Apple On Track to Ship Intel Macs by June 2006

[BGil]

That pic at the bottom shows two 50w (average or 75% capacity) dual core procs running at 1.5Ghz. AMD is selling 1.8ghz Dual cores at 55W absolute maximum, right now. Intel promises to have a dual core 65w proc for desktops and AMD already have 2 different A64 cores that run cooler than that. They also promis single core laptop chips under 15w average and AMD has the Turion MT which runs much cooler than that as well.

Then take into account that all those processors from AMD are 64-bit (20% performance boost) and much faster than the current Intels and future ones as well (the next-gen processors are actually a decrease in performance for Intel). Intel couldn't even beat AMD in 32 bit so once you load 64-bit on to that AMD processor it's no contest.

There’s a reason why AMD told Anandtech they weren’t going to release some :”new thing” to combat Intel; because they released it already.

 

[BGil]

*pounds his head into the keyboard in frustration*

89h3e8934w2q43i8I!I*I*8IIKMUIK43I34IU,KEWR, ,. ,K K,I


God this has to be the most misunderstood "fact" ever to grace the Apple community. The chip in the XBOX is not a G5. Why do you think they can have multiple cores in the thing at such a high freq?

I didn't say it was a G5. It is pretty obvious that it's fairly closely related though. It's close enough that MS sent PowerMacs with modified video cards out as developer boxes and that they ported their DirectX stack to it knowing it would be an easy transition to the Xbox 360 processor.

Carmack said the Xbox 360 run unoptimized x86 code as half as fast as the fastest x86 procs. Nothing Apple sells for anywhere near the price can do that. I imagine it would take very little work on the part of IBM to produce a slightly lower clocked version of that proc that was more tuned for general performance computing. There's no reason IBM can produce great stuff for Nvidia, ATI, Microsoft, Sony, and Nintendo while struggling to provide similar stuff to Apple. For goodness sakes, the Xbox 360 ships with a configuration not too dissimilar from the Mini (HD, RAM,DVD drive, Ethernet, Wifi) plus it has water-cooling, a kick-ass proc and a high end video card... and it's profitable at $400.
It should be obvious that IBM doesn't want to make processors for Apple.

 

[AidenShaw]

time to raise the "troll" flag

Then take into account that all those processors from AMD are 64-bit (20% performance boost) and much faster than the current Intels and future ones as well...

Surely you must know that even the current Intel Celerons are 64-bit...

That "20% boost" applies to all of Intel's chips as well....except for the Dothans that the Lord God Jobs is lusting after!

 

[AidenShaw]

can you explain ISA ??

I didn't say it was a G5. It is pretty obvious that it's fairly closely related though.

Anyone who understands what "ISA" means would realize that otherwise completely unrelated systems that shared the same ISA could be used more-or-less interchangeably.

P3, P4, P5, P6, K6, K7 and Netburst are pretty much unrelated, except for the fact that they all implement the x86 ISA.

 

[BGil]

Time to re-read the thread before you post!

Surely you must know that even the current Intel Celerons are 64-bit...

That "20% boost" applies to all of Intel's chips as well....except for the Dothans that the Lord God Jobs is lusting after!

Current chips are 64-bit but the next-gen chips (the subject of the thread) are not. The chips I was comparing to AMD's current lineup are not.

 

[AidenShaw]

right, I replied out of context

Current chips are 64-bit but the next-gen chips (the subject of the thread) are not. The chips I was comparing to AMD's current lineup are not.

Except for the Dothan, current Intel chips are 64-bit - and except for the Yonah, next-gen Intel chips are 64-bit.

Your 20% performance boost, however, isn't realistic. You have to be running a 64-bit operating system, and be running 64-bit applications.

The 64-bit O/S is there for Windows, but none of the common applications are 64-bit - so you won't get the 20% in real life.

Apple is 32-bit only for OSX86 - so it's not possible to get the 20% at all!

 

[EvilMole]

as I think they will help break Intel's "losing streak" and set them apart once again. 😉

Hmmm yeah. 81.7% of the market, compared to AMD's 16.9%. $7.5 billion profit in 2004, compared to AMD's $91 milllion.

Some losing streak 😉

 

[~Shard~]

Hmmm yeah. 81.7% of the market, compared to AMD's 16.9%. $7.5 billion profit in 2004, compared to AMD's $91 milllion.

Some losing streak 😉

I meant from a technology standpoint wrt to the P4. 😎

 

[BGil]

Except for the Dothan, current Intel chips are 64-bit - and except for the Yonah, next-gen Intel chips are 64-bit.

