Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.sys.mac.system > #53929 > unrolled thread

[NEWS] Mac Pro CPUs can be replaced

Back to article view | Back to comp.sys.mac.system

Contents

  [NEWS] Mac Pro CPUs can be replaced Your Name <YourName@YourISP.com> - 2013-12-28 19:17 +1300
    Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-28 04:25 -0500
      Re: [NEWS] Mac Pro CPUs can be replaced Your Name <YourName@YourISP.com> - 2013-12-29 09:49 +1300
        Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-28 18:36 -0500
          Re: [NEWS] Mac Pro CPUs can be replaced Jolly Roger <jollyroger@pobox.com> - 2013-12-29 04:23 +0000
            Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-29 01:23 -0500
            Re: [NEWS] Mac Pro CPUs can be replaced Lewis <g.kreme@gmail.com.dontsendmecopies> - 2013-12-30 04:32 +0000
              Re: [NEWS] Mac Pro CPUs can be replaced nospam <nospam@nospam.invalid> - 2013-12-30 00:30 -0500
              Re: [NEWS] Mac Pro CPUs can be replaced Julian Vrieslander <MY_FIRST_NAME@MY_LAST_NAME.com> - 2013-12-31 15:32 -0800
                Re: [NEWS] Mac Pro CPUs can be replaced Lewis <g.kreme@gmail.com.dontsendmecopies> - 2014-01-01 00:05 +0000
                Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-31 19:35 -0500
                  Re: [NEWS] Mac Pro CPUs can be replaced Your Name <YourName@YourISP.com> - 2014-01-01 13:45 +1300
                    Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-31 21:24 -0500
                      Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2013-12-31 21:33 -0500
                        Re: [NEWS] Mac Pro CPUs can be replaced dempson@actrix.gen.nz (David Empson) - 2014-01-01 21:07 +1300
                          Re: [NEWS] Mac Pro CPUs can be replaced JF Mezei <jfmezei.spamnot@vaxination.ca> - 2014-01-01 17:43 -0500
                            Re: [NEWS] Mac Pro CPUs can be replaced dempson@actrix.gen.nz (David Empson) - 2014-01-02 22:48 +1300
                              Re: [NEWS] Mac Pro CPUs can be replaced Lewis <g.kreme@gmail.com.dontsendmecopies> - 2014-01-02 20:20 +0000
                        Re: [NEWS] Mac Pro CPUs can be replaced Lewis <g.kreme@gmail.com.dontsendmecopies> - 2014-01-01 09:31 +0000
                          Re: [NEWS] Mac Pro CPUs can be replaced Lewis <g.kreme@gmail.com.dontsendmecopies> - 2014-01-01 11:59 +0000
                      Re: [NEWS] Mac Pro CPUs can be replaced Michelle Steiner <michelle@michelle.org> - 2013-12-31 19:48 -0700

Page 1 of 2  [1] 2  Next page →

#53929 — [NEWS] Mac Pro CPUs can be replaced

Apparently the CPUs on the Mac Pro motherboard are NOT soldered in
place, so they are able to be user-swapped at a future date. This is
from MacRoumors.com ...

     New Mac Pro Confirmed to Have Removable CPU
     -------------------------------------------
     With the first new Mac Pro units having made their
     way to reviewers and even some early online orders
     arriving in the closing days of 2013, more details
     on the new machines are continuing to surface.

     Other World Computing (OWC) has performed a quick
     teardown on one of the new machines, and while
     documentation currently consists of only a small 
     set of photos, the company has confirmed that the
     Intel Xeon E5 processor found in the Mac Pro is 
     indeed removable, allowing for future upgrades. 
     All four available CPUs for the new Mac Pro use 
     the same LGA 2011 socket standardized on the Mac 
     Pro's motherboard.

     The main processor is one of the most significant 
     variables in the cost of the new Mac Pro, with 
     the four available CPU options spanning $3500 in
     upgrade charges. Pricing relative to the stock 
     3.7GHz quad-core Intel Xeon E5 with 10MB of L3 
     cache is as follows:

     - 3.5GHz 6-Core Intel Xeon E5 
         with 12MB of L3 cache:            +$500

     - 3.0GHz 8-core Intel Xeon E5 
         with 25MB of L3 cache:           +$2000

     - 2.7GHz 12-core Intel Xeon E5
         with 30MB of L3 cache:           +$3500

     Using a removable socketed processor rather than
     the soldered processors found in most of Apple's 
     Macs means that users may be able to upgrade their
     machines in the future as their needs change and/or
     chip prices decline.



