Intel Core Ultra 5 338H vs NVIDIA RTX Spark (MediaTek AHJ11488B) Comparison

FAQ

Q: How does the Intel Core Ultra 5 338H compare to the NVIDIA RTX Spark in terms of benchmark availability?

A: The database contains 17 benchmark scores for the Intel Core Ultra 5 338H, while the NVIDIA RTX Spark has no recorded benchmark scores. The Intel part achieves an average benchmark score of 33989, placing it in the 84th percentile of all CPUs, while the NVIDIA part has a percentile ranking of 50 with an average score of 0.

Q: What are the core and thread counts for each processor?

A: The Intel Core Ultra 5 338H has 12 cores and 12 threads. The NVIDIA RTX Spark has 20 cores and 20 threads. This gives the NVIDIA part a 66.7% higher core count and thread count compared to the Intel part.

Q: How do the cache configurations differ between the two processors?

A: The Intel Core Ultra 5 338H features a shared L3 cache of 18 MB, with 192 KB of L1 per core and 2.5 MB of L2 per core. The NVIDIA RTX Spark has a shared L3 cache of 8 MB, with 128 KB of L1 per core and 2 MB of L2 per core. The Intel part has 10 MB more shared L3 cache.

Q: What is the process node and foundry for each processor?

A: Both processors are built on a 3 nm process node. The Intel Core Ultra 5 338H is manufactured by Intel, while the NVIDIA RTX Spark is manufactured by TSMC. The NVIDIA part has a reported die size of 208 mm², while the Intel part's die size is not recorded in the database.

Q: What are the memory specifications for each processor?

A: Both processors support LPDDR5X memory. The Intel Core Ultra 5 338H uses a dual-channel memory bus with a bandwidth of 136.5 GB/s. The NVIDIA RTX Spark uses a quad-channel memory bus with a bandwidth of 273.1 GB/s, which is exactly double the Intel part's bandwidth.

Q: What is the production status of each processor?

A: The Intel Core Ultra 5 338H is listed as active in production, with a release date of January 4, 2026. The NVIDIA RTX Spark is listed as unreleased, with a release date of December 31, 2025.

Architecture Differences

The two processors come from entirely different design philosophies. The Intel Core Ultra 5 338H belongs to the Core Ultra Series 3, built on the Panther Lake architecture with the Panther Lake-H generation designation. It uses Intel's own 3 nm process node and is designed for mobile platforms, fitting into the Intel BGA 2540 socket. The part number is SA4REQ9EW, and its integrated graphics are the Arc B370.

The NVIDIA RTX Spark, carrying the part number GSE1-675-A1, uses the N1X codename with a generation label of Spark (GB10). Its architecture field is not specified in the database, but it is manufactured by TSMC on a 3 nm process node, with a die size of 208 mm². Its integrated graphics are the GB20B "Blackwell". The socket field is not recorded for this part.

Core configuration differs significantly. The Intel part provides 12 cores and 12 threads, meaning no hyperthreading is present in the recorded data. The NVIDIA part provides 20 cores and 20 threads, again with a 1:1 core-to-thread ratio. This gives the NVIDIA part a 66.7% advantage in raw core count.

Clock speeds show a different picture. The Intel Core Ultra 5 338H has a base clock of 1.90 GHz and a boost clock of 4.70 GHz. The NVIDIA RTX Spark has a base clock of 1.70 GHz and a boost clock of 4.00 GHz. The Intel part boosts 0.70 GHz higher, while the NVIDIA part has a 0.20 GHz lower base clock.

Cache hierarchies are structured differently. The Intel part allocates 192 KB of L1 per core and 2.5 MB of L2 per core, with a shared 18 MB L3 cache. The NVIDIA part allocates 128 KB of L1 per core and 2 MB of L2 per core, with a shared 8 MB L3 cache. Despite having fewer cores, the Intel part's larger per-core L1 and L2 allocations and its larger shared L3 cache give it a cache capacity advantage.

