Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Snow Leopard in September, Xserve in June, Mac Pro Speed Boost, and More?
- Thread starter MacRumors
- Start date
- Sort by reaction score
So... no quad core MacBook Pros till Q4 at earliest? Sigh... thats all I want... that and 8 gigs of ram...
I am just glad that after the announcement by Steve Jobs several years ago that the IBM Power PC G5 will be 3 ghz, that now sometime by late 2009, we will have an Intel Nahelem 3+ ghz chip.
Better never than late I always say.
Rocketman
Better never than late I always say.
Rocketman
Yup. It was probably about three months too. I'm really hopping they add the 3.2 GHz soon. I'm looking to buy and I'm willing to pay for it. I really don't think this will be a bump, just a new BTO option. I'm assuming the 3.2 GHz isn't there because they couldn't get the chips and didn't want to backlog a ton of orders for a couple months.
You seem to be assuming that all the northbridge did was talk to memory.
If you look at several Northbridge chips they also have duties talking to highbandwidth PCI. From example http://en.wikipedia.org/wiki/Direct_Media_Interface has a few links to some Intel Northbridge chips.
The Nehalem era X58 is still a northbridge. It didn't disappear. It has fewer duties but it isn't gone. Primarily, the X4x series also have the PCI-e 8x connectivity in addintion to memory.
Wouldn't putting your graphics on the other side of DMI limit your bandwidth. Granted not going to have multi GPU kinds of data movement but this is same southbridge where all the other I/O is coming from too. Putting EVERYTHING (graphics + graphics memory , disks , etc. ) on the other side of DMI seem to unbalance the system significantly. Doing heavy disk I/O a the same time as heavy graphics going through the same pipe ( a PCI 4x sized one) ? [ Like pull and decode HD video for example. ]
Likewise, the 9400M has a huge chunk dedicated to graphics. Those Watts aren't going to disappear from a system that doesn't have one.
Likewise for AMD ( which has had emedded memory controllers for a even longer time)
http://en.wikipedia.org/wiki/Comparison_of_AMD_chipsets
Note the column that outlines the name of the Southbridge. Some of nVidia design might have collaspsed into one ( or they just sell them bundled. )
Does Nvidia have ability to do DMI ( or just the CPU/Northbridge one)? I thought it is the QPI interface that they are suing on?
The memory controllers are integrated in the CPU; not the Northbridge.
Discrete GPUs hooked to what? If there is only a Southbridge left you'd have to hook the those to the Southbridge. So some discrete GPU that gets some fraction of a PCI-e 4x worth of bandwidth.
Or maybe even nothing hidden. (of course there is something glitzy hidden so can get oohs and aahs at WWDC's demo). It isn't like Mac OS X hasn't been delayed before due to additional resources needed to get iPhone OS out the door.
October seems reasonable if want to actually get lots of developer feedback about their applications before going out the door. Folks with lots of extra resources can spare folks to do builds on early OS build, but if want folks to get serious would have to tell them when the OS is mostly complete (no more curve balls coming that might effect their code.)
If freeze in September can go build DVDs and packages. 3 months after WWDC is September. Otherwise if folks find broken stuff at WWDC or shortly after not really going to have much time to ship fixes and get back confirmations (of both the fix and no regressions).... assuming that a few of this initial fixes don't work and have iterate some.
Redoing Quicktime is a crossplatform situation also. Probably need to get Quicktime X working on Windows 7 too. So can release at same time.
Plus whatever Windows 7 hype blitzkriegh Microsoft puts on.... best not to compete head-on-head with that.
Was planning to buy a MBP this summer... oh well... guess it will have to wait until November now.
I guess if they still had "some" of the Shake dev team,maybe they could do a kind of service update.I think then though we know they are working on Phenomenon.
October is in fact reasonable though I don't think we'll have to wait that long. Apple is moving along nicely with a new ADC 10.6 seed with 64-bit kernel support.
Let's see...April May June (WWDC could deliver feature complete SL) then another 2-3 months of polishing sounds about right. I'm guessing that Aug/Sept Golden Master dates are probably correct. If Apple's got the 64-bit kernel ready for testing most of the big bugs should be gone by WWDC.
