Intel Core Ultra 5 245 vs Intel Core Ultra X9 378H Comparison

Where Each One Wins

The benchmark split between these two processors is heavily lopsided, but the wins that matter for each side are telling. The Intel Core Ultra 5 245 takes 13 of the 17 recorded head-to-head tests, and it does so across nearly every major workload category. Its cleanest victories come in data compression, extended instructions, floating-point math, and random string sorting, where it leads by margins ranging from 2.2% to 10.1%. Those are the kinds of tasks that benefit from raw cache bandwidth and per-core efficiency, and the desktop part clearly has the edge there.

The Intel Core Ultra X9 378H, despite losing the overall count, wins the two most interesting individual tests. It takes PassMark physics by a massive 24.5% margin, scoring 3404 against 2569, which is the single largest delta in the entire comparison. It also edges out the Ultra 5 in integer math by 1.5%, posting 92603 versus 91187, and it wins both single-thread PassMark tests by 1.3% with a score of 4453 against 4394. That last result matters: the mobile chip is slightly faster on single-threaded PassMark workloads, even though the desktop chip wins every Cinebench single-core test.

Looking at the average benchmark scores, the Ultra 5 245 sits at 48995, which places it in the 90th percentile of all CPUs in the database. The Ultra X9 378H averages 47468, good for the 89th percentile. Those are close positions, but the desktop part holds a small overall lead. The nearest rivals for the Ultra 5 include the AMD Ryzen 7 PRO 5755G at 49196 (0.4% ahead) and the AMD Ryzen 9 7900 at 49228 (0.5% ahead), while the Intel Core i5-14600K trails by 0.8%. For the Ultra X9, the AMD Ryzen 9 PRO 5945 sits 0.1% ahead, the Intel Core Ultra 7 265T leads by 0.5%, and the Intel Core i7-13700KF and i9-12900F trail by 0.3% and 0.6% respectively. Both chips are firmly in the same performance class, just with different strengths.

Architecture Differences

The two processors come from different foundries and different generations. The Ultra 5 245 is built on TSMC's 3 nm process and belongs to the Arrow Lake family, specifically Arrow Lake-S, under the Core Ultra Series 2 banner. The Ultra X9 378H uses Intel's own 3 nm process, carries the Panther Lake codename, and belongs to the Core Ultra Series 3. That foundry split is notable: the desktop part relies on an external fab, while the mobile part is made in-house.

Cores and threads differ as well. The Ultra 5 has 14 cores and 14 threads, meaning no hyperthreading, while the Ultra X9 has 16 cores and 16 threads, also without hyperthreading. The desktop chip has the higher base clock at 3.50 GHz versus 2.00 GHz, and the higher boost clock at 5.10 GHz versus 5.00 GHz. The TDP figures are dramatically different: 65 watts for the Ultra 5, 25 watts for the Ultra X9. That is a 40-watt gap that explains why one is a desktop part and the other is a mobile part.

Cache layouts differ in the L2 and L3 tiers. Both use 192 KB of L1 per core, but the Ultra 5 has 3 MB of L2 per core versus 2.5 MB for the Ultra X9. The shared L3 is 24 MB on the desktop chip and 18 MB on the mobile chip. The die sizes and transistor counts are only listed for the Ultra 5: 17,800 million transistors on a 243 mm² die. The Ultra X9 has no recorded transistor or die size in the database.

Memory support is another differentiator. The Ultra 5 uses DDR5 on a dual-channel bus with 102.4 GB/s of bandwidth and supports ECC memory. The Ultra X9 uses LPDDR5X, also dual-channel, but with a higher listed bandwidth of 153.6 GB/s and no ECC support. PCIe lanes differ sharply: the Ultra 5 offers 20 Gen 5 lanes from the CPU, while the Ultra X9 provides only 4 Gen 5 lanes. Integrated graphics are also different, with the Ultra 5 carrying Arc Xe-LPG Graphics with 64 execution units, while the Ultra X9 has the newer Arc B390.

Head-to-Head Benchmarks

The biggest desktop win is in PassMark random string sorting, where the Ultra 5 scores 49140 against 44648, a 10.1% advantage. That is a memory-latency-sensitive workload, and the larger L3 cache plus higher boost clock likely explains the gap. The next largest desktop win is in extended instructions, where the Ultra 5 posts 33304 versus 31315, a 6.4% lead. Floating-point math follows at 5.3%, with scores of 120548 against 114500.

The Cinebench results are all close, with the Ultra 5 winning every test by roughly 1.1% to 1.3%. In Cinebench R23 multi-core, the score is 32924 versus 32553, a 1.1% lead. In single-core R23, it is 4648 versus 4595, a 1.2% lead. The R20 and R15 tests show the same pattern: the desktop chip wins by 1.1% to 1.3% across the board. Data compression goes to the Ultra 5 by 3.7%, with scores of 400942 against 386591. Data encryption is a 1.3% win, and finding prime numbers is a 2.2% win.

