Intel CPU's for Apple Products
I'm anticipating the following Product / CPU combinations when Apple switches to Intel Architecture.
Mac Mini = Intel Yonah
iBook = Intel Yonah
iMac = Intel Merom (or Possibly Sossaman)
PowerBook = Intel Merom
PowerMac = AMD Opteron (Intel x86 Compatible)
XServe = AMD Opteron (Intel x86 Compatible)
Mac Mini = Intel Yonah
iBook = Intel Yonah
iMac = Intel Merom (or Possibly Sossaman)
PowerBook = Intel Merom
PowerMac = AMD Opteron (Intel x86 Compatible)
XServe = AMD Opteron (Intel x86 Compatible)
Comments
Don't hold your breath on Opteron's. They're not coming now or in the near future. Apple has hyped performance/watt. Opteron's give you performance but at the wattage problem remains. Hell you may as well stick with a G5.
Here's how I see it.
Mac mini- Single Core Yonah
iBook- Single Core Yonah
iMac Dual Core Yonah
Powerbook Dual Core Yonah(Q3 2006 adds a Merom config)
Powermac 2007- Dual Core Conroe in 2MB and 4MB configs
Xserve 2007- Dual Core Woodcrest in 4MB and 8MB configs.
Some of ngmapple's assumptions are right, others not.
1- most of us believe that only INTEL chips will be used (at least for some time)
2- Sossaman is a server chip (the server version of yonah)
3- Merom is a laptop/small desktop chip, while Conroe is the desktop equivalent.
Unless Conroe is too hot, that's the chip that we will see in the Intel based iMac.
4- Yonah is a transition chip, it will be replaced by Merom at some point
5- Yonah and Sossaman and 32-bit chips, Merom, Conroe and Woodcrest are 64-bit chips
6- Woodcrest is a multi-processor chip (it can be used in a dual (or more) chip configs).
Conroe can only be used in a mono config. Sossaman can be used in dual chip config also.
I agree with most of hmurchison's ideas, expect for:
- the iMac getting only a Yonah (32-bit while the current iMac has a 64-bit G5)
- the PMac getting only a Conroe (can't do a dual dual-core config)
So here is, my forecasted roadmap including the lastest info/rumors:
PowerBook#1= dual-core Yonah (Q1)
iBook = single core Yonah (Q2) *maybe a dual-core version too (or later)
Mac Mini = single core Yonah (Q2/Q3) *maybe a dual-core version too (in a new bigger design/form factor...)
PowerBook#2 = dual-core Merom (Q3/Q4)
iMac = dual-core Conroe (Q4)
PowerMac = up to dual dual-core Woodcrest (Q1 2007)
XServe = up to dual dual-core Woodcrest *maybe a up to dual dual-core Sossaman in a very low-cost model.
2. Yup. I don't see where Sossaman fits other than blade servers.
3. Merom sounds like the bet. Can't wait.
4. Conroe won't be in an iMac. Its TDP is 65 watts. A wee bit hot for a new slim iMac i'm thinking DC Yonah is a winner here.
5. Conroe. Hmmm good point Mjteix. Conroe's going to unneccesarily limit Quads for DC SMP systems. Hopefully Woodcrest pricing allows for affordable Quad systems.
6. The iMac is a useless 64-bit computer. It has only 1 ram slot which limits the primary benefit of 64-bit computing.
7. Wooohoooo Woodcrest Powermacs! I'll take the one with 8MB L2 cache!!
I'd like Apple to device a nice small case that support two 3.5" drives or four 2.5" drives with a SAS/SATA backplane.
That way people that need low cost storage can use SATA and larger drives and people that need RAID have 4 drives for RAID 5/6 configs.
Add a Zero Channel RAID card BTO for a low cost hardware RAID solution as well.
Forget AMD...you get better chip cost from Intel by offering them exclusivity.
Originally posted by hmurchison
Why wouldn't the Powerbook have a Yonah processor? Merom isn't due until 2H 2006 and we don't know what the pricing is so initially it may be high end only.
Yonah does not meet the performance standards Apple is looking for the power line. Apple was pressuring Intel into releasing Merom sooner, there has been no information regarding Apple and Yonah thus far. A 64-bit PowerBook is long overdue and Intel's Yonah does not support 64-bit math (64-bit addressing is not a major issue in a book, but 64-bit math certainly is for scientific and video applications).
Quote:
Don't hold your breath on Opteron's. They're not coming now or in the near future. Apple has hyped performance/watt. Opteron's give you performance but at the wattage problem remains. Hell you may as well stick with a G5.
