Intel Core i5-680 vs Intel Pentium Silver J5005 Comparison

Two Intel processors separated by roughly seven years sit side by side in the database with nearly identical aggregate scores, and the recorded data tells a story of how CPU design changed over that span. The Pentium Silver J5005, a mobile chip from the Gemini Lake family, and the Core i5-680, a desktop Clarkdale part, land within a fraction of a point of each other on average benchmark score: 776 versus 772. Yet the head-to-head record is a clean sweep, five wins to zero, all for the J5005. How can two chips rank as equals overall while one dominates their direct matchups? The answer lies in what each competitor pool looks like, and it is worth unpacking.

Where Each One Wins

Strictly from the head-to-head data, the J5005 wins everywhere. It leads the Core i5-680 by 18.6 percent in Cinebench R15 multi-core, 18.6 percent in R20 multi-core, 18.9 percent in R20 single-core, 18.6 percent in R23 multi-core, and 18.7 percent in R23 single-core. The pattern is striking in its consistency: whether the workload stresses one core or all of them, the delta hovers just under nineteen percent in the newer chip's favor.

That uniformity raises a question. Usually a four-core processor without SMT beating a two-core, four-thread part would show a much larger multi-core gap than single-core gap, since parallel workloads should scale with core count. Here they do not. The recorded data implies the J5005's per-core throughput advantage, driven by newer architecture, roughly accounts for the entire lead, and its four physical cores without hyperthreading merely hold that lead steady against the i5-680's two cores with four threads. In other words, SMT on the older chip appears to compensate almost exactly for having half the physical cores.

So where does the i5-680 win? On nothing in the recorded head-to-head suite. Its case rests on platform factors instead: it is a desktop socketed part on Socket 1156 with sixteen Gen 2 PCIe lanes, versus the J5005's soldered BGA 1090 package with only six lanes. For anyone needing a discrete GPU or add-in cards, the data hints the i5-680 platform offers more expansion headroom even though the CPU itself loses every benchmark.

Architecture Differences

The architectural gulf between these two is wide. The J5005 uses Intel's Goldmont Plus architecture, codenamed Gemini Lake, built on a 14 nm process at Intel's own foundry. The i5-680 uses Westmere architecture, codenamed Clarkdale, on a much older 32 nm process, also from Intel. The database shows no transistor count for the J5005, but the i5-680 packs 382 million transistors onto an 81 mm² die, while the J5005's die measures 93 mm².

Core configurations differ fundamentally. The J5005 has four cores and four threads, no hyperthreading. The i5-680 has two cores and four threads, meaning each physical core handles two threads. Clock behavior points in opposite directions: the i5-680 runs a high base clock of 3.60 GHz boosting to 3.87 GHz, while the J5005 starts at 1500 MHz and boosts to 2.80 GHz. Despite clocking far lower, the J5005 still wins every test, which speaks to the per-clock efficiency gains across those process generations.

Cache layouts diverge too. The J5005 carries 56 KB of L1 per core and a shared 4 MB L2, with no L3 at all. The i5-680 has 64 KB of L1 per core, 256 KB of L2 per core, and a shared 4 MB L3. Both end up with a 4 MB shared pool, but the J5005's sits at L2 while the older chip places it at L3 with a small per-core L2 in front, a classic older-generation hierarchy.

Power draw is perhaps the sharpest contrast: the J5005 is rated at 10 watts TDP, the i5-680 at 73 watts. Neither chip has an unlocked multiplier. Both carry integrated graphics, UHD Graphics 605 on the J5005 and Intel HD on the i5-680. Both are end-of-life products; neither supports ECC memory.

Head-to-Head Benchmarks

Walking through the recorded results, the margins are remarkably stable. In Cinebench R15 multi-core, the J5005 scores 268 against 226, an 18.6 percent win. R20 multi-core reads 1117 versus 942, again 18.6 percent. R23 multi-core is 2660 against 2243, 18.6 percent once more.

Single-core follows suit. R20 single-core: 157 for the J5005, 132 for the i5-680, a 18.9 percent gap. R23 single-core: 375 versus 316, 18.7 percent. The i5-680 has no R15 single-core result recorded, so the single-core comparison rests on two tests.

