Intel Merom (64-bit Mobile Processor) Due in September?

[Anonymous Freak]

According Intel's roadmap, it's really Conroe, which is the star (Spring IDF 2006).

According to Intel, Conroe will give 40% more performance while using 40% less power.

Merom, will be "at a constant battery life compared to the Core Duo T2600", while providing 20% more performance.
That means, according to Intel, battery life will not be any 'better' by itself than that of the current Yonah, while we'll see a moderate performance increase.

The basic comparison is that the existing Pentium 4/Pentium D is a power hog. Core Duo isn't. That's because Core Duo is based on a long-established core that was designed from the ground up for low power consumption. Pentium 4 wasn't. Unfortunately, while the P4 core can scale well in terms of clock speed, it has run into a THERMAL barrier. People using extreme cooling have gotten Pentium D's up to over 4 GHz no problem; but it requires more extreme cooling than the liquid cooling used in the G5.

So what is Conroe/Merom? It's a redesigned core based loosly on the same core as Core Duo! Conroe will basically be a Core Duo with looser power/thermal requirements to it can scale to higher clockspeeds on the desktop where every last Watt isn't as important; while Merom is basically just a simple upgrade of Core Duo.

As for Merom being pin-compatible with Core Duo? While the physical socket may be the same, and it may even have the same power requirements, it will almost certainly require a new chipset. Intel has been notorious for this in the past. (The Pentium 4 used Socket 478 for a long time, and had three major upgrades to the core in that time. Each core upgrade required a new chipset, even though the physical socket was technically 'pin compatible'.)

So I wouldn't hold your breath waiting to hear that the just-released mini can be upgraded to a Merom. If it does, take it as a happy surprise.

 

[milo]

As for Merom being pin-compatible with Core Duo? While the physical socket may be the same, and it may even have the same power requirements, it will almost certainly require a new chipset. Intel has been notorious for this in the past. (The Pentium 4 used Socket 478 for a long time, and had three major upgrades to the core in that time. Each core upgrade required a new chipset, even though the physical socket was technically 'pin compatible'.)

Intel is saying Merom will run on the same chipset.

http://www.xbitlabs.com/news/cpu/display/20051019183430.html

 

[Anonymous Freak]

Intel is saying Merom will run on the same chipset.

http://www.xbitlabs.com/news/cpu/display/20051019183430.html

Double check your linked article. It says that it is 'pin-compatible', which I explain in my previous post. The closest to saying that it runs on the same chipset is some pure speculation from X-bit:

... which may mean that systems originally designed for Yonah may be upgraded to support the future chips by installing a new BIOS. Provided that code-named Napa platform intended for Yonah also supports Merom, the latter should operate using 667MHz Quad Pumped Bus.

To clarify, 'Napa' is the 845M chipset family developed for the Yonah (a.k.a. Core Duo) processor. Napa has already been released, as it's in the MacBook Pro, iMac, and Mac mini right now. They are saying that IF the current chipset was built to support Merom, then it means that Merom would likely use the same bus speed. It is not a definite yes from Intel, it is pure speculation in an attempt to figure out how fast the bus speed of Merom will be. It's all speculation at this point. I honestly think that the bus speed for Merom will go up to at least 800 MHz; but that's not based on any inside information. (I'll try to get some inside information later this week.)

And 'pin-compatible' has been done by Intel in the past with chipsets, too. It may require a new chipset to support Merom, but the chipset may very well be pin-compatible with the old chipset, meaning that a manufacturer just has to swap out both chips with minimal effort. Easy for a manufacturer, but still impossible for an end-user.

P.S. The X-bit article is from last October, it's outdated now. I'm trying to find current info, but we'll have to wait for more complete reports from IDF...

 

[paz117]

whatever those specs are hot

 

[airkarol]

Can't wait all of the panic-induced "should I sell my current macbook/powerbook/whatever for the new Macs?" threads.

im way ahead of everyone on that haha, except for the thread part, i sold it around a month ago, and got $1525, which seems pretty good.

 

[wrxsti86]

So aft all .... NOT really DUO

http://www.amd.com/us-en/assets/content_type/DownloadableAssets/top_10_v1.pdf

What can I say?

 

[DVK916]

Intel is saying Merom will run on the same chipset.

http://www.xbitlabs.com/news/cpu/display/20051019183430.html

It doens't say that, it says it uses the same pin not that the chipset will be compatible. I doubt Napa will work with Merom, there is a reason intel is designing Napa64 Chipset. because the current Napa chip set doesn't support 64 bits.