The Opteron's have been shown to offer far superior performance to Intel's Xeons and Itaniums, and now even beat them on performance/watt ratios. Intel's Xeon is really not a high performance contender to the Opeteron at this point or near future. In regards to the G5, the new 970FX LV's have a pretty decent performance/watt ratio. I read an article that suggests the performance of the G5 is far greater under Linux than OS X, but I would expect that to change as Apple works to speed up X. Best way to compare G5 performance to x86 performance is Linux-Linux since you keep the OS constant for the test.
Yonah does not meet the performance standards Apple is looking for the power line. Apple was pressuring Intel into releasing Merom sooner, there has been no information regarding Apple and Yonah thus far. A 64-bit PowerBook is long overdue and Intel's Yonah does not support 64-bit math (64-bit addressing is not a major issue in a book, but 64-bit math certainly is for scientific and video applications).
Laughable considering a 2Ghz Pentium M beats Apple's fastest notebook today. Add in another 2Ghz core and a shared 2MB cache and you have a nice speed increase for Powerbook users. Everything has been rumor regarding Apple's choice for notebook processor. 64-bit is the most overhyped BS. Laptops have 2 RAM slots meaning you can't even break the 4GB 32-Bit RAM ceiling let along a 64-bit Petabyte limit. OS X doesn't support 64-bit well enough to matter in a portable configuration. I'm sure there are just millions of Mathmatica users pissed off right now about lack of 64-bit Powerbooks<sarcasm>
The Opteron's have been shown to offer far superior performance to Intel's Xeons and Itaniums, and now even beat them on performance/watt ratios. Intel's Xeon is really not a high performance contender to the Opeteron at this point or near future
As they should be. Intel cancelled Netbursts future over a year ago. 2006 is where we see the fruits of their labor. Intel's first to 65nm with Yonah and then Merom. I like AMD but their performance advantage means squat to the Enterprise who still feels "comfortale" with the Wintel duopoly.
We're going to be lucky to get Merom on time. Hell if Apple can get Merom into the Powerbooks early then by all means do it but reality is a dual core Yonah is going to crush whatever Powerbook Apple foists on us until they hop aboard Yonah or Merom is ready.
Originally posted by ngmapple
Intel's Yonah does not support 64-bit math (64-bit addressing is not a major issue in a book, but 64-bit math certainly is for scientific and video applications).
Actually, virtually all PC CPUs today have, and for a very very long time supported, 64-bit precision (or greater) floating point math, native in hardware.
What you're talking about is 64-bit integer math, but that really isn't in much high demand. The term "64-bit" really is about 64-bit integers which typically equate to larger memory addressing and the aforementioned 64-bit integer math. The 4+ GB memory support is very important for those scientific and video applications.
The Opteron's have been shown to offer far superior performance to Intel's Xeons and Itaniums, and now even beat them on performance/watt ratios. Intel's Xeon is really not a high performance contender to the Opeteron at this point or near future.
Far superior is a rather excessive term. I find it hard to believe that Opteron actually beats Itaniums in FPU. It probably doesn't, and loses by quite a margin. Xeons it outperforms, yes.
In regards to the G5, the new 970FX LV's have a pretty decent performance/watt ratio.
Unfortunately, the highest clock rate for those low-wattage 970fx appears to be 1.6 GHz. Perhaps they have a 30 Watt 2 GHz version, but I'll believe it when the documentation comes out.
At one point in time, I believed it would have been possible for IBM to put out a 90nm 30W 1.8 GHz 970fx, then a 65nm 30W 1.8 GHz 970mx (dual core) to put in Apple's laptops, but that was in late 2003. The 2 year wait for the low watt 970fx pretty much killed those hopes.
MWSF 06
New iTMS video content
Leopard UI preview and approximate release date
March 06
Mac mini: single-core Yonah, 1.6 GHz, no fan
iBook: single-core Yonah, 1.8 GHz, 1 fan
April 06
Powerbook: dual-core Yonah, 2 GHz, 2 fans, 1" thick form factor
May 06
Xserve: 2.3 GHz 970mp update, quad-core in $4000+ config
WWDC 06
iMac: dual-core Yonah, 2.2 GHz, 2 fans w/thinner form factor
July 06
PowerMac: price reduction on 2.5 GHz 970mp quad, 2.3 GHz quad replaces dual 2.3, price reduction 2 Ghz dual
October 06
Mac mini: dual-core LV Yonah, 1.6 GHz, no fan
iBook: dual-core ULV Yonah, 1.8 GHz, 1 fan
November 06
Leopard release
Powerbook: dual-core Merom, 2.2 GHz, 2 fans
MSWF 07
PowerMac: Woodcrest, up to 2.8 GHz, duals and quads
Xserve: Woodcrest, up to 2.6 GHz, duals and quals
iMac: Conroe, 2.4 GHz
Originally posted by THT
I'll one-up the predictions with clock rates and other things:
MWSF 06
New iTMS video content
Leopard UI preview and approximate release date
Your assumption that no hardware whatsoever will be released at MWSF is unlikely at best.
