Ok so they are comparing a 3.2ghz (harpertown) to a 2.93ghz (nehalem). Most hte tests are roughly 1.5x faster.
So there is a 300mhz difference (give or take). And the speed is comparable. Well lets look at slower models.
If the 300mhz difference is roughly 1.5x faster and we look at the 2.26 (nehalem) vs the 2.8 (harpertown) we are now dealing with a 540mhz difference. That makes them relatively even using apple's own benchmarks.
You can pick up a octo 2.8 refurb for around 2200, that's almost a 1k difference for the same performance. I'm really considering just picking up a refurb instead.
That's true. But there are other things to think about.
The older model is Express 1 and 2. Two slots each. The new one is all Express 2.
I hope you don't actually claim to be knowledgeable about this stuff, because that analogy is so far from the truth its pathetic. The Xeon 5xxx series and Core i7 ARE CUT FROM THE SAME LITERAL SILICON DIE. While they may pluck out the most stable and efficient units during verification to sell as server-grade Xeons and perhaps tweak the CPU, Cache, and QPI clocks, they are otherwise identical...
The 3000-series are server branded versions of desktop CPUs. The 5000-series are twin socket derivatives and the 7000-series is for more than two sockets.
Quote:
We'll see about that considering MANY "prosumers" and enthusiasts buy the Mac Pro (or did).
Yes we will. Apple has put the Mac Pro out of reach for all but the highest echelons of pros. We'll see how many settle for an iMac (and be less than happy with it), how many stick with their current Mac a bit longer, and how many come to the conclusion that Apple no longer serves their needs and leave the platform.
Quote:
Years ago? How about 2 months ago??? The new Mac Pro has exorbitantly increased in price while using components that are similar in cost -- particularly as the CPUs are actually CHEAPER than the old Harpertowns when they were released.
Not only that, but going back to late 2005, the last 4 PowerMac/MacPro updates have had a price hike. ($500, $150, $150, $200). In the last four years, the entry price for a MacPro has risen $1000 from $1499 to $2499. The funny thing is the component cost for the new $2499 MacPro might not be much more than the old $1499 Mac Pro.
Quote:
Originally Posted by emig647
Can you prove this? Are you are you speculating? I wasn't aware any of these prices had been announced yet?
They were released a couple months back. A bit more expensive, but not so much to justify the price hikes as we got them. In fact, the both the 2.66ghz 3500 and the 2.26ghz 5500 are roughly $300 cheaper per chip than the 2.83ghz 5400 series xeon they replaced. Prices for the dual socket x58 boards that are trickling out are similar to 5400 chipset boards. There is something else at play whether it be Apple getting greedy or intel charging more for either getting them early or making Apple pay full price if they were giving them discounts before.
Can you prove this? Are you are you speculating? I wasn't aware any of these prices had been announced yet?
speculating? Did you read the thread? I just posted in the prices..
Quote:
Originally Posted by melgross
You don't know any of that. That's just something you are guessing at.
Nope. We know the general costs of the CPU, RAM, and GPU, and the margins are much higher this time around. Refer to the pricing chart from the post of mine on page 4 or 5:
Quote:
Originally Posted by winterspan
"Well, for everyone that wants to make excuses about the enormous price increase of mid-level Mac Pro configurations (bumping solely the CPU to dual 2.66Ghz from a base config gets you to $5000), I made a little chart. I knew despite the platform switch and all the hype that the new CPUs shouldn't cost much more than the older Harpertown Xeons did (when they were new). What I didn't realize is that the new Nehalems Xeons that the Mac Pro is using are even CHEAPER than their Apple equivalent from the old Mac Pro!
Disclaimers:
1) These prices represent unit cost in batches of 1000 at the time of initial release. Remember, Apple does NOT lower their prices over the lifecycle of a certain model even though the CPU costs may drop somewhat. Regardless this point is moot anyhow as server grade CPUs don't change price much if at all until a new, faster model/series is released.
2) The single CPU xeon configuration (at least the base model) uses a different CPU from the Xeon 3xxx series. These prices are for the dual-CPU machines.
It doesn't matter that the basic chip is the same. The extra functions give it extra performance, they always do....
extra functions? Again, these are the same damn chips! The Core i7 extreme editions have literally IDENTICAL specifications of the Xeon 55xx series save for one of the QPI links being disabled. Same exact frequency multiplier, same exact Quickpath frequency, save exact memory controller and speed, etc.. They even all use DDR3 now!
Quote:
Originally Posted by melgross
....The point though is not even that, it's simply the point of calling things by their correct names. I just don't know why that's so difficult for you.
You must be arguing with the wrong person here. I agree with you on this point, and have never used Core i7 and Xeon 55xxx/Gainestown interchangeably...
Quote:
Originally Posted by emig647
Keep in mind these new cpu's have hyperthreading. So if you get a 2.26 you are getting 16 virtual cores. If I were you, I'd go with the new 8 core 2.26... it is going to last you much longer with snow leopard's GCD and OpenCL. Also the new cpu's offer virtualization... so better compatibility with VMWare running windows if that is a factor to you. It is to me because I have to test ie6 against web apps.