     Earlier this week, OWC also shared a photo showing
     the new Mac Pro driving a total of six 27-inch 
     displays, each with a horizontal resolution of at
     least 2500 pixels. The new Mac Pro offers six 
     Thunderbolt 2 ports for connectivity, allowing for
     a number of combinations of displays and other
     peripherals.
  <http://cdn.macrumors.com/article-new/2013/12/mac_pro_6_displays.jpg>

[toc] | [next] | [standalone]

#53932

On 13-12-28 01:17, Your Name wrote:
> 
> Apparently the CPUs on the Mac Pro motherboard are NOT soldered in
> place, so they are able to be user-swapped at a future date. This is
> from MacRoumors.com ...

MacPro 2009 also had remopvable CPUs (from a obese daughtercard complete
with its own fans). Yet, when Apple updated the MacPro in 2010, we were
not allowed to buy the new CPU to be plugged into our 2009 Macs. So the
CPU upgradability was there but never enabled by Apple.

However, at the socket level, it may mean that you could source your
CPUs elsewhere at lower price than Apple's initial upgrade price.

From a manufacutring point of view, since they are apparently built to
order, it would make sense to not solder the CPUs in since they can then
have vanilla mothermoards mass produced once, and then fit then with a
variety of user options when the machine is being assembled for a
sepcific customer.

[toc] | [prev] | [next] | [standalone]

#53963

In article <52be9923$0$22399$c3e8da3$f6268168@news.astraweb.com>, JF
Mezei <jfmezei.spamnot@vaxination.ca> wrote:
> On 13-12-28 01:17, Your Name wrote:
> > 
> > Apparently the CPUs on the Mac Pro motherboard are NOT soldered in
> > place, so they are able to be user-swapped at a future date. This is
> > from MacRoumors.com ...
> 
> MacPro 2009 also had remopvable CPUs (from a obese daughtercard complete
> with its own fans). Yet, when Apple updated the MacPro in 2010, we were
> not allowed to buy the new CPU to be plugged into our 2009 Macs. So the
> CPU upgradability was there but never enabled by Apple.
> 
> However, at the socket level, it may mean that you could source your
> CPUs elsewhere at lower price than Apple's initial upgrade price.
> 
> From a manufacutring point of view, since they are apparently built to
> order, it would make sense to not solder the CPUs in since they can then
> have vanilla mothermoards mass produced once, and then fit then with a
> variety of user options when the machine is being assembled for a
> sepcific customer.

All Apple's other computers use soldered parts and yet are still "Build
to Order" customisable, so I doubt that is the reason.

[toc] | [prev] | [next] | [standalone]

#53986

On 13-12-28 15:49, Your Name wrote:

> All Apple's other computers use soldered parts and yet are still "Build
> to Order" customisable, so I doubt that is the reason.

The iMac and laptops are mass produced with educated guesses on demand
for each soldered config.  With such high volumes they can afford to
have different configs mass produced.

With the MacPro being low volume, they probbaly don't have enough
information to predict demand for each config so it becomes most cost
effective to put a socket in and configured on-demand.

(And it is also possible that for higher end CPUs, sockets are standard,
does anyone know ?)

[toc] | [prev] | [next] | [standalone]

#53999

On 2013-12-28, JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:
> On 13-12-28 15:49, Your Name wrote:
>
>> All Apple's other computers use soldered parts and yet are still "Build
>> to Order" customisable, so I doubt that is the reason.
>
> The iMac and laptops are mass produced with educated guesses on demand
> for each soldered config.  With such high volumes they can afford to
> have different configs mass produced.
>
> With the MacPro being low volume, they probbaly don't have enough
> information to predict demand for each config so it becomes most cost
> effective to put a socket in and configured on-demand.
>
> (And it is also possible that for higher end CPUs, sockets are standard,
> does anyone know ?)

More blind speculation from the peanut gallery...

-- 
E-mail sent to this address may be devoured by my ravenous SPAM filter.
I often ignore posts from Google. Use a real news client instead.

JR

[toc] | [prev] | [next] | [standalone]

#54005

On 13-12-28 23:23, Jolly Roger wrote:

> More blind speculation from the peanut gallery...

It is not home brewed speculation that the MacPro is coming with
socketed CPUs. It comes from from a reliable trade news outlet.

Once you take the "fact" of socketed CPUs on MacPro, one can make many
educated guesses on why Apple chose that route. Sure it is speculation
until Apple explains why (which it won't). But doesn't mean it is
"blind" speculation.