The rumored timeline looks feasible, no?
So not feature complete till WWDC/June, but spilling a bit more this month.
Attachments
I think we have a misunderstanding of what Nehalem chips we are talking about. First for clarity Nehalem is not a chip it's an architecture. Specific chips include Bloomfield, Gainestown, Clarksfield, and Arrandale (which is really a 32nm shrink as part of the Westmere architecture update).
Current Nehalem chips are Bloomfield used in the Core i7 and the Xeon 35xx series (in the single processor Mac Pro) and Gainestown. These both feature QPI links and don't have an integrated northbridge only an integrated memory controller. They require a separate northbridge now called an I/O hub that contains PCIe links.
Mainstream desktop and all mobile chips will come in the form of Clarksfield and Arrandale which actually integrated the whole northbridge. This includes memory controller, PCIe links, and an IGP in the case of Arrandale. Discrete graphics cards can connect directly to Clarksfield and Arrandale through standard PCIe links as always. These chips don't have QPI since the northbride is already onboard. They only have a DMI link which is low-bandwidth and has always been used to connect southbridges.
And now onto the nVidia-Intel licensing issue. My view is that the entire thing is pointless from an implementation perspective.
Mainstream desktop and mobile chips already have the northbridge onboard. I don't see what advantage nVidia's southbridge could have over Intel's southbridge when their feature-sets are pretty much standard anyways. If anything, Intel's southbridge for mobile Nehalem would probably be the first to adopt USB3.0 seeing that Intel lead it's development.
The entire nVidia-Intel Nehalem chipset debate is a non-issue for Apple and Macs. No mobile Nehalem chips will have QPI links or need a northbridge and therefore will not support an nVidia IGP. It doesn't matter whether nVidia wins a license or not, there aren't the QPI links to connect it.
nVidia has never manufactured chipsets for Xeons and seeing that the whole point of Xeons and it's higher prices is tight platform integration for guaranteed stability over performance, there is little advantage with going with nVidia. It's not like the Mac Pro needs an IGP. I doubt nVidia is arguing that their license extends to Nehalem Xeons seeing that it never extended to any previous Xeons to begin with.
The only narrow Nehalem niche for nVidia chipsets is high-end desktop chips which don't have an integrated northbridge and have QPI links. However, Apple has never used Intel desktop chips. And the idea of better integration of GPUs and chipsets for Hybrid Power is now dead on desktops, with nVidia no longer supporting Hybrid Power with the GTX285 and GTX295, while the previous GTX280 had support.
The nVidia-Intel chipset licensing war is largely irrelevent since they are going to end up fighting over making southbridges which is hardly a product differentiator. Instead of focusing on making an IGP chipset with no processor to connect to nVidia and Apple should be focused on powerful, low-cost GPUs which would be better than the IGPs that Intel can come up with anyways.
If Clarkfield is circa October not so sure this "end of 2009" is really going to hold up ( expect for getting the hardware vendors numerous copies for extensive testing.)
http://www.ubergizmo.com/zoom.php?dir=2009/2/Intel-Westmere/&page=1
Arrandale has fewer cores in part because it nukes 2 cores to move the digital portion of graphics ( and PCIe interface ) onto the die. That somewhat solves the problem of moving the GPU to the other side of a relatively limited DMI interface. However, now need to get the video signal out of the package containing the CPU too. So pulling the graphics apart a bit.
http://www.ubergizmo.com/zoom.php?dir=2009/2/Intel-Westmere/&page=5
So it isn't till the next shrink that get close to the Northbridge really disappearing and that is at the sacrifice of a couple of cores.
Which makes the Clarkfield somewhat of an oddball chip. Presumably "need" 4 cores on your laptop because need high performance, but then choke down the on the I/O pipe. Or they are assuming that mostly reading everything out of memory so most of of the I/O pipe can be used for graphics.
Will waiting another 8 months be worth an upgrade to the Nehalem Processor type?
How much of a difference do you think it will make in the notebooks?
