Original Link: https://www.anandtech.com/show/16614/xiaomi-mi-11-lite-5g-performance-report-first-taste-of-the-snapdragon-780
Xiaomi Mi 11 Lite 5G Performance Report: First Taste of the Snapdragon 780G
by Andrei Frumusanu on April 13, 2021 9:00 AM ESTIt’s only been a couple of weeks since Qualcomm first announced its new “premium” range SoC for 2021: the new Snapdragon 780G. The chip is a successor to the Snapdragon 765 which ended up being a quite successful and popular chip among 2020 mid-range and “premium” segment devices, powering many devices such as for example the Pixel 5 and 4a 5G.
Also a couple of weeks ago, Xiaomi had made a flurry of device announcements among the Mi 11 Ultra as the new mega-flagship device for 2021. One device we didn’t have time to cover was the new Mi 11 Lite 5G. Beyond the super interesting form-factor of a 6.8mm thick device and weighing in only 159g despite its larger 160.5 x 75.5mm footprint, as well as higher quality OLED display, the phone differentiated itself by being the very first Snapdragon 780G phone in the market.
We’ve got our hands on a review sample, and wanted to dedicate a piece focusing a bit more on the SoC, as the new chip will be powering not only the Mi 11 Lite 5G, but undoubtedly a whole lot of other mid-range and premium devices in the coming months.
Qualcomm Snapdragon Premium SoCs | ||||
SoC | Snapdragon 765 Snapdragon 765G |
Snapdragon 768G |
Snapdragon 780G |
|
CPU | 1x Cortex-A76 @ 2.3GHz (non-G) @ 2.4GHz (765G) 1x Cortex-A76 @ 2.2GHz 6x Cortex-A55 @ 1.8GHz |
1x Cortex-A76 @ 2.8GHz 1x Cortex-A76 @ 2.4GHz 6x Cortex-A55 @ 1.8GHz |
1x Cortex-A78 @ 2.4GHz 3x Cortex-A78 @ 2.2GHz 4x Cortex-A55 @ 1.9GHz |
|
GPU | Adreno 620 |
Adreno 620 +15% perf over 765G |
Adreno 642 +50% perf over 768G |
|
DSP / NPU | Hexagon 696 HVX + Tensor 5.4TOPS AI (Total CPU+GPU+HVX+Tensor) |
Hexagon 770 Scalar+Tensor+Vector 12TOPs AI (Total CPU+GPU+DSP) |
||
Memory Controller |
2x 16-bit CH @ 2133MHz LPDDR4X / 17.0GB/s |
|||
ISP/Camera | Dual 14-bit Spectra 355 ISP 1x 192MP or 1x 36MP ZSL or 2x 22MP with ZSL |
Triple 14-bit Spectra 570 ISP 1x 192MP or 1x 84MP ZSL or 2x 64+20MP ZSL or 3x 25MP ZSL |
||
Encode/ Decode |
2160p30, 1080p120 H.264 & H.265 10-bit HDR pipelines |
|||
Integrated Modem | Snapdragon X52 Integrated (LTE Category 24/22) DL = 1200 Mbps 4x20MHz CA, 256-QAM UL = 210 Mbps 2x20MHz CA, 256-QAM (5G NR Sub-6 4x4 100MHz + mmWave 2x2 400MHz) DL = 3700 Mbps UL = 1600 Mbps |
Snapdragon X53 Integrated (LTE Category 24/22) DL = 1200 Mbps 4x20MHz CA, 256-QAM UL = 210 Mbps 2x20MHz CA, 256-QAM (5G NR Sub-6 4x4 100MHz) DL = 3300 Mbps UL = ? Mbps |
||
Mfc. Process | Samsung 7nm (7LPP) |
Samsung 5nm (5LPE) |
On paper, the new Snapdragon 780G is a major upgrade from the Snapdragon 765G. The new SoC features not only upgraded CPU microarchitectures, actually jumping from a Cortex-A76 generation to a newer Cortex-A78 generation, but also shifts from a 1+1+4 core setup to a 1+3+4 setup, meaning the new SoC has double the amount of performance CPU cores. The clock frequency remains at a peak 2.4GHz for the fastest core in the configuration.
