Intel Celeron N5105

The Intel Celeron N5105 is a quad-core mobile processor from the Tremont architecture, also known as Jasper Lake, manufactured on Intel's 10 nm process. It occupies a specific niche in the benchmark database: its average benchmark score of 1174 places it at the 33rd percentile of all CPUs, indicating that it outperforms roughly one-third of the processors tracked. The data shows a chip designed for efficiency and basic computing tasks, with a die size of 63.8 mm² and a shared 4 MB L3 cache, positioning it as an entry-level part that still offers modern platform features. This analysis examines its platform compatibility, thermal profile, benchmark performance against direct rivals, and the workloads for which it is best suited, based strictly on the available factual data.

Platform and Compatibility

The N5105 uses the Intel BGA 1338 socket, which is a soldered, non-upgradeable platform. This means the processor is permanently attached to the motherboard, and the upgrade path is effectively non-existent; users must replace the entire board to change the CPU. The chip supports DDR4 and LPDDR4 memory in a dual-channel configuration, with a maximum memory bandwidth of 46.9 GB/s. This bandwidth figure is a fixed hardware characteristic, and the dual-channel support is important for integrated graphics performance, as the UHD Graphics 24EU shares this memory bandwidth.

PCIe support is limited to Gen 3 with 8 lanes available from the CPU. This is a modest allocation compared to desktop platforms, but it is sufficient for basic NVMe storage and standard peripherals. The platform does not support ECC memory, which rules out the chip for error-correcting workloads like NAS systems that require data integrity safeguards. The architecture is based on the Tremont core design, and the processor is part of the Jasper Lake generation, which is a mobile-focused lineup. The production status is listed as end-of-life, so new designs should not be based on this chip, but existing systems will continue to function with the given specifications.

The memory bus is dual-channel, and the 46.9 GB/s bandwidth is a theoretical maximum that is shared between the CPU cores and the integrated UHD Graphics 24EU. For a quad-core part, this bandwidth is adequate for typical office and light media tasks, but it may become a bottleneck for memory-intensive applications or when the iGPU is under load. The lack of ECC memory support and the fixed BGA socket make this a closed platform, appropriate for pre-built systems rather than DIY builders. The PCIe Gen 3, 8-lane configuration is a clear indicator of the chip's intended role: it can handle a single fast SSD and a few expansion cards, but not a multi-GPU setup or heavy I/O expansion.

Power and Thermals

The N5105 has a TDP of 10 watts, which classifies it as an ultra-low-power part. This thermal design point is remarkably low, and it directly implies a passive cooling solution or a very small, low-speed fan is sufficient. The data shows that no dedicated liquid cooler or large tower cooler is necessary; a simple heatsink integrated into a compact chassis is adequate. The 10-watt TDP also means that the chip generates minimal heat, which is a critical advantage for fanless designs, thin-and-light laptops, and embedded systems where space and noise are at a premium.

The low power draw also affects sustained performance. With a base clock of 2000 MHz and a boost clock of 2.90 GHz, the chip can reach its maximum frequency without hitting thermal limits in most enclosures. However, the boost clock is likely to be sustained only for short bursts, as the power delivery and cooling in ultra-low-power designs are often constrained. The 10 nm process node from Intel contributes to the efficiency, allowing the four Tremont cores to operate within this tight power envelope. For system integrators, the 10-watt TDP simplifies power supply requirements, as even a small 19V DC adapter can provide ample headroom. The thermal management strategy is straightforward: the chip is designed to run cool and quiet, with no complex thermal throttling behavior expected under normal conditions.

Benchmark Performance

Benchmark results place the N5105 squarely in a performance tier that is competitive with older desktop processors. In Cinebench R23, the multicore score is 3413 and the single-core score is 481. These numbers are the most recent and comprehensive in the data set. The average benchmark score across all tests is 1174, which is nearly identical to the Intel Core i5-3550, which scores 1173, a delta of 0.1% in favor of the N5105. The comparison to the Intel Core i5-4440 shows a delta of -0.2%, meaning the N5105 is 0.2% slower than that chip. These deltas are essentially negligible, indicating that the N5105 delivers performance on par with a mid-range desktop processor from several generations ago.

