More Leaked Benchmarks from Intel's Ivy Bridge

[kyuuketsuki]

[citation][nom]DavidC1[/nom]You are talking about the official slide results right? You will see your gains, but only when both of your setups are using integrated graphics. It seems that Sandy Bridge loses few % by using integrated graphics, and Ivy Bridge could fix that. So the results are true, just not using integrated graphics.[/quote]
I honestly don't know what you're talking about here.

It's unlikely that Turbo frequencies will account for more than 1-2% gains in average because desktop CPUs don't give big gains using Turbo Mode.[/citation]
On what basis do you make this assertion? Turbo frequencies make a difference on any lightly threaded workload where the processor has thermal/power headroom, which really make up the majority of an average or even enthusiast user's real-world usage.

Granted, it was probably overstating to say *most* of the improvement comes from the higher turbo frequencies; there's also the larger (the size probably makes a nominal difference except in thread heavy situations) and faster cache and other minor architectural differences. But, unless I misread/misinterpreted, the Ivy Bridge used in those benchmarks had the same base clock, but had a higher maximum turbo frequency. That would mean it was likely running at a higher frequency most of the time, and therefore those benchmarks don't exclusively demonstrate IPC gains. That's all I was really getting at.

 

[blazorthon]

[citation][nom]Kyuuketsuki[/nom]On what basis do you make this assertion? Turbo frequencies make a difference on any lightly threaded workload where the processor has thermal/power headroom, which really make up the majority of an average or even enthusiast user's real-world usage.Granted, it was probably overstating to say *most* of the improvement comes from the higher turbo frequencies; there's also the larger (the size probably makes a nominal difference except in thread heavy situations) and faster cache and other minor architectural differences. But, unless I misread/misinterpreted, the Ivy Bridge used in those benchmarks had the same base clock, but had a higher maximum turbo frequency. That would mean it was likely running at a higher frequency most of the time, and therefore those benchmarks don't exclusively demonstrate IPC gains. That's all I was really getting at.[/citation]

Yes, Ivy Bridge's performance gains in purely IPC are supposed to be something like 6% according to the last time it was checked (I think Anand did the checking, sorry about no link to back up the claim). Any other performance improvements are likely the turbo.