[P]

This news is reported all over the place, but Anandtech does a decent job of summarising the points. Basically, the bus dispute that prevented NVIDIA from making chipsets for Arrandale has been resolved with Intel paying $1.5 billion for NVIDIA to walk away from its chipset business. 320M will likely be the last chipset NVIDIA ever makes - they can still make new Core 2 chipsets, but it seems unlikely.

The long and short of it is that Intel just killed all competition for integrated graphics for its processors. The answer now is either discrete graphics, or whatever Intel supplies.

[Lateralus]

I miss PowerPC, dammit.

[Spheric Harlot]

Yeah, six-month shipping delay and ending up with a slower processor for the same price…those were the days.

*sigh*

[Lateralus]

Didn't that only happen once, over a decade ago and with only one product line?

Say what you will about PowerPC; expandable Macintosh desktops started well below $2500 (they were also updated more than once every year and a half) and the GPUs in notebooks and iMacs weren't dictated by IBM or Motorola's elephant-in-the-room industry politics.

And there's also the fact that a 4 year old 13" MacBook is in the same CPU performance tier as the machine that exists now at the same price point... Things were never this bad in the PowerPC era.

And before anybody tries to claim that Penryn offers an all-trumping IPC improvement over Merom; it does not.

[Spheric Harlot]

I take it you don't remember the "3 GHz within a YEAR from today" bit either, then?

$2500 in 2000 is the equivalent of $3200 today.
$2500 in 2005 is the equivalent of $2800 today.

Current Mac Pros start at $2500.

The other thing is that while the market share for desktop towers has shrunken dramatically since ten years ago, the "consumer" desktops have become powerful and flexible enough to cover for *much* more of the "production" needs that forced people to buy a tower back then. Every pixel jockey had a tower back then; today, a $700 Mac mini completely suffices for a whole market segment that had no choice back then.

Also, are you saying that a current 2.4 GHz MacBook, with the current architecture, is not as much faster than a 1.83 GHz Core Duo MacBook than a 2005 1.33 GHz G4 iBook is than a 500 MHz G3 iBook of 2001?

I'd think the difference is roughly the same.

[P]

Originally Posted by Lateralus 

Say what you will about PowerPC; expandable Macintosh desktops started well below $2500 (they were also updated more than once every year and a half) and the GPUs in notebooks and iMacs weren't dictated by IBM or Motorola's elephant-in-the-room industry politics.

Oh, Apple can use which ever GPU they want, as long as they're happy to use a discrete one - which is what they had to do all the time in the PPC era.

Originally Posted by Lateralus 

And there's also the fact that a 4 year old 13" MacBook is in the same CPU performance tier as the machine that exists now at the same price point... Things were never this bad in the PowerPC era.

And before anybody tries to claim that Penryn offers an all-trumping IPC improvement over Merom; it does not.

Actually no. In early 2007, the cheapest Macbook was $1099 - it is $999 now. Also note that the clockspeed has gone from 1.83 to 2.4 GHz, which is a boost of 31%. It's hard to call that the same performance tier. Penryn also has turbo, even if Intel calls it something else, so it boosts 200 MHz more when one core is inactive. You're correct that the IPC didn't make a huge steps forward, but at least it didn't go down the way the G4 did at one point. The FSB has also gone from 166 MHz to 266 MHz, a boost of 60% which helps performance.

Your focus on CPU speed is also quite unfair. Apple could have opted to pick an Arrandale CPU - which would have boosted CPU performance while sacrificing graphics performance compared to the current setup - or they could have used Arrandale and a low-end discrete graphics chip to boost both even more, but sacrifice performance. I might as well make the comparison that the latest Macbook quadruples stated battery running time (from 2.5 to 10 hours) compared to the 4 year old one while the iBook never improved the battery life at all.

[Lateralus]

The May 2007 MacBook revision was available at 2.0 or 2.16GHz
The November revision was available at 2.0 or 2.2GHz

The last time a 1.83GHz MacBook was offered was in the late 2006 revision.

And no, Apple can't use whatever discrete GPU they want as they could have in the PPC days because politics combined with availability has screwed them; they could go nVidia, but nVidia can't seem to put together a worthwhile kit small enough to be worked into a mini, Air or 13" MacBook. They could go AMD/ATi, in theory, but I'm sure Intel has winked that 'special pricing' will change should that happen since there hasn't been an ATi GPU in a small form factor Mac since the PowerPC days - and there's no good reason for it outside of Intel politics.

