Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

rMBP 2.6 Vs. 2.7 Difference

H

HighEndMac

macrumors regular
Original poster
Mar 30, 2011
229
1
Can someone please tell me the difference between the 2.6 vs the 2.7 ?

When will you see the performance boost?

Why do they even offer such a small bump?
 
cstm said:
8mb cache on 2.7, 6mb cache on 2.6
Click to expand...

Pretty much this. Many would argue that for a lot of applications the 2.6 -> 2.7 could be more effective than the 2.3 -> 2.6.

But it all depends on your usage, so YMMV. Google "L3 cache" if you want to figure out what 6MB -> 8MB actually means.
 
Thanks for the feedback, and I did Google "L3 cache" but I am still having a hard time understanding the difference in real world applications.

Anybody mind breaking it down in terms a 6th grader can understand?
 
HighEndMac said:
Thanks for the feedback, and I did Google "L3 cache" but I am still having a hard time understanding the difference in real world applications.

Anybody mind breaking it down in terms a 6th grader can understand?
Click to expand...

Basically, when a computer looks for something, it goes through this pattern:

L1 cache -> L2 cache -> L3 cache -> RAM -> HD

Each further step in this progression takes more and more CPU cycles (ex. HD takes much longer to access than RAM). However, each step in the progression is also several times larger size-wise than the former.

So that extra 2MB of L3 cache will basically let your computer to store more running processes there and make it faster to call them (in the same way a bigger RAM means you don't need to page out to the HD).

That's my take, anyway - most of this is from self-study.
 
Slivortal said:
Basically, when a computer looks for something, it goes through this pattern:

L1 cache -> L2 cache -> L3 cache -> RAM -> HD

Each further step in this progression takes more and more CPU cycles (ex. HD takes much longer to access than RAM). However, each step in the progression is also several times larger size-wise than the former.

So that extra 2MB of L3 cache will basically let your computer to store more running processes there and make it faster to call them (in the same way a bigger RAM means you don't need to page out to the HD).

That's my take, anyway - most of this is from self-study.
Click to expand...

Nice, thanks buddy.

Now would you really notice the difference between the 2.6 vs the 2.7 if the rest of the configuration is the same? (16RAM/512GB SSD)

When would you notice the difference using the 2.7? (Like only large video editing programs or does it affect something like web browsing?)
 
HighEndMac said:
Nice, thanks buddy.

Now would you really notice the difference between the 2.6 vs the 2.7 if the rest of the configuration is the same? (16RAM/512GB SSD)

When would you notice the difference using the 2.7? (Like only large video editing programs or does it affect something like web browsing?)
Click to expand...

As I said, I'm not an expert. The times when you'd most recognize the L3 cache is when you have some kind of program that has a few key processes that it keeps needing to call over and over, yet already exhaust the L1 and L2 caches. The frequency with which it would call that extra 2MB of L3 cache will determine how much you'll actually benefit.
 
For the most part, 2MB L3 cache will not make a noticable difference. Even if the L3 cache were entirely disabled, the performance penalty would only amount to about 100-200MHz.
 
The difference you will see is time save at full performance, the 2.7 is useful to those that rely on their Mac`s for businesses, if you can reduce your render times by 10%, you can moniterize the time saving, increase your productivity etc. 2.7 with the 8Mb L3 will really only be of significant benefit to certain applications such as video rendering, essentially you will see no benefit unless applications can utilise the additional cache.

For the average user the 2.7 and even the 2.6 will offer little if any real world increase in performance, a few fps in a game etc, even the base 2.3 is an extremely powerful machine by portable standards. The 2.7 is simply not going to kick in and "smoke" the other CPU`s, dont get me wrong the 2.7 is faster the only question is will you ever notice that difference? I know it`s very cliched, however if you need to ask, you likely dont need the performance increase.

L1 Cache is the fastest Cache, each core has its own L1 cache its the smallest but its the fastest and the first one to accesses by the Core.

L2 Cache is slower than L1, L2 cache is the 2nd cache that the core search's for its data, and bigger than L1.

L3 Cache is shared between all of the cores, in the Core i7 for example, is shared between all 4 cores, and bigger than both L1 and L2.

Just wait till Haswell as it`s looking like L4 cache will go mainstream :p
 
Last edited:
There is virtually no difference in real-world computing between 6MB and 8MB cache on the Sandy/Ivy Bridge Intel chips.

If you look at 2500K and 2600K benchmarks from last year, there is virtually no difference in any tests at the same clockspeed, and when there was a measurable difference, it was due to hyperthreading rather than L3 cache. (and that doesn't apply here)

If you absolutely must have the fastest machine available, by all means go for it, but you're paying 10% extra for less than a 5% performance gain.
 
jcpb said:
If you have to ask whether something is worth the extra cost, it usually isn't.
Click to expand...

I can see that and in most cases you are right, but not this time.