I disagree. I would recommend getting the 2.66 Ghz CPUs. Nehalem is indeed fast, but based on the Core i7 reviews I've seen, I'd be surprised if the 2.26Ghz is to going to outrun the older 3.2Ghz Harpertown. I may be wrong however. I'll get back to you on this after some Googling..
The 3000-series are server branded versions of desktop CPUs. The 5000-series are twin socket derivatives and the 7000-series is for more than two sockets.
Yep.. Am I missing something here? Did I make a typo somewhere?
Thats true, but you must consider the size of the files. That's the most important factor.
If the file is 20 MB in size, any Mac product is going to be plenty fast, Quartz, Open CL, Grand Central, or whatnot.
In fact, the slowest computer Apple made a year ago is more then fast enough for file around that size. People are getting themselves tied in knots worrying about speeds.
You are so right, and I try to tell people the same thing. You know what's crazy, is how I have delegated my machines now days. I used to separate projects I used to work on to whatever computers were available. But, now days there is a big difference on my Mac Pro, I got up to a 25 GB MPEG-2 file this week, not bragging, but I wouldn't touch that with my G4s or my MacBook. Even if I had another 25 GB file in a queue I would just wait for the Mac Pro to free up. I wouldn't think of putting it on the other machines...
So now days there are systems for thick files and thin files... and thick is not something everybody gets into. There is a big difference between a 2 Core Mac with 4 GB of RAM and 16-Core Mac (Virtually) with 32 GB of RAM.
Laters...
p.s. bored while mac pro is rendering/compressing, ugh
Nope. We know the general costs of the CPU, RAM, and GPU, and the margins are much higher this time around. Refer to the pricing chart from the post of mine on page 4 or 5:
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price.
You also don't know the cost to produce the machine itself, or whatever R&D cost they may add in. Then there is the overhead.
I would love to know these numbers, but none of us do. It's possible that Apple is selling less, so they have to charge more as component pricing isn't as low as it would otherwise be. We are just speculating.
[QUOTE=winterspan;1385409]extra functions? Again, these are the same damn chips! The Core i7 extreme editions have literally IDENTICAL specifications of the Xeon 55xx series save for one of the QPI links being disabled. Same exact frequency multiplier, same exact Quickpath frequency, save exact memory controller and speed, etc.. They even all use DDR3 now!/quote[
I pointed out other differences. They do exist.
Quote:
You must be arguing with the wrong person here. I agree with you on this point, and have never used Core i7 and Xeon 55xxx/Gainestown interchangeably...
Well, possibly, so many are doing this its getting hard to sort them all out.
Quote:
I disagree. I would recommend getting the 2.66 Ghz CPUs. Nehalem is indeed fast, but based on the Core i7 reviews I've seen, I'd be surprised if the 2.26Ghz is to going to outrun the 3.2Ghz Harpertown from last generation. I may be wrong however.
I agree on this as well. It's one reason I ordered the dual 2.66 model.
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price. You also don't know the cost to produce the machine itself, or whatever R&D cost they may add in. Then there is the overhead.
I would love to know these numbers, but none of us do. It's possible that Apple is selling less, so they have to charge more as component pricing isn't as low as it would otherwise be. We are just speculating.
I will agree to a very small extent. Certainly I'm not claiming to know their exact figures, but unless there are some very strange circumstances, I can't see how their bill of materials for the new model could come anywhere close to justifying the massive price hike. I'm sure they do factor in R&D and other things, but at the moment I am only comparing the new Mac Pro to the old model, not to PC workstation manufacturers..
Quote:
Originally Posted by melgross
I pointed out other differences. They do exist.
What are they? I acknowledged that depending on the particular generation, there can be differences in clock, cache, FSB/Quickpath clock, memory support, etc. BUT, in this case the Core i7 extreme editions have literally the exact same specs as their Xeon (35xx) counterparts. The only difference both of them have with the Xeon 55xx is the fact that Intel disabled the second QPI link after manufacturing. There is no other difference whatsoever other than to be binned differently in testing for power consumption and stability. There should not be any major difference in performance..
Quote:
Originally Posted by melgross
Well, possibly, so many are doing this its getting hard to sort them all out.
understood -- I know the feeling
BTW, I apologize if I come across as an ass sometimes. I always visit these forums after a long day at work and can seem quite aggro. I honestly mean no harm, and greatly enjoy the discussion.
Yes based on what I can see the Quad Mac Pro is a $2k computer with a hefty margin appied. It's going to get ugly when the PC World rags of the world start showing the basic $1200 Core I7 pc beating the Mac pro in performance
I don't see that happening. Besides, I'm surprised at you. You know that the Mac Pro is not a home machine. Compare it to comparable machines when they do come out.
Mel', I'm not suprised at you. I agree with Hmurch' completely.
I answered that in the post above the one you made to me.
Sure, I'll go back.
Quote:
Originally Posted by melgross
There are more differences than that ECC memory is one. The possible use of FB-DIMMs in four channels is another, though it's not likely, at least right now, that Apple will be using Becton. Additionally, the i7 parts currently are all 130 watt parts, while Gainstown, except for the not yet available 3.2 GHz part are either 80 or 95 watt parts.
Also 2 QPI links are used, not just one as in i7.
We're comparing to the xeon 3500 series here, not to future chips intel isn't shipping and apple isn't using.
So no four channel memory.