The Cinebench R20 multicore score of 1433 and single-core score of 202 show a similar pattern. The multicore score is competitive with the nearest rivals, but the single-core score is lower, which is a characteristic of the low-power Tremont architecture. The Cinebench R15 multicore score of 343 is the oldest benchmark in the set, and it confirms the general performance level. The percentile rank of 33 means that the N5105 outperforms 33% of all CPUs in the database, which is a low ranking but expected for a power-sipping mobile part. Compared to the AMD Opteron 6220, the N5105 is 0.5% faster, and it is 0.4% faster than the Intel Core i7-2960XM. These deltas are all under 1%, so the N5105 is essentially in a dead heat with these older, higher-power parts.

The data reveals that the N5105's multicore performance is its stronger suit relative to its single-core performance. The R23 multicore score of 3413 is more than seven times the single-core score of 481, which is a typical ratio for a quad-core without hyper-threading. This suggests that workloads that can utilize all four cores will see reasonable throughput, while single-threaded tasks will be limited by the lower per-core performance.

FAQ

Q: What is the socket type for the Intel Celeron N5105?

A: The N5105 uses the Intel BGA 1338 socket, which is a soldered, non-upgradeable platform.

Q: Does the N5105 support ECC memory?

A: No, the data indicates that ECC memory is not supported.

Q: What is the TDP of the N5105 and what cooling does it require?

A: The TDP is 10 watts, which implies that a passive cooler or a very small, low-speed fan is sufficient for thermal management.

Q: How does the N5105 compare to the Intel Core i5-3550 in average benchmark score?

A: The N5105 has an average score of 1174, which is 0.1% higher than the Core i5-3550's score of 1173.

Q: What is the memory bandwidth of the N5105?

A: The memory bandwidth is 46.9 GB/s, using dual-channel DDR4 or LPDDR4 memory.

Q: What is the production status of the N5105?

A: The production status is listed as end-of-life.

Who Should Consider It

The benchmark data suggests that the N5105 is suitable for users whose primary workload is basic office productivity, web browsing, and light media consumption. The multicore score of 3413 in Cinebench R23 is adequate for spreadsheet calculations, document editing, and video playback, but it is not designed for heavy content creation. For gaming, the integrated UHD Graphics 24EU is present, but the low single-core score of 481 in Cinebench R23 will limit performance in modern titles that rely heavily on single-thread performance. The chip is not a viable option for AAA gaming, but it can handle older or indie games at low settings.

For creation workloads like video editing or 3D rendering, the multicore performance is the limiting factor. The R23 multicore score of 3413 is comparable to the Intel Core i5-4440, which is a desktop part from 2013, so it is not suitable for professional rendering or complex simulations. The chip is better suited for embedded systems, thin clients, and basic home servers where low power consumption is more important than raw performance. The 10-watt TDP makes it ideal for always-on devices, and the PCIe Gen 3 lanes allow for a single fast NVMe drive. Users who need a silent, low-heat system for file serving, light Docker containers, or as a secondary office machine will find the N5105 adequate. Those who require high single-thread performance for responsive desktop use should look elsewhere, as the data shows the chip is only at the 33rd percentile globally.

Single-Thread vs Multi-Thread Behavior

The split between single-thread and multi-thread performance is stark and informative. In Cinebench R23, the single-core score is 481, while the multicore score is 3413, giving a ratio of roughly 7:1. This ratio is expected for a four-core, four-thread processor with no hyper-threading, but the absolute values reveal the chip's character. The single-core score of 481 is low, indicating that each Tremont core is not particularly fast on its own. This directly impacts everyday responsiveness, as most operating system tasks and application launches are single-threaded. The data suggests that the N5105 will feel less snappy than a chip with a higher single-core score, even if the multicore score is similar to older desktop parts.

In contrast, the multicore score of 3413 shows that the four cores can work together effectively to handle parallel workloads. The Cinebench R20 multicore score of 1433 and R15 score of 343 follow the same pattern. The performance delta against the nearest rivals is negligible in average score, but the single-core performance is where the N5105 lags behind. The Intel Core i5-3550 and i5-4440, which have similar average scores, likely have higher single-core scores due to their higher clock speeds and older but more powerful core architectures. The N5105 compensates with efficient multi-core scaling, but the low per-core performance means that heavily threaded tasks like batch image processing will see decent throughput, while interactive workloads will be less responsive. For real-world use, this means the chip is best suited for tasks that can be parallelized, such as encoding video in the background, while any foreground interaction with the system will be limited by the single-thread performance.