And no, my CPU argument isn't unfair because the blame for this lies at Intel's feet. It was a tough choice, I'm sure Apple deliberated over it and I don't envy their being caught in the middle. Regardless of who caused who to choose what, the highest volume Macs are being saddled with old crap CPUs because Intel doesn't want nVidia playing on the same field... yet they're unwilling to put any laudable effort themselves into designing something that can compete with, let alone best, what nVidia offers.

I will not give Intel credit for any processor advancements when their own greed and paranoia has set the field in such a way that if your hardware is designed to actually be portable, then you have to saddle the graphics performance of your future hardware by downgrading to their crap graphics in order to get a newer/faster CPU.

[Lateralus]

Originally Posted by Spheric Harlot 

I take it you don't remember the "3 GHz within a YEAR from today" bit either, then?

$2500 in 2000 is the equivalent of $3200 today.
$2500 in 2005 is the equivalent of $2800 today.

Current Mac Pros start at $2500.

The other thing is that while the market share for desktop towers has shrunken dramatically since ten years ago, the "consumer" desktops have become powerful and flexible enough to cover for *much* more of the "production" needs that forced people to buy a tower back then. Every pixel jockey had a tower back then; today, a $700 Mac mini completely suffices for a whole market segment that had no choice back then.

Also, are you saying that a current 2.4 GHz MacBook, with the current architecture, is not as much faster than a 1.83 GHz Core Duo MacBook than a 2005 1.33 GHz G4 iBook is than a 500 MHz G3 iBook of 2001?

I'd think the difference is roughly the same.

And I take it you conveniently don't remember the fact that Intel, AMD and IBM all hit the wall at the same time and for similar periods after going 64-bit? Of course not.

Your math seems consistently fuzzy throughout this post, so I hesitate to even reply. But I'll bite;

You're only proving my point with that pricing;

Power Mac G5s started at $1499 only 5 years ago. Inflation isn't that bad. And then there was that good while where the Power Mac G4 was simultaneously available starting at what, $1199?

But regardless - bringing inflation into a computer cost debate is ludicrous. The industry, by its nature, pushes hardware costs down year over year, making things more and more accessible. But feel free to use the 1980s Mac IIfx price of $10,000 as a reason to praise Mac Pro pricing...

I compared a MacBook available in 2011 to what was debuted in 2007, which you for some reason stretched all the way back to a 2005 MacBook first-gen/32bit Core Duo... So that's 6 years from my 4. And then you bring up a 4-year iBook evolution as comparison? What?

And yes, a 1.33GHz G4 iBook would be a bigger leap over a 500MHz G3 than a 2.4GHz Penryn C2D is over a 1.83GHz Core-anything-Duo... and it was accomplished over a period of time that, as you pointed out for me, was only 2/3 as long.

[mduell]

Apple could use modern Intel chips with any discrete GPU they want... but they'd end up with a bigger logic board for a 3 chip setup, so a thicker machine, and oh noes must have the thinnnnnesssssss.

[Spheric Harlot]

Originally Posted by Lateralus 

And I take it you conveniently don't remember the fact that Intel, AMD and IBM all hit the wall at the same time and for similar periods after going 64-bit? Of course not.

Oh - so the same sorts of things happen to PowerPC that happen under Intel?

That kind of obviates your point I was responding to. Nice that we agree.

Originally Posted by Lateralus 

Your math seems consistently fuzzy throughout this post, so I hesitate to even reply. But I'll bite;

You're only proving my point with that pricing;

Power Mac G5s started at $1499 only 5 years ago. Inflation isn't that bad. And then there was that good while where the Power Mac G4 was simultaneously available starting at what, $1199?

Your memory is a little off. That 2005 $1500 PowerMac was basically the 2003 model.

Yes, the PMac G5 started at $1499 in 2005. Except that they had to basically REintroduce it at that price, because the others - starting at $2000 - simply weren't moving, and the performance/feature gap down to the iMac was too large.

That's changed rather drastically, now that you can get quad-core iMacs with Firewire, multiple-monitor support, and gigabit ethernet, and now that the cheapest Mac you can get will do an adequate job of running Photoshop with multiple monitors, in many cases.

Originally Posted by Lateralus 

But regardless - bringing inflation into a computer cost debate is ludicrous. The industry, by its nature, pushes hardware costs down year over year, making things more and more accessible. But feel free to use the 1980s Mac IIfx price of $10,000 as a reason to praise Mac Pro pricing...

Who exactly is "praising Mac Pro pricing"?

Bringing inflation into computer cost debate is only ludicrous if it's completely ludicrous to start a cost debate at all.