And from what I have read, it looks like the xeon 3500s only have one QPI link, just like the i7. If you find documentation otherwise, I'll stand corrected.
So the only differences are ECC and using a bit less power, neither of which are going to make a difference in real world performance - the two machines will run a given app the same, the user is just paying twice as much for ECC and to use a bit less power. Oh, and for twice the price you get the ability to use LESS ram.
Quote:
Originally Posted by melgross
You have every right to complain that Apple doesn't MAKE a consumer tower. You would be right in that. I bemoan that fact as well.
Funny, with a ram limit lower than an $899 dell and about the same performance, it sure looks like the Mac "Pro" quad IS a consumer tower. Apple is just charging double for it.
And I'm not asking for a consumer machine, I AM asking for a pro machine. I just don't define "pro" as a machine built with sturdier parts but consumer level performance (or worse).
I'd be happy with a consumer machine at a consumer price. Or with a pro machine at a pro price. Apple doesn't seem to hit either market.
Quote:
Originally Posted by melgross
Most of these boxes are bought by companies or other organizations, not individuals.
Either way, it is bought for someone to sit in front of it and run apps. You really think that the majority of these buyers are willing to pay twice as much for "workstation parts" but don't care if performance is no better than a machine half the price?
Quote:
Originally Posted by melgross
The truth is that very few pros need 8 Gb RAM. People seem to like to think they do, but they don't.
But don't "very few" people really need a workstation class machine in the first place? You insist there are enough people who need ECC and whatever else allegedly makes this a Workstation...yet you don't seem to think that that is the same group that may need more than 8 gigs?
When it comes to ECC ram and the other excuses for the pricing, this machine is aimed at the small group of Professionals who can't settle for a consumer machine and insist on only the best.
But when it comes to ram limitations, then the market for this machine suddenly becomes those guys who really don't need anything that special and can settle for something LESS than what some cheap consumer boxes are offering?
I don't see how that's not a contradiction.
And for the record, I am one of those who can take advantage of more than eight gigs of ram, I have more than that in my current machine already.
Quote:
Originally Posted by melgross
It doesn't matter that the basic chip is the same. The extra functions give it extra performance, they always do.
I'm dying to see your response when the benchmarks are out and the i7 and xeon 3500 have negligible differences at the same clock speed. ECC isn't going to boost performance, is it? Nor lower power use?
So what IS going to make the 3500 faster than the i7 at the same clock speed?
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
hips! The Core i7 extreme editions have literally IDENTICAL specifications of the Xeon 55xx series save for one of the QPI links being disabled.
And in the case of the 35xx series, if what I have read is correct, that xeon has one QPI link disabled as well.
Quote:
Originally Posted by melgross
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price.
True. But when the price to the public is lower, it's a safe assumption that the price to Apple drops as well.
If a new chip sells publicly for less, but Apple ends up paying more than the previous generation, then they are doing something wrong.
Quote:
Originally Posted by melgross
I pointed out other differences. They do exist.
Well, one of the differences doesn't exist in shipping chips. A second one doesn't seem to exist in the 3500 series. And the other differences aren't ones that would have any effect on performance.
Quote:
Originally Posted by winterspan
What are they?
Aside from the ECC you mentioned and the 2 qpi (which looks like it's 55xx only, not in the 35xx), it looks like the only other difference is lower power use.
And he listed more memory channels...which is only in a chip that apple is NOT using yet.
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
The 5500 allows use of ECC memory, it has two QPI channels instead of one, and it uses less power. I also assume the i7 is only supported in single chip (4 core) configurations, but I could be wrong.
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
Core i7: Bloomfield 1 quickpath link Nehalem core without ECC support.
Xeon 3500: Bloomfield 1 quickpath link Nehalem core with ECC support. Same price as the core i7 branded chips.
Xeon 5500: Gainestown. Essentially bloomfield with a second quickpath link for multi-processing. Intel is also making these in more clock speeds than the bloomfield chips.
The 5500 allows use of ECC memory, it has two QPI channels instead of one, and it uses less power. I also assume the i7 is only supported in single chip (4 core) configurations, but I could be wrong.
Otherwise the chips are identical.
OK, so does that mean the i7 will find its way into the iMacs and Mac minis?
Does the i7 have a series number like the Nehalem (5500) and Woodcrest (5400)?
If the i7 and Nehalem are similar, does that the mean the Xeon 5400s are similar to the Core 2 Duos?
What will be in the future MacBooks? The Atom? The i7?
The Core 2 Duos that are in the current iMacs are a mobile (low power low heat) proc, right?
Does the i7 have a low heat/mobile version?
This wiki is a good source of info, but it doesnt explain Apple's strategy and branding.
I will agree to a very small extent. Certainly I'm not claiming to know their exact figures, but unless there are some very strange circumstances, I can't see how their bill of materials for the new model could come anywhere close to justifying the massive price hike. I'm sure they do factor in R&D and other things, but at the moment I am only comparing the new Mac Pro to the old model, not to PC workstation manufacturers..
One thing I've found over the past five or six year or so is that the higher end machines are getting more expensive from every manufacturer, not just Apple. Workstations have risen in price and servers have risen in price.
