Intel Core i7-5820K vs Intel Core i7-8700T Comparison

The Intel Core i7-8700T and the Intel Core i7-5820K are two six-core, twelve-thread desktop CPUs that land within a hair of each other in the database rankings, but they arrive there by very different routes. The 8700T is a low-power Coffee Lake part that wins seven of eight head-to-head tests, while the 5820K is a quad-channel Haswell-E platform that still takes one multi-core win and offers an unlocked multiplier. Neither part is current: both are end-of-life, and both sit almost exactly at the median of all recorded CPUs, at the 50th and 49th percentiles respectively.

FAQ

Q: Which CPU is faster overall? A: The Core i7-8700T. It wins 7 of the 8 recorded head-to-head benchmarks, including every Cinebench test in R15, R20, and R23, and it posts the higher average benchmark score, 2701 versus 2642.

Q: Does the Core i7-5820K win anything? A: Yes. It takes the Geekbench multi-core test with 5464 points against 5168 for the 8700T, a 5.4 percent advantage that likely reflects its quad-channel memory subsystem delivering 68.3 GB/s of bandwidth versus 42.7 GB/s on the 8700T's dual-channel setup.

Q: How big is the single-core gap? A: It is the largest gap on the board. The 8700T scores 1343 in Geekbench single-core against 1174 for the 5820K, a 14.4 percent lead, and it leads by roughly 4 percent in every Cinebench single-core run as well.

Q: How do these chips compare against the broader CPU landscape? A: Both are mid-pack. The 8700T sits in the 50th percentile against all CPUs in the database, and the 5820K sits in the 49th. Their average scores are nearly identical to rivals like the Intel Core i7-1185G7E (2703 average) and the Intel Xeon E-2274G (2643 average).

Q: Which one is more power efficient? A: The 8700T, decisively. Its rated TDP is 35 W while the 5820K is rated at 140 W, a fourfold difference, and the 8700T still delivers the higher average benchmark score.

Q: Can either CPU be overclocked? A: Only the 5820K. Its multiplier is unlocked, while the 8700T's is locked, so the 5820K is the only one of the two with officially sanctioned tuning headroom.

The Verdict

The data points one direction for most buyers: the Core i7-8700T is the stronger all-around performer. It wins seven of eight tests, posts a 4 to 4.2 percent advantage across the entire Cinebench suite in both single-core and multi-core modes, and delivers its Geekbench single-core win by a commanding 14.4 percent. It does all of this at a 35 W TDP versus 140 W, making it the clear pick where efficiency matters.

The 5820K is not without its case. It wins Geekbench multi-core by 5.4 percent, it has 15 MB of shared L3 cache against 12 MB, it offers 28 PCIe 3.0 lanes against 16, and its quad-channel memory bus delivers 68.3 GB/s of bandwidth versus 42.7 GB/s. Its unlocked multiplier adds tuning flexibility the 8700T cannot match. For memory-bandwidth-sensitive or expansion-heavy workloads, those platform traits matter more than a 4 percent Cinebench deficit.

In pure benchmark terms, though, the verdict is straightforward: the 8700T leads in average score, wins the head-to-head series 7-1, and does it at a quarter of the rated power.

Head-to-Head Benchmarks

The 8700T's dominance is remarkably consistent across the Cinebench family. In R15 it scores 876 multi-core and 123 single-core points against 841 and 118 for the 5820K, wins of 4.2 percent in both cases. R20 repeats the pattern almost exactly: 3653 versus 3508 multi-core (4.1 percent) and 515 versus 495 single-core (4.0 percent). R23 continues it with 8698 versus 8354 multi-core (4.1 percent) and 1228 versus 1179 single-core (4.2 percent). Six tests, six wins, all clustered tightly around 4 percent, which suggests a genuine architectural advantage in clock-for-clock rendering throughput rather than a single outlier result.

Geekbench splits the picture. The 5820K's best moment is Geekbench multi-core, where it scores 5464 against 5168, a 5.4 percent win that stands out as the only red mark on the 8700T's record. The quad-channel memory controller is the plausible explanation, since Geekbench workloads tend to exercise memory more heavily than Cinebench.

