Deutsch
Español
Français
Italiano
Nederlands
Polski
Português
Русский
Türkçe
Svenska
Chinese
MagyarMSI GS66 Stealth 10UG Laptop Review: How Fast is a 95 W GeForce RTX 3070?
MSI has been quick to jump in on the mobile GeForce RTX 30 bandwagon with plenty of updated models to choose from. The GS66, GE76, and GE66 are now all available with the latest Nvidia Ampere GPUs to replace last year's Turing options. Our GS66 10UG configuration comes with a GeForce RTX 3070 which MSI has confirmed with us to be targeting a 95 W TGP. There is also a new Core i7-10870H CPU to supplement the existing Core i7-10875H.
Other than the new CPU and GPU, the chassis remains visually identical to previous GS66 models. We recommend checking out our existing reviews on the GS66 to learn more about the physical features of the laptop as they all apply here as well. The GS66 competes directly with other ultra-thin high-performance 15.6-inch laptops like the Razer Blade 15, Lenovo Legion 7, Gigabyte Aero 15, and the Asus Zephyrus series.
More MSI reviews:
potential competitors in comparison
Rating | Date | Model | Weight | Height | Size | Resolution | Best Price |
|---|---|---|---|---|---|---|---|
| 84 % | 02/2021 | MSI GS66 Stealth 10UG i7-10870H, GeForce RTX 3070 Laptop GPU | 2.1 kg | 19.8 mm | 15.60" | 1920x1080 | |
| 84 % | 05/2020 | MSI GS66 10SGS i7-10750H, GeForce RTX 2080 Super Max-Q | 2.2 kg | 20 mm | 15.60" | 1920x1080 | |
| 82.4 % | 02/2021 | Asus TUF Dash F15 FX516PR i7-11370H, GeForce RTX 3070 Laptop GPU | 2 kg | 20 mm | 15.60" | 1920x1080 | |
| 86.9 % | 08/2020 | Razer Blade 15 RTX 2070 Super Max-Q i7-10875H, GeForce RTX 2070 Super Max-Q | 2.1 kg | 17.8 mm | 15.60" | 1920x1080 | |
| 84.5 % | 01/2021 | Aorus 15P WB i7-10750H, GeForce RTX 2070 Max-Q | 2.1 kg | 23 mm | 15.60" | 1920x1080 | |
| 84.8 % | 07/2020 | Alienware m15 R3 P87F i7-10750H, GeForce RTX 2070 Mobile | 2.3 kg | 19.9 mm | 15.60" | 1920x1080 |
Case
Connectivity
Communication
Webcam
Maintenance
MSI has made no major internal changes outside of the CPU and GPU swap. The cooling solution and components appear identical to what we observed on last year's Turing-powered GS66.
Accessories and Warranty
The retail box include a large velvet cloth in addition to the usual AC adapter and manuals. The standard one-year limited warranty applies with additional options if purchased from Xotic PC.
Input Devices
Display
The 300 Hz AU Optronics B156HAN12.0 IPS panel on our GS66 10UG is the same one as on last year's GS66 10SGS. Strangely, however, maximum brightness is noticeably dimmer this time at just 264 nits compared to 331 nits on the older model even after double-checking. We're not sure why this is happening as we can identify no automatic brightness controls or limiters, either.
Aside from the brightness discrepancy, other characteristics of the panel remain intact including the very fast response times, >1000:1 contrast ratio, and wide color space. Users should keep in mind that the 300 Hz refresh rate is useful even if you're unlikely to ever reach such high frame rates when gaming.
