Intel Core i3 2100 Sandy Bridge vs. Core i3 7100 Kabylake Performance

[carewolf]

Yeah. I wonder if there is any way to factor that out. I guess not.

Though he could have at least have had the same amount of memory in each. That probably affected the compile-time benchmarks which were the only really surprising thing.

[strtj]

That scimark performance is baffling. Even using the exact same executable on both CPUs (which seems to have been the case, no -march=native or the like) I would at least expect an increase commensurate with the clock speed increase. Are the results for the scimark subtests available?

[mbello]

So is Moore's law dead or not?

Experts from the foundry industry say Moore's law is alive and doing well. Cost per transistor charts shows that costs are still going down with each new node which should have enabled for much faster improvements in Intel's processors over the past 6 years.

At its worse on the benchmarks Sandy Bridge was running at about half the speed. So Kabylake is twice as fast as Sandy Bridge... after 6 years. That is not what Moore's law is about. But if the process technology is doing well why are Intel's processors no longer evolving as they used to?

Some hypothesis:
1. Lack of competition;
2. Lower gains from architecture improvements... there is only so much optimization one can do;
3. Intel is increasing their gross margins (is the manufacturing cost of a USD 120 Intel processor today lower than it was 6 years ago?);

[triangle]

Yeah. I wonder if there is any way to factor that out. I guess not.

Maybe the DDR4 could be under clocked to match the DDR3 memory? I don't know if it is possible or not.

[aschmidtm]

cj.wijtmans
Oh, you don’t get it. Sorry.