This setup is quite interesting as it’s quite similar to the middle CPU cores of the Snapdragon 888 flagship SoC, which employs the same microarchitecture at roughly the same clocks, and both chips are manufactured on Samsung’s 5nm 5LPE process node. There are differences in the memory setup as the Snapdragon 780G remains on a LPDDR4X interface that’s only 2x16b wide, which is smaller than the LPDDR5 4x16b bus that the flagship SoC gets. There’s also differences in caches such as the L3 cache being much smaller.
CPU Performance & Efficiency
Comparing the performance and power efficiency of the new chip compared to not only its predecessor, but also its current generation flagship sibling, we should be able to paint a better picture of how the new chip sits in the competitive landscape.
In SPEC2006, the new chip – albeit running at the same frequency as the Snapdragon 765G, is showcasing significant performance leaps compared to its predecessor. We’re seeing roughly +34% performance improvements in both integer and floating point test suites, not quite 40% as Qualcomm had proclaimed, but that’s also an “up to” figure depending on the workload.
In the performance positioning, we indeed see that the performance core of the Snapdragon 780 roughly matches up to the middle cores of the Snapdragon 888, which is expected given their similar specifications. The S888 wins out ahead due to a stronger memory subsystem.
In terms of power and energy efficiency, it’s an interesting situation. In terms of power usage, we see that the performance increases did come with a cost, as the new chip is roughly +20% more power hungry as its predecessor in single-threaded tasks. Due to the performance increase being higher than the power increase however, the energy efficiency of the new chip is actually higher than its predecessor, showcasing 8 to 13% less energy usage to complete a task, which should have direct effects on the battery life of devices.
In general, the larger performance increase combined with a reduction in energy usage of the new SoC means that it’s a pretty significant generational leap against the Snapdragon 765.
A bit unrelated to the Snapdragon 780, last month Google had finally pushed AArch64 versions of Chrome and the system WebView packages, signifying an important shift away from AArch32 32-bit era. This actually has a larger impact on benchmarks as newer Arm Cortex microarchitectures have more significant differences (relative to past ones) between the two execution modes, with the AArch64 mode being more performant and more efficient. For this reason, we’re starting with a new clean slate of test results with explicitly stated (64b) in the graphs.
We’re comparing the Mi 11 Lite 5G to the Pixel 5 here as the Mi 10 Lite 5G with the Snapdragon 765G unfortunately is still being deployed 32-bit packages due to its lower 6GB RAM, and we still want an apples-to-apples comparison.
In the JetStream 2 and Speedometer 2.0 browser benchmarks we’re seeing +36-43% performance increases compared to the Snapdragon 765G, which are significant boosts. The Snapdragon 780G still clearly lags behind the performance of the flagship SoCs, signifying that there’s still plenty of performance differentiation to be had between the segments.
Excellent User Experience
In general, the performance leap of the Snapdragon 780G is quite significant. The new CPU microarchitecture delivers large generational IPC boosts and the fact that the new chip has double the big performance cores also helps out multi-threaded workloads a lot.
Truth be told, it’s a vast shift in user experience as the device definitely feels much faster and snappier than Snapdragon 765G devices. It roughly lands in between the Snapdragon 855 and Snapdragon 865 flagship generations in terms of overall device experience, which is not surprising as that’s also exactly where the new chip benchmarks itself at.
GPU Performance & Power
The GPU on the Snapdragon 780 is interesting – Qualcomm calls it an Adreno 642, and had promised performance increases over +50% compared to the Snapdragon 768G. This latter chip wasn’t employed as much, but it had promised +15% increases over the 765G. The combined math of the two generations should signify that the Snapdragon 780G should be around +72% faster than the Snapdragon 765G. It’s also interesting that Qualcomm called the GPU an Adreno 642 which is a larger numerical jump from the Adreno 620 of the previous generations – we asked ourselves how this would position itself against the Adreno 640 of the Snapdragon 855 flagship from a few years back.
Starting off with Basemark GPU, we’re off to a good start with peak performance increases of +50% versus the Xiaomi Mi 10 Lite 5G with the Snapdragon 765G.
