Wouldn't be allowed, right? That would position Intel as absolute monopolist on desktop, server and mobile CPU.
Intel CPU cost already way to much due to lacking competition from the only left x64 vendor AMD which is years behind. And Intel slowed down the release of faster CPUs. Where are the 8 and 16 core CPUs? Where are the 5+ GHz CPUs? Why hasn't the single core performance advanced that much (if you discount the advance of memory speed) since 2004 (Pentium 4 era).
Are you sure about the single thread performance? Because the jump from Pentium 3 to Pentium 4 was not that large, but the jump from Pentium 4 to Core 2 Duo (2006) was. All the jumps after that (Nehalem etc.) were larger than the jumps we have nowdays, and the jump to Sandy Bride (2011) was really large. Only since then has the single thread performance stagnated.
I'm not sure that cinebench is a good single thread benchmark because its basically a benchmark of how many vector ops you can run in parallel. But using it allows people to print pretty charts like you have there, as will benchmarks doing AES encryption, or any other "special purpose" instructions that have been added to the CPU in recent history.
A better traditional single thread benchmarks might be spec int base minus libquantum. That tests things like how good the branch predictors are relative to pipeline depth, how fast dependent instructions or pointer chasing can be handled, etc.
Yes, I was not really sure how representable Cinebench is, hence my question how sure the parent is. But the benchmarks were easy to come by (back when I posted them already a while back). The only other benchmark I could find that had CPUs from Pentium 4 up to Haswell was Geekbench.
Would love to see someone do a benchmark like that (would do it myself if I had enough spare money), with equal clock and memory bandwith etc.
EDIT: after reading the critique about libquantum on wikipedia I'm even more interested in the benchmarks.
"In the SpecInt2006 benchmarks, the 462.libquantum benchmark is highly vectorizable. The baseline computer for all benchmarks is a 1997 Sun Ultrasparc server computer. Whereas most of the spec sub-benchmarks turn in a performance improvement of about 5x to 80x times faster than the Ultrasparc, the particular 462.libquantum sub-benchmark turns in a result that is up to 4082 times faster than the Sun Ultrasparc .[3] This suggests that for this sub-benchmark, most of the improvements over the Ultrasparc are due to vectorizing compiler improvements, NOT due to CPU hardware improvements, since 1997."
> Why hasn't the single core performance advanced that much (if you discount the advance of memory speed) since 2004 (Pentium 4 era).
It has, but you have to use AVX and other new instruction sets. Singlethreaded, a fifth-gen Core m5 tablet CPU can be as fast as a first-gen Core i Xeon workstation CPU, and a sixth gen i7 desktop CPU can be twice as fast as both: https://admin.tao.at/post/evolution-of-cpu-efficiency/
(Sadly I don't have any Pentium 4s left to compare on. And of the Core2 generation only laptops.)
It seems like a lot of new tech went into AVX and increasing both width and number of vector registers. Kind of makes you doubt whether single-threaded performance is just a mask for wider execution pipe underneath.
Intel CPU cost already way to much due to lacking competition from the only left x64 vendor AMD which is years behind. And Intel slowed down the release of faster CPUs. Where are the 8 and 16 core CPUs? Where are the 5+ GHz CPUs? Why hasn't the single core performance advanced that much (if you discount the advance of memory speed) since 2004 (Pentium 4 era).