With consumer machines, we see that companies make cheaper, smaller machines all the time, and that makes it seem as though prices are going down, but they are not. Those cheaper machines are lower spec'ed all around. But the $1,000 machines are still there. The $1,500 machines are still there, etc, all the way up. Alien still has consumer grade machines for $3,000, as do others.
I'm surprised the new Mac Pros have gone up as much as they have. I expected a small increase. but I doubt if Apple is being nasty. When other comparable Xeon machines arrive, they will also be expensive. This has been true for Harpertown machines, and those further back.
Look at workstan manufacturers to see this. I mentioned a well known one Boxx. You idn't respond to that. This is what Apple is competing with, and their prices are higher than Apple's are by mre than a small bit.
Even looking at the home built model shown here doesn't use the best parts. There are much more expensive cases, power supplies etc.
Quote:
What are they? I acknowledged that depending on the particular generation, there can be differences in clock, cache, FSB/Quickpath clock, memory support, etc. BUT, in this case the Core i7 extreme editions have literally the exact same specs as their Xeon (35xx) counterparts. The only difference both of them have with the Xeon 55xx is the fact that Intel disabled the second QPI link after manufacturing. There is no other difference whatsoever other than to be binned differently in testing for power consumption and stability. There should not be any major difference in performance..
The difference in the 35xx parts are marginal, I know, except for ECC support. But that just shows the base cost of the machine itself. Again, these are workstations, built to a much higher standard. Why do they cost more? WE don't know. You can't assume that Apple simply decided to raise prices for no reasons of cost to them. If that were true, then their competitors, who are not the consumer manufacturers, come out with their machines, Apple will be at a competitive disadvantage. That hasn't happened in the workstation space before, and I see no reason why it would be true today.
Quote:
understood -- I know the feeling
BTW, I apologize if I come across as an ass sometimes. I always visit these forums after a long day at work and can seem quite aggro. I honestly mean no harm, and greatly enjoy the discussion.
Believe it or not we ALL come across as asses at times. It's just that some people don't want to admit it about themselves, so you get credit where they don't.
We're comparing to the xeon 3500 series here, not to future chips intel isn't shipping and apple isn't using.
So no four channel memory.
And from what I have read, it looks like the xeon 3500s only have one QPI link, just like the i7. If you find documentation otherwise, I'll stand corrected.
I'm looking at the entire family, not just the low end chip. but I grant that there isn't too much difference there.
Quote:
So the only differences are ECC and using a bit less power, neither of which are going to make a difference in real world performance - the two machines will run a given app the same, the user is just paying twice as much for ECC and to use a bit less power. Oh, and for twice the price you get the ability to use LESS ram.
ECC is very important to more than a few users, so don't denigrate that. Lower power makes for a cooler machine, a quieter machine, and less total power use. This is important where there are installations of a number of them.
Quote:
Funny, with a ram limit lower than an $899 dell and about the same performance, it sure looks like the Mac "Pro" quad IS a consumer tower. Apple is just charging double for it.
We don't yet know if that 8 GB is a hard limit, or just marketing. As I pointed out earlier, Apple has done this many times in the past. PC manufacturers are quick to tout every new feature and capacity increase. Apple is much more conservative.
Quote:
And I'm not asking for a consumer machine, I AM asking for a pro machine. I just don't define "pro" as a machine built with sturdier parts but consumer level performance (or worse).
Well, performance certainly won't be worse. But many institutions that buy these machines are not looking for ultimate performance, or they would buy the more expensive machines, but rather reliability, which these machines do provide.
[quote]
I'd be happy with a consumer machine at a consumer price. Or with a pro machine at a pro price. Apple doesn't seem to hit either market.[quote]
While I agree with the first sentence, I don't agree with the last.
Quote:
Either way, it is bought for someone to sit in front of it and run apps. You really think that the majority of these buyers are willing to pay twice as much for "workstation parts" but don't care if performance is no better than a machine half the price?
Not all workstations have people sitting in front of them. Many are used for rendering, or other purposes, headless. It's not likely the low end machine will be used for that, but the dual machines certainly will.
Quote:
But don't "very few" people really need a workstation class machine in the first place? You insist there are enough people who need ECC and whatever else allegedly makes this a Workstation...yet you don't seem to think that that is the same group that may need more than 8 gigs?
You don;t get points for that, because as I said, most of these machines are bought by companies or other institutions. Price is less of a problem for them, because the competition is at the same price level.
As far as the RAM goes, we're back to the "we don't yet know". So let's drop that until we do, ok? Even so, not every use needs that much RAM. That's a myth.
[quote]
When it comes to ECC ram and the other excuses for the pricing, this machine is aimed at the small group of Professionals who can't settle for a consumer machine and insist on only the best.
But when it comes to ram limitations, then the market for this machine suddenly becomes those guys who really don't need anything that special and can settle for something LESS than what some cheap consumer boxes are offering?
I don't see how that's not a contradiction.[/quote}
PLEASE, drop the RAM argument for now.
Quote:
And for the record, I am one of those who can take advantage of more than eight gigs of ram, I have more than that in my current machine already.
Good.
Quote:
I'm dying to see your response when the benchmarks are out and the i7 and xeon 3500 have negligible differences at the same clock speed. ECC isn't going to boost performance, is it? Nor lower power use?