Memory architecture also diverges. The Intel part uses a dual-channel LPDDR5X bus delivering 136.5 GB/s of bandwidth. The NVIDIA part uses a quad-channel LPDDR5X bus delivering 273.1 GB/s, exactly double the bandwidth. PCIe support differs as well: the Intel part provides Gen 5 with 4 lanes (CPU only), while the NVIDIA part lists PCIe as N/A.

Power characteristics are recorded differently. The Intel Core Ultra 5 338H has a TDP of 25, while the NVIDIA RTX Spark has a TDP of 38. Neither processor supports ECC memory, and neither has an unlocked multiplier. The Intel part is active in production, while the NVIDIA part is unreleased.

Head-to-Head Benchmarks

The database contains no direct head-to-head benchmark entries for these two processors. The winsA and winsB fields are both 0. However, the Intel Core Ultra 5 338H has a full set of recorded benchmark scores, while the NVIDIA RTX Spark has none. This absence of comparative data means any performance analysis must rely on the Intel part's absolute scores and its position relative to other CPUs in the database.

Looking at the Intel part's scores, the Cinebench R15 multi-core test records 2504 points, with single-core at 305 points. Cinebench R20 shows 10213 multi-core and 1441 single-core. Cinebench R23 records 16331 multi-core and 2044 single-core. These scores indicate a processor with substantial multi-threaded capability relative to its single-threaded performance.

PassMark results for the Intel part show a multi-thread score of 28717 and a single-thread score of 4180. The data compression test records 276539, data encryption records 21367, and extended instructions records 23906. Floating-point math scores 84067, integer math scores 64934, and find prime numbers scores 304. Random string sorting records 34082, and physics scores 2697.

The Intel part's average benchmark score of 33989 places it in the 84th percentile of all CPUs in the database. Its nearest rivals provide context for this performance. The Intel Core Ultra 7 165H has an average score of 34083, which is 0.3% higher than the 338H. The Intel Core i7-12800HX scores 33875, which is 0.3% lower. The Intel Xeon 6353P scores 33844, 0.4% lower, and the AMD EPYC 4244P scores 34220, 0.7% higher. This cluster of scores within a 1% range indicates that the 338H sits in a very tight competitive band around the 34000 average score mark.

Since the NVIDIA RTX Spark has no benchmark scores, the database provides no direct measurement of its performance. Its percentile ranking of 50 and average score of 0 reflect the absence of recorded data rather than actual performance characteristics. Any head-to-head comparison must therefore be limited to architectural and specification differences rather than measured benchmark outcomes.

The Verdict

The recorded data shows a clear asymmetry between these two processors. The Intel Core Ultra 5 338H is a measured, active product with 17 benchmark scores, an 84th percentile ranking, and a 33989 average score. The NVIDIA RTX Spark is an unreleased part with no benchmark scores, a 50th percentile ranking, and a 0 average score.

For users requiring validated performance data, the Intel Core Ultra 5 338H is the only option with measurable results. Its position among nearest rivals, all within 0.7% of its average score, indicates that it delivers performance comparable to established Intel and AMD parts in the 33800 to 34200 score range.

The NVIDIA RTX Spark offers architectural advantages on paper: 20 cores versus 12, quad-channel memory with 273.1 GB/s bandwidth versus dual-channel 136.5 GB/s, and a 208 mm² die from TSMC. However, without benchmark scores, the database cannot confirm whether these specifications translate into actual performance advantages.

The Intel part's higher boost clock of 4.70 GHz versus 4.00 GHz, larger shared L3 cache of 18 MB versus 8 MB, and lower TDP of 25 versus 38 suggest it may have advantages in single-threaded responsiveness and power efficiency. The NVIDIA part's higher core count and doubled memory bandwidth suggest potential advantages in heavily parallel workloads, but this remains unverified by recorded measurements.

Given the production status, the Intel Core Ultra 5 338H is the viable choice for current deployment. The NVIDIA RTX Spark's unreleased status and lack of benchmark data mean it cannot be recommended based on the database's recorded information.

