Here is the speed difference (CPU wise)

[slayerizer]

Here is the average Geekbench scores of the different cpu configuration

Imac 21.5" i5 2.7ghz: ~ 8200
Imac 21.5" i5 2.9ghz: ~ 9100
Imac 21.5" i5 3.1ghz: ~ 12400
Imac 27" i5 2.9ghz: unknown
Imac 27" i5 3.2ghz: unknown
Imac 27" i7 3.4ghz: ~ 12800

link

I tried to be careful not picking up scores of hackintosh machines.
It doesn't measure the GPU or HDD/SDD.

The GPU benchmarks are listed in this post: http://forums.macrumors.com/showthread.php?t=1499259

[Donza]

Based on these benchmarks we could assume that the highend 27" i5 -> i7 CPU upgrade would offer very marginal performance gains.

[alexdd]

Imac 21.5" i5 3.1ghz: ~ 12400---->i7 3.1ghz is the correct

[madrag]

Originally Posted by alexdd

Imac 21.5" i5 3.1ghz: ~ 12400---->i7 3.1ghz is the correct

Is that with the 8GB RAM?
It doesn't seem much more than the Mini i7 at 2.6Ghz...

[slayerizer]

Originally Posted by madrag

Is that with the 8GB RAM?
It doesn't seem much more than the Mini i7 at 2.6Ghz...

16gb ram

take note this is an i5 processor, not an i7 like the reference to the mini..
http://browser.primatelabs.com/geekbench2/1340639

[alexdd]

Originally Posted by madrag

Is that with the 8GB RAM?
It doesn't seem much more than the Mini i7 at 2.6Ghz...

yes not much difference...less than 10%..my mac mini (i7 2.6Ghz) is 11900 with 16Gb Ram

[madrag]

Originally Posted by slayerizer

16gb ram

take note this is an i5 processor, not an i7 like the reference to the mini..
http://browser.primatelabs.com/geekbench2/1340639

I just followed that link and it is an i7: Intel Core i7-3770S
Also there are no iMac 2012 with the i5 at 3.1Ghz... It is indeed an i7.

[slayerizer]

Originally Posted by alexdd

yes not much difference...less than 10%..my mac mini 2.6Ghz is 11900 with 16Gb Ram

my 2008 imac is around ~ 2900
Not that I have 6GB of ram in it, it's way faster. It doesn't feel slow for basic tasks. Movie editing is the only part where I need to leave the computer because it takes way too long!

[Laucian Nailor]

Originally Posted by slayerizer

Here is the average Geekbench scores of the different cpu configuration

Imac 21.5" i5 2.7ghz: ~ 8200
Imac 21.5" i5 2.9ghz: ~ 9100
Imac 21.5" i5 3.1ghz: ~ 12400
Imac 27" i5 2.9ghz: unknown
Imac 27" i5 3.2ghz: unknown
Imac 27" i7 3.4ghz: ~ 12800

Something appears a little odd about the results.

i5 0.2ghz increase (2.9 to 3.1) - which is a 7% increase in clock speed gives a 36% increase in the score (12400-9100), however

architecture upgrade (i5 to i7) plus 9.7% clock speed increase (3.1 to 3.4) only gives a 3.3% score increase (12400 to 12800)

Are they really saying the increase in screen size destroys the performance as much as the results appear to be indicating??

[slayerizer]

Originally Posted by Laucian Nailor

Something appears a little odd about the results.

i5 0.2ghz increase (2.9 to 3.1) - which is a 7% increase in clock speed gives a 36% increase in the score (12400-9100), however

architecture upgrade (i5 to i7) plus 9.7% clock speed increase (3.1 to 3.4) only gives a 3.3% score increase (12400 to 12800)

Are they really saying the increase in screen size destroys the performance as much as the results appear to be indicating??

screen size should not affect the Geekbench results.. but background tasks can influence.

[Laucian Nailor]

Originally Posted by slayerizer

screen size should not affect the Geekbench results.. but background tasks can influence.

In which case, this makes me believe its definitely flawed if its just an architecture v speed comparison. (I wouldn't think that 'clean' install models would have such different background tasks as to effect the % performance by as much as the numbers are stating).

----------

Oh, I get it now. Just looked at the source data......basically anybody can run this (irrespective of what they are also doing on the machine at the same time) and then post a score.

I see that identical spec machines are showing in some cases a 20% variation in their scores. I thought for a second this may have been a 'controlled' comparison, but guess not

[slayerizer]

Originally Posted by Laucian Nailor

In which case, this makes me believe its definitely flawed if its just an architecture v speed comparison. (I wouldn't think that 'clean' install models would have such different background tasks as to effect the % performance by as much as the numbers are stating).

----------

Oh, I get it now. Just looked at the source data......basically anybody can run this (irrespective of what they are also doing on the machine at the same time) and then post a score.

I see that identical spec machines are showing in some cases a 20% variation in their scores. I thought for a second this may have been a 'controlled' comparison, but guess not

well be careful some of them are hackintosh (pc hacked to run OSX).

You can tell by looking at several items, first indicator is Bios name and model, then you look at the ram, should be DDR1600mhz and the cpu is also an indicator!

I was careful picking them..

[Galatian]

Originally Posted by Laucian Nailor

Something appears a little odd about the results.

i5 0.2ghz increase (2.9 to 3.1) - which is a 7% increase in clock speed gives a 36% increase in the score (12400-9100), however

architecture upgrade (i5 to i7) plus 9.7% clock speed increase (3.1 to 3.4) only gives a 3.3% score increase (12400 to 12800)

Are they really saying the increase in screen size destroys the performance as much as the results appear to be indicating??

This is a common misconception...i5 and i7 is the same architecture. The i7 just has a few more featured available such as:

• integrated HD4000 vice HD2500 on i5
• SMT - creating a "virtual" core for every physical core essentially appearing as an 8 core machine
• slightly more L3 cache

Most people really won't need the small increase of the i7. See this superb Ars Technica article for a more in depth view at the issue.

[bflowers]

Are these scores a true measure? I'm wondering if the benchmark software is written to take advantage of hyperthreading and other features of each processor. Probably not.

Even if it is, someone already pointed out that background tasks will be using cycles, so unless you test everything straight from the box, or manually kill tasks until both are identical at the start.

[Laucian Nailor]

Originally Posted by Galatian

This is a common misconception...i5 and i7 is the same architecture. The i7 just has a few more featured available such as:

• integrated HD4000 vice HD2500 on i5
• SMT - creating a "virtual" core for every physical core essentially appearing as an 8 core machine
• slightly more L3 cache

Most people really won't need the small increase of the i7. See this superb Ars Technica article for a more in depth view at the issue.

Apologies for the tech burp!! Is it permitted to say it's a 'slightly improved' variance on the same architecture?

Can you tell me then, even assuming the same architecture, why the variation in performance increases (of the 2.9 to 3.1 compared to the 3.1 to 3.4) that I've quoted. If i7 does not contribute anything to the score compared to i5, it makes the results look slightly better but not by a factor of 10.

[Galatian]

Originally Posted by Laucian Nailor

Apologies for the tech burp!! Is it permitted to say it's a 'slightly improved' variance on the same architecture?

Can you tell me then, even assuming the same architecture, why the variation in performance increases (of the 2.9 to 3.1 compared to the 3.1 to 3.4) that I've quoted. If i7 does not contribute anything to the score compared to i5, it makes the results look slightly better but not by a factor of 10.

Please to read the Ars Technica article I've linked in my prior post then...bottom line: the i5/i7 nomenclature is pretty confusing...