As a note, these results also include updated performance numbers of the Pixel 5 which features the same S765G chipset; Google’s latest April patch seems to have finally (after 6 months….) resolved the crippling GPU performance behaviour which the device suffered from since launch.
In GFXBench Aztec high, the performance increases are much higher, with the Mi 11 Lite 5G showcasing a +107% performance boost versus its predecessor, which is quite astounding.
GFXBench Aztec High Offscreen Power Efficiency (System Active Power) |
||||
Mfc. Process | FPS | Avg. Power (W) |
Perf/W Efficiency |
|
iPhone 12 Pro (A14) π₯ Throttled | N5 | 28.36 | 3.91 | 7.24 |
Mate 40 Pro (Kirin 9000) π Power-Save | N5 | 23.71 | 3.35 | 7.07 |
iPhone 11 Pro (A13) π₯ Throttled | N7P | 26.14 | 3.83 | 6.82 |
Galaxy S21U (Snapdragon 888) π₯ Throttled | 5LPE | 18.94 | 2.81 | 6.71 |
iPhone 12 Pro (A14) βοΈ Peak | N5 | 37.40 | 5.57 | 6.64 |
iPhone 11 Pro (A13) βοΈ Peak | N7P | 34.00 | 6.21 | 5.47 |
Galaxy S20U (Snapdragon 865) | N7P | 20.35 | 3.91 | 5.19 |
Mi 11 Lite 5G (Snapdragon 780G) | 5LPE | 17.68 | 3.46 | 5.10 |
Mate 40 Pro (Kirin 9000) π₯ Throttled | N5 | 27.37 | 5.39 | 5.07 |
iPhone XS (A12) π₯ Throttled | N7 | 19.32 | 3.81 | 5.07 |
Reno3 5G (Dimensity 1000L) | N7 | 11.93 | 2.39 | 4.99 |
Galaxy S21U (Exynos 2100) π₯ Throttled | 5LPE | 18.55 | 3.73 | 4.96 |
iPhone XS (A12) βοΈ Peak | N7 | 26.59 | 5.56 | 4.78 |
Mi 10 Lite 5G (Snapdragon 765G) | 7LPP | 8.53 | 1.93 | 4.42 |
Mate 40 Pro (Kirin 9000) βοΈ Peak | N5 | 37.22 | 8.53 | 4.36 |
ROG Phone III (Snapdragon 865+) | N7P | 22.34 | 5.35 | 4.17 |
Mate 30 Pro (Kirin 990 4G) | N7 | 16.50 | 3.96 | 4.16 |
Galaxy S21U (Snapdragon 888) βοΈ Peak | 5LPE | 29.82 | 8.10 | 3.68 |
Galaxy S21U (Exynos 2100) βοΈ Peak | 5LPE | 28.04 | 7.69 | 3.64 |
Galaxy S20+ (Exynos 990) | 7LPP | 20.20 | 5.02 | 3.59 |
Galaxy S10+ (Snapdragon 855) | N7 | 16.17 | 4.69 | 3.44 |
Galaxy S10+ (Exynos 9820) | 8LPP | 15.59 | 4.80 | 3.24 |
Looking at power and power efficiency between the two generations, we’re seeing that the new chip roughly uses +80% power versus the S765G, and because the performance increase is larger than the power increase, there’s a perf/W increase in efficiency as well.
