Comparing processors with same clocks and cores

[Coldblackice]

What kind of performance difference would one see between two different processor architectures running at the same clocks and cores?

For example, an i7 930 @ 4.2GHz compared to a 4770k.


Practically, I'm wondering if I should upgrade, and what performance benefits I might see. Theoretically, I'm wondering what the benefit of a newer processor would be if it's only running at the same clocks/cores of the previous generation (setting aside newer abilities like on-die gfx processing). Isn't CPU performance mostly a function of top clock speed and parallel-processing/cores abilities (hinging on software's ability to spread across cores)?

I guess I'm getting hung up on the relative measure of performance across CPU's based on clock speed. I gather you probably can't reliably compare performance between CPU archs based on clock speed alone, and so I'm wondering if there's some sort of objective measure of comparison for such. Is there some way to "standardize" what the mean performance increase would be between CPU's and architectures?

 

[evilsofa]

Here's some chatter about i7 930 vs i7 4770k here on HardOCP:

http://hardforum.com/showthread.php?t=1764826&page=4

 

[Warrior]

The only reasons I upgraded from my i7 920 is SATA3 (I now have 1GB/s RAID0) and WAY lower power draw from an OCed CPU. And just those benefits alone are worth the upgrade, but if these don't matter to you, then you won't notice much.

 

[teh_chem]

IIRC, Haswell has up to ~20% higher IPC vs. Bloomfield CPUs (though some sources state only 10%). People have reported significant performance improvements in going to haswell from bloomfield--for a lot of reasons (instruction sets, IPC improvement). But most notably, if cost isn't the biggest factor, you can hit as-high if not higher clocks with haswell with less-extreme cooling; simple air cooling can easily get you 4.3-4.5GHz, and lower power draw. But you'll also get SATA3, PCI-e 3.0, USB3.0, etc.

There probably wasn't as much pressure to upgrade from Bloomfield up until Haswell (before SATA3, PCI-e 3, USB3 became standard features).

Whether you care to pay more for those features and mild (10-20%) performance improvements is a completely different matter. Do Bloomfield CPUs generally overclock to over 4.0GHz with relative ease?

 

[Warrior]

Do Bloomfield CPUs generally overclock to over 4.0GHz with relative ease?

Depends on stepping. The famous D0 would do 4GH+ all day. Mine was doing 4.4Ghz on water.

 

[Skripka]

IIRC, Haswell has up to ~20% higher IPC vs. Bloomfield CPUs (though some sources state only 10%). People have reported significant performance improvements in going to haswell from bloomfield--for a lot of reasons (instruction sets, IPC improvement). But most notably, if cost isn't the biggest factor, you can hit as-high if not higher clocks with haswell with less-extreme cooling; simple air cooling can easily get you 4.3-4.5GHz, and lower power draw. But you'll also get SATA3, PCI-e 3.0, USB3.0, etc.

There probably wasn't as much pressure to upgrade from Bloomfield up until Haswell (before SATA3, PCI-e 3, USB3 became standard features).

Whether you care to pay more for those features and mild (10-20%) performance improvements is a completely different matter. Do Bloomfield CPUs generally overclock to over 4.0GHz with relative ease?

Depends on stepping. The famous D0 would do 4GH+ all day. Mine was doing 4.4Ghz on water.

That being said...X58 ran hot. Further your choice and amount of RAM, as well as how it was configured, influenced your thermals and your OCs quite a bit. Whereas SB/IB/HW don't give a shit how large your RAM pool is and how many DIMMs....further they run a bit cooler, and not having the power-vampire that was the X58 PCH the system overall runs quite a bit cooler.