And more importantly, it does not imply that Apple will provide CPU
upgrades.

[toc] | [prev] | [next] | [standalone]

#54031

In message <bi9mdhFflh8U4@mid.individual.net> 
  Jolly Roger <jollyroger@pobox.com> wrote:
> On 2013-12-28, JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:
>> On 13-12-28 15:49, Your Name wrote:
>>
>>> All Apple's other computers use soldered parts and yet are still "Build
>>> to Order" customisable, so I doubt that is the reason.
>>
>> The iMac and laptops are mass produced with educated guesses on demand
>> for each soldered config.  With such high volumes they can afford to
>> have different configs mass produced.
>>
>> With the MacPro being low volume, they probbaly don't have enough
>> information to predict demand for each config so it becomes most cost
>> effective to put a socket in and configured on-demand.
>>
>> (And it is also possible that for higher end CPUs, sockets are standard,
>> does anyone know ?)

> More blind speculation from the peanut gallery...

The MacPro processors are not socketed.

-- 
I AM ZOMBOR! (kelly) ZOMBOR!

[toc] | [prev] | [next] | [standalone]

#54039

In article <slrnlc1trh.c4g.g.kreme@amelia.local>, Lewis
<g.kreme@gmail.com.dontsendmecopies> wrote:

> The MacPro processors are not socketed.

wrong. 

<http://www.macrumors.com/2013/12/27/new-mac-pro-confirmed-to-have-remov
able-cpu/>
  ...the company has confirmed that the Intel Xeon E5 processor found
  in the Mac Pro is indeed removable, allowing for future upgrades.

[toc] | [prev] | [next] | [standalone]

#54160

In article <slrnlc1trh.c4g.g.kreme@amelia.local>,
 Lewis <g.kreme@gmail.com.dontsendmecopies> wrote:

> The MacPro processors are not socketed.


2013 Mac Pro CPUs are socketed.

http://www.ifixit.com/Teardown/Mac+Pro+Late+2013+Teardown/20778

-- 
Julian Vrieslander

[toc] | [prev] | [next] | [standalone]

#54163

In message <MY_FIRST_NAME-E98034.15322131122013@news.giganews.com> 
  Julian Vrieslander <MY_FIRST_NAME@MY_LAST_NAME.com> wrote:
> In article <slrnlc1trh.c4g.g.kreme@amelia.local>,
>  Lewis <g.kreme@gmail.com.dontsendmecopies> wrote:

>> The MacPro processors are not socketed.


> 2013 Mac Pro CPUs are socketed.

That's odd, I was told by two people it wasn't. Good to know.

> http://www.ifixit.com/Teardown/Mac+Pro+Late+2013+Teardown/20778



-- 
Amazingly Beautiful Creatures Dancing Excites the Forest Glade, in my
Heart how I do Jump like the Kudo Listen to the Music so Nice the Organ
Plays. Quietly Rests the Sleepy Tiger Under the Vine tree at the Water's
side and X marks the spot 'neath the Yellow moon where the Zulu king and
I did hide.

[toc] | [prev] | [next] | [standalone]

#54165

On 13-12-31 18:32, Julian Vrieslander wrote:

> 2013 Mac Pro CPUs are socketed.
> 
> http://www.ifixit.com/Teardown/Mac+Pro+Late+2013+Teardown/20778


Which will lead to an endless series of "how many Mac heads does it take
to screw in a CPU in a Mac Pro" jokes :-)

[toc] | [prev] | [next] | [standalone]

#54166

In article <52c362e1$0$46047$c3e8da3$88b277c5@news.astraweb.com>, JF
Mezei <jfmezei.spamnot@vaxination.ca> wrote:
> On 13-12-31 18:32, Julian Vrieslander wrote:
> > 
> > 2013 Mac Pro CPUs are socketed.
> > http://www.ifixit.com/Teardown/Mac+Pro+Late+2013+Teardown/20778
> 
> Which will lead to an endless series of "how many Mac heads does it take
> to screw in a CPU in a Mac Pro" jokes :-)

Apparently only one ... as long as they know what they're doing.  ;-)
The cylinder Mac Pro, like previous models, is a lot easier to get into
than Apple's other computers and devices. It received an 8 out of 10
repairability score from iFixtIt.com:
<http://www.macrumors.com/2013/12/31/teardown-of-new-mac-pro-reveals-sur
prising-amount-of-accessibility-circular-daugtherboard/>

[toc] | [prev] | [next] | [standalone]

#54170

On 13-12-31 19:45, Your Name wrote:

> The cylinder Mac Pro, like previous models, is a lot easier to get into
> than Apple's other computers and devices. It received an 8 out of 10
> repairability score from iFixtIt.com:


That is because iFixit has become so used to Apple going out of its way
to make dismantling impossible, this one looks easy. But consider how
much disassembly is required to get to the CPU.

