 9750H, in this case 19% faster, but can't keep up to the 9880H, ending up 3% behind. The 10875H is also substantially slower than the Ryzen 9 4900HS. Like with the previous two examples, Intel doesn't have the performance to match AMD in long term multithreaded tasks and doesn't come close in terms of efficiency.
7-Zip compression is a great example of a short term multi-thread workload that runs entirely in the CPU's boost state. Thanks to a generous 80W PL2 power limit, the 10875H has enough power to feed all 8 cores and pull substantially ahead of the 9750H, coming in 26% faster in decompression and 34% faster in compression. It also manages to match the performance of the 9880H.
Comparing Intel and AMD in this test does show AMD pulling well ahead in decompression, which is likely going to be the more used workload of the two. However, Intel is 10% faster for compression, we suspect helped out significantly by its larger cache, at 16MB versus just 8MB on the AMD side.
Excel with large calculations is another benchmark that is favorable to the new Core i7-10875H. While still a bit slower than the i9-9880H, it does come in 18% faster than the 9750H, and 9% faster than the Ryzen 9 4900HS. This is a short multi-threaded test and with Intel blasting away nicely in its boost state taking the win.
For lighter productivity and general app usage, Intel's Core i7-10875H does well but isn't the runaway winner. In PCMark's Productivity workload, the 10875H is around 6% slower than the 4900HS, while handily beating the 9750H, while in the Essentials test which includes app loading and light web browsing, the 4900HS and 10875H offer equivalent performance. Both 8 core offerings here are a step above previous H-series CPUs, so you can expect around a 10 percent improvement.
One particularly interesting case for the 10875H is MATLAB R2020a using the built-in benchmark. Previously we found that the Core i9-9880H was able to take the crown over the Ryzen 9 4900HS in this engineering tool. However, the 10875H can't keep up with the 9880H falling about 8 percent behind.
Acrobat PDF exporting is an easy win for the Core i7-10875H, blasting through this single-core test in record time thanks to a combination of a large cache size and high frequencies. This is one of the worst results for Ryzen, it just can't match the burst speed Intel is offering for this export-to-image workload.
However this flips for AES-256 cryptography. SiSoft Sandra does show the 10875H slotting in between the 9880H and 9750H for cryptography bandwidth, but well behind the Ryzen 9 4900HS. This means that for encryption or decryption tasks, Ryzen CPUs should be able to finish ahead.
Now we get into some more compute-heavy workloads, so there's a bit going on here that needs to be discussed. We'll start here with Adobe Photoshop's Iris Blur test, which is mostly CPU limited when you have a half-decent discrete GPU, like is the case with most H-series laptops. It's a tight battle at the top between the 8-core options, the 10875H is 5% behind the 9880H but 6% ahead of the Ryzen 9 4900HS.
And we see similar margins in Puget's Photoshop workload. The 10875H isn't the outright fastest CPU for Photoshop, but it's a very capable chip in this app and notably ahead of the Ryzen 9 4900HS. It's also a decent 16% faster than the 9750H, Photoshop likes frequency but it also likes cores depending on the task you are performing, and the 10875H is faster in both regards.
Time for some Adobe Premiere results. Intel benefits in this application from hardware accelerated encoding via QuickSync, any configuration that doesn't support this which includes all AMD processors and our i9-9880H test laptop falls a bit behind in this test.
For example, in the Puget Export Test, the Ryzen 9 4900HS is equivalent to a QuickSync enabled i7-9750H, while the 10875H comes in 11% faster, albeit with a faster GPU as well. However in this particular workload we aren't often GPU limited, Intel does see better performance here by virtue of supporting QuickSync and in many cases this makes it a better choice for those that like to export videos in Premiere.
While Intel configurations can be up to 25% faster in specific QuickSync accelerated tests, Ryzen is faster for software encoding, for example when performing a 2 pass H.264 encode using the settings we use for rendering YouTube videos.
Intel also falls behind in Puget's Live Playback test: while the 10875H is 2% faster than the 9880H, it's 9% behind the Ryzen 9 4900HS for viewing footage in the timeline. It's also significantly slower for running lightly-threaded effects like the Warp Stabilizer: the 10875H is 3% faster than a 9750H for stabilizing footage, but 15% slower than AMD's Zen 2 CPU. This isn't a full single-threaded test, it uses 1-2 cores with a single instance, and it seems that Intel is only seeing gains here when 100% single-threaded.
Here's a look at how the Core i7-10875H stacks up against the Core i9-9880H in clock speeds across the first ten minutes of our Handbrake x265 test. Both CPUs fluctuate between 2.7 GHz and 2.8 GHz all-core for the most part, which of course is well below the rated 4.3 GHz all-core turbo for this processor. But mobile processors generally don't get anywhere near those all-core numbers long term, due to their power limits.
With clock speeds so close between these two processors, this suggests Intel have made no improvements to 14nm efficiency between the 9th and 10th generations. In this test the 9880H is faster, about 1.5% faster, and that seems down to very slightly higher clock speeds on average with the 9880H, in the first 10 minutes the 9880H was clocked 1% higher on average.
