RTX Spark With 20-Core CPU Closes Multi-Core Performance Gap With 14-Core M4 Max; New Leak Shows NVIDIA's ARM-Based SoC Is 11% Slower

Wccftech's Omar Sohail reports that fresh Geekbench 7 leaks show NVIDIA's RTX Spark 20-core CPU closing the multi-core performance gap with Apple's 14-core M4 Max — though the ARM-based SoC remains 11 percent slower in multi-core and trails by a notable 24.5 percent in single-core. Two RTX Spark laptops with 64 GB of unified memory were spotted in the Geekbench 7 database, both operating under the "Balanced" power profile: the 20-core model (identified as "OEMQAJ 766_MIS Product Name DV") scored 2,570 in single-core and 23,126 in multi-core, while the 18-core version ("OEMQAJ OEMQAJ Product") scored 2,541 and 21,776 respectively.

In comparison, Apple's 14-core M4 Max posted 3,404 in single-core and 25,957 in multi-core — meaning the flagship RTX Spark is 11 percent slower than the M4 Max in multi-threaded work, with the gap widening to 24.5 percent in single-core performance. Between the two RTX Spark variants, the 20-core CPU is 1.14 percent faster in single-core and 6.2 percent faster in multi-core than the 18-core chip. Wccftech notes this marks an improvement over earlier pre-announcement N1X engineering-sample scores on Geekbench 6, which failed to beat Apple's 2023-era M3 Max. The report also reiterates that per NVIDIA's developer driver for Windows 11 ARM64, the 18-core version's GPU packs 5,120 CUDA cores while the top-end model has 6,144 — the same count as a desktop RTX 5070 or RTX 3080, though there is no confirmation yet that graphics performance will match those cards.

For Canadian readers evaluating RTX Spark ahead of its Fall 2026 launch, the Geekbench 7 results add a second independent data point to the Cinebench 2026 leaks from mid-July, painting a more nuanced picture of the platform's CPU. The multi-core result closing to within 11 percent of the M4 Max suggests NVIDIA's 20-core Grace CPU is competitive for productivity workloads, while the larger single-core deficit indicates the platform will lean on its Blackwell GPU and unified memory — its strongest assets — rather than raw CPU throughput. Canadian buyers comparing RTX Spark systems against Apple Silicon laptops should weigh these engineering-sample numbers (recorded under the Balanced power profile, with unfinished drivers) against the expectation that final retail units with mature firmware will deliver improved performance.