What is not clear to me is how realistic it is to expect upgradable
CPUs. Would the firmware be coded to allow only one specific CPU type ?
(would not be s surprise for Apple to do that).

[toc] | [prev] | [next] | [standalone]

#54171

BTWL: Generic question:

With current memory interfaces, is there any performance differences
between having 1 12 core CPU vs 2 6 core CPUs ?

Do the current intel processors have shared cache between cores ?

Would 1 12 core processor with 1 shared cache memory be different than 2
6 core processors with 2 separate caches (twice the cache size, but not
shared by all cores)

If the first 6 cores deal with one application and the second 6 cores
deal with different app/data, then having separate caches would be an
advantage since you have overall more cache. Burt if you do a video that
makes use of all 12 cores, then having a single cache would make sense.

Any comments on that logic ?

[toc] | [prev] | [next] | [standalone]

#54180

JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:

> BTWL: Generic question:
> 
> With current memory interfaces, is there any performance differences
> between having 1 12 core CPU vs 2 6 core CPUs ?

Not sure without doing research. (Moot point since Apple doesn't make a
dual CPU configuration of the 2013 Mac Pro.)

> Do the current intel processors have shared cache between cores ?

The L3 cache is shared between cores. L1 and L2 caches are separate for
each core.

> Would 1 12 core processor with 1 shared cache memory be different than 2
> 6 core processors with 2 separate caches (twice the cache size, but not
> shared by all cores)
>
> If the first 6 cores deal with one application and the second 6 cores
> deal with different app/data, then having separate caches would be an
> advantage since you have overall more cache.

You are making false assumptions about cache sizes. In the Intel Xeon E5
v2 series (as used by the 2013 Mac Pro), the 12-core processor has more
than twice as much L3 cache as the 6-core processor. That is mainly to
compensate for its lower clock speed with the same number of cores
active.

> Burt if you do a video that makes use of all 12 cores, then having a
> single cache would make sense.
> 
> Any comments on that logic ?

Since Apple doesn't make a Mac Pro with two Xeon E5 v2 CPUs, the
question of relative performance doesn't seem relevant except for
academic interest, unless you want to compare two generations of the Mac
Pro, or compare a Mac Pro with some other brand that supports two of
these processors.

If there was a hypothetical dual CPU 2013 Mac Pro, a dual 6-core CPU
configuration (12 cores total) would probably be somewhat faster than a
single 12-core CPU configuration, due to the faster clock speed of the
6-core processor for any number of active cores, despite having less
total L3 cache.

See Anandtech's review, particularly page 4.

http://www.anandtech.com/show/7603/mac-pro-review-late-2013

Here is the page on Intel's web site to access technical details of the
Xeon E5 v2 processors.

http://ark.intel.com/products/family/78582/Intel-Xeon-Processor-E5-v2-Family

The chips Apple is using are:

4-core:  Xeon E5-1620 v2
6-core:  Xeon E5-1650 v2
8-core:  Xeon E5-1680 v2
12-core: Xeon E5-2697 v2

-- 
David Empson
dempson@actrix.gen.nz

[toc] | [prev] | [next] | [standalone]

#54222

On 14-01-01 03:07, David Empson wrote:

> You are making false assumptions about cache sizes. In the Intel Xeon E5
> v2 series (as used by the 2013 Mac Pro), the 12-core processor has more
> than twice as much L3 cache as the 6-core processor. 

In the couple of pages of the anandtech article you pointed to, I didn't
see mention of cache differences between the various options.

> Since Apple doesn't make a Mac Pro with two Xeon E5 v2 CPUs, the
> question of relative performance doesn't seem relevant except for
> academic interest,

It is relevant in terms of evaluating Apple's decision to allocate space
for only one chip. In other words, was this done because of Jony Ive's
industrial design requirements (beauty/small), or has the chip
technology really gotten to a point where there is no longer the need
for multiple chips.

Based on the anandtechn opinion/tests, it appears that stopping at 6-8
cores is better because you benefit from higher clock rate.


(I am also surprised how well the 2009 Mac Pro fares in his tests -
although he has an 8 core one, I have only a 4 core one).