GFXBench Aztec Normal Offscreen Power Efficiency (System Active Power) |
||||
Mfc. Process | FPS | Avg. Power (W) |
Perf/W Efficiency |
|
iPhone 12 Pro (A14) π₯ Throttled | N5 | 77.44 | 3.88 | 19.95 |
iPhone 12 Pro (A14) βοΈ Peak | N5 | 102.24 | 5.53 | 18.48 |
iPhone 11 Pro (A13) π₯ Throttled | N7P | 73.27 | 4.07 | 18.00 |
Galaxy S21U (Snapdragon 888) π₯ Throttled | 5LPE | 51.81 | 2.93 | 17.67 |
Mate 40 Pro (Kirin 9000) π Power-Save | N5 | 53.49 | 3.10 | 17.25 |
iPhone 11 Pro (A13) βοΈ Peak | N7P | 91.62 | 6.08 | 15.06 |
iPhone XS (A12) π₯ Throttled | N7 | 55.70 | 3.88 | 14.35 |
Galaxy S20U (Snapdragon 865) | N7P | 54.09 | 3.91 | 13.75 |
iPhone XS (A12) βοΈPeak | N7 | 76.00 | 5.59 | 13.59 |
Reno3 5G (Dimensity 1000L) | N7 | 27.84 | 2.12 | 13.13 |
Mi 11 Lite 5G (Snapdragon 780G) | 5LPE | 49.61 | 3.80 | 13.05 |
Galaxy S21U (Exynos 2100) π₯ Throttled | 5LPE | 46.29 | 3.85 | 12.02 |
Mate 40 Pro (Kirin 9000) π₯ Throttled | N5 | 63.56 | 5.37 | 11.84 |
Mi 10 Lite 5G (Snapdragon 765G) | 7LPP | 23.48 | 2.07 | 11.70 |
ROG Phone III (Snapdragon 865+) | N7P | 58.77 | 5.34 | 11.00 |
Mate 40 Pro (Kirin 9000) βοΈ Peak | N5 | 82.74 | 7.95 | 10.40 |
Mate 30 Pro (Kirin 990 4G) | N7 | 41.68 | 4.01 | 10.39 |
Galaxy S20+ (Exynos 990) | 7LPP | 49.41 | 4.87 | 10.14 |
Galaxy S10+ (Snapdragon 855) | N7 | 40.63 | 4.14 | 9.81 |
Galaxy S21U (Snapdragon 888) βοΈ Peak | 5LPE | 81.77 | 8.40 | 9.73 |
Galaxy S20U (Exynos 2100) βοΈ Peak | 5LPE | 71.53 | 8.10 | 8.83 |
Galaxy S10+ (Exynos 9820) | 8LPP | 40.18 | 4.62 | 8.69 |
The figures in Aztec Normal showcase a similar story: Almost double the peak performance at roughly +80% power.
The generational comparisons here are interesting, and indeed the Adreno 642 of the Snapdragon 780 here lands ahead of the Adreno 640 of the Snapdragon 855, both in performance, as well as lower power, for a much higher efficiency. It’s actually comparable to the Snapdragon 865 from last year.
GFXBench Manhattan 3.1 Offscreen Power Efficiency (System Active Power) |
||||
Mfc. Process | FPS | Avg. Power (W) |
Perf/W Efficiency |
|
iPhone 12 Pro (A14) π₯ Throttled | N5 | 103.11 | 3.90 | 26.43 |
Galaxy S21U (Snapdragon 888) π₯ Throttled | 5LPE | 75.62 | 2.91 | 25.98 |
iPhone 12 Pro (A14) βοΈ Peak | N5 | 137.72 | 5.63 | 24.46 |
iPhone 11 Pro (A13) π₯ Throttled | N7P | 100.58 | 4.21 | 23.89 |
Mate 40 Pro (Kirin 9000) π Power-Save | N5 | 95.01 | 4.35 | 21.83 |
Galaxy S20U (Snapdragon 865) | N7P | 88.93 | 4.20 | 21.15 |
iPhone 11 Pro (A13) βοΈPeak | N7P | 123.54 | 6.04 | 20.45 |
iPhone XS (A12) π₯ Throttled | N7 | 76.51 | 3.79 | 20.18 |
Reno3 5G (Dimensity 1000L) | N7 | 55.48 | 2.98 | 18.61 |
Galaxy S21U (Exynos 2100) π₯ Throttled | 5LPE | 72.66 | 4.04 | 17.98 |
Mate 40 Pro (Kirin 9000) π₯ Throttled | N5 | 87.31 | 4.98 | 17.54 |
iPhone XS (A12) βοΈPeak | N7 | 103.83 | 5.98 | 17.36 |
Mi 11 Lite 5G (Snapdragon 780G) | 5LPE | 65.57 | 3.78 | 17.34 |
Mi 10 Lite 5G (Snapdragon 765G) | 7LPP | 36.98 | 2.61 | 17.28 |
ROG Phone III (Snapdragon 865+) | N7P | 93.58 | 5.56 | 16.82 |
Galaxy S20U (Exynos 2100) βοΈ Peak | 5LPE | 115.20 | 7.62 | 15.11 |
Mate 40 Pro (Kirin 9000) βοΈPeak | N5 | 124.69 | 8.28 | 15.05 |
Mate 30 Pro (Kirin 990 4G) | N7 | 75.69 | 5.04 | 15.01 |