 

[sjl]

So besides 64-bit based software and more possible RAM, does 64-bit capabilities make the chip any faster, or does that depend just on clock speed/cache speed/FSB speed?

It's already been covered by others, but to give you a bit more detail: there are currently three major platforms available in both 32 bit and 64 bit forms: SPARC, PowerPC, and x86. Sparc has 32 general purpose registers. PowerPC also has 32 general purpose registers. For both those CPUs, the number of available registers doesn't change depending on whether you're in 32 bit or 64 bit mode.

x86 has just 6 general purpose registers in 32 bit mode: EAX, EBX, ECX, EDX, ESI, and EDI. There's also a couple of stack registers, which I'll ignore, as they're tied up with other duties, and can't really be used for general purposes. ESI and EDI are often used for memory indexing, which means you can't really treat them as general purpose, unless you're very careful about the assembly opcodes you emit.

When you look at x86-64, though, AMD added another eight general purpose registers. These are unavailable in 32 bit mode -- only code compiled for 64 bit mode can use them.

In terms of speed, there is a definite heirarchy in the computing world. Slowest to access (locally; I'm ignoring networks) is the hard drive. Next is the main memory. Then the various levels of cache, and finally, the registers, which are physically part of the main CPU. So if you have a function that manipulates a lot of data, it's going to benefit from having access to a lot of registers.

With PowerPC and Sparc, because the number of registers doesn't change from 32 bit to 64 bit mode, there's actually a slowdown when going to 64 bit mode: addresses are twice the size, so need more space in the cache, and so on. As a result, you'll only see apps compiled to 64 bit mode on these platforms when they need to be. On x86, though, the additional registers are a major win, which offsets -- in a major way -- the performance hit from the greater amount of data being thrown around.

So: on PowerPC, 64 bit only really buys you more memory, at the cost of a little speed; the loss of speed is insignificant for those apps that need the memory (as they'd otherwise be swapping data in and out of memory to keep it within the limits), but is noticeable otherwise. On x86, 64 bit buys you both more memory, and more registers, which makes it a win for every app.

Me? I have a PowerBook G4 (1.25), and a first-release Mac Mini. I'll be upgrading my nearly three-year-old PowerBook when Apple releases a 64 bit x86 laptop for just the reasons above; I can limp along with the G4 until then. OS X may not be compiled in 64 bit mode yet, and hence won't really take advantage, but long term, it's going to be a better buy. The fact that it's going to take me all that time to save up the necessary dollars is just a side point. I'd advise anybody that can hang fire until the 64 bit laptops come out to do so, as well (whilst acknowledging that some people can't do that; for them, the current MacBook Pro is a decent buy.)

 

[MarkCollette]

And to think one day I'll be stairing at the new XServes with (Woodcrest?). *shudder of joy*

Please don't throw your XServe down the stairs.

I was skeptical at first but it does seem that they are increasing specifications at an alarming rate, when compared to PowerPC which as far as I know has not progressed much at all?

Since the MacBook Pros were marketed at one speed, and then shipped at a faster speed, a difference that eclipses G4 speedbumps, I'd say they already have a better track record for improvement

I wouldn't hold your breath - just because Merom might be available earlier now than expected, that doesn't mean you'll see it in a MacBook anytime soon. Look for the initial batch of Intel iBooks (MacBooks) to sport similar configurations to that of the new Mac mini (1.5 Solo, 1.66 Duo), and look for the MacBook Pros to receive the Merom well before the MacBooks ever will.

Maybe. I guess it comes down to cost, but since Merom will have better battery life than Core Duos, it's probably more likely that MacBooks will get them soon after MacBook Pros, but just lag in MHz. Nah, who am I kidding. I can totally see Apple shooting itself in the foot just to protect "Pro" sales.

Early "retirement" for 32-bit Intel support, as companies realize that they won't make enough money from 32-bit MacIntels to cover the costs of the extra fat. Rapidly dropping resale value for Yonah as soon as Merom appears, and especially when OSx64 shows up.

I think that most companies, when trying to reduce the number of versions of software, and faced with the choice of less performance and more potential users, versus more performance and less potential users, will inevitably choose more paying users.

 

[AidenShaw]

Originally Posted by AidenShaw
Early "retirement" for 32-bit Intel support, as companies realize that they won't make enough money from 32-bit MacIntels to cover the costs of the extra fat.​

I think that most companies, when trying to reduce the number of versions of software, and faced with the choice of less performance and more potential users, versus more performance and less potential users, will inevitably choose more paying users.