I really wanted to know the tech side of it.
 
Queen6 said:
The difference you will see is time save at full performance, the 2.7 is useful to those that rely on their Mac`s for businesses, if you can reduce your render times by 10%, you can moniterize the time saving, increase your productivity etc. 2.7 with the 8Mb L3 will really only be of significant benefit to certain applications such as video rendering, essentially you will see no benefit unless applications can utilise the additional cache.

For the average user the 2.7 and even the 2.6 will offer little if any real world increase in performance, a few fps in a game etc, even the base 2.3 is an extremely powerful machine by portable standards. The 2.7 is simply not going to kick in and "smoke" the other CPU`s, dont get me wrong the 2.7 is faster the only question is will you ever notice that difference? I know it`s very cliched, however if you need to ask, you likely dont need the performance increase.

L1 Cache is the fastest Cache, each core has its own L1 cache its the smallest but its the fastest and the first one to accesses by the Core.

L2 Cache is slower than L1, L2 cache is the 2nd cache that the core search's for its data, and bigger than L1.

L3 Cache is shared between all of the cores, in the Core i7 for example, is shared between all 4 cores, and bigger than both L1 and L2.

Just wait till Haswell as it`s looking like L4 cache will go mainstream :p
Click to expand...

You won't save 10% on render times. And if render times were that important to your business - you wouldn't be doing them on a laptop.

For the most part, even an SSD can move data faster than the CPU can render the video. So that slight increase in cache will have pretty much no effect.

It will have a bigger impact on things like doing massive number calculations (finding prime numbers, digits of pi, etc...) than things like video rendering and such.

Will the CPU frequency increase help? Sure. But we're talking a 3-5% difference there.
 
Last edited:
terraphantm said:
You won't save 10% on render times

For the most part, even an SSD can move data faster than the CPU can render the video. So that slight increase in cache will have pretty much no effect.

It will have a bigger impact on things like doing massive number calculations (finding prime numbers, digits of pi, etc...) than things like video rendering and such.
Click to expand...

Good 2 Know! Thx, just reordered my rMBP, but this time 2.6GHz/16GB RAM/512GB SSD (W/ ML preinstalled + Pages and Numbers + Apple Care) vs a maxed out 2.7 model.
 
terraphantm said:
You won't save 10% on render times. And if render times were that important to your business - you wouldn't be doing them on a laptop.

For the most part, even an SSD can move data faster than the CPU can render the video. So that slight increase in cache will have pretty much no effect.

It will have a bigger impact on things like doing massive number calculations (finding prime numbers, digits of pi, etc...) than things like video rendering and such.

Will the CPU frequency increase help? Sure. But we're talking a 3-5% difference there.
Click to expand...

10% is simply an arbitrary figure, "if you can" read the post the 2.7 is for those that have a specialised need and the applications that can take advantage of the additional cache, the exact time save will clearly be dependent on the software implemented. For the majority it`s money ill spent.

Some will and do require a high end portable solution, Apple provides that regardless of the usage; be it rendering 3D or crushing numbers, as for video the faster you can render the faster you will complete, SSD will be waiting on CPU not vice versa, it`s simply an example not verbatim, cast in stone...

----------
HighEndMac said:
Good 2 Know! Thx, just reordered my rMBP, but this time 2.6GHz/16GB RAM/512GB SSD (W/ ML preinstalled + Pages and Numbers + Apple Care) vs a maxed out 2.7 model.
Click to expand...

Good for you, you will get far more use out out the additional software than you would have likely gotten from the 2.7 CPU, and you will have $$$ in your hand.
 
Queen6 said:
10% is simply an arbitrary figure, "if you can" read the post the 2.7 is for those that have a specialised need and the applications that can take advantage of the additional cache, the exact time save will clearly be dependent on the software implemented. For the majority it`s money ill spent.

Some will and do require a high end portable solution, Apple provides that regardless of the usage; be it rendering 3D or crushing numbers, as for video the faster you can render the faster you will complete, SSD will be waiting on CPU not vice versa, it`s simply an example not verbatim, cast in stone...

----------



Good for you, you will get far more use out out the additional software than you would have likely gotten from the 2.7 CPU, and you will have $$$ in your hand.
Click to expand...

My point is, the examples you cite are one of the examples where the money is ill spent. It's only useful in the rare case that you have a process that's too large to fit in 6MB, but small enough to fit within 8MB. It's mainly number crunching that will benefit. Otherwise you're going to have to be constantly cycling through that memory anyway.

And unless the CPU can render 2K+ video significantly faster than real time, the SSD will be able to feed the data to the CPU fast enough that the cache will essentially be irrelevant.