How much of a difference do you think it will make in the notebooks?
http://images.brighthub.com/B2/4/B2437A8CCD3F0A2194EB3D5ED8D618955F892A4F_large.jpg
Maybe this will make things more clear for you. Clarksdale is 4-cores and integrates the whole northbridge. Memory and PCIe graphics will connect directly to Clarksdale. There is only a DMI link for the southbridge.
HardMac posted a collection of rumors they've received. They don't seem particularly confident about the information but do provides some believable tidbits:
[snip]
- Final Cut Studio 3, DVD Studio Pro, Shake, Logic updates.
Shake was EOL'ed many years ago. Its Captain Awesome, Ron Brinkmann, left for the Foundry where he worked on Nuke. Nuke is orders of magnitude superior to Shake (except for the paint tools & the price).
They COULD be talking about Phenomenon - the rumoured Shake successor - but that project seems so unlikely to me.
Predicting the other updates its a bit of a truism. They're due for updates.
Final Cut Pro is ALREADY compatible with the RED cameras - unless they're specifically talking about Scarlet and Epic, but even that seems 'well, duh' as they'd be based on the same redcode codec.
End broadcast.
Quote:
Originally Posted by gmcalpin View Post
In the bank, you'll earn a massive 1% interest. If you're lucky.
Hilarious!
Originally Posted by gmcalpin View Post
In the bank, you'll earn a massive 1% interest. If you're lucky.
Hilarious!
Intel does some funky things with Mutli chip modules ( e.g., throw to dual cores out there as a quad core ). That smells much more of Northbridge + die in a MCM where you just can't externally see the QPI between the Northbrdige/IOHub and "CPU" than they actually integrated. [edit: this seems to support that. http://www.theinquirer.net/inquirer/news/926/1050926/intel-talks-32nm-westmere ]
I didn't see the earlier parts about having an "on board" PCIe that makes the Clarksfield not look so oddball. I saw intel slides on the Westmere stuff.
However, finally tracked this one down.
http://en.expreview.com/2008/07/16/intel-roadmap-indicates-multiple-havendale-incoming-q309.html
Still you are only getting a very limited subset of PCI lanes. ( what if you need a fast Ethernet / RAID / etc. card or something. ) Either DMI has more pep or these will be limited to machines where folks don't push them very hard. Laptops that's OK but this is also the desktop.
The Core5 chips still have PCIe slots... the report mentions Kings Creek having 2 PCI 8x http://www.dailytech.com/Intel+Targ...on+With+Core+i5+New+Chipsets/article14228.htm
So if the Northbridge is on the CPU package why are there more on this P5x chip? perhaps there is just one PCI 8x slot just for graphics? I can see 8 wire/pins for a slot. What not so sure about is how get 2-3 8x slots coming out along with the rest.
Doesn't make much sense to have those on the other side of DMI (if it is bascially a 4x-ish )
And now onto the nVidia-Intel licensing issue. My view is that the entire thing is pointless from an implementation perspective.
....
nVidia has never manufactured chipsets for Xeons and seeing that the whole point of Xeons and it's higher prices is tight platform integration for guaranteed stability over performance,
[/QUOTE]
But to some sense Intel was standardizing on QPI for a situation similar to what AMD did with HyperTransport.
Not too surprising that Intel would flip the game to a MCM approach to squeeze both AMD/ATI and nVidea more completely out of the integrated graphics space. They've gotten their butts kicked so far in that space when it comes to performance so just eliminate the competition. Then they won't look as bad.
Not so sure the "southbridges" are going to be so low performing unless the multislot PCIe boxes disappears. ( the desktops all move to iMac style kinds of expansion. )
What is perverse is that if Intel goes to a model where they put a IGP on every single option you can choose, even if nVidia does come up with a better solution Intel still gets paid for the IGP that you aren't going to use.
Just gives them more money to keep shooting at Intel.
Hope that OpenCL allows the GPUs to compete with Intels CPU as far as where folks put their additional money for "more power". Otherwise very long term going to end up with a Duopoly. Or perhaps folks just move off to a newer platform where there is more competition.