The 8700T's best moment is Geekbench single-core: 1343 against 1174, a 14.4 percent gap that dwarfs every other result in the series. Its 4.00 GHz boost clock against the 5820K's 3.60 GHz, combined with the newer architecture, explains a lead of that size. The final tally is 7 wins for the 8700T and 1 for the 5820K.

Specification Differences

The two chips share a great deal: both are six-core, twelve-thread desktop CPUs from Intel with 64 KB of L1 and 256 KB of L2 cache per core, DDR4 memory support, no ECC, Gen 3 PCIe, and end-of-life production status. The differences are where the story lives.

The clocks differ sharply. The 5820K runs a 3.30 GHz base and 3.60 GHz boost; the 8700T runs a 2.40 GHz base but a 4.00 GHz boost. The base clock favors the 5820K, the boost behavior favors the 8700T, and the benchmark data shows the boost advantage wins out in practice.

TDP is the widest gap in the table: 35 W for the 8700T against 140 W for the 5820K. The sockets are incompatible, Socket 1151 versus Socket 2011-3, so neither drops into the other's board. The 5820K carries 15 MB of shared L3 cache versus 12 MB, 28 PCIe 3.0 lanes versus 16, a 356 mm² die versus 154 mm², and 2,600 million transistors, a figure not specified for the 8700T. The memory bus is quad-channel at 68.3 GB/s on the 5820K versus dual-channel at 42.7 GB/s on the 8700T.

Finally, the 8700T includes UHD Graphics 630 integrated graphics; the 5820K has none and requires a discrete GPU. The 5820K's multiplier is unlocked, the 8700T's is not. Release dates differ too, with the 5820K arriving in 2014 and the 8700T in 2018.

Architecture Differences

The generational gap underpins everything in the benchmark table. The 5820K is Haswell-E on Intel's 22 nm process; the 8700T is Coffee Lake on 14 nm. Four years and two process shrinks separate them, and the results show it: the 8700T's 4.00 GHz boost exceeds the 5820K's 3.60 GHz despite the 8700T operating inside a 35 W envelope that is a quarter the size of the 5820K's 140 W rating.

The 5820K counters with sheer silicon. Its 356 mm² die is well over twice the 8700T's 154 mm², and it carries 2,600 million transistors plus 15 MB of shared L3 cache against 12 MB. As an HEDT-class part on Socket 2011-3, it was built for platform bandwidth: quad-channel DDR4 and 28 PCIe 3.0 lanes, both of which exceed the mainstream-oriented 8700T's dual-channel bus and 16 lanes.

The cache hierarchy below L3 is identical between them, 64 KB L1 and 256 KB L2 per core, so the meaningful cache difference is the 3 MB of extra L3 on the 5820K. Feature-wise, the 8700T adds integrated UHD Graphics 630 while the 5820K ships with no iGPU at all. Neither supports ECC memory, and both are locked to DDR4.

Where Each One Wins

The 8700T owns responsiveness and efficiency. Its 14.4 percent Geekbench single-core lead and consistent 4 percent Cinebench single-core wins make it the better choice for lightly threaded applications, general desktop work, and any scenario where the 35 W TDP matters, such as compact or thermally constrained systems. Its integrated UHD Graphics 630 also lets it run without a discrete GPU, something the 5820K cannot do.

The 5820K owns bandwidth and expansion. Its lone benchmark win, a 5.4 percent edge in Geekbench multi-core, aligns with its 68.3 GB/s of quad-channel memory bandwidth, and its 28 PCIe lanes and 15 MB of L3 cache suit multi-GPU setups, storage-heavy configurations, and workstation workloads that saturate a dual-channel bus. Its unlocked multiplier makes it the only option of the two for tuning.

The overall split: the 8700T wins the benchmark series 7-1, holds a higher average score (2701 versus 2642), sits a percentile point higher (50 versus 49), and does it at a fraction of the rated power. The 5820K is the platform play, competitive in memory-bound multi-core work and richer in lanes, cache, and tuning freedom, but it concedes the performance crown on the strength of the 8700T's single-thread and rendering consistency.