So what IS going to make the 3500 faster than the i7 at the same clock speed?
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
Comments
You know, I retract my last statement.... after looking at these benchmarks I have some other thoughts....
http://www.apple.com/macpro/performance.html
Ok so they are comparing a 3.2ghz (harpertown) to a 2.93ghz (nehalem). Most hte tests are roughly 1.5x faster.
So there is a 300mhz difference (give or take). And the speed is comparable. Well lets look at slower models.
If the 300mhz difference is roughly 1.5x faster and we look at the 2.26 (nehalem) vs the 2.8 (harpertown) we are now dealing with a 540mhz difference. That makes them relatively even using apple's own benchmarks.
You can pick up a octo 2.8 refurb for around 2200, that's almost a 1k difference for the same performance. I'm really considering just picking up a refurb instead.
That's true. But there are other things to think about.
The older model is Express 1 and 2. Two slots each. The new one is all Express 2.
The new one has four FW 800 ports, etc.
It's not just the Cpus.
I hope you don't actually claim to be knowledgeable about this stuff, because that analogy is so far from the truth its pathetic. The Xeon 5xxx series and Core i7 ARE CUT FROM THE SAME LITERAL SILICON DIE. While they may pluck out the most stable and efficient units during verification to sell as server-grade Xeons and perhaps tweak the CPU, Cache, and QPI clocks, they are otherwise identical...
The 3000-series are server branded versions of desktop CPUs. The 5000-series are twin socket derivatives and the 7000-series is for more than two sockets.
We'll see about that considering MANY "prosumers" and enthusiasts buy the Mac Pro (or did).
Yes we will. Apple has put the Mac Pro out of reach for all but the highest echelons of pros. We'll see how many settle for an iMac (and be less than happy with it), how many stick with their current Mac a bit longer, and how many come to the conclusion that Apple no longer serves their needs and leave the platform.
Years ago? How about 2 months ago??? The new Mac Pro has exorbitantly increased in price while using components that are similar in cost -- particularly as the CPUs are actually CHEAPER than the old Harpertowns when they were released.
Not only that, but going back to late 2005, the last 4 PowerMac/MacPro updates have had a price hike. ($500, $150, $150, $200). In the last four years, the entry price for a MacPro has risen $1000 from $1499 to $2499. The funny thing is the component cost for the new $2499 MacPro might not be much more than the old $1499 Mac Pro.
Can you prove this? Are you are you speculating? I wasn't aware any of these prices had been announced yet?
They were released a couple months back. A bit more expensive, but not so much to justify the price hikes as we got them. In fact, the both the 2.66ghz 3500 and the 2.26ghz 5500 are roughly $300 cheaper per chip than the 2.83ghz 5400 series xeon they replaced. Prices for the dual socket x58 boards that are trickling out are similar to 5400 chipset boards. There is something else at play whether it be Apple getting greedy or intel charging more for either getting them early or making Apple pay full price if they were giving them discounts before.
http://en.wikipedia.org/wiki/List_of...5B1.5D_.5B2.5D
Can you prove this? Are you are you speculating? I wasn't aware any of these prices had been announced yet?
speculating? Did you read the thread? I just posted in the prices..
You don't know any of that. That's just something you are guessing at.
Nope. We know the general costs of the CPU, RAM, and GPU, and the margins are much higher this time around. Refer to the pricing chart from the post of mine on page 4 or 5:
"Well, for everyone that wants to make excuses about the enormous price increase of mid-level Mac Pro configurations (bumping solely the CPU to dual 2.66Ghz from a base config gets you to $5000), I made a little chart. I knew despite the platform switch and all the hype that the new CPUs shouldn't cost much more than the older Harpertown Xeons did (when they were new). What I didn't realize is that the new Nehalems Xeons that the Mac Pro is using are even CHEAPER than their Apple equivalent from the old Mac Pro!
Disclaimers:
1) These prices represent unit cost in batches of 1000 at the time of initial release. Remember, Apple does NOT lower their prices over the lifecycle of a certain model even though the CPU costs may drop somewhat. Regardless this point is moot anyhow as server grade CPUs don't change price much if at all until a new, faster model/series is released.
2) The single CPU xeon configuration (at least the base model) uses a different CPU from the Xeon 3xxx series. These prices are for the dual-CPU machines.
It doesn't matter that the basic chip is the same. The extra functions give it extra performance, they always do....
extra functions? Again, these are the same damn chips! The Core i7 extreme editions have literally IDENTICAL specifications of the Xeon 55xx series save for one of the QPI links being disabled. Same exact frequency multiplier, same exact Quickpath frequency, save exact memory controller and speed, etc.. They even all use DDR3 now!
....The point though is not even that, it's simply the point of calling things by their correct names. I just don't know why that's so difficult for you.
You must be arguing with the wrong person here. I agree with you on this point, and have never used Core i7 and Xeon 55xxx/Gainestown interchangeably...
Keep in mind these new cpu's have hyperthreading. So if you get a 2.26 you are getting 16 virtual cores. If I were you, I'd go with the new 8 core 2.26... it is going to last you much longer with snow leopard's GCD and OpenCL. Also the new cpu's offer virtualization... so better compatibility with VMWare running windows if that is a factor to you. It is to me because I have to test ie6 against web apps.