If I was nVidia I'd try to repair the AMD relationship rather than try from keeping Intel from slamming the door shut in their face. HT (or the equivalent) to the their classic kinds of Northbridge / Southbridge set up is much more open to competition and gets you better classic desktops. ( for the all-in-one market the intel will probably work better longrun... if AMD gets back in the game. )
The entire Northbridge is 2 slots?
This one has up to 4.
http://en.wikipedia.org/wiki/File:X58_Block_Diagram.png
As commented in another reply ... perhaps all mainstream PCs are heading to the iMac and/or 10 lb laptop model. But wonder where all the "I need an expandable minitower" folks are going to go then.
edit: More think about this looks like they are farming NB functionality partially to the CPU package and partially to the SB package. The chip package count is going down, but the actual entities as chip dies not so much.
Mobile Nehalem!
Will Apple get early access to the mobile Nehalem chips much like they did with the Nehalem Xeons?
Just dreaming...
Will Apple get early access to the mobile Nehalem chips much like they did with the Nehalem Xeons?
Just dreaming...
Mobile Nehalem!
Will Apple get early access to the mobile Nehalem chips much like they did with the Nehalem Xeons?
Just dreaming...
Not likely. Read somewhat ( can't remember) that Apple got Xeons for Mac Pros early because the volume of expected Mac Pros sold is so low they could keep up with Apple's initial demands with pre-full scale production abilities.
Apple sells laptops in much higher volumes. Maybe if the 17" went early, but since it came out last in the last update round that doesn't seem likely since it is least due for a refresh.
You are making still far more complicated than it is, probably because you aren't clear what the capabilities of the traditional northbridge and southbridge are.The entire Northbridge is 2 slots?
This one has up to 4.
http://en.wikipedia.org/wiki/File:X58_Block_Diagram.png
As commented in another reply ... perhaps all mainstream PCs are heading to the iMac and/or 10 lb laptop model. But wonder where all the "I need an expandable minitower" folks are going to go then.
edit: More think about this looks like they are farming NB functionality partially to the CPU package and partially to the SB package. The chip package count is going down, but the actual entities as chip dies not so much.
Even in Pre-Nehalem chipsets, not all PCIe lanes come from the Northbridge. The PCIe links from the northbridge are traditionally for graphics cards, ie. high-bandwidth applications. And links don't necessarily equate to slots. Most northbridges have 16 links so that motherboard makers can make 1 PCIe x16 slot or 2 PCIe x8 slots. High-end northbridges have 32 links that can be implemented as 2 PCIe x16 slots or 4 PCIe x8 slots.
This is functionality will be replicated in Nehalem chips. Specifically Bloomfield and Gainestown are high-end chips and so they have external northbridges with at least 32 links just like previous high-end chipsets. Clarksfield and Arrandale which are mainstream processors will have integrated northbridges with 16 links, just as was common before.
Additional PCIe links for expansion cards like a sound card have always been provided by the southbridge connected through DMI. This doesn't change with Nehalem.
With Nehalem, you aren't losing I/O bandwidth or functionality in # of slots. Performance-wise you are gaining from reduced latency. The only thing you are losing is some flexibility since you can't really select your northbridge anymore. But realistically, mainstream Nehalems are already coupled with a mainstream chipset, so it's only the people who buy a high-end chipset with a mainstream CPU looking to upgrade the CPU later that are disadvantaged, which I guess Intel figures isn't the majority.
In regards to being forced to pay for Intel IGPs, the IGP will only be included in Arrandale which is a dual core part and would be pretty low-end by the time it's released in 2010. In that segment, IGP usage would be common anyways so for most people it isn't a big deal. And for people that don't, Intel's IGPs do support GPU switching with discrete nVidia and ATI GPUs, in notebooks at least, so you do get the functionality of using the IGP to save power and switching to a discrete for more performance. So it isn't completely useless. An Intel was part of the working group that ratified OpenCL and they have pledged to support it in products sooner or later. If the IGP in Arrandale does support OpenCL, you could use it as a processing target while your discrete GPU focuses on graphics related stuff or vice-versa.