But, "better performance" can mean "more paying users" - especially if your competitors start advertising how much faster they are on the latest 64-bit PowerMacIntels. (Remember the typical 20% performance improvement just from recompiling for 64-bit....)

Your reasoning makes sense, but you also need to consider that there really won't be that many MacIntels sold during the "9 months of Yonah" - compared to the number sold this fall after 64-bit chips are available on every MacIntel from the low end to the high end. And, the MacIntels that are currently being sold are primarily lower-end systems - the 64-bit systems will be higher-end.

It's all speculation at this point, and Apple could change the equation by refusing to actually support 64-bit in any meaningful way (like they've done with Tiger - the 64-bit PPC support is lame).

I'll stand by my prediction/opinion that the Yonah systems will be quickly marginalized by Merom in the current lines, by Conroe in a new mini-tower line, and by Woodcrest in the full PowerMac tower line.

 

[MarkCollette]

But, "better performance" can mean "more paying users" - especially if your competitors start advertising how much faster they are on the latest 64-bit PowerMacIntels. (Remember the typical 20% performance improvement just from recompiling for 64-bit....)

I totally agree that this will create pressure to add 64 bit support. I just would be surprised at companies that use this as a justification to drop 32 bit support.

But of course, all it takes is for that one peice of software that you totally rely on, to drop 32 bit support, and then bam, you're screwed.

I'll stand by my prediction/opinion that the Yonah systems will be quickly marginalized by Merom in the current lines, by Conroe in a new mini-tower line, and by Woodcrest in the full PowerMac tower line.

Ahh you slipped in a new Apple product rumour/prediction there I really really wish that Apple would release something like a double height mini, that would use desktop parts, and maybe even allow for GPU upgrading. THAT would be a real switcher machine.

 

[Anonymous Freak]

It's all speculation at this point, and Apple could change the equation by refusing to actually support 64-bit in any meaningful way (like they've done with Tiger - the 64-bit PPC support is lame).

What do you mean that the 64-bit PPC support is lame?

You can have an app address more than 4 GB of RAM, you can have an app use 64-bit wide data. How much more '64-bit' do you get? Remember, on PPC, the chip can mix 32-bit an 64-bit seamlessly. There is no reason to run the entire OS in 64-bit, only the bits that deal with 64-bit data structures, or memory management.

On Intel, there will be. You take a severe performance hit for trying to switch between 64-bit mode and 32-bit mode. So on Intel, it WILL make sense to run everything in 64-bit; not just for the 20% speed increase, but to cut down on mode switching.

 

[Val-kyrie]

Battery Life and LEDs

No unfortunately, the CPU while a major drain isn’t the only power hungry device. Memory, GPU, Display, media, they all suck power... That said, we should still see substantial increases in battery life.

Actually, Apple (and other pc manufacturers) could boost battery life in laptops by implementing LED technology in their displays. LEDs (compared to current LCDs) are brighter, consume less energy, and have a more dynamic color range. A couple laptops already use LED technology (can't find the link at the moment). This would excite me.

 

[AidenShaw]

What do you mean that the 64-bit PPC support is lame?

You can have an app address more than 4 GB of RAM, you can have an app use 64-bit wide data. How much more '64-bit' do you get?

Write an app that uses 64-bit addressing and has a GUI or uses any of the Cocoa frameworks. You can't.

Windows 64-bit doesn't force you to run a 64-bit app from a command line in a terminal window - Apple does.

That's lame.

ps: Windows 32-bit can use 64-bit and 128-bit wide data no problem. "64-bit" means that an app has flat addressing for more than 32-bit (4 GiB) addresses. 64-bit wide data has been around since FORTRAN appeared during the middle ages.

 

[sjl]

Write an app that uses 64-bit addressing and has a GUI or uses any of the Cocoa frameworks. You can't.

Windows 64-bit doesn't force you to run a 64-bit app from a command line in a terminal window - Apple does.

That's lame.

You obviously don't understand how Unix works (or what Apple's intent is). The idea is very straightforward: write a nice GUI that displays whatever needs displaying. When you need to do work that requires the 64 bit addressing, you launch, from within the 32 bit "shell", the 64 bit code as a subprocess. Unix has a very nice IPC (inter-process communication) framework; OS X, having all the Unix APIs available, gets that free.

That way, you get the best of both worlds: the GUI stays uncluttered with 64-bit address spaces that it really doesn't need, so stays nice and snappy, and the sub-system can do the grunt work in its 64 bit environment. I can't think of any reason why you would need to do the grunt work in the same address space as the GUI code; all you would need to pass back would be the end results, and they should take significantly less space.