Specification Differences

The following specifications differ between the Intel Core Ultra 5 338H and the NVIDIA RTX Spark:

• Cores: 12 (Intel) versus 20 (NVIDIA)
• Threads: 12 (Intel) versus 20 (NVIDIA)
• Base clock: 1.90 GHz (Intel) versus 1.70 GHz (NVIDIA)
• Boost clock: 4.70 GHz (Intel) versus 4.00 GHz (NVIDIA)
• TDP: 25 (Intel) versus 38 (NVIDIA)
• Socket: Intel BGA 2540 (Intel) versus not recorded (NVIDIA)
• Architecture: Panther Lake (Intel) versus not recorded (NVIDIA)
• Codename: Panther Lake (Intel) versus N1X (NVIDIA)
• Generation: Ultra 5 (Panther Lake-H) (Intel) versus Spark (GB10) (NVIDIA)
• Foundry: Intel (Intel) versus TSMC (NVIDIA)
• Die size: not recorded (Intel) versus 208 mm² (NVIDIA)
• L1 cache: 192 KB per core (Intel) versus 128 KB per core (NVIDIA)
• L2 cache: 2.5 MB per core (Intel) versus 2 MB per core (NVIDIA)
• L3 cache: 18 MB shared (Intel) versus 8 MB shared (NVIDIA)
• Memory bus: Dual-channel (Intel) versus Quad-channel (NVIDIA)
• Memory bandwidth: 136.5 GB/s (Intel) versus 273.1 GB/s (NVIDIA)
• PCIe: Gen 5, 4 lanes CPU only (Intel) versus N/A (NVIDIA)
• Integrated graphics: Arc B370 (Intel) versus GB20B "Blackwell" (NVIDIA)
• Production status: Active (Intel) versus unreleased (NVIDIA)
• Part number: SA4REQ9EW (Intel) versus GSE1-675-A1 (NVIDIA)

Both processors share the 3 nm process node, LPDDR5X memory support, lack of ECC memory support, and locked multipliers.

Where Each One Wins

Based strictly on the recorded data, the Intel Core Ultra 5 338H wins in every measured category because it is the only part with benchmark scores. Its 84th percentile ranking and 33989 average score provide a quantifiable baseline that the NVIDIA RTX Spark cannot match, as the latter has no scores at all.

The Intel part also wins on clock speed specifications. Its 4.70 GHz boost clock exceeds the NVIDIA part's 4.00 GHz by 0.70 GHz, and its 1.90 GHz base clock is 0.20 GHz higher. This suggests an advantage in single-threaded workloads that depend on raw clock frequency.

Cache capacity favors the Intel part as well. The 18 MB shared L3 cache is 10 MB larger than the NVIDIA part's 8 MB. The per-core L1 allocation of 192 KB versus 128 KB and per-core L2 allocation of 2.5 MB versus 2 MB further favor Intel. For workloads with high data locality, the Intel part provides more on-die storage.

Power efficiency appears to favor Intel. The 25 TDP is 13 lower than the NVIDIA part's 38 TDP. For mobile deployments where thermal and power budgets are constrained, the Intel part draws less power in the recorded specifications.

The NVIDIA RTX Spark wins on core count with 20 cores versus 12, a 66.7% advantage. Thread count follows the same ratio at 20 versus 12. For workloads that scale linearly with core count, the NVIDIA part has a theoretical advantage, though no benchmark data confirms this.

Memory bandwidth clearly favors the NVIDIA part. The quad-channel bus delivers 273.1 GB/s, exactly double the Intel part's 136.5 GB/s. Applications that saturate memory bandwidth, such as large data streaming or certain scientific workloads, could benefit from this specification.

The NVIDIA part also holds the process node advantage through TSMC's 3 nm fabrication, though both parts use 3 nm nodes. Its 208 mm² die size is recorded, while the Intel part's die size is not, so no direct comparison is possible.

The integrated graphics differ in naming but neither has benchmark scores for GPU tasks. The Intel Arc B370 and NVIDIA GB20B "Blackwell" remain specification entries without measured performance data.

For production readiness, the Intel part wins decisively. Its active status and release date of January 4, 2026 mean it is available in the market. The NVIDIA part's unreleased status and December 31, 2025 release date indicate it has not yet shipped, making any real-world deployment impossible at the time of this analysis.