amd is always fast until u bench it. thread for thread cache for cache sandy bridge is faster
While I think I understand what you're trying to say (that Zen and its derivatives generally perform better in productivity/compute than gaming), that statement literally makes no sense. How can something "be fast until you bench it"? You can't know if it's fast until you bench it. Also, the IPC numbers I mentioned are based on actual IPC testing, i.e. a wide range of benchmarks (for the Zen 2 v. CFL/SKL IPC test AnandTech ran the full SPEC2006SPEED test suite) on a single core/thread at normalized clock speeds. I would call that "benching it", wouldn't you? No, it's not a gaming benchmark, but that's entirely besides the point. Intel generally overperforms vs. AMD in gaming applications compared to general compute benchmarks, though at best at a high single digit percentage when looking at Zen 2. (It's also worth mentioning that despite the IPC deficit the tested 9900K achieved overall higher scores across the majority of the benchmark suite compared to the 3700X it was compared to thanks to its higher clock speeds - it's the sum of IPC x clock speed that matters in the end, after all.) So, of course, we could take my statement of
a 3.9GHz Zen 2 chip should be about on par with a 4.8GHz Haswell chip
and subtract 5-10% to account for Intel's lead in gaming loads, making it instead "a 3.9GHz Zen 2 chip should be about on par with a 4.32-4.56GHz Haswell chip." Not that big of a difference, eh? Still faster than the
vast majority of Haswell chips out there, let alone Sandy Bridge. And that statement is meant to imply at the same core and thread count in case that wasn't obvious, as IPC is thread count agnostic.
Also, saying something is faster "thread for thread, cache for cache" is rather meaningless for three reasons:
1: IPC denominates single-threaded performance normalizing for clock speed, so a chip with lower IPC
can't be faster thread for thread.
2: Cache isn't variable (unless you're using a 286 or something similar with cache chips on the motherboard), so what's the point of attempting to normalize for cache?
3: Not taking clock speeds into account for a comparison like that makes it utterly meaningless. Sandy Bridge is able to stay somewhat relevant only thanks to its OC ability - at stock even Zen (1) beats it soundly. But even a 5GHz Sandy Bridge chip is
easily beaten even in gaming by a Zen 2 chip with the same number of cores and threads at significantly lower clock speeds.