Galaxy S20+ (Exynos 990) | 7LPP | 85.66 | 5.90 | 14.51 |
Galaxy S10+ (Snapdragon 855) | N7 | 70.67 | 4.88 | 14.46 |
Galaxy S21U (Snapdragon 888) βοΈ Peak | 5LPE | 120.32 | 8.34 | 14.42 |
Galaxy S10+ (Exynos 9820) | 8LPP | 68.87 | 5.10 | 13.48 |
Galaxy S9+ (Snapdragon 845) | 10LPP | 61.16 | 5.01 | 11.99 |
Mate 20 Pro (Kirin 980) | N7 | 54.54 | 4.57 | 11.93 |
Galaxy S9 (Exynos 9810) | 10LPP | 46.04 | 4.08 | 11.28 |
Galaxy S8 (Snapdragon 835) | 10LPE | 38.90 | 3.79 | 10.26 |
Galaxy S8 (Exynos 8895) | 10LPE | 42.49 | 7.35 | 5.78 |
The Manhattan 3.1 figures are both equivalent in the performance and power increases at around +77% - with efficiency roughly staying flat because of that. It’s to be noted that this is still extremely impressive as usually performance increases of such a degree come with efficiency hits – however the new GPU and process node are doing their job to keep the efficiency high.
GFXBench T-Rex Offscreen Power Efficiency (System Active Power) |
||||
Mfc. Process | FPS | Avg. Power (W) |
Perf/W Efficiency |
|
iPhone 12 Pro (A14) π₯ Throttled | N5 | 260.28 | 4.08 | 63.97 |
Galaxy S21U (Snapdragon 888) π₯ Throttled | 5LPE | 172.67 | 2.70 | 63.74 |
iPhone 11 Pro (A13) π₯ Throttled | N7P | 289.03 | 4.78 | 60.46 |
iPhone 12 Pro (A14) βοΈ Peak | N5 | 328.50 | 5.55 | 59.18 |
iPhone 11 Pro (A13) βοΈ Peak | N7P | 328.90 | 5.93 | 55.46 |
Galaxy S20U (Snapdragon 865) | N7P | 205.37 | 3.83 | 53.30 |
Mate 40 Pro (Kirin 9000) π₯ Throttled | N5 | 147.13 | 2.92 | 50.38 |
iPhone XS (A12) π₯ Throttled | N7 | 197.80 | 3.95 | 50.07 |
Mate 40 Pro (Kirin 9000) π Power-Save | N5 | 201.85 | 4.10 | 49.22 |
ROG Phone III (Snapdragon 865+) | N7P | 224.48 | 4.92 | 45.60 |
Mi 10 Lite 5G (Snapdragon 765G) | 7LPP | 95.17 | 2.58 | 45.09 |
Mi 11 Lite 5G (Snapdragon 780G) | 5LPE | 151.91 | 3.37 | 44.98 |
iPhone XS (A12) βοΈPeak | N7 | 271.86 | 6.10 | 44.56 |
Galaxy 10+ (Snapdragon 855) | N7 | 167.16 | 4.10 | 40.70 |
Galaxy S21U (Exynos 2100) π₯ Throttled | 5LPE | 153.28 | 3.80 | 40.30 |
Reno3 5G (Dimensity 1000L) | N7 | 139.30 | 3.57 | 39.01 |
Mate 40 Pro (Kirin 9000) βοΈ Peak | N5 | 235.04 | 6.11 | 38.46 |
Galaxy S20+ (Exynos 990) | 7LPP | 199.61 | 5.63 | 35.45 |
Mate 30 Pro (Kirin 990 4G) | N7 | 152.27 | 4.34 | 35.08 |
Galaxy S21U (Snapdragon 888) βοΈ Peak | 5LPE | 279.39 | 7.98 | 35.01 |
Galaxy S9+ (Snapdragon 845) | 10LPP | 150.40 | 4.42 | 34.00 |
Galaxy 10+ (Exynos 9820) | 8LPP | 166.00 | 4.96 | 33.40 |
Galaxy S9 (Exynos 9810) | 10LPP | 141.91 | 4.34 | 32.67 |
Galaxy S8 (Snapdragon 835) | 10LPE | 108.20 | 3.45 | 31.31 |
Galaxy S20U (Exynos 2100) βοΈ Peak | 5LPE | 237.71 | 8.02 | 29.61 |
Mate 20 Pro (Kirin 980) | N7 | 135.75 | 4.64 | 29.25 |
Galaxy S8 (Exynos 8895) | 10LPE | 121.00 | 5.86 | 20.65 |
Finally, in T-Rex, the performance has also gone up by +60%. This is a workload that we expect to be more fill-rate bound by now, but the fact that the new SoC is able to manage such a larger boost even though it retains the exact same LPDDR4X-2133 2x16b memory interface is again quite impressive – it’s actually faster and more efficient than the flagship Exynos 2100 when it has to throttle down after prolonged periods.
