After a product cycle in 2021 and 2022 that was a bit more interesting than Qualcomm might have liked, 2023 has been a much simpler year for the company that makes a lot of SoCs and cellular modems. The Snapdragon 8 Gen 2 was the first of the company’s Gen 2 parts to come out earlier this year. With the Snapdragon 7+ Gen 2, the company is getting ready to move to the next level of its product line. Aimed at what has become Qualcomm’s traditional “premium” market segment of $400 to $600, which focuses on flagship-level features with more modest performance and costs, Qualcomm wants to give the Snapdragon 7+ Gen 2 a big performance boost.
The Snapdragon 7 Gen 2 is meant to replace the Snapdragon 7 Gen 1 from last year. In general, this year’s version of the Snapdragon 7 focuses more on improving performance than adding new features. Last year’s Gen 1 part added support for mmWave and new CPU and GPU architectures, especially Armv9 architecture CPU cores. This year, however, there are only a few new features. In its place is what Qualcomm says is one of the biggest performance improvements they’ve ever made for the Snapdragon 7 family. This is made possible in large part by a welcome switch from Samsung’s troubled 4nm process to TSMC’s 4nm process. This is similar to the switch Qualcomm made last year for the well-received mid-cycle Snapdragon 8+ Gen 1 part.
Also new this year is that Qualcomm is dropping hints that this will not be the only Snapdragon 7 Gen 2 part we see this year. They did this by naming their first Gen 2 part the 7+ instead of the 7. In short, releasing it as a Snapdragon 7+ part gives Qualcomm room to release a regular Snapdragon 7 part later. Certainly, Qualcomm isn’t announcing such a part right now, but there’s not much reason to launch a 7+ first unless they had plans for something below it. Otherwise, they could have launched it as a 7 part, like the Snapdragon 7 Gen 1, which was always a one-chip stack.
In terms of how the CPU is set up, the Snapdragon 7+ Gen 2 has the same 1+3+4 CPU core setup as the last few generations of the Snapdragon 7 family. The big news here is that the top-performing Prime core is getting a big performance boost. This is because Qualcomm is switching from using a mid-core with a slightly higher clock speed to using a CPU architecture that is better all around.
So, Qualcomm is using one of Arm’s Cortex-X cores for the Prime core for the first time ever in a Snapdragon 7 part. Cortex-X2 is Arm’s design from the previous generation, so it won’t interfere with the Snapdragon 8 Gen 2 and its Cortex-X3 core. But compared to the A710 core used in the Prime core of the 7 Gen 1 and the mid-cores of the 7+ Gen 2, the Cortex-X2 is a big step up in both IPC and clockspeed. As a result, the Prime core’s top clockspeed has gone from 2.4GHz to 2.91GHz. This makes the IPC gains of the more complex core even more important.
Qualcomm says that the 7+ Gen 2 CPU is “up to” 50% faster than the 7 Gen 1. Almost all of this improvement comes from the new Prime core.
The trade-off is that since there is only one Cortex-X2 core, you can only get such a big performance boost when you only have one task going at once. The three mid-performance cores are still based on the Cortex-A710 architecture and are only 2% faster than they were before. So, the 7+ Gen 1 won’t see big improvements when doing a lot of tasks at once. The better power efficiency of TSMC’s 4nm process should help, but some of those gains have been used to make the power-hungry Cortex-X2 workable in terms of battery life.
On the other hand, the 7+ Gen 2 also has a faster Adreno GPU. As has been the case with Qualcomm’s integrated GPUs for a few generations now, the company isn’t giving it a product number or telling us much about its architecture, so we can’t say much about it. Based on the list of features, it doesn’t look like this uses the newer GPU architecture from the 8 Gen 2. Instead, it looks like Qualcomm used a bigger version of their existing GPU and probably gave it a good boost in clockspeed.
No matter what happens, GPU performance expectations for the new SoC are high. Qualcomm says that the new chip will be two times faster than the 7 Gen 1 platform, which was only 20% faster than its predecessor. Even though these aren’t flagship-class SoCs, Qualcomm still markets the Snapdragon 7 series as a good fit for gaming smartphones, especially in China. This is why it’s not too surprising that Qualcomm is putting so much effort into GPU performance.
All in all, Qualcomm says that the 7 Gen 2 uses 13% less power than the 7 Gen 1, at least for “extended daily use.” The switch to TSMC’s 4nm process should pay off in a big way, as shown by last year’s 8+ Gen 1 part, but it’s clear that Qualcomm has been putting a lot of those gains towards improving performance as a whole.
Feeding the dragon is an LPDDR5 memory controller with 32 bits (two 16-bit parts). Unlike the Snapdragon 8 Gen 2, the 7+ Gen 2 won’t be able to use faster LPDDR5X memory. This means that the Snapdragon 7 family will stay the same. In this case, this means that memory speeds up to LPDDR5-6400 are supported, which is equal to a memory bandwidth of 25.6GB/second. In contrast to the big improvements in CPU and GPU performance, Qualcomm’s cache and memory subsystem will be under a lot more pressure to keep all of the processing blocks fed.
Speaking of which, there have been big performance improvements in more than just the CPU and GPU blocks. The performance of Qualcomm’s Hexagon DSP/AI engine block has also been improved a lot, almost as much as the GPU. Qualcomm didn’t give many technical details about this, but in our briefing, we didn’t hear anything about INT4 or micro-tiling, which are two of the most important features of the next-generation Hexagon block on the 8 Gen 2. It seems likely that this is a much more powerful version of the Hexagon block used on the 7 Gen 1.
The triple 18-bit Spectra ISP from the Snapdragon 8 is, however, making its way down to the Snapdragon 7. The 18-bit unit on the 7+ Gen 2 will replace the 14-bit unit in earlier generations of the platform. It will allow for triple exposure computational HDR video capture as well as better low-light photography, which Qualcomm calls their Mega Low Light feature. As a result, the 7+ Gen 2 can capture at higher resolutions when using the “zero shutter lag” feature, and when combined with the updated GPU, it can now record 4K video at up to 60fps, which is twice as fast as the 7 Gen 1’s maximum of 30fps.
Last but not least, Qualcomm’s Snapdragon X62 integrated modem is used again to finish off the package. Like the SoC from last year, this one is a mmWave + Sub-6 Release 16 design with a maximum theoretical download speed of 4.4Gbps. The design for this year is different, though. It supports dual SIM dual active (DSDA), which is another first for the Snapdragon 7 platform. Both of 7+ Gen 2’s active radios support 5G and 4G communications. This means that users with two SIM cards can use either radio to connect to almost any network they want. This is another high-end feature that was only available on Qualcomm’s Snapdragon 8 platform up until now.
The 7+ Gen 2 uses a FastConnect 6900 radio system to connect to networks that are not cell phones. This is a small update to the older 6700 radio. It supports Bluetooth version 5.3 and boosts the peak bandwidth of the 2×2 stream Wi-Fi 6E radio to 3.6Gbps by adding support for dual-band simultaneous (DBS).
In the end, we can say that the Snapdragon 7+ Gen 2 will be on the market very soon. Qualcomm says that phones with the SoC will be available later this month. Redmi and Realme are two of the OEMs that are planning to release phones with the new chip.