Originally posted by hmurchison
4. Conroe won't be in an iMac. Its TDP is 65 watts. A wee bit hot for a new slim iMac i'm thinking DC Yonah is a winner here.
6. The iMac is a useless 64-bit computer. It has only 1 ram slot which limits the primary benefit of 64-bit computing.
Apple can always underclock Conroe to fit in the iMac (since Conroe is very similar to Merom and Merom has 35W TDP, you'd think Intel could produce a 35W Conroe also).
If 64-bit is free, Apple can take it and just not use it. Just like me and my Opteron that runs 32-bit games on 32-bit Windows XP.
Isn't it amazing how we're nitpicking about something that's a year away?
According to Intel :
Thermal Design Power (TDP) represents the maximum amount of power the thermal solution is required to dissipate. The thermal solution should be designed to dissipate the TDP without exceeding the maximum Tjunction specification.
TDP does not represent the power delivery and voltage regulation requirements for the processor
You must be aware that some parts of the chip have more watt consumption than others (and therefore produce more heat), thus the TDP take into account the hotter zones of the chips and not the average watt consumption of the chip.
If 10 % of the chip produce 50 % of the chip, the TDP will be 5 time more important than the max power consumption of the chip.
As we don't know the hot and cold part of a chip, it's impossible to bet what is the maximum watt consumption of a chip, with the TDP alone.
From all accounts the Intel version of the OS seems pretty much there. And Intel hardware in one flavor or another has, by all accounts, been available for some time.
So while it might be useful for Apple to join the club at the point when a shiny new processors appear, it is not strictly necessary.
However if you sell a pro computer you have to have pro applications to run on it. That is to say you need proper universal binary versions of those applications, not just stuff running under Rosetta.
It would simply not be credible for Apple to launch Intel-based Powermacs until FCP, Motion, Logic, Shake, Maya and ideally Photoshop are all ported and debugged and singing sweetly on the new platform.
My guess is that it will be the shipping of this strategic software and not specific chip availability which will influence release date.
Carni
Originally posted by Powerdoc
You must be aware that some parts of the chip have more watt consumption than others (and therefore produce more heat), thus the TDP take into account the hotter zones of the chips and not the average watt consumption of the chip.
If 10 % of the chip produce 50 % of the heat, the TDP will be 5 times more important than the max power consumption of the chip.
NetBurst chips have hot spots, but that's why they also have integrated heat spreaders. From my measurements, actual power consumption is close to TDP.
Originally posted by wmf
NetBurst chips have hot spots, but that's why they also have integrated heat spreaders. From my measurements, actual power consumption is close to TDP.
You may be true, I did not checked it my self, but an another guy on Mac Bidouilles claiming that his job was to do such measurements said otherwise. So I am open on the subject.
The problem of TDP wich is a measurement for OEM companies, is that it differ from a manufacturer to an another.
It's surprising that companies do not communicate in term of power consumption. But it seems with the coming of the Merom, and the new intel credo power/watt consumption, companies will communicate more on the subject.
It seems that the merom should be much better than the Yonah in this aspect. Thus I will recommand to switch to the intel mac powerbook, only when the merom will be released.
PS : my explanation about the hot spots of chips, was an over simplification. The point was to emphasize that TDP is not powerconsumption. And even with powerconsumption , there is max powerconsumption (not very easy to do : you have to find the way to stress a max a CPU) , typical powerconsumption, consumption under word ...
That's why it's very difficult to compare one chip to an another.
Originally posted by Powerdoc
It's surprising that companies do not communicate in term of power consumption.
Well, the processor data sheets will give you an Imax and Vmax, which you could multiply to get the worst-case max instantaneous power, but this number is even higher than TDP.
PS : my explanation about the hot spots of chips, was an over simplification. The point was to emphasize that TDP is not powerconsumption. And even with powerconsumption , there is max powerconsumption (not very easy to do : you have to find the way to stress a max a CPU) , typical powerconsumption, consumption under word ...
That's why it's very difficult to compare one chip to an another.
You're right about that. But the good news is that hardware reviewers are starting to include power measurements in their reviews.