| |||||||||||||||||||||||||
Brightness Distribution: 90 %
Center on Battery: 259.5 cd/m²
Contrast: 1038:1 (Black: 0.25 cd/m²)
ΔE Color 2.52 | 0.59-29.43 Ø5.2, calibrated: 1.93
ΔE Greyscale 1.7 | 0.57-98 Ø5.5
98.8% sRGB (Argyll 1.6.3 3D)
63.3% AdobeRGB 1998 (Argyll 1.6.3 3D)
69.1% AdobeRGB 1998 (Argyll 2.2.0 3D)
98.4% sRGB (Argyll 2.2.0 3D)
67.7% Display P3 (Argyll 2.2.0 3D)
Gamma: 2.04
| MSI GS66 Stealth 10UG AU Optronics B156HAN12.0, IPS, 15.60, 1920x1080 | MSI GS66 10SGS AU Optronics B156HAN12.0, IPS, 15.60, 1920x1080 | Asus TUF Dash F15 FX516PR LQ156M1JW26 (SHP1532), IPS, 15.60, 1920x1080 | Razer Blade 15 RTX 2070 Super Max-Q AU Optronics B156HAN12.0, IPS, 15.60, 1920x1080 | Aorus 15P WB BOE HF NV156FHM-N4G, IPS, 15.60, 1920x1080 | Alienware m15 R3 P87F AU Optronics B156HAN, IPS, 15.60, 1920x1080 | |
|---|---|---|---|---|---|---|
| Display | 2% | -1% | 2% | -7% | 1% | |
| Display P3 Coverage | 67.7 | 69.9 3% | 66.9 -1% | 69.4 3% | 63 -7% | 68.6 1% |
| sRGB Coverage | 98.4 | 99.1 1% | 98.1 0% | 98.8 0% | 91.3 -7% | 98.9 1% |
| AdobeRGB 1998 Coverage | 69.1 | 71.1 3% | 67.9 -2% | 70.5 2% | 64.6 -7% | 69.8 1% |
| Response Times | 18% | 16% | 55% | 30% | 40% | |
| Response Time Grey 50% / Grey 80% * | 14.4 ? | 11.2 ? 22% | 12 ? 17% | 3.2 ? 78% | 8.8 ? 39% | 4 ? 72% |
| Response Time Black / White * | 11.6 ? | 10 ? 14% | 10 ? 14% | 8 ? 31% | 9.2 ? 21% | 10.8 ? 7% |
| PWM Frequency | 23580 ? | |||||
| Screen | 14% | -3% | -12% | -7% | -6% | |
| Brightness middle | 259.5 | 339.4 31% | 286 10% | 345.4 33% | 275 6% | 326.7 26% |
| Brightness | 264 | 331 25% | 275 4% | 342 30% | 268 2% | 316 20% |
| Brightness Distribution | 90 | 92 2% | 91 1% | 90 0% | 91 1% | 87 -3% |
| Black Level * | 0.25 | 0.24 4% | 0.25 -0% | 0.34 -36% | 0.33 -32% | 0.33 -32% |
| Contrast | 1038 | 1414 36% | 1144 10% | 1016 -2% | 833 -20% | 990 -5% |
| Colorchecker dE 2000 * | 2.52 | 1.58 37% | 2.43 4% | 3.18 -26% | 2.72 -8% | 2.4 5% |
| Colorchecker dE 2000 max. * | 5.55 | 3.8 32% | 6.33 -14% | 6.57 -18% | 4.45 20% | 6.06 -9% |
| Colorchecker dE 2000 calibrated * | 1.93 | 1.47 24% | 0.83 57% | 1.77 8% | 2.09 -8% | 1.51 22% |
| Greyscale dE 2000 * | 1.7 | 2.3 -35% | 3.5 -106% | 3.8 -124% | 2.3 -35% | 3.2 -88% |
| Gamma | 2.04 108% | 2.24 98% | 2.32 95% | 2.15 102% | 2.266 97% | 2.13 103% |
| CCT | 6414 101% | 6952 93% | 7145 91% | 7109 91% | 6933 94% | 7129 91% |
| Color Space (Percent of AdobeRGB 1998) | 63.3 | 64.8 2% | 68 7% | 64.3 2% | 65 3% | 63.7 1% |
| Color Space (Percent of sRGB) | 98.8 | 99.4 1% | 91 -8% | 99 0% | 90 -9% | 99.2 0% |
| Total Average (Program / Settings) | 11% /
13% | 4% /
-0% | 15% /
-1% | 5% /
-3% | 12% /
1% |
* ... smaller is better
Color space is identical to last year's high refresh rate 1080p panels at approximately 63 percent AdobeRGB and 99 percent sRGB. MSI is offering no 4K options at the moment for its GeForce RTX 30 GS66 refresh.
The display is well-calibrated against the sRGB standard out of the box with average grayscale and ColorChcker deltaE values of just 1.7 and 2.52, respectively. These values would improve just marginally after calibrating the panel even further ourselves.
Display Response Times
| ↔ Response Time Black to White | ||
|---|---|---|
| 11.6 ms ... rise ↗ and fall ↘ combined | ↗ 7.6 ms rise |  |
| ↘ 4 ms fall | ||
| The screen shows good response rates in our tests, but may be too slow for competitive gamers. In comparison, all tested devices range from 0.4 (minimum) to 240 (maximum) ms. » 21 % of all devices are better. This means that the measured response time is better than the average of all tested devices (22.7 ms). | ||
| ↔ Response Time 50% Grey to 80% Grey | ||
| 14.4 ms ... rise ↗ and fall ↘ combined | ↗ 6.8 ms rise |  |
| ↘ 7.6 ms fall | ||
| The screen shows good response rates in our tests, but may be too slow for competitive gamers. In comparison, all tested devices range from 0.25 (minimum) to 636 (maximum) ms. » 18 % of all devices are better. This means that the measured response time is better than the average of all tested devices (35.8 ms). | ||
Screen Flickering / PWM (Pulse-Width Modulation)
| Screen flickering / PWM not detected |  | ||
In comparison: 53 % of all tested devices do not use PWM to dim the display. If PWM was detected, an average of 19466 (minimum: 5 - maximum: 3846000) Hz was measured. | |||
Outdoor visibility is worse than most Ultrabooks as the display is simply not as bright. The latest Dell XPS 15, for example, is over two times brighter than our MSI. For such a thin and light gaming laptop, we were hoping for a more vibrant display when traveling and working out of office.