Take a look at this
http://www.tomshardware.com/reviews/athlon-l3-cache,2416-8.html

That's on an AMD CPU (Phenom has 8MB L3, and Athlon has no L3 or shared memory at all). Even in that case, the L3 cache has little to no effect on video rendering. A drop from 8MB to 6MB will essentially be irrelevant.
 
terraphantm said:
My point is, the examples you cite are one of the examples where the money is ill spent. It's only useful in the rare case that you have a process that's too large to fit in 6MB, but small enough to fit within 8MB. It's mainly number crunching that will benefit. Otherwise you're going to have to be constantly cycling through that memory anyway.

And unless the CPU can render 2K+ video significantly faster than real time, the SSD will be able to feed the data to the CPU fast enough that the cache will essentially be irrelevant.

Take a look at this
http://www.tomshardware.com/reviews/athlon-l3-cache,2416-8.html

That's on an AMD CPU (Phenom has 8MB L3, and Athlon has no L3 or shared memory at all). Even in that case, the L3 cache has little to no effect on video rendering. A drop from 8MB to 6MB will essentially be irrelevant.
Click to expand...

It`s an example nothing more, nothing less, you are simply reading far too much into it, once again the 2.7 is only useful for very specific applications that is the point.
 
I find it amazing how small a cache size we're talking about here. We're always on about gigabytes, and usually hundreds of them in hard drives, yet here we're increasing 6 MB to 8MB! Is cache storage extremely expensive or hard to engineer or just that very small amounts lead to very low latency? I wonder if/when Level caches will ever reach the gigabytes!
 
Good for you, you will get far more use out out the additional software than you would have likely gotten from the 2.7 CPU, and you will have $$$ in your hand.
Click to expand...

with $250 dollar savings, just add another $50 plus tax you can buy yourself a final cut pro X. That can take advantage of your new machine. If you already own one good for you.
 
Dangerous Theory said:
I find it amazing how small a cache size we're talking about here. We're always on about gigabytes, and usually hundreds of them in hard drives, yet here we're increasing 6 MB to 8MB! Is cache storage extremely expensive or hard to engineer or just that very small amounts lead to very low latency? I wonder if/when Level caches will ever reach the gigabytes!
Click to expand...

Yes, quite expensive. L1-L3 caches all reside on the same die for the CPU's in question. The die is the piece of silicon contains the entire processor. These caches have to be really close to the CPU, since at the clock rates we're talking about (2.6 GHz), light would only travel 11.5cm (4.5 inches), and in that time, not only do the electrons have to move through lots of tiny wires, lots of gates have to turn on and off. There ain't no time to travel off the die.

Here is a a diagram of the a Sandy Bridge processor: http://www.brightsideofnews.com/news/2009/7/5/intels-anti-amd-fusion-sandy-bridge-cpu-tapes-out.aspx

Notice that about 1/4 of the space is spent on the L3 cache (the L1 caches are even closer, and part of each Core). That's 1/4 of the space actually costs more than 1/4 of the fabrication cost of that die, because as die size increases, price increases exponentially. If you care, this is due to the fabrication yield:
http://en.wikipedia.org/wiki/Semiconductor_device_fabrication

Processors, in order to hit the high clock rate and keep electricity usage down, tend to use the smallest transistor sizes that are available. This increases costs in two ways: 1) The fabrication process is newer, and less available. 2) The silicon and fabrication process has to be even more error free, since smaller defects will cause problems.

This is all to say, yes. It's expensive. Those 32GB micro SD cards are much easier to make than that tiny 3MB of L3 cache.
 
Penooker said:
Yes, quite expensive. L1-L3 caches all reside on the same die for the CPU's in question. The die is the piece of silicon contains the entire processor. These caches have to be really close to the CPU, since at the clock rates we're talking about (2.6 GHz), light would only travel 11.5cm (4.5 inches), and in that time, not only do the electrons have to move through lots of tiny wires, lots of gates have to turn on and off. There ain't no time to travel off the die.

Here is a a diagram of the a Sandy Bridge processor: http://www.brightsideofnews.com/news/2009/7/5/intels-anti-amd-fusion-sandy-bridge-cpu-tapes-out.aspx

Notice that about 1/4 of the space is spent on the L3 cache (the L1 caches are even closer, and part of each Core). That's 1/4 of the space actually costs more than 1/4 of the fabrication cost of that die, because as die size increases, price increases exponentially. If you care, this is due to the fabrication yield:
http://en.wikipedia.org/wiki/Semiconductor_device_fabrication

Processors, in order to hit the high clock rate and keep electricity usage down, tend to use the smallest transistor sizes that are available. This increases costs in two ways: 1) The fabrication process is newer, and less available. 2) The silicon and fabrication process has to be even more error free, since smaller defects will cause problems.

This is all to say, yes. It's expensive. Those 32GB micro SD cards are much easier to make than that tiny 3MB of L3 cache.
Click to expand...
All very interesting information. I suspect I'll learn this a bit more since I start my EEE MEng in October.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back