[toc] | [prev] | [next] | [standalone]

#54258

JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:

> On 14-01-01 03:07, David Empson wrote:
> 
> > You are making false assumptions about cache sizes. In the Intel Xeon E5
> > v2 series (as used by the 2013 Mac Pro), the 12-core processor has more
> > than twice as much L3 cache as the 6-core processor. 
> 
> In the couple of pages of the anandtech article you pointed to, I didn't
> see mention of cache differences between the various options.

Page 4 (CPU Choices) has a table near the top titled "Mac Pro (Late
2013) CPU Choices" which has a line "L3 Cache" that lists the cache
sizes for the four processor options:

4 core: 10 MB
6 core: 12 MB
8 core: 25 MB
12 core: 30 MB

> > Since Apple doesn't make a Mac Pro with two Xeon E5 v2 CPUs, the
> > question of relative performance doesn't seem relevant except for
> > academic interest,
> 
> It is relevant in terms of evaluating Apple's decision to allocate space
> for only one chip. In other words, was this done because of Jony Ive's
> industrial design requirements (beauty/small), or has the chip
> technology really gotten to a point where there is no longer the need
> for multiple chips.

Or keeping within reasonable power/heat requirements. Having two 6-core
processors instead of one 12-core processor would increase the maximum
power consumption and heat output without having a major performance
benefit. (It would be cheaper, but cost doesn't seem to be a
consideration.)

> Based on the anandtechn opinion/tests, it appears that stopping at 6-8
> cores is better because you benefit from higher clock rate.

Agreed, unless the application will be able to keep more than 8 cores
busy. Some server applications, for example.

I'm expecting Ars Technica to post their 2013 Mac Pro review soon, which
is likely to go further into benchmark comparisons.

This article might also have some useful information, though it doesn't
cover some of the variants Apple is using.

http://www.anandtech.com/show/7285/intel-xeon-e5-2600-v2-12-core-ivy-bridge-ep

> (I am also surprised how well the 2009 Mac Pro fares in his tests -
> although he has an 8 core one, I have only a 4 core one).

Single core performance isn't improving much between generations.

For multicore operations, the 12-core 2013 Mac Pro was about 70% faster
than the 8-core 2009 model.

I saw an article a while ago which did theoretical benchmarks, with
guesses as to which processors Apple was going to use in the 2013 Mac
Pro (the guesses turned out to be right).

http://www.primatelabs.com/blog/2013/11/estimating-mac-pro-performance/

This suggests a 4-core 2013 Mac Pro should almost match a fast 6-core or
slow 8-core 2010 Mac Pro in multi-core tasks; the 8-core 2013 Mac Pro
should beat almost all configurations of earlier Mac Pros.

-- 
David Empson
dempson@actrix.gen.nz

[toc] | [prev] | [next] | [standalone]

#54295

In message <1lev3zd.zfygrj147g0ecN%dempson@actrix.gen.nz> 
  David Empson <dempson@actrix.gen.nz> wrote:
> JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:

>> On 14-01-01 03:07, David Empson wrote:
>> 
>> > You are making false assumptions about cache sizes. In the Intel Xeon E5
>> > v2 series (as used by the 2013 Mac Pro), the 12-core processor has more
>> > than twice as much L3 cache as the 6-core processor. 
>> 
>> In the couple of pages of the anandtech article you pointed to, I didn't
>> see mention of cache differences between the various options.

> Page 4 (CPU Choices) has a table near the top titled "Mac Pro (Late
> 2013) CPU Choices" which has a line "L3 Cache" that lists the cache
> sizes for the four processor options:

> 4 core: 10 MB
> 6 core: 12 MB
> 8 core: 25 MB
> 12 core: 30 MB

>> > Since Apple doesn't make a Mac Pro with two Xeon E5 v2 CPUs, the
>> > question of relative performance doesn't seem relevant except for
>> > academic interest,
>> 
>> It is relevant in terms of evaluating Apple's decision to allocate space
>> for only one chip. In other words, was this done because of Jony Ive's
>> industrial design requirements (beauty/small), or has the chip
>> technology really gotten to a point where there is no longer the need
>> for multiple chips.

> Or keeping within reasonable power/heat requirements. Having two 6-core
> processors instead of one 12-core processor would increase the maximum
> power consumption and heat output without having a major performance
> benefit. (It would be cheaper, but cost doesn't seem to be a
> consideration.)

>> Based on the anandtechn opinion/tests, it appears that stopping at 6-8
>> cores is better because you benefit from higher clock rate.