I disagree. I would recommend getting the 2.66 Ghz CPUs. Nehalem is indeed fast, but based on the Core i7 reviews I've seen, I'd be surprised if the 2.26Ghz is to going to outrun the older 3.2Ghz Harpertown. I may be wrong however. I'll get back to you on this after some Googling..
The 3000-series are server branded versions of desktop CPUs. The 5000-series are twin socket derivatives and the 7000-series is for more than two sockets.
Yep.. Am I missing something here? Did I make a typo somewhere?
Thats true, but you must consider the size of the files. That's the most important factor.
If the file is 20 MB in size, any Mac product is going to be plenty fast, Quartz, Open CL, Grand Central, or whatnot.
In fact, the slowest computer Apple made a year ago is more then fast enough for file around that size. People are getting themselves tied in knots worrying about speeds.
You are so right, and I try to tell people the same thing. You know what's crazy, is how I have delegated my machines now days. I used to separate projects I used to work on to whatever computers were available. But, now days there is a big difference on my Mac Pro, I got up to a 25 GB MPEG-2 file this week, not bragging, but I wouldn't touch that with my G4s or my MacBook. Even if I had another 25 GB file in a queue I would just wait for the Mac Pro to free up. I wouldn't think of putting it on the other machines...
So now days there are systems for thick files and thin files... and thick is not something everybody gets into. There is a big difference between a 2 Core Mac with 4 GB of RAM and 16-Core Mac (Virtually) with 32 GB of RAM.
Laters...
p.s. bored while mac pro is rendering/compressing, ugh
s
Nope. We know the general costs of the CPU, RAM, and GPU, and the margins are much higher this time around. Refer to the pricing chart from the post of mine on page 4 or 5:
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price.
You also don't know the cost to produce the machine itself, or whatever R&D cost they may add in. Then there is the overhead.
I would love to know these numbers, but none of us do. It's possible that Apple is selling less, so they have to charge more as component pricing isn't as low as it would otherwise be. We are just speculating.
I pointed out other differences. They do exist.
You must be arguing with the wrong person here. I agree with you on this point, and have never used Core i7 and Xeon 55xxx/Gainestown interchangeably...
Well, possibly, so many are doing this its getting hard to sort them all out.
I disagree. I would recommend getting the 2.66 Ghz CPUs. Nehalem is indeed fast, but based on the Core i7 reviews I've seen, I'd be surprised if the 2.26Ghz is to going to outrun the 3.2Ghz Harpertown from last generation. I may be wrong however.
I agree on this as well. It's one reason I ordered the dual 2.66 model.
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price. You also don't know the cost to produce the machine itself, or whatever R&D cost they may add in. Then there is the overhead.
I would love to know these numbers, but none of us do. It's possible that Apple is selling less, so they have to charge more as component pricing isn't as low as it would otherwise be. We are just speculating.
I will agree to a very small extent. Certainly I'm not claiming to know their exact figures, but unless there are some very strange circumstances, I can't see how their bill of materials for the new model could come anywhere close to justifying the massive price hike. I'm sure they do factor in R&D and other things, but at the moment I am only comparing the new Mac Pro to the old model, not to PC workstation manufacturers..
I pointed out other differences. They do exist.
What are they? I acknowledged that depending on the particular generation, there can be differences in clock, cache, FSB/Quickpath clock, memory support, etc. BUT, in this case the Core i7 extreme editions have literally the exact same specs as their Xeon (35xx) counterparts. The only difference both of them have with the Xeon 55xx is the fact that Intel disabled the second QPI link after manufacturing. There is no other difference whatsoever other than to be binned differently in testing for power consumption and stability. There should not be any major difference in performance..
Well, possibly, so many are doing this its getting hard to sort them all out.
understood -- I know the feeling
BTW, I apologize if I come across as an ass sometimes. I always visit these forums after a long day at work and can seem quite aggro. I honestly mean no harm, and greatly enjoy the discussion.
Originally Posted by hmurchison
Yes based on what I can see the Quad Mac Pro is a $2k computer with a hefty margin appied. It's going to get ugly when the PC World rags of the world start showing the basic $1200 Core I7 pc beating the Mac pro in performance
I don't see that happening. Besides, I'm surprised at you. You know that the Mac Pro is not a home machine. Compare it to comparable machines when they do come out.
Mel', I'm not suprised at you.
Lemon Bon Bon.
I answered that in the post above the one you made to me.
Sure, I'll go back.
There are more differences than that ECC memory is one. The possible use of FB-DIMMs in four channels is another, though it's not likely, at least right now, that Apple will be using Becton. Additionally, the i7 parts currently are all 130 watt parts, while Gainstown, except for the not yet available 3.2 GHz part are either 80 or 95 watt parts.
Also 2 QPI links are used, not just one as in i7.
We're comparing to the xeon 3500 series here, not to future chips intel isn't shipping and apple isn't using.
So no four channel memory.
And from what I have read, it looks like the xeon 3500s only have one QPI link, just like the i7. If you find documentation otherwise, I'll stand corrected.
So the only differences are ECC and using a bit less power, neither of which are going to make a difference in real world performance - the two machines will run a given app the same, the user is just paying twice as much for ECC and to use a bit less power. Oh, and for twice the price you get the ability to use LESS ram.
