MediaTek’s Dimensity 9600M And Dimensity 9600 Pro Go Official; New “2 + 3 + 3” CPU Cluster With LPDDR6 RAM Support, Capable Of Handling 30B Local AI Models & More - Wccftech





As for the NPU, the Dimensity 9600 Pro features a dual AI Engine called NPU 1090 paired with a second-generation Ultra-Efficient NPU, plus Generative AI Engine 3.0 and Agentic AI Engine 3.0. With all of these components combined, MediaTek claims that the Dimensity 9600 Pro delivers a 21 percent larger L2 cache, bringing a 17 percent uplift in single-core performance alongside a 15 percent boost in multi-core throughput. Like the Dimensity 9600 Pro, the Dimensity 9600M is mass-produced on the same 2nm process, but there are several differences between the two SoCs.
Apple’s A20 Pro isn’t going to be the only SoC treated to TSMC’s 2nm process because MediaTek didn’t take long to announce its latest and greatest silicon offerings, the Dimensity 9600M and Dimensity 9600 Pro. As expected, the Taiwanese firm has adopted a similar approach to Qualcomm’s, bringing a multitude of upgrades to the “Pro” version, while also ensuring that the regular SoC isn’t a slouch either. Here are all the details.
It’s important that we talk about the silicon that will take on Qualcomm’s upcoming Snapdragon 8 Elite Gen 6 Pro, and with the new “2 + 3 + 3” cluster, we should expect major single-core and multi-core performance. You get two Arm C2-Ultra cores running at 4.55GHz, three Arm C2-Pro tweaked at 4.35GHz, and the remaining three Arm C2-Pro cores operating at 3.10GHz, making it an All Big Core architecture.
MediaTek has also incorporated 34.5MB of L2, L3, and SLC cache, with LPDDR6 RAM and UFS 5.0 support. As for the GPU, the Dimensity 9600 Pro sports an Arm Mali G2-Ultra NX, which adopts the company’s MediaTek Neural Network Rendering Architecture, 3rd-Gen Ray Tracing Engine, Dimensity OMM Ray Tracing Acceleration, Frame Rate Converter 4.0, and Ray Tracing Pipeline Technology. Before we forget, the GPU brings 185FPS gaming to the table.
As for the NPU, the Dimensity 9600 Pro features a dual AI Engine called NPU 1090 paired with a second-generation Ultra-Efficient NPU, plus Generative AI Engine 3.0 and Agentic AI Engine 3.0. With these improvements, MediaTek claims that the Dimensity 9600 Pro can handle up to 30B local AI MoE (Mixture-of-Experts) models while running completely on-device, with double the INT4 compute performance boost.
For the camera, we get an Imagiq 1290, bringing 4K 240FPS ultra-fast slow-motion video recording, CS-ISP 4K 60FPS movie mode, and more. Other additions, like Panel OverDrive multi-frame compensation and a 38 percent wider range than DCI-P3, have been added. As for wireless connectivity, we get sub-6GHz and mmWave 5G support, with Tri-SIM and Tri-Standby support, along with Bluetooth 6.2.
With all of these components combined, MediaTek claims that the Dimensity 9600 Pro delivers a 21 percent larger L2 cache, bringing a 17 percent uplift in single-core performance alongside a 15 percent boost in multi-core throughput. These speed gains are paired with notable power reductions, achieving 37 percent higher single-core efficiency and up to 61% lower multi-core peak power consumption.
With AI being all the rage these days, generative performance gets a similar bump, with local AI model launches being 27 percent faster and increasing LLM token generation speeds by 40 percent. These processing upgrades are further supported by next-generation memory and storage interfaces, with LPDDR6 memory bringing 33 percent higher effective bandwidth and UFS 5.0 storage doubling the read and write speeds of standard two-channel UFS 4.0.
The chipset’s dedicated NPU 1090 accelerates LLM prefill performance by 51 percent and achieves a 55 percent increase in token generation efficiency-per-watt. Complementing the high-performance unit, the low-power Super Efficient NPU 2.0 operates with 40 percent lower energy draw to handle continuous background sensing and contextual tasks.
Graphics processing receives an upgrade through MediaTek’s Neural Graphics Architecture, boosting peak GPU performance by 27 percent and peak power efficiency by 24 percent while accelerating hardware ray tracing capabilities by up to 18 percent. Fortunately, we’ve only wrapped the Dimensity 9600 Pro’s specifications, but expect a lot of MediaTek partners to adopt the Dimensity 9600M in devices later this year and in 2027.
Like the Dimensity 9600 Pro, the Dimensity 9600M is mass-produced on the same 2nm process, but there are several differences between the two SoCs. For instance, the Dimensity 9600M doesn’t support the new CPU cluster and adopts a “1 + 3 + 4” configuration in an octa-core layout.
You get one Arm C1-Ultra core with 2MB of L2 cache, three Arm C1-Premium cores with 1MB of L2 cache each, and four Arm C1-Pro cores with 512KB of L2 cache each, backed by a 16MB L3 cache and a 10MB System Level Cache (SLC). The older LPDDR5X memory with up to 10,667MT/s speeds and 4-lane UFS 4.1 support is present.
For AI processing, there’s the MediaTek NPU 990, which supports both Generative and Agentic AI workloads. For graphics and media processing, the chip adds an Arm Mali-G1 Ultra MC12 GPU capable of driving WQHD+ displays at up to 180Hz, including support for tri-port MIPI tri-fold display configurations.
The SoC can also handle 8K60 10-bit decoding across HEVC, AVC, VP9, and AV1 formats, 8K30 10-bit encoding for HEVC and AVC, and 8K60 8-bit HEVC encoding. The ISP can handle single sensors up to 320MP alongside maximum video recording modes that include 8K at 60FPS, 4K at 120FPS with EIS (Electronic Image Stabilization), and a dedicated 4K60 cinematic mode.
Cellular connectivity is handled through a 3GPP Release 17 5G modem equipped with MediaTek 5G UltraSave technology, supporting sub-6GHz 5CC downlink aggregation at 350MHz with 4 x 4 MIMO delivering downlink speeds of up to 7.4Gbps. You also get Wi-Fi 7 support, Bluetooth 6.0, and the usual standards. The first wave of Dimensity 9600 Pro and Dimensity 9600M-powered smartphones is expected to arrive later this year, so let us keep our fingers crossed and see how big a leap we get compared to last year.
About the author: Omar Sohail is a reporter and analyst for Wccftech's mobile section, specializing in the technology and business of the mobile industry. His expertise lies in the intricate hardware supply chain, covering developments in semiconductor manufacturing, chip lithography, and camera sensor technology.
Follow Wccftech on Google to get more of our news coverage in your feeds.