So far all 64-bit chips from now and until after the Yonah is replaced by the Merom are Netburst chips. In other word's think Prescott and hotter. So you're looking at the second half of 2006 at least, before energy effcient 64-bit chips come out of Intel. As everyone here is so sold that Apple isn't going to use anything based on Netburst (I think they will) then that means 3rd or 4th quarter 2006.

Your 20% performance boost, however, isn't realistic. You have to be running a 64-bit operating system, and be running 64-bit applications.

The 64-bit O/S is there for Windows, but none of the common applications are 64-bit - so you won't get the 20% in real life.

That's not true. I'm sure you've seen the After Effects scores from digitalvideoediting.com and other places, right? Most applications that use globs of memory like AE and Photoshop see improvements even in 32-bit on 64-bit Windows.

 

[AidenShaw]

So far all 64-bit chips from now and until after the Yonah is replaced by the Merom are Netburst chips. In other word's think Prescott and hotter. So you're looking at the second half of 2006 at least, before energy effcient 64-bit chips come out of Intel. As everyone here is so sold that Apple isn't going to use anything based on Netburst (I think they will) then that means 3rd or 4th quarter 2006.

IMO, Apple will do a Netburst-based system (other than the one they already have) in order to have a multi-CPU system for developers. It won't be their mainstream, and it will be sold in parallel with PPC Macs.

That's not true. I'm sure you've seen the After Effects scores from digitalvideoediting.com and other places, right? Most applications that use globs of memory like AE and Photoshop see improvements even in 32-bit on 64-bit Windows.

Interesting - the charts at
http://www.creativemac.com/articles/viewarticle.jsp?id=32951-1 do clearly show an advantage to running 32-bit apps on 64-bit windows. Nice to see that quad CPU Windows desktops are out on the market, too.

64-bit XP is actually based off Windows 2003 - so inside it's quite a change from 32-bit XP. In particular, it has better scheduling support for HT and for the NUMA memory on the Opteron.

However, the dual-core Pentium has similar speedups from 64-bit, so it looks like either memory management improvements in XP64/2k3 are responsible, or perhaps that significant amounts of processing are occurring in system libraries that are 64-bit.

Of course, it's merely an academic issue, since OSX86 is 32-bit only for the announced future. In my mind, that's the really puzzling question - why didn't Apple wait for Merom so that they could do a pure 64-bit only OSX86. Why put the developers through both an Intel 32-bit port and then later require another port to 64-bit x64? Why, Steve?

(Also, why port to Dothan rather than Yonah, since everyone expects Yonah to be here long before any MacIntels....)

Very interesting info about 64-bit XP, thanks for the tip.

 

[BGil]

64-bit XP is actually based off Windows 2003 - so inside it's quite a change from 32-bit XP. In particular, it has better scheduling support for HT and for the NUMA memory on the Opteron.

Yep. XP SP2 has NUMA support as well but it sucks from what I can tell. I can't find a single real-world benchmark where it actually improves anything. I'm so glad Vista is built on Win2k3.

 

[AidenShaw]

Yep. XP SP2 has NUMA support as well but it sucks from what I can tell. I can't find a single real-world benchmark where it actually improves anything. I'm so glad Vista is built on Win2k3.

Vista is built on the main Windows codebase - XP is a snapshot of that codebase from around 2002.

Development continued, and Windows 2003 is a later branch, and Vista is the current main path.

XP64 is a modification of the Win2k3 branch, not an update to the older XP branch. (XP SP2 does pull a few things from the later branches back to the older branch.)
______________

I thought of another possibility for the XP64 performance advantage - related to system code.

All 64-bit x64 systems have at least SSE2 support. It's possible that system libraries and APIs in XP64 are using SSE2 far more widely - since it's guaranteed to be there. XP might be using simpler instructions from the base set.

This could make a big difference in data moving tasks - like getting file data from the file system cache into the users' buffers.

It would also help floating point, since SSE2 is faster at scalar floating point than the standard x87 FPU.

 

[widgetguy]

G5 INTEL MAC here I come hehehe!

 

[BGil]

Vista is built on the main Windows codebase - XP is a snapshot of that codebase from around 2002.

Development continued, and Windows 2003 is a later branch, and Vista is the current main path.

That's not right. The Longhorn "reset" moved everything to Windows Server 2003 SP1 code just like XP64 bit.