You mentioned an arbitrary price point as historical reference.
If you do that, you'll have to live with people pointing out that in historical context, it doesn't mean what your claim implies.

Originally Posted by Lateralus 

I compared a MacBook available in 2011 to what was debuted in 2007, which you for some reason stretched all the way back to a 2005 MacBook first-gen/32bit Core Duo... So that's 6 years from my 4. And then you bring up a 4-year iBook evolution as comparison? What?

The original MacBook was released in May of 2006. The Core 2 Duo was released in November.

I leveled the playing field by comparing FOUR years to FOUR (and a half) years.

Originally Posted by Lateralus 

And yes, a 1.33GHz G4 iBook would be a bigger leap over a 500MHz G3 than a 2.4GHz Penryn C2D is over a 1.83GHz Core-anything-Duo... and it was accomplished over a period of time that, as you pointed out for me, was only 2/3 as long.

As pointed out above, you err in regard to timeframes.

However, Geekbench says you're right:

$1000 MacBook 2.4 GHz mid 2010: 3365

$1100 MacBook 1.83 GHz mid 2006: 2286

a delta of about 50% over four years.


$1000 iBook G4 1.33 GHz mid 2005: 686

$1300 iBook G3 500 MHz mid 2001: 222

a delta of about 200% over four years.

Mac Benchmarks

I'm not entirely sure how much sense the comparison makes, since I rather doubt that the PowerPC platform has seen 200% performance increase over the past four years, either, but whatever.

[P]

Originally Posted by Lateralus 

The May 2007 MacBook revision was available at 2.0 or 2.16GHz
The November revision was available at 2.0 or 2.2GHz

The last time a 1.83GHz MacBook was offered was in the late 2006 revision.

Correct, so the low-end Macbook available in January 2007 was the one launched in late 2006. The one available in May 2007 would have to be compared to the May 2011 model, which would ruin your argument if Apple updates the Macbook as everyone expects. Sandy Bridge will be the biggest performance boost for Intel laptops since Merom at the very least, and it will be even bigger for the 13" Macbooks as they skipped Arrandale.

Originally Posted by Lateralus 

And no, Apple can't use whatever discrete GPU they want as they could have in the PPC days because politics combined with availability has screwed them; they could go nVidia, but nVidia can't seem to put together a worthwhile kit small enough to be worked into a mini, Air or 13" MacBook.

It wouldn't be larger than the discrete graphics in the old PPC machines. The difference is that Apple has used the space savings from integrated graphics to increase the volume taken by the battery

Originally Posted by Lateralus 

They could go AMD/ATi, in theory, but I'm sure Intel has winked that 'special pricing' will change should that happen since there hasn't been an ATi GPU in a small form factor Mac since the PowerPC days - and there's no good reason for it outside of Intel politics.

The iMacs are pretty small - they have ATi graphics now. There hasn't been any integrated ATi graphics since the Intel transition, but that's because ATi doesn't make any integrated graphics for Intel chips anymore. In all fairness, there wasn't integrated ATi graphics back in the PPC days either, because there wasn't integrated graphics at all.

Originally Posted by Lateralus 

And no, my CPU argument isn't unfair because the blame for this lies at Intel's feet. It was a tough choice, I'm sure Apple deliberated over it and I don't envy their being caught in the middle. Regardless of who caused who to choose what, the highest volume Macs are being saddled with old crap CPUs because Intel doesn't want nVidia playing on the same field... yet they're unwilling to put any laudable effort themselves into designing something that can compete with, let alone best, what nVidia offers.

I will not give Intel credit for any processor advancements when their own greed and paranoia has set the field in such a way that if your hardware is designed to actually be portable, then you have to saddle the graphics performance of your future hardware by downgrading to their crap graphics in order to get a newer/faster CPU.

But how on Earth was the PPC better? On PPC, you could never have integrated graphics because noone made them. On x86, you can have integrated graphics but you are quite limited in their performance if you want to use them. Intel's politics regarding the DMI bus makes the situation less than ideal, but then it's still better than it was in the PPC days. The PPC option - low-end discrete graphics - is still there.

While Intel's politics don't help, they are only accelerating the inevitable. The memory controller is moving onto the CPU, because that improves CPU performance enough that the change is worth it. That means that the old option, of integrating the GPU with the memory controller, is gone unless you use Intel graphics. What you might have done is to copy the setup AMD/ATi use for it's integrated graphics (separate northbridge and sideport memory), and Intel's politics block that, but it would have been a 1 year parenthesis between Penryn and Sandy Bridge.

