Today, let's take a look at what is probably the most popular 10th generation processor of the Intel H series, the Core i7-10750H. This six-core processor is the successor to the very popular Core i7-8750H and Core i7-9750H, which are almost the universal choice for gaming laptops in the $ 1,000 to $ 1,800 range and often play a leading role in productivity-oriented 15- play. Inch ultraportables too.
This is not the only Core i7 processor in the Intel Comet Lake range. There are several other options, including the new 8-core i7-10875H that we tested a few weeks ago. However, after previous generations, anything above the 10750H is usually limited to premium options, while the Core i5 is shown here. which, surprisingly, are still quad cores – are designed for the most basic entry-level designs.
However, during this test there was one important question that I kept asking myself: is the Core i7-10750H actually better than the 9750H or the 8750H that stood before it? Finally, they are all based on a similar Skylake-derived architecture built using the same 14nm technology, and contain all 6 cores in a standard 45W power envelope. Can the third iteration of this processor without significant improvements to the underlying technology actually offer something new?
No doubt we will find that out today from a comprehensive comparison of the 10750H with the 9750H and 8750H, along with a look at how it can assert itself against AMD's newly shaped competitors in this price range, the Ryzen 7 with eight cores 4800H. This review focuses on productivity performance. In a separate article that we are preparing, we will then deal with games, as there are some interesting things going on with this latest generation of GPUs as well.
Let's quickly return to the technical diagram to take a look at the Core i7-10750H and what we get. The basic specifications are six cores and twelve threads, a 45 W standard TDP and 12 MB L3 cache, just like in the last two generations.
What has increased is memory support, now up to DDR4-2933, as well as clock speeds. The base remains at 2.6 GHz, but the boost clocks for single-core turbo with Thermal Velocity Boost have risen to 5.0 GHz, which means that we can achieve this frequency when the processor is running below 65 ° C. At higher temperatures, with which most laptops are operated, this frequency drops by 200 MHz to sit with a 4.8 GHz turbo. This is still a 7% increase over the Core i7-9750H's 4.5 GHz turbo.
All-core turbos have also increased from 4.0 GHz to 4.3 GHz. However, we believe that this is influenced to some extent by Thermal Velocity Boost. In practice, we've seen all-core turbos in the 4.1 to 4.2 GHz range in most cases, and the sustained frequencies are of course much lower. Don't expect your 10750H CPU to sit on all cores at 4.3 GHz all the time.
Test setup: MSI GS66 Stealth
The test system for today's review is the MSI GS66 Stealth. This fantastic laptop came with the i7-10750H and a discrete 80W RTX 2060 GPU. By default there are also 16 GB of dual-channel DDR4-2666 (which we exchanged for DDR4-2933 to get the most out of the CPU), a 512 GB SSD and a 1080p 240Hz display, which is very nice.
We are big fans of the MSI GS stealth series. We think they look great and this is no exception. We won't fully test this laptop today, but we can tell you that it has a very nice case, a great display, a huge 99.9 Wh battery, and a generally decent keyboard. This remains one of my favorite ultraportables in terms of design and functionality.
In terms of performance and performance limitations, we don't get top-notch hardware features due to the slim and light design. The CPU can easily operate at the standard 45W limit, but reaches a maximum of 53W in Turbo modes. We have seen that some thicker designs can deliver up to 70W or more. The GPU is okay with 80W, no worries.
The CPU, which only achieves a power of 53 W in CoolerBoost mode, is basically irrelevant for this test, since we operate all CPUs with standard settings. In this case, this is a long-term PL1 power limit of 45 W for the Core i7-10750H. The MSI GS66 has been configured to immediately use a huge 135 W short-term PL2. In practice, however, the CPU was not very often over 70 W. However, this is slightly higher than the 56-60 W that we have normally seen with 9750H systems.
The reason we use default settings and similar performance limits is that we can compare processor performance at a particular performance level. In contrast to desktops, the power consumption of laptops is of crucial importance: more power-hungry parts require larger coolers and therefore larger laptops. By comparing chips at an equivalent level of performance, we can see how they would behave in an equivalent design. The more a CPU can do within a certain performance limit, the more efficient it is and the more you can do with a smaller, lighter design.
Of course, this is one of the many challenges when testing laptops. Given the different configurations, it is quite difficult to make comparisons between apples. That is why we try our best to bring everything onto a balanced playing field. The following performance graphs provide averages from equivalent hardware configurations to provide a general overview of the performance of a particular CPU. The full list of laptops we tested can be found here.