You have every right to complain that Apple doesn't MAKE a consumer tower. You would be right in that. I bemoan that fact as well.
Funny, with a ram limit lower than an $899 dell and about the same performance, it sure looks like the Mac "Pro" quad IS a consumer tower. Apple is just charging double for it.
And I'm not asking for a consumer machine, I AM asking for a pro machine. I just don't define "pro" as a machine built with sturdier parts but consumer level performance (or worse).
I'd be happy with a consumer machine at a consumer price. Or with a pro machine at a pro price. Apple doesn't seem to hit either market.
Most of these boxes are bought by companies or other organizations, not individuals.
Either way, it is bought for someone to sit in front of it and run apps. You really think that the majority of these buyers are willing to pay twice as much for "workstation parts" but don't care if performance is no better than a machine half the price?
The truth is that very few pros need 8 Gb RAM. People seem to like to think they do, but they don't.
But don't "very few" people really need a workstation class machine in the first place? You insist there are enough people who need ECC and whatever else allegedly makes this a Workstation...yet you don't seem to think that that is the same group that may need more than 8 gigs?
When it comes to ECC ram and the other excuses for the pricing, this machine is aimed at the small group of Professionals who can't settle for a consumer machine and insist on only the best.
But when it comes to ram limitations, then the market for this machine suddenly becomes those guys who really don't need anything that special and can settle for something LESS than what some cheap consumer boxes are offering?
I don't see how that's not a contradiction.
And for the record, I am one of those who can take advantage of more than eight gigs of ram, I have more than that in my current machine already.
It doesn't matter that the basic chip is the same. The extra functions give it extra performance, they always do.
I'm dying to see your response when the benchmarks are out and the i7 and xeon 3500 have negligible differences at the same clock speed. ECC isn't going to boost performance, is it? Nor lower power use?
So what IS going to make the 3500 faster than the i7 at the same clock speed?
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
hips! The Core i7 extreme editions have literally IDENTICAL specifications of the Xeon 55xx series save for one of the QPI links being disabled.
And in the case of the 35xx series, if what I have read is correct, that xeon has one QPI link disabled as well.
You don't know the cost to Apple or any other company in the amounts they buy them. All you know is the single bin price.
True. But when the price to the public is lower, it's a safe assumption that the price to Apple drops as well.
If a new chip sells publicly for less, but Apple ends up paying more than the previous generation, then they are doing something wrong.
I pointed out other differences. They do exist.
Well, one of the differences doesn't exist in shipping chips. A second one doesn't seem to exist in the 3500 series. And the other differences aren't ones that would have any effect on performance.
What are they?
Aside from the ECC you mentioned and the 2 qpi (which looks like it's 55xx only, not in the 35xx), it looks like the only other difference is lower power use.
And he listed more memory channels...which is only in a chip that apple is NOT using yet.
Please clarify...
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
The 5500 allows use of ECC memory, it has two QPI channels instead of one, and it uses less power. I also assume the i7 is only supported in single chip (4 core) configurations, but I could be wrong.
Otherwise the chips are identical.
Please clarify...
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
Core i7: Bloomfield 1 quickpath link Nehalem core without ECC support.
Xeon 3500: Bloomfield 1 quickpath link Nehalem core with ECC support. Same price as the core i7 branded chips.
Xeon 5500: Gainestown. Essentially bloomfield with a second quickpath link for multi-processing. Intel is also making these in more clock speeds than the bloomfield chips.
The 5500 allows use of ECC memory, it has two QPI channels instead of one, and it uses less power. I also assume the i7 is only supported in single chip (4 core) configurations, but I could be wrong.
Otherwise the chips are identical.
OK, so does that mean the i7 will find its way into the iMacs and Mac minis?
Does the i7 have a series number like the Nehalem (5500) and Woodcrest (5400)?
If the i7 and Nehalem are similar, does that the mean the Xeon 5400s are similar to the Core 2 Duos?
What will be in the future MacBooks? The Atom? The i7?
The Core 2 Duos that are in the current iMacs are a mobile (low power low heat) proc, right?
Does the i7 have a low heat/mobile version?
This wiki is a good source of info, but it doesnt explain Apple's strategy and branding.
http://en.wikipedia.org/wiki/Nehalem_(microarchitecture)
I will agree to a very small extent. Certainly I'm not claiming to know their exact figures, but unless there are some very strange circumstances, I can't see how their bill of materials for the new model could come anywhere close to justifying the massive price hike. I'm sure they do factor in R&D and other things, but at the moment I am only comparing the new Mac Pro to the old model, not to PC workstation manufacturers..
One thing I've found over the past five or six year or so is that the higher end machines are getting more expensive from every manufacturer, not just Apple. Workstations have risen in price and servers have risen in price.
With consumer machines, we see that companies make cheaper, smaller machines all the time, and that makes it seem as though prices are going down, but they are not. Those cheaper machines are lower spec'ed all around. But the $1,000 machines are still there. The $1,500 machines are still there, etc, all the way up. Alien still has consumer grade machines for $3,000, as do others.
I'm surprised the new Mac Pros have gone up as much as they have. I expected a small increase. but I doubt if Apple is being nasty. When other comparable Xeon machines arrive, they will also be expensive. This has been true for Harpertown machines, and those further back.