FYI: Sandy Bridge integrated graphics beats the 320M in the current MBP.

[Lateralus]

Originally Posted by P 

The iMacs are pretty small - they have ATi graphics now. There hasn't been any integrated ATi graphics since the Intel transition, but that's because ATi doesn't make any integrated graphics for Intel chips anymore. In all fairness, there wasn't integrated ATi graphics back in the PPC days either, because there wasn't integrated graphics at all.

The iMac is pretty small in the way that a 17" PC notebook is small - it's not.

And I thought we were talking ATi about discrete options?

I'll let you have your thread back though as I don't feel like debating with you for the next several pages. Though I stand by my position that the main reason for the Intel switch was notebook advancement, and it has not paid off in the way most had hoped or even assumed.

[P]

Originally Posted by Lateralus 

I'll let you have your thread back though as I don't feel like debating with you for the next several pages.

I'm not even really debating yet, because I honestly don't understand how a PPC Macbook would be better here. Perhaps I'm not understanding you correctly, but... All PPC laptops were at least 4 chip solutions (CPU, Northbridge, Southbridge, GPU), while the newest 13" laptops are 2 chip solutions (one CPU chip and one chip that combines the other three chips in one). Intel's latest platform combines the chips a bit differently, but it's still 2 chips. If you make a 3 or 4 chip solution, you can select your GPU freely. If you want to stay at 2 chips, you're limited to whatever combinations are available on the market. As far as I know, there are no 2 chip solutions available on the PPC market, nor has there ever been, so how is that better?

Originally Posted by Lateralus 

Though I stand by my position that the main reason for the Intel switch was notebook advancement, and it has not paid off in the way most had hoped or even assumed.

IMO, it paid off on the day of the switch.

[Lateralus]

You're trying to turn this into an apples to apples conversation, which I never intended to start. My original point is that Intel politics have hampered Apple's notebook advancements more than they ever should have - what is available in how many chips on the x86 side vs what we may have had by now under PowerPC is something that could never be more than conjecture so it isn't worth the effort of bringing into the conversation. My other point is that Apple's relationship with ATi has been backburnered as an appeasement to Intel; do I have proof of this? No. But there seems to be a lot of practical evidence to suggest it.

But the situation we're in now is not conjecture; Intel doesn't want to play nice with companies that offer better solutions for their products than even they're willing to develop themselves. And so to save face and avoid a GPU downgrade, Apple is sticking with C2D and 320m long enough for Intel's IGPs to catch up. The situation does not benefit Apple in their quest to grab more of the low-to-mid range market and it does not benefit the consumers buying those machines in the configurations they're offered in now.

[P]

Less than ideal, certainly, but Intel's actions around the DMI bus are not purely political, and Apple wasn't forced to do what they did. Apple accepted a much bigger graphics downgrade when moving from the Radeon 9550 to GMA950 with the first Macbook. The Arrandale graphics (GMA HD) isn't the huge downgrade from 9400M that it's sometimes portrayed as. Apple could also have elected to use the 325M from the 15" in the 13" as well, but they judged the extra battery capacity to be more important.

It is also important to note that a traditional integrated graphics of the 9400M/320M type is not possible with Arrandale. The memory controller is on the package, and there is a good technical reason for that. A graphics chip in the PCH would have three options for memory access:

1) Use main memory for the framebuffer. This would add latency to all GPU operations, as you'd now have to add the latency of going over the PCIe x16 bus to the on-package memory controller. ATI does this, but they use HyperTransport instead of PCIe.
2) Use a separate memory space for the framebuffer, connected directly to the GPU. This would add cost because you have to add more memory chips, but it would potentially be faster than the old setup. Cost and performance would be like a low-end discrete graphics adapter, but saving one chip on the board.
3) Do both - small framebuffer memory backed by a bigger chunk of main memory for textures etc. ATI does this and calls it sideport memory. The technology for the split memory space is used by NVIDIA as TurboCache (ATI calls it HyperMemory).

1 makes little sense - you'd likely lose performance. 2 is basically a discrete chip except with the space saving of not having a separate PCH (but still extra memory), in return for adding quite a bit of cost. 3 is what makes some sense, and sure, NVIDIA might have done that if Intel hadn't denied them the DMI bus, but there is nothing stopping them from making that into a low-end discrete chip next to the PCH. That cost is much smaller than having to stay with Core 2. There is something more here than Intel throwing a hissy fit.

[Cool_Beanz]

Sucks that NVIDIA is in bed with Intel...