One last quick note on the under tension before we turn to the benchmark results. As we discussed in our Core i7-10875H test, more and more OEMs are blocking the underdevelopment of this generation, which is likely to prevent Plundervolt exploits. This was the case with the MSI GS66 Stealth. However, you can reactivate the undervoltage using the advanced options in the MSI BIOS.
Let's start with a look at Cinebench R20, a favorite for testing multithreading, although Intel thinks this is pretty irrelevant. The Core i7-10750H doesn't offer much of an improvement over the Core i7-9750H when we look at the standard 45W power limits. In the multithread test, the new part of the 10th generation is less than 2 percent faster – error rate of the result type -, while in the single thread there is a slight increase in performance of 3 percent.
It is not much better than the 8750H: 4% faster in the multithread test and 12% faster in the single thread test thanks to an increase in the clock rate from a maximum of 4.1 GHz to 4.8 GHz. In both tests, it is easily beaten by the higher-tier Core i7-10875H with eight cores, but also by the Ryzen 7 4800H, which is in a similar tier. Ryzen is not only over 60 percent faster in the multi-threaded area, but also has a lead of 7 percent in the single-core area.
Cinebench R15 is not much different. In this case, the 10750H is on average slower than the 9750H in the multithread test, but still has a lead in the single thread. All of these margins are single digits, so there are minor changes between them.
With a long-term workload like Handbrake, we don't see much improvement compared to the 10750H versus the 9750H. Ryzen has an absolute lead over the most popular Intel Core i7 and ends this x265 transcode over 50% faster.
Blender delivers results almost identical to Handbrake when testing the current and last generation of six-core processors from Intel. With this step you simply do not gain anything essential. We don't think that's a surprise since it's basically the same silicon.
Code compilation was one of the best results we've seen for the 10750H. In this GCC compilation, which is a mix of single and multithreaded sections, the 10750H was 10 percent ahead of the 9750H, which in some situations is a combination of higher boost and clock rates. However, it still loses slightly against the Ryzen 7 4800H, which in most cases is faster for both multi-thread and light-thread workloads.
Microsoft Excel was a workload in which the 10750H of the laptops we tested lagged slightly behind the 9750H on average. This is caused by some deviations in the boost behavior. Since Excel largely consists of multithreading, the performance is in double digits behind the 10875H and the 4800H.
In PCMark 10's lighter workloads, slightly improving single-thread performance helps in some situations. We see a performance test in the productivity test that corresponds to that of the 9750H, although it lags behind by around 10 percent in both the 10875H and the 4800H. In the Essentials test, this increase in clock speed contributes to 3% better performance, but again falls behind the faster 8-core model from Intel and the new Ryzen 7 from AMD.
7-Zip shows that the performance between 10750H, 9750H and 8750H is very similar, not much here. Since this test is a multithread test, the performance is slightly behind the 8-core 10875H from Intel, which is up to 25% slower, and also behind the Ryzen 7 4800H.
MATLAB shows no real performance gain with the 10750H compared to the 9750H. Similar to other tests we just looked at, it's two digits slower than 8 core CPU options. Given the results of Ryzen 7, six cores look a bit dated at this price.
One of the most significant successes for the 10750H is the export of Acrobat PDF, which is a single-threaded workload. The 10750H was 7% faster in this test with 7% higher clock speeds for the 10750H compared to the 9750H and the ability of both CPUs to work with these clocks within the performance limit on a single core. This enables him to claim one of the few performance advantages over the Ryzen 7 4800H that I have seen.
Again, we don't see a big difference in AES performance between the 9750H and the 10750H or even the 10875H. This is because it is a hardware accelerated function and there doesn't seem to be any advantage here, probably because the architecture is the same. Cryptography is another area where Ryzen is much faster in laptops, 35% faster in multithreading.
Photoshop is one of the rare workloads where the needle has shifted slightly in favor of the new 10th generation CPU. In Puget's benchmark, we see 9% higher performance with the same RTX 2060 GPU. This allowed him to match the Ryzen 7 4800H configuration with a lower-class GTX 1660 Ti, which is reasonable at par until you find a cheaper Ryzen configuration outperforms it.
DaVinci Resolve Studio with the Puget workload is another rare case where our Core i7-10750H configuration was slower than our Core i7-9750H system with the same RTX 2060 GPU. It wasn't massively slower, but it was slower overall, which is a strange result. Nevertheless, both CPU options fell behind the Ryzen 7 4800H typical for video coding applications.