Extremely Large Performance Boosts
In general, the GPU of the Snapdragon 780 is extremely impressive due to the fact how large a boost it makes in comparison to the Snapdragon 765G last year. I admit, we were never really too convinced about the S765G performance as it really only matched quite old flagship SoCs such as the Snapdragon 835.
The new Snapdragon 780 essentially jumps a few generations, and you can essentially expect to outperform Snapdragon 855 flagships, falling short of the Snapdragon 865 devices of last year.
Qualcomm was able to achieve this through a big increase in the power usage of the GPU. The Snapdragon 765 was extremely conservative with tiny power envelopes of around 2-2.6W. The Snapdragon 780G now is falling around 3.5-3.8W. This is still perfectly reasonable and well within the power envelope capacity of most devices without throttling. On the Mi 11 Lite 5G I saw some minor throttling, but I’m not sure the device is all that thermally optimised given its thinner form-factor.
First Impressions & Conclusion
2020 had been a quite successful year for the Snapdragon 765 and it had seen usage in a lot of devices, some of which we reviewed or covered:
- The Google Pixel 5: A Mini-Review - Small Package, Small Value?
- OPPO's Reno3 5G vs Reno3 Pro vs Reno3 Pro 5G: Why Don't We See More MediaTek Dimensity 1000 Phones?
- The LG V60 and VELVET Review: A Classic & A Design Restart
The commonality between all of them was that while they promised the same features as the flagship counterparts that year, with a specific focus on 5G connectivity, the performance reminded us more of flagship devices of several generations ago. Sometimes I would wonder if it’s not actually better to purchase an older flagship device as in many cases the experience would have possibly ended up being better.
The new Snapdragon 780 goes a long way to resolve such performance discrepancies and bridging the gap to the current flagship generation. The performance uplifts to the Snapdragon 765G are quite massive, both in CPU and in particular GPU, resulting in a new end-user experience that is much different and much improved compared to the last generation.
Qualcomm also managed to achieve the performance uplifts all while improving energy efficiency, which is sometimes a hard balance to achieve when making big leaps like these.
We’ll be focusing on a review of the Mi 11 Lite 5G at a later date, but the device is a perfect example of what can be achieved with a Snapdragon 780G. At 369€ MSRP, the phone’s performance positions itself extremely competitively in terms of value proposition against flagship SoC devices, besides it being a breath of fresh air in terms of device design and form-factor.
In general, the Snapdragon 780 is a big leap forward for this device segment, much improving the device experiences you can expect from a “high mid-range” or “premium” range device, and should prove itself a solid foundation for many other 2021 devices.
Related Reading:
- Qualcomm Announces Snapdragon 780G: New 5nm 765 Successor
- Qualcomm Announces Snapdragon 768G: Higher-bin 765 up to 2.8GHz
- Qualcomm Announces Snapdragon 865 and 765(G): 5G For All in 2020, All The Details
- The Snapdragon 888 vs The Exynos 2100: Cortex-X1 & 5nm - Who Does It Better?
- Qualcomm Details The Snapdragon 888: 3rd Gen 5G & Cortex-X1 on 5nm