Look at workstan manufacturers to see this. I mentioned a well known one Boxx. You idn't respond to that. This is what Apple is competing with, and their prices are higher than Apple's are by mre than a small bit.
Even looking at the home built model shown here doesn't use the best parts. There are much more expensive cases, power supplies etc.
What are they? I acknowledged that depending on the particular generation, there can be differences in clock, cache, FSB/Quickpath clock, memory support, etc. BUT, in this case the Core i7 extreme editions have literally the exact same specs as their Xeon (35xx) counterparts. The only difference both of them have with the Xeon 55xx is the fact that Intel disabled the second QPI link after manufacturing. There is no other difference whatsoever other than to be binned differently in testing for power consumption and stability. There should not be any major difference in performance..
The difference in the 35xx parts are marginal, I know, except for ECC support. But that just shows the base cost of the machine itself. Again, these are workstations, built to a much higher standard. Why do they cost more? WE don't know. You can't assume that Apple simply decided to raise prices for no reasons of cost to them. If that were true, then their competitors, who are not the consumer manufacturers, come out with their machines, Apple will be at a competitive disadvantage. That hasn't happened in the workstation space before, and I see no reason why it would be true today.
understood -- I know the feeling
BTW, I apologize if I come across as an ass sometimes. I always visit these forums after a long day at work and can seem quite aggro. I honestly mean no harm, and greatly enjoy the discussion.
Believe it or not we ALL come across as asses at times. It's just that some people don't want to admit it about themselves, so you get credit where they don't.
Mel', I'm not suprised at you.
Lemon Bon Bon.
You have the right to agree with anything you think you should.
We're comparing to the xeon 3500 series here, not to future chips intel isn't shipping and apple isn't using.
So no four channel memory.
And from what I have read, it looks like the xeon 3500s only have one QPI link, just like the i7. If you find documentation otherwise, I'll stand corrected.
I'm looking at the entire family, not just the low end chip. but I grant that there isn't too much difference there.
So the only differences are ECC and using a bit less power, neither of which are going to make a difference in real world performance - the two machines will run a given app the same, the user is just paying twice as much for ECC and to use a bit less power. Oh, and for twice the price you get the ability to use LESS ram.
ECC is very important to more than a few users, so don't denigrate that. Lower power makes for a cooler machine, a quieter machine, and less total power use. This is important where there are installations of a number of them.
Funny, with a ram limit lower than an $899 dell and about the same performance, it sure looks like the Mac "Pro" quad IS a consumer tower. Apple is just charging double for it.
We don't yet know if that 8 GB is a hard limit, or just marketing. As I pointed out earlier, Apple has done this many times in the past. PC manufacturers are quick to tout every new feature and capacity increase. Apple is much more conservative.
And I'm not asking for a consumer machine, I AM asking for a pro machine. I just don't define "pro" as a machine built with sturdier parts but consumer level performance (or worse).
Well, performance certainly won't be worse. But many institutions that buy these machines are not looking for ultimate performance, or they would buy the more expensive machines, but rather reliability, which these machines do provide.
[quote]
I'd be happy with a consumer machine at a consumer price. Or with a pro machine at a pro price. Apple doesn't seem to hit either market.[quote]
While I agree with the first sentence, I don't agree with the last.
Either way, it is bought for someone to sit in front of it and run apps. You really think that the majority of these buyers are willing to pay twice as much for "workstation parts" but don't care if performance is no better than a machine half the price?
Not all workstations have people sitting in front of them. Many are used for rendering, or other purposes, headless. It's not likely the low end machine will be used for that, but the dual machines certainly will.
But don't "very few" people really need a workstation class machine in the first place? You insist there are enough people who need ECC and whatever else allegedly makes this a Workstation...yet you don't seem to think that that is the same group that may need more than 8 gigs?
You don;t get points for that, because as I said, most of these machines are bought by companies or other institutions. Price is less of a problem for them, because the competition is at the same price level.
As far as the RAM goes, we're back to the "we don't yet know". So let's drop that until we do, ok? Even so, not every use needs that much RAM. That's a myth.
[quote]
When it comes to ECC ram and the other excuses for the pricing, this machine is aimed at the small group of Professionals who can't settle for a consumer machine and insist on only the best.
But when it comes to ram limitations, then the market for this machine suddenly becomes those guys who really don't need anything that special and can settle for something LESS than what some cheap consumer boxes are offering?
I don't see how that's not a contradiction.[/quote}
PLEASE, drop the RAM argument for now.
And for the record, I am one of those who can take advantage of more than eight gigs of ram, I have more than that in my current machine already.
Good.
I'm dying to see your response when the benchmarks are out and the i7 and xeon 3500 have negligible differences at the same clock speed. ECC isn't going to boost performance, is it? Nor lower power use?
So what IS going to make the 3500 faster than the i7 at the same clock speed?
I'm dying to see yours as well.
Please clarify...
Is the Nehalem 5500 Xeon proc the same as the "i7"? I see both names referencing the same CPU class. If you Google Nehalem you will see a lot of confusing info.
I thought the i7 was the replacement of the Core 2 Duo proc class, not the Xeon proc class, right?
Yes, it is.