And finally, we come to our Premiere workloads. When exporting, there is no real difference between the 10750H and 9750H when paired with the same GPU. With the latest Premiere Beta version, which adds hardware acceleration for Nvidia GPUs, the error rate is examined. As with DaVinci Resolve, the Ryzen 7 4800H is a better choice here because it can chew through the CPU-limited parts of the encoding faster.
Live playback performance is pretty similar for most of these options as well. The Ryzen 7 4800H is slightly, but not much. If you are like us and stick to pure software coding for maximum quality, there is not much to gain between 10750H and 9750H again. We noticed an improvement of 6% here, which is actually on the higher side, but 25% lower performance than a Ryzen 7 4800H configuration, it is just not as impressive overall.
And finally, we have the warp stabilizer effect, which is slightly threaded. No increase in performance compared to the Core i7-9750H, and as we mentioned in apparently every single benchmark in this video, it simply cannot keep up with the Ryzen 7 4800H in this test. When running a single warp stabilizer instance, the 4800H was 22% faster.
What we have learned
The main reason we don't get better performance with this generation is the simple fact that Intel hasn't changed architecture or manufacturing technology in years. If there are no IPC gains and efficiency is only minimally improved, you can stay with a performance-constrained form factor like a laptop with practically no performance gains. This is most obvious when you are viewing clock rates.
In Cinebench R20, the 10750H works almost exactly like a Core i7-9750H, which itself works like a Core i7-8750H. These two processors of the last generation have a long-term performance of 3.1 GHz over six cores over a period of 45 W within the 45 W power limit. The 10750H clocks around 3.2 GHz in the long term, which in this case means an increase of 100 MHz. With a clock speed of 3%, it's no surprise that we've often seen improvements in multi-core performance in the 2 to 3 percent range, which is negligible.
If we look at the overall performance summary, at best we see a 10% increase in apps with low threading or single threading. However, this depends on the test and for the most part we would say that the processors are a game. This leads to limited 8750H tests. Multi-core performance has only improved a few percent in Cinebench. The greatest advantages result from increasing the clock rate for single core.
Compared to the other Core i7 CPU from Intel, which offers 8 cores, the Core i7-10875H, the 10750H is generally slower as expected. Sometimes it is 10 percent slower for single-thread workloads, but 20 to 25 percent slower for multi-thread workloads. Given that most 10750H high-end laptops have a more expensive 10875H option, or even something with the Core i9-10980HK, this is the kind of performance you're missing out on.
If you then compare the Core i7-10750H with the Ryzen 7 4800H, it is a bloodbath for the Intel processor. Since both CPUs are limited to 45 W, Ryzen offers better multicore performance and better single-core performance in almost all cases. If you have productivity workloads that largely consist of multiple cores, the performance advantage that Ryzen offers with 8 Zen 2 cores at 7 nm is very significant.
Our general thoughts on the Core i7-10750H can be summed up in one word: inconspicuous. Intel offers the same performance as the last generation i7-9750H with some very minor improvements. This practically corresponds to the performance of the Core i7-8750H from two generations ago. Productivity performance has no reason to consider upgrading.
The only way to get more performance with a new Intel chip in this segment is to buy a laptop that can run the CPU at a higher performance limit. If you had a 45 W laptop and can upgrade to something with sufficient cooling for 60 W or 70 W, you will see an improvement of 15 to 20 percent. However, this is not an actual performance gain over the 10th generation CPU, but a performance gain from a laptop with a better cooler.
Since Intel has been offering performance at the same price for two years, AMD has shaken the laptop market with a much faster processor in this category. The Ryzen 7 4800H is mostly two digits faster with the same power consumption, and this lead can be 50% or more if a workload is fed with all 8 cores.
This conversation can get a bit cloudy if you consider all the performance limits and configurations that OEMs offer. Given that Intel still can't beat AMD's 45W processor with a 90W 8-core option.
The obvious reality of this generation is that any laptop using a Core i7-10750H could very likely be faster and better for developer productivity or workload if a Ryzen 7 4800H were used instead. Even with this MSI GS66 Stealth, and we'll be honest, it's a very nice laptop, but it lacks Ryzen's 50% productivity boost.
In other words, in the portable H-series category, the MSI GS66 is easily outperformed by the smaller and cheaper Asus Zephyrus G14 in terms of productivity. The G14 is $ 300 cheaper, a 14-inch design instead of 15 inches, and weighs 500 grams less. This is the real advantage AMD currently offers, albeit in a limited number of systems.