Performance
The Core i7-10870H is exactly the same as the Core i7-10875H with just a couple of minor differences. Firstly, base and maximum clock rates are each 100 MHz slower than the Core i7-10875H at 2.2 GHz vs. 2.3 GHz and 5 GHz vs. 5.1 GHz, respectively. Secondly, the CPU is not compatible with certain Intel security features including vPro and Trusted Execution that dedicated gaming laptops would likely never utilize, anyway. It was rumored that Intel introduced the Core i7-10870H to alleviate the supply issues some laptop makers were experiencing with the Core i7-10875H late last year.
Unless otherwise noted, our benchmarks below were performed on Extreme Performance mode. We recommend users become familiar with the MSI Dragon Center software as important performance and fan controls can be found here. Note that the Dragon Center real-time monitoring tool is prone to freezing whenever the CPU is at 100 percent utilization and so it is not always reliable.
Processor
Raw CPU performance is where we expect it to be according to CineBench benchmarks with scores that are just a few percentage points behind the Core i7-10875H. However, maximum Turbo Boost is not sustainable for long periods as exemplified by our CineBench R15 xT loop test. Performance would start out strong at almost 1650 points before falling to 1480 points and lower over time. Results would not change even after a retest. Note that we had set our system to Cooler Boost Turbo mode prior to starting the loop test and so the GS66 ought to be performing at its best under these conditions.
Despite the dip in clock rates, the Core i7-10870H can still offer 30 to 35 percent faster multi-thread performance than the Core i7-10750H on last year's GS66 to be roughly in between the competing AMD Ryzen 5 4600H and Ryzen 7 4800H.
Cinebench R15: CPU Multi 64Bit | CPU Single 64Bit
Blender: v2.79 BMW27 CPU
7-Zip 18.03: 7z b 4 -mmt1 | 7z b 4
Geekbench 5.4: Single-Core | Multi-Core
HWBOT x265 Benchmark v2.2: 4k Preset
LibreOffice : 20 Documents To PDF
R Benchmark 2.5: Overall mean
Cinebench R15: CPU Multi 64Bit | CPU Single 64Bit
Blender: v2.79 BMW27 CPU
7-Zip 18.03: 7z b 4 -mmt1 | 7z b 4
Geekbench 5.4: Single-Core | Multi-Core
HWBOT x265 Benchmark v2.2: 4k Preset
LibreOffice : 20 Documents To PDF
R Benchmark 2.5: Overall mean
* ... smaller is better
System Performance
PCMark results are within single-digit percentage points of other GeForce RTX-powered laptops including those from the older Turing series. In other words, it doesn't appear that the gen-to-gen GPU performance advantage of mobile Ampere is wide enough to really impact PCMark scores.
| PCMark 8 | |
| Home Score Accelerated v2 | |
| Razer Blade 15 RTX 2070 Super Max-Q | |
| Alienware m15 R3 P87F | |
| MSI GS66 Stealth 10UG | |
| MSI GS66 10SGS | |
| Aorus 15P WB | |
| Work Score Accelerated v2 | |
| Razer Blade 15 RTX 2070 Super Max-Q | |
| MSI GS66 Stealth 10UG | |
| Alienware m15 R3 P87F | |
| MSI GS66 10SGS | |
| Aorus 15P WB | |
| PCMark 8 Home Score Accelerated v2 | 4792 points | |
| PCMark 8 Work Score Accelerated v2 | 6300 points | |
| PCMark 10 Score | 6121 points | |
Help | ||
DPC Latency
LatencyMon shows no major latency issues when opening multiple browser tabs of our homepage. 4K video playback at 60 FPS is smooth but not perfect with 8 dropped frames during our minute-long video test.