So it's not as big a deal as you're making out. IPC is really not that hard; hell, if you're doing threaded programming, you're already making use of a fair chunk of IPC. All that differs is that the address space of the processes is not shared; most Unixes include the ability to create shared memory, or you can just use one of the other mechanisms available (such as pipes).

 

[Drummerboy]

Assuming 64-bit would be adopted by Mac within the next 12 months (seems to be the common assumption) would they only push the 64-bit to the Pro product line or all products?

If they moved all products to 64-bit, how long would it take before new software, including games, would only be 64-bit?

I'm concerned with my yet-to-ship MBP that it will be able to use new software within the next 3 years minimum (hoping to get 5 years out of the MBP).

Edit:
After mulling it over I figured that I shouldn't be too concerned with lack of software in the years to come considering Apple hasn't even flushed out the rest of their products yet. Assuming the MBP is supposedly the mobile "cream of the crop" (barring the 17") MBP owners (including myself) should be fine.

 

[oli84]

What does Intel's roadmap currently look like right now? Does it go this way:

1. Merom - 65nm running on the Napa platform
2. Merom - 65nm running on the Santa Rosa platform with 800mhz bus
3. Merom - 45nm iteration

??

If so, how big would you say is the jump from 65nm to 45nm in terms of performance and clock speed?

 

[capone2]

i dont know if anyone gave you any kind of answer, but......

Why does anyone want a 64-bit notebook? With Intel X86 architecture the only thing "64-bit" adds is the ability to address more than 4GB of RAM. Unless you actually install and use 8GB or 16GB in the notebook the 64-bit stuff is usless. Not much changes betwen the 32 and 64 bit CPUs except the size of pointers. They all do 80-bit floating point math and 32 bit integers.

The way to buy a computer is always to wait until the last possable minute and then buy whatever is available at your price point at that time. There is _always_ something better comming out in a few months so wait as long as you can but not a day longer

64bit question ? some big reasons 1) its the near future of computing
2) you will be able to boot Windows Vista
3) perfomance, graphics
4) Ram

 

[Anonymous Freak]

What does Intel's roadmap currently look like right now? Does it go this way:

1. Merom - 65nm running on the Napa platform
2. Merom - 65nm running on the Santa Rosa platform with 800mhz bus
3. Merom - 45nm iteration

??

If so, how big would you say is the jump from 65nm to 45nm in terms of performance and clock speed?

Based on past similar refreshes, you'll see no performance increase simply because of the die shrink, (at same clockspeeds,) but you will see higher clockspeeds, and lower power consumption at equal clockspeeds. Intel ususally aims for 50% clock speed increase with each die shrink (from the then-present max speed, usually 100% from the initial max speed.) So, for example, Pentium 4 hit 2.0 GHz on its original 180 nm process, 130 nm hit 3.2 GHz. 90 nm was a bit of a disappointment, only increasing that to 3.8 GHz. But with some of those process changes, other changes were introduced, too.

If Merom's 45 nm die shrink is PURELY a die shrink, then clock speed is the only improvement. And with Intel, usually if it's more than just a die shrink, the chip gets a new codename. So if it really is 'Merom 45 nm', then it will just be a die shrink, with no new features.

 

[RoboCop001]

I'm looking to get a MacBook Pro this year. But with everything moving towards 64-bit, I'm wondering if I should wait for the 64 MBP.

I'd like for it to last around 5 years. I'd be upgrading from this 350MHz G3 iMac. It's 5 years old, maybe 6, and other than the expected slowness with certain apps, it's running very well on 10.3.9. That's how I'd like my MBP to be in 5 years. Although the MBP I'd get would be the top-of-the-line model, whereas the iMac was the bottom of the line. So I think it might even last a bit more than 5 years.

But in terms of software, I'll deffinetly be using at least 1 pro application. If I got a 32-bit MacBook Pro this month, would it last me 5 years software-wise?

Or are companies, especially Apple, going to start making 64-bit exclusive applications soon after the 64-bit MBP arrives? Chances are I'd be using Apple pro apps like Logic and FinalCut, for example.

 

[iHeartTheApple]

I'm looking to get a MacBook Pro this year. But with everything moving towards 64-bit, I'm wondering if I should wait for the 64 MBP.

