Yes, there have been instances where a direct refresh improved CPU performance noticeably.
Yes, there have been instances where a direct refresh improved CPU performance noticeably.
The changes described represent minor architectural updates rather than substantial performance improvements. Most upgrades fall within a 10% threshold or less, so differences like these don’t meaningfully impact results. Significant gains are rare, and most reviews highlight incremental rather than revolutionary progress. The idea of consistent double-digit improvements is uncommon across generations.
But have there ever been instances where a significant upgrade truly changed the game? Like when buyers weren’t keen on CPU X before, but the refreshed version of X v2 at the same cost convinced them it was worth it? I get it—discounts can help, but improvements from newer generations like Kaby Lake to Coffee Lake usually drive interest more. The i3 8100 was a decent swap for its time, mainly because of the price reduction rather than major performance upgrades.
You're clarifying what makes the silicon functionally the same despite differences. If you're focusing on core architecture rather than node size, then Apple's M2 and M3 chips share the same 4P + 4E configuration. The M3 offers about a 22% performance boost, runs faster with higher clock speeds, and uses a 3nm process versus the M2's 5nm. Performance tests show the M3 outperforms the M2 by 18% on multicore tasks and 9% on single-core tasks, surpassing the 10% improvement target.
I mentioned that node shrinks counted in the same place. The i7 3770K wasn't just a direct refresh from the i7 2600K, even though Ivy Bridge is essentially Sandy Bridge on 22nm. All the same node details—cache size, core count, architecture—were identical. That would have been considered a straight refresh.
This wouldn't be a new design, it would be the same chip again. Intel used 14nm for a time, but each generation got tweaked slightly. Perhaps the X299 models are the closest match, though even they had minor adjustments, mainly in terms of Skylake-X performance. Your question is about whether any overclocked version existed. If yes, then it was the Intel KS series CPUs. I believe the 9900KS was the pioneer, with a maximum turbo boost only about a 10% improvement over the 9900K overall: https://www.techpowerup.com/review/intel...ks/21.html
A refresh means minor adjustments and improvements, which is typical as it often leads to a +10% boost in performance.
Using both the i3 10300 and i7 6700K at identical clock speeds shows minimal variation. The gains Intel achieved from Skylake to Comet Lake were mainly due to increased core count, larger cache, and higher frequencies. Despite the same basic design, Intel’s strategy from 2016 to 2020 mostly involved reusing the same chip. If you skip the gaming benchmarks, the overall performance boost at stock for the 9900KS is around 10%. But when considering real-world tests, the numbers tend to be lower. The 9900KS sometimes outperforms by more than 10%, yet many enthusiasts could easily push a regular 9900K over its limits, which means it appealed less to those seeking stable, stock-level results.