Performance Comparisons
Before wrapping this one up, let's take a look at some performance comparisons. We'll start here with the i7-10875H versus the i9-9880H, a battle of two very similar CPUs. The 10875H has a clear advantage in single-threaded workloads, but outside of this, the 9880H is faster.
Core i7-10875H vs. Core i9-9880H
Slightly faster in long term multi-threaded tasks, and moderately faster in shorter term tasks where the higher PL2 limit for our 9880H system leads to an advantage. From an efficiency standpoint, both CPUs seem identical, and that's basically because they are.
Core i7-10875H vs. Core i7-9750H
The Core i7-10875H brings a performance gain to the table in this comparison. In single-thread workloads this could be a 15-20 percent gain, while for multi-thread we're looking at 10 to 20 percent depending on the test. The 10875H is universally faster as it has both more cores and higher single-core turbo frequencies within the same power limit. As it can run more cores at a relatively small decrease to frequency, it ends up sitting in a more efficient point on the voltage frequency curve which is why we see higher performance despite the same 45W power limit.
As for increasing the TDP of this processor from its default 45W up to the maximum 62W supported by our Gigabyte Aorus 15 (outside of using XTU), you can expect to see up to an 18 percent performance improvement in long term multi-thread workloads. However, gains for most other workloads are limited as they're either single-core or only use the boost period, which remain unchanged.
Core i7-10875H vs. Ryzen 9 4900HS
The Core i7-10875H is generally slower than the Ryzen 9 4900HS while consuming more power. This is especially true for any long-term multi-thread workloads, where the 10875H is 25 to 30 percent slower. In lightly threaded tests, the 10875H tends to be equivalent to AMD's efficient Zen 2 design or marginally slower. However, the 10875H takes the lead in pure single-thread tests, as well as anything cache limited, like Excel or Photoshop. Workloads that make use of QuickSync like Adobe Premiere exports are also faster on the Intel CPU.
Despite being able to easily increase the power limit on the 10875H, which is something our Ryzen 9 4900HS test system in the Asus Zephyrus G14 did not, the 10875H is still generally slower in most workloads at 62W, especially anything long term or multi-threaded. Unless you can find a 10875H system that really ups that power budget to above 90W, we expect most 10875H configurations to come in behind a similar Ryzen system in these sorts of tests.
What We Learned
This is the first 10th-gen Comet Lake H processor we've tested and arguably the most interesting of the bunch. Some may look at the Core i7-10875H with skepticism with Intel keeping the 14nm process and Skylake architecture. This is a valid outlook, but we do think there are some positives, so we'll talk about those first.
The biggest one is Intel bringing 8 cores down into the Core i7 space for the first time. Spending quite a lot of money to get a Core i9 system if you wanted an 8-core processor was not worth it most of the time, but now thanks to competition, Intel's hand was forced into offering former Core i9 performance at a lower price point. In practice we are getting 10 to 20 percent more performance than a 9750H CPU.
Intel also holds the single-thread performance crown. It's not because they're hitting 5 GHz – because in most circumstances they aren't – but even at a reasonably high 4.8 GHz, they have the best single-thread performance. Combined with a larger cache and some niche advantages like QuickSync support in applications, there are some workloads where Intel's 8-core CPU is the fastest.
Now, take into consideration that the 8-core 10875H is not a drop-in replacement for the Core i7-9750H, which was used across a wide range of systems. At least in the short term, expect to see those more affordable 9750H laptops get an upgrade to the six-core 10750H, not the 10875H.
We've been able to gather some pricing information from manufacturers like MSI and Gigabyte, and the Core i7-10875H will still sit a class above the 10750H. Not quite up there with Core i9s of old, but still reasonably expensive. MSI's cheapest 10875H laptop, for example, starts at $1,800 with an RTX 2060 GPU, while the 10750H is available in $1,200 systems. Directly comparing their Creator 17 systems shows the 10875H laptop is a $300 add-in option. Also, most of the laptops that use the 10875H are your premium RTX 2070-class units.
What has changed from years prior is that AMD is joining the party and they are doing so with a more aggressive pricing scheme. Right now you can get an Asus TUF Gaming A15 with the Ryzen 7 4800H and RTX 2060 for just $1,200. Intel laptops that also feature the RTX 2060 and compete on price are using a six-core CPU, not this new 8-core.
It's possible the Core i7-10875H will see direct competition from the Ryzen 9 4900HS which is significantly faster in anything long term and multi-threaded. They are about a match in lighter productivity and burst type applications, while Intel does take the lead in pure single-thread workloads. In this regard, what remains to be seen is how well AMD can cater to laptop makers, after all it's a newcomer to this segment where Intel has ruled alone for a long time. A majority of OEMs have opted to launch high-end H-series laptops with 10th-gen Intel parts and only a few offer the option to swap to Ryzen.
The next step is testing which CPU is faster for high-end discrete GPU gaming. That's something we'll have to explore when we get the right laptops on hand. We anticipate this will be a close battle, but we're not wiling to call it either way until we run the benchmarks. Look out for that over the coming weeks.