The mobile chip's wins are fewer but sharper. The physics test is the standout: 3404 versus 2569, a 24.5% margin that suggests a fundamentally different scheduling or power behavior under that specific workload. Integer math goes to the Ultra X9 by 1.5%, 92603 to 91187. The two single-thread PassMark tests both go to the Ultra X9 by 1.3%, with identical scores of 4453 against 4394. Notably, the Ultra X9 wins single-thread PassMark while losing every single-core Cinebench test, which indicates the two suites measure slightly different aspects of single-threaded performance.

Specification Differences

The socket is the most obvious difference: the Ultra 5 uses Intel Socket 1851, while the Ultra X9 uses Intel BGA 2540, which is a soldered mobile package. Process nodes are both 3 nm, but the foundries differ as described above. Clock speeds favor the desktop part by 1.50 GHz at base and 0.10 GHz at boost. TDP favors the mobile part by 40 watts, which is the entire point of the X9's existence.

Cache totals favor the desktop part: 3 MB versus 2.5 MB per core in L2, and 24 MB versus 18 MB in L3. Memory bandwidth favors the mobile part at 153.6 GB/s versus 102.4 GB/s, though the desktop part supports ECC and the mobile part does not. PCIe connectivity heavily favors the desktop part with 20 lanes versus 4 lanes. The integrated GPU is different, with the Ultra 5 having 64 execution units and the Ultra X9 having the Arc B390. The Ultra 5 has a recorded launch MSRP of $270, while the Ultra X9 has no listed MSRP. Release dates differ: the Ultra 5 launched in January 2025, the Ultra X9 in April 2026.

FAQ

Q: Which processor is faster overall?

A: The Intel Core Ultra 5 245 has a higher average benchmark score of 48995 versus 47468 for the Ultra X9 378H, and it wins 13 of the 17 head-to-head tests. It also sits in the 90th percentile of all CPUs versus the 89th percentile for the mobile part.

Q: Is the mobile chip ever faster?

A: Yes, in four tests. The Ultra X9 378H wins PassMark physics by 24.5%, integer math by 1.5%, and the two single-thread PassMark tests by 1.3% each.

Q: How do these chips compare to their nearest rivals?

A: The Ultra 5 245 sits within 0.8% of the AMD Ryzen 7 PRO 5755G, AMD Ryzen 9 7900, Intel Xeon Gold 5318H, and Intel Core i5-14600K. The Ultra X9 378H sits within 0.6% of the AMD Ryzen 9 PRO 5945, Intel Core i7-13700KF, Intel Core Ultra 7 265T, and Intel Core i9-12900F.

Q: Why does the mobile chip have a lower TDP?

A: The Ultra X9 378H is designed for mobile use with a 25-watt TDP, while the Ultra 5 245 is a desktop part rated at 65 watts. That 40-watt difference allows the mobile chip to fit in laptops, but it also means lower base and boost clocks.

Q: Do both chips support the same memory?

A: No. The Ultra 5 245 supports DDR5 with ECC on a dual-channel bus at 102.4 GB/s. The Ultra X9 378H supports LPDDR5X without ECC, also dual-channel, but at a higher 153.6 GB/s.

Q: Which chip has more cores?

A: The Ultra X9 378H has 16 cores and 16 threads, while the Ultra 5 245 has 14 cores and 14 threads. Neither supports hyperthreading.

The Verdict

The data points to a clear split by intended use case. If you are building a desktop system with room for a 65-watt part, the Intel Core Ultra 5 245 is the better choice. It wins the majority of benchmarks, has a larger L3 cache, more PCIe lanes, ECC memory support, and a higher boost clock. Its 13 wins include every Cinebench test and most PassMark workloads, making it the more consistent performer for general computing, content creation, and memory-intensive tasks.

If you are shopping for a laptop, the Intel Core Ultra X9 378H is the only one of the two that makes sense physically, given its BGA 2540 socket and 25-watt TDP. The fact that it nearly matches the desktop chip in average score, with a 47468 average against 48995, is remarkable for a part running at 2.00 GHz base. Its 24.5% physics win and slight integer math advantage show that it is not simply a scaled-down version, it has genuine strengths of its own.

The practical advice is straightforward: for a desktop build, take the Ultra 5 245. For a thin-and-light laptop that still delivers near-desktop multi-core performance, the Ultra X9 378H is the part to look for. Neither chip is a bad buy in its segment, and the benchmark data confirms that Intel has two very capable processors aimed at different form factors. The desktop part is the overall performance leader, but the mobile part wins the efficiency argument by a wide margin, and its single-thread PassMark win suggests it is no slouch even in lightly threaded tasks.