What do these absolute numbers mean in context? Both chips sit at the 21st percentile versus all CPUs in the database, so neither is competitive with modern performance tiers. The J5005's nearest rivals by average score are the AMD Athlon X4 760K at 777 (within 0.1 percent), the Intel Core i3-3250T at 775, the Celeron G5920 at 775, and the Pentium G4520 at 778, all effectively tied. The i5-680's rival list includes the Core i7-3687U at 773, the Celeron G5920, the Core i3-3250T, and the J5005 itself at a 0.5 percent delta. The two comparison sets overlap almost completely, which explains the near-identical aggregate scores even as the J5005 wins every direct matchup: they beat the same opponents by slightly different margins.

Specification Differences

The fields where these two diverge, per the database:

• Cores/threads: 4C/4T (J5005) versus 2C/4T (i5-680)
• Base clock: 1500 MHz versus 3.60 GHz
• Boost clock: 2.80 GHz versus 3.87 GHz
• TDP: 10 W versus 73 W
• Socket: Intel BGA 1090 (soldered, mobile) versus Intel Socket 1156 (desktop)
• Architecture: Goldmont Plus (Gemini Lake) versus Westmere (Clarkdale)
• Process node: 14 nm versus 32 nm
• Die size: 93 mm² versus 81 mm², with 382 million transistors recorded only for the i5-680
• Cache: 56 KB L1 per core with 4 MB shared L2 and no L3, versus 64 KB L1 per core, 256 KB L2 per core, and 4 MB shared L3
• Memory: DDR4 at 38.4 GB/s on dual-channel versus DDR3 at 21.3 GB/s on dual-channel
• PCIe: Gen 2 with 6 CPU lanes versus Gen 2 with 16 CPU lanes
• Integrated graphics: UHD Graphics 605 versus Intel HD
• Market segment: Mobile versus Desktop
• Release dates: 2017-12-10 versus 2010-04-18
• Launch MSRP: $161 for the J5005, $294 for the i5-680
• Part numbers: SR3S3 versus SLBTM

Shared traits include the Intel foundry, dual-channel memory buses, no ECC support, locked multipliers, and end-of-life status.

FAQ

Q: Which CPU is faster in benchmarks?

A: The Pentium Silver J5005 wins all five recorded head-to-head tests, by margins of 18.6 to 18.9 percent, covering both single-core and multi-core Cinebench R15, R20, and R23 workloads.

Q: How can their average scores be so close if one wins every matchup?

A: Both average within a point (776 versus 772) and both sit at the 21st percentile, because each is rated against the whole database. Their rival lists overlap heavily, including the Celeron G5920 and Core i3-3250T on both, so aggregate placement converges even though direct results favor the J5005.

Q: Does the i5-680's hyperthreading close the multi-core gap?

A: The data suggests yes, largely. Its four threads on two cores keep multi-core deficits at 18.6 percent, essentially identical to its 18.7 to 18.9 percent single-core deficits, implying thread count offsets the core-count disadvantage almost exactly.

Q: Which uses less power?

A: The J5005, by a wide margin: 10 watts TDP against 73 watts for the i5-680.

Q: Can either chip be overclocked?

A: No. Both have locked multipliers according to the recorded data.

Q: Which platform has more PCIe connectivity?

A: The i5-680, with sixteen Gen 2 CPU lanes versus six Gen 2 CPU lanes on the J5005.

The Verdict

Judged purely on the recorded data, the J5005 is the stronger processor. It wins every benchmark, offers newer DDR4 memory with substantially higher bandwidth (38.4 GB/s versus 21.3 GB/s), runs on a far newer process, and does all of this at a fraction of the power draw. The i5-680 counters only on platform grounds: a socketed desktop design, more PCIe lanes for expansion, and higher raw clock speeds that nonetheless fail to translate into wins.

Neither is a performance pick by modern standards, both occupying the 21st percentile and both end-of-life. But for the matchup itself, the data is unambiguous: the low-power mobile chip from 2017 outperforms the desktop chip from 2010 across the board, and the question the numbers provoke is less about which to choose than about how much architecture and process evolution can achieve even at a tenth of the power budget.