| DPC Latencies / LatencyMon - interrupt to process latency (max), Web, Youtube, Prime95 | |
| Razer Blade 15 RTX 2070 Super Max-Q | |
| MSI GS66 10SGS | |
| Alienware m15 R3 P87F | |
| Asus TUF Dash F15 FX516PR | |
| Aorus 15P WB | |
| MSI GS66 Stealth 10UG | |
* ... smaller is better
Storage Devices
| MSI GS66 Stealth 10UG WDC PC SN730 SDBPNTY-1T00 | MSI GS66 10SGS Samsung SSD PM981 MZVLB512HAJQ | Asus TUF Dash F15 FX516PR SK Hynix HFM001TD3JX013N | Razer Blade 15 RTX 2070 Super Max-Q Samsung SSD PM981a MZVLB512HBJQ | Aorus 15P WB ADATA XPG SX8200 Pro (ASX8200PNP-512GT) | Alienware m15 R3 P87F 2x Micron 2200S 512 GB (RAID 0) | |
|---|---|---|---|---|---|---|
| AS SSD | -11% | -20% | -9% | -18% | -40% | |
| Seq Read | 2198 | 1086 -51% | 2671 22% | 1957 -11% | 2241 2% | 2058 -6% |
| Seq Write | 2596 | 1901 -27% | 1501 -42% | 2092 -19% | 2121 -18% | 2316 -11% |
| 4K Read | 43.81 | 51 16% | 39.65 -9% | 50.2 15% | 52.3 19% | 39.72 -9% |
| 4K Write | 115.5 | 117.3 2% | 69 -40% | 96.5 -16% | 116.2 1% | 62.9 -46% |
| 4K-64 Read | 1692 | 1503 -11% | 1209 -29% | 1594 -6% | 713 -58% | 1334 -21% |
| 4K-64 Write | 2162 | 1721 -20% | 1650 -24% | 2005 -7% | 827 -62% | 1271 -41% |
| Access Time Read * | 0.056 | 0.059 -5% | 0.093 -66% | 0.057 -2% | 0.054 4% | 0.062 -11% |
| Access Time Write * | 0.098 | 0.1 -2% | 0.055 44% | 0.112 -14% | 0.031 68% | 0.388 -296% |
| Score Read | 1956 | 1662 -15% | 1516 -22% | 1840 -6% | 989 -49% | 1580 -19% |
| Score Write | 2537 | 2029 -20% | 1869 -26% | 2310 -9% | 1155 -54% | 1565 -38% |
| Score Total | 5513 | 4588 -17% | 4105 -26% | 5103 -7% | 2659 -52% | 3968 -28% |
| Copy ISO MB/s | 2236 | 2155 -4% | 1793 -20% | 2192 -2% | 1685 -25% | |
| Copy Program MB/s | 690 | 715 4% | 573 -17% | 576 -17% | 669 -3% | |
| Copy Game MB/s | 1546 | 1506 -3% | 1418 -8% | 1062 -31% | 1493 -3% | |
| CrystalDiskMark 5.2 / 6 | -10% | -13% | 3% | 3% | -7% | |
| Write 4K | 163.1 | 132.5 -19% | 129.6 -21% | 139.8 -14% | 131.7 -19% | 67.3 -59% |
| Read 4K | 48.97 | 45.03 -8% | 44.71 -9% | 48.62 -1% | 55.7 14% | 41.47 -15% |
| Write Seq | 2607 | 1987 -24% | 1929 -26% | 2260 -13% | 2273 -13% | 2239 -14% |
| Read Seq | 2316 | 1434 -38% | 2035 -12% | 2135 -8% | 2533 9% | 2112 -9% |
| Write 4K Q32T1 | 338 | 467.5 38% | 425.9 26% | 499.3 48% | 475.1 41% | 488.9 45% |
| Read 4K Q32T1 | 475.6 | 536 13% | 433.4 -9% | 552 16% | 590 24% | 557 17% |
| Write Seq Q32T1 | 3090 | 2003 -35% | 1390 -55% | 2960 -4% | 2373 -23% | 2654 -14% |
| Read Seq Q32T1 | 3421 | 3306 -3% | 3505 2% | 3269 -4% | 3068 -10% | 3137 -8% |
| Total Average (Program / Settings) | -11% /
-10% | -17% /
-17% | -3% /
-5% | -8% /
-10% | -24% /
-28% |
* ... smaller is better
Disk Throttling: DiskSpd Read Loop, Queue Depth 8
GPU Performance
The mobile GeForce RTX 3070 here is about 20 percent faster than the GeForce RTX 2080 Max-Q in last year's GS66 10SGS or 30 percent faster than the 2070 Super Max-Q in the GS66 10SFS. The gen-to-gen gains are nowhere near the desktop RTX 3070 as the desktop version offers about 50 percent faster graphics power than our mobile version. Overall gaming performance is neck-to-neck with the desktop RTX 2070.
| 3DMark 11 Performance | 25114 points | |
| 3DMark Cloud Gate Standard Score | 49338 points | |
| 3DMark Fire Strike Score | 19754 points | |
| 3DMark Time Spy Score | 9140 points | |
Help | ||