Nvidia continues to improve its upcoming ARM-based processor, known as RTX Spark, and a recent leak from the Geekbench 7 database shows better multi-core performance. After its first Geekbench 6 results failed to beat Apple’s M3 Max, the new results put the SoC just 11% behind the 14-core M4 Max.
The processor, developed together with MediaTek, was tested in two laptop configurations with 64 GB of unified memory, both running under the “Balanced” power profile. The more powerful version, featuring a 20-core CPU, scored 2,570 points in single-core and 23,126 points in multi-core. Meanwhile, the 18-core CPU version recorded 2,541 and 21,776 points, respectively.
Nvidia RTX Spark Closes the Gap With Apple M4 Max in Geekbench 7
In a direct comparison (thanks Wccftech), Apple‘s M4 Max still has a clear advantage in single-core workloads, scoring 3,404 points, making it 24.5% faster than Nvidia’s 20-core option. However, with 25,957 points in multi-core, the gap with Nvidia’s processor has become smaller, making RTX Spark a very competitive option for multi-threaded workloads.
Beyond the CPU, one of the biggest attractions of this platform is its integrated graphics based on the Blackwell architecture. According to leaked development drivers for Windows 11 ARM64, the 20-core model will feature a GPU with 6,144 CUDA cores, while the 18-core version will have 5,120 CUDA cores. In terms of core count, these figures are comparable to desktop graphics cards such as the RTX 5070 or RTX 3080, although real-world tests will still be needed to see its final graphics and thermal performance.
With its launch getting closer, the RTX Spark SoC is positioning Nvidia as a major contender in the high-end ARM laptop market.