Good point. I'm wondering this, too. Is Apple planning on switching all hardware (even the new Yonah-based macs) to the 64-bit merom or conroe/woodcrest processors by the end of this year? Apologies if this has already been asked/answered in this thread, but it's moving too fast for me to keep up. I'm not one of those people complaining about the fact that Apple may be pushing new CPUs into their compies too fast (I'm all for it! ) , I just want to know what the plan is so I can decide whether I should wait for them or buy one of the new Yonah macs now.

 

[mrfop]

The general impression that I've got from this thread is to hold fire if you can. However, I want to use (and buy) Logic Pro for a mac now, and I might as well get an Intel machine to do it using Logic 7.2. I'm reasoning that it would be a good idea to buy a 17inch iMac until Merom comes on the scene, and then at the end of the year buy a Merom equipped MBP. See my thread here detailing my thoughts.

In short, there are a number of noise issues with the current MBP that may hinder my audio work (yes many have no problem, but there appears to be an element of chance with this). There is a wait on the MBP (3-4 weeks on BTO shipping), and Merom is creeping ever closer whilst we wait. And the idea that software developers (for me audio software is crucial) may not support 32 bit a few years down the line, well, I shudder to think ... I expect my MBP to last a fair few years!

The new iMac's are quiet (not a buzz in sight! With or without the use of iSight ), plus if I get a 17 inch with 1GB Ram, I save nearly £800 as compared to 2.0 Ghz, 7200, 256 Graphix, 1GB MBP I was going to purchase. Plus the iMac would reach me a few days and might be handy in the future as a seperate node for Logic!

Let me know you thoughts, I challenge anyone to convince me to buy a MBP now!

Laters.

 

[generik]

But, "better performance" can mean "more paying users" - especially if your competitors start advertising how much faster they are on the latest 64-bit PowerMacIntels. (Remember the typical 20% performance improvement just from recompiling for 64-bit....)

I'm not sure about the Conroe, but to date no Intel gets improved performance from 64 bit compilation. Intel's IA64 is merely a 64 bit implementation in name, with no hardware support to boot.

 

[Anonymous Freak]

I'm not sure about the Conroe, but to date no Intel gets improved performance from 64 bit compilation. Intel's IA64 is merely a 64 bit implementation in name, with no hardware support to boot.

What are you talking about? It is an exact duplicate of AMD's 64-bit extension to the x86 architecture. That includes the extra registers. The extra registers alone give up to a 20% boost in speed. (See other posts and links ad nauseum.)

The desktop Pentium 4/Pentium D does indeed support more than 4 GB of RAM.

Now, to make it even worse, you're using terminology for a completely different horse. IA64 is the name for the Itanium's architecture, a completely different architecture than x86, which is probably one of the most efficient 64-bit platforms out there. It just suffers from extremely high power draw, horrible marketing, and little software support. But at 1.5 GHz, it can outrun just about any other chip out there in floating point. (Yes, it even crushes IBM's POWER series.)

IA32 = x86, the line from the 386 through the latest Pentiums, Athlons, and now Core Duo.
IA64 = Itanium, a much-maligned, but technically very good architecture.
AMD64 = x86 with 64-bit bolted on, Intel calls this EM64T.

The confusion comes from the fact that when AMD came out with their 64-bit extensions, Intel immediately dismissed it as 'not really 64-bit', because they were pushing Itanium. Well, Intel changed their mind, under pressure from low-end server proponents, and introduced AMD's extensions as 'Extended Memory 64-bit Technology', or EM64T, to downplay the '64-bitness' of it as solely being memory-addressing. Well, it's more than that, and now Intel has wholeheartedly embraced it. Too bad it's too late (for their marketing droids, anyway,) to take back their dismissal of it as 'just memory addressing'.

 

[Drummerboy]

Attempting to put two and two together:
I keep hearing the Merom and Conroe referred to as processors with 64-bit extensions.
Are these extensions similar to the MMX extensions of long ago?
The case with MMX was that you could still use new software that was coming out. The software would check if you had MMX and would inturn use the optimized MMX code (no MMX caused lower quality graphics for games).
Does this mean that the OS's will remain compatible with the 32-bit Core?
If yes, does that mean that software only needs to be retooled if it wants to take advantage of the 64-bit extentions?
So in turn does that mean that based on current information software for at while should still run on the Core Duo & Solo hardware, just that it won't be as accelerated as it could be?

Edit:
If the "EM64T" extension turns out to be similar to "MMX" extensions then its not a full-architecture change requiring full OS and software changes and is no different than any other speed increase Intel does. Therefore waiting for this is no different than waiting to buy a computer because a faster processor will be coming soon.