> Agreed, unless the application will be able to keep more than 8 cores
> busy. Some server applications, for example.

> I'm expecting Ars Technica to post their 2013 Mac Pro review soon, which
> is likely to go further into benchmark comparisons.

> This article might also have some useful information, though it doesn't
> cover some of the variants Apple is using.

> http://www.anandtech.com/show/7286/intel-xeon-e5-2600-v2-12-core-ivy-bridge-ep

The real benchmarks with the MacPro will be the Final Cut benchmarks, as
it appears this machine is designed around making Final Cut as fast as
possible.

The fact is, for normal day-to-day usage, it's not much faster than the
top-of-the-line iMac.

>> (I am also surprised how well the 2009 Mac Pro fares in his tests -
>> although he has an 8 core one, I have only a 4 core one).

> Single core performance isn't improving much between generations.

> For multicore operations, the 12-core 2013 Mac Pro was about 70% faster
> than the 8-core 2009 model.

> I saw an article a while ago which did theoretical benchmarks, with
> guesses as to which processors Apple was going to use in the 2013 Mac
> Pro (the guesses turned out to be right).

> http://www.primatelabs.com/blog/2013/11/estimating-mac-pro-performance/

> This suggests a 4-core 2013 Mac Pro should almost match a fast 6-core or
> slow 8-core 2010 Mac Pro in multi-core tasks; the 8-core 2013 Mac Pro
> should beat almost all configurations of earlier Mac Pros.

And still, it really depends on what you are doing with those cores. If
you are reading email, loading web pages, watching videos, or even most
things that computers are used for you will not notice a difference. If
you are throwing around very large video files and applying complex
transformations then you will certainly see the difference.

I doubt you will even see much difference on the usual sate of
'developer' tasks, like building apps in XCode, for example.

-- 
And sometimes there's a short cut. A door or a gate. Some standing
stones. A tree cleft by lightning, a filing cabinet. Maybe just a spot
on some moorland somewhere... A place where THERE is very nearly HERE...
If some people knew where such a spot was, if they had experience of
what happens when here and there become entangled, then they might - if
they knew how - mark such a spot with certain stones. In the hope that
enough daft buggers would take it as a warning and keep away.  (Lords and
Ladies)

[toc] | [prev] | [next] | [standalone]

#54183

In message <52c37e93$0$9689$c3e8da3$aae71a0a@news.astraweb.com> 
  JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:
> BTWL: Generic question:

> With current memory interfaces, is there any performance differences
> between having 1 12 core CPU vs 2 6 core CPUs ?

Yes. The 12 core has more than double the L3 Cache (25MB instead of
12MB, IIRC)

> Do the current intel processors have shared cache between cores ?

Between cores on the CPU? I believe so. The L2 cache is not shared, but
the L3 is.

> Would 1 12 core processor with 1 shared cache memory be different than 2
> 6 core processors with 2 separate caches (twice the cache size, but not
> shared by all cores)

I suspect the difference of having a single 25MB L3 cache is significant
versus 2 12MB L3 caches, but that's just my feeling.

> If the first 6 cores deal with one application and the second 6 cores
> deal with different app/data, then having separate caches would be an
> advantage since you have overall more cache. Burt if you do a video that
> makes use of all 12 cores, then having a single cache would make sense.

That's not how multi-processor/multi-core systems work, and certainly
not on OS X. GCD makes the decisions on where to send work, not the
program.

-- 
'Now what?' it said.  IT'S UP TO YOU. IT'S ALWAYS UP TO YOU. --Maskerade

[toc] | [prev] | [next] | [standalone]

#54186

In message <slrnlc7o4f.n71.g.kreme@amelia.local> 
  Lewis <g.kreme@gmail.com.dontsendmecopies> wrote:
> In message <52c37e93$0$9689$c3e8da3$aae71a0a@news.astraweb.com> 
>   JF Mezei <jfmezei.spamnot@vaxination.ca> wrote:
>> BTWL: Generic question:

>> With current memory interfaces, is there any performance differences
>> between having 1 12 core CPU vs 2 6 core CPUs ?

> Yes. The 12 core has more than double the L3 Cache (25MB instead of
> 12MB, IIRC)

30MB instead of 12MB, in fact.

-- 
They all have husbands and wives and children and houses and dogs, and
you know, they've all made themselves a part of something and they can
talk about what they do. What am I gonna say? "I killed the president of
Paraguay with a fork. How've you been?"

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | comp.sys.mac.system

csiph-web