The code base for Longhorn will be the same as Microsoft is using for the Service Pack 1 release of Windows Server 2003, an update slated for the first half of next year. Allchin said the decision to use that code base was made some months ago.
http://beta.news.com.com/Microsoft+revamps+its+plans+for+Longhorn/2100-1016_3-5327150.html?
http://en.wikipedia.org/wiki/Windows_Longhorn

http://channel9.msdn.com/ShowPost.aspx?PostID=119593#119593

 

[AidenShaw]

really a minor semantic difference

That's not right. The Longhorn "reset" moved everything to Windows Server 2003 SP1 code just like XP64 bit.

It can't be the same branch, or it would be the same system.

I think that if you read more carefully, you'll realize that the "reset" moved much code from the 2k3 SP1 branch to the mainline, and sidelined a lot of previous vista development work.

I think that my statements are still accurate - 2k3 is a branch, SP1 is on that branch, and vista is the mainline. 2k3 SP2 will be further out on the 2k3 branch than SP2 - but not on the mainline.

My description fits even if they completely pruned the old vista branch at the "reset" and started the vista mainline from SP1. The mainline is vista, 2k3 and SP1 would be older points on the mainline, further 2k3 development would branch from the SP1 point. (If this is the case, certainly a lot of code from the pruned longhorn branch would be pulled into the vista mainline.)

 

[BGil]

It can't be the same branch, or it would be the same system.

I think that if you read more carefully, you'll realize that the "reset" moved much code from the 2k3 SP1 branch to the mainline, and sidelined a lot of previous vista development work.

I think that my statements are still accurate - 2k3 is a branch, SP1 is on that branch, and vista is the mainline. 2k3 SP2 will be further out on the 2k3 branch than SP2 - but not on the mainline.

My description fits even if they completely pruned the old vista branch at the "reset" and started the vista mainline from SP1. The mainline is vista, 2k3 and SP1 would be older points on the mainline, further 2k3 development would branch from the SP1 point. (If this is the case, certainly a lot of code from the pruned longhorn branch would be pulled into the vista mainline.)

That's not at all what they are saying. Win2k3 SP1 is the main branch from here on out. The XP SP2 code is dead. Win2k3 doesn't have a SP2 it's got a second release called R2 but that it just some Vista/Longhorn compatible stuff and some interface enhancements being shipped with Win2k3. It's not a separate codebase at all.

Both Windows Vista, Windows XP Pro 64-bit, and Longhorn Server are based on Windows Server 2003 SP1 code. They didn't move win2k3 SP1 code to the main line, they ditched the XP SP2 mainline completely. That's part of the reason why it took so long to release XP 64-bit and why Vista Beta 1 is so much different than Longhorn Alpha.

 

[AidenShaw]

let's just disagree

That's not at all what they are saying. Win2k3 SP1 is the main branch from here on out. The XP SP2 code is dead. Win2k3 doesn't have a SP2 it's got a second release called R2 but that it just some Vista/Longhorn compatible stuff and some interface enhancements being shipped with Win2k3. It's not a separate codebase at all.

Somehow I get the feeling that you've never used a modern codebase maintenance system that allows multiple branching, merging, and similar features necessary for large scale system development and maintenance.

How does MS issue a patch to XP SP2? They check out the *not*dead* XP SP2 version of the source files, fix them, and check them back into the XP SP2 branch. If needed, the code maintenance tools will help propogate the fix to other branches (maybe the same fix needs to be in w2k3 R2, and in the mainline Vista).

How do they patch 2k3 SP1 ? Same thing - check out the modules from the 2k3 SP1 branch, fix, and update. Maybe merge (to XP64, Vista,...).

They didn't move win2k3 SP1 code to the main line, they ditched the XP SP2 mainline completely.

When XP shipped, it became a branch. XP SP* are developments on the branch. XP SP2 is based on XP SP1, with obvious merges back to the mainline and other branches.

That's part of the reason why it took so long to release XP 64-bit and why Vista Beta 1 is so much different than Longhorn Alpha.

XP.64 had to wait for w2k3 SP1 - that has been widely reported.

Vista Beta 1 is different due to the "reset"...

Let's just disagree - because all we're arguing about is the semantics of what "mainline" and "branch" mean, not really any distinct change in the parentage of the code.

Note the following from http://www.softpanorama.org/Unixification/index.shtml:

"Then there are of course a lot of improvements on the device drivers, the way Windows handles drivers, wireless conectivity etc. The main code branch is built upon Windows Server 2003 SP1."