Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Also, do not be confused by Intel's naming protocols. The 2.8 GHZ core i7 on offer is NOT a Nehelem chip, but rather uses the same Lynnfield architecture of the 2.6 core i5 chip. Both use dual channel memory controllers (as opposed to the faster tri-channel memory for Nehelem chips), are capable of only half the video bandwidth of the Nehelem chip and also utilizes a higher latency, slower interconnect bus than the Nehelem core i7's.

Nehalem is not the name of Intel's first generation i7 900 CPU and its Xeon derivative (they were called Bloomfield and Gainestown respectively) but the name for the entire new architecture. So, Lynnfield is merely another member of the Nehalem family.

Read it up here.

Furthermore, the i7 9xx is not much faster in most applications than the i7 8xx. Only in very specialised benchmarks like specInt or specFP you'll see the benefits of the triple-channel memory controller. And even then it's not more than 10-12%. In real world application the i7 900 is about 0-2% faster than the i7 800 at the same clock speed.
The faster interconnect of the i7 900 is only relevant for multi-CPU systems and since neither i7 900 nor i7 800 are available in multi-CPU configurations, this is simply a "virtual" disadvantage.
SLI systems might suffer from the lack of PCI express lanes, but that will concern only a tiny minority of people. AFAIK, SLI is not even possible on a Mac.

The power consumption of the new model both idle and under load is MUCH lower. That will outweigh the tiny performance disadvantage for the vast majority of people.
 
Thanks for this insightful quote. Interestingly, the Apple Store (NL at least) says the i7 iMac has a 'Nehalem' processor. Mistake on Apple's part?

Well not everything that people write here in lenghty postings and that sounds technical is in the end correct.

Of course, the Apple Store is correct in calling the i7 a Nehalem processor.
 
So is the core i5 even for games better than the 3,33 Ghz core 2 duo?


No it's not. Nearly all games are single-threaded, dual-threaded at best. They crave for CPU clock, not for more cores. Also, if you read up on the benchmarks on the net, you will find that a Core CPU is not slower than a Core i5 or i7 CPU at the same clock speed in the gaming department.

Thus, a Core 2 Duo at 3.33 GHz is the fastest you can get for games.
 
Long time Mac user and Enthusiast here.

My first mac was a G4 Dual 450, with dual 17" studio displays. $3,500 :D Most I've ever spent on a computer.

Before that computer I was a mac hater (for no intelligent reason). Probably because I enjoyed building and tinkering with computers and macs took that away.

About 2 months ago I made a dell mini run Leopard (haven't upgraded to snow on it yet) Hackbook mini. It's a great computer, no KP issues at all, and it's perfect for sitting in bed and surfing around. Plus it was fun too tinker again. :)

After building this computer it made me want to take a stab at a Hack Pro. Before committing myself to expensive equipment I decided to limit my budget to $400. With that budget I built the computer I'm typing this on. It is a Gigabyte G31 ES2L, 4GB RAM with an Intel E5200 Dual Core 2.5 OC'd to 3.4, and Geek benching at 5500. I was very very impressed.

I decided I would build an i7, thinking I could score an easy 9,000 ish, with modest settings. I ordered a motherboard and processor. Then the new imacs came out. :D

I knew the i7's wouldn't be available but I ran to Best Buy to run Geekbench on the 3.06's I couldn't get a score above 4500, Made me feel pretty good about my home build. And made me think the i7 imac would probably score around a 7500.

Now my i7 build could be had for around $1200, and Benchmark substantially faster than the C2D imacs, and I've got a nice display 23" 1920X1200, but that 27" had me drooling. For me to get anything close to that monitor My build would cost $200 more than the imac i7.

Benchmarks came out. Over 9,000!! Wow... I sold my i7 mobo and processor yesterday. There is no way I can build a computer that looks as nice as that imac and has a comparable display. I can't wait until mine gets here :)

Of course I'm sure I'll still build several little hack boxes because it's fun to tinker. And just so everyone knows, I buy Retail Snow Leopard disks for my hack installations. Hackintoshes have come a long way, and believe it or not the community is fully supportive of apple. It's just a bunch of apple enthusiasts who love to tinker too. Most of us own Factory macs, and our own hacks.

I Personally own:
i7 imac on the way
(2007) imac 20" C2D
Macbook 2.0 C2D
Macbook pro LED display 15"
ibook G4
Mac mini G4s
Powermac G4 Dual 450 (can't bring myself to ditch the machine that started it all)
Powermac G5 Dual 1.8 (with OSX server)

Apple is getting plenty of love from me.
 
It's clear even from the early benchmarks that the new iMacs represent a closing of the gap between the iMacs and Mac Pros. There's no longer a crying need for a "mini tower" version of the Mac Pro.
Except that there you only can get mid-range performance in an all-in-one. Arguably, Apple lacks a decent mid-range computer-only solution between Mac Mini and Mac Pro.

Perf.---Computer---All-in-One
Low----mini C2D---iMac C2D
Mid-----??-----------iMac i5/i7
High----Pro----------[Not required]
 
The Chips are differebt

Haha where are you getting this nonsense? Lynnfield is Nehalem. Just like Bloomfield is nehalem. And just like Beckton and Gainstown are Nehalem. Nehalem is an architecture, not a specific CPU. The bloomfield chips are the high-end socket 1366 chips with 3 memory channels. Lynnfield only has two memory channels but tests have show there is no significant decrease in performance as you rarely saturate 3 channels of memory bandwith, so its a moot point. Also I have no idea where you are getting this "video bandwidth" figure from..nor do I even know what exactly video bandwidth is or how its CPU related. Also Lynnfield integrates the PCIexpress controller so there is A LOT lower latency to your videocards/gpus..not higher. Lynnfields are a 1 chip solution because of this and dont have a northbridge and there is no need for the same Quickpath bus as on Bloomfields. Its unnecessary and a moot point as you will never have a 2 socket lynnfield system and the high speed bus, quickpath is unneeded. Also Lynnfields officially support ddr3-1333 as opposed to ddr-1066 on the higher end bloomfields, so lynnfield actually has more memory bandwidth ( this is official support tho because you can easily run higher memory on both). Also hyper-threading does not cause the cpu to overheat and does it rarely cause worse performance. Get your facts strait before you post man. ;)



Was just going to say the same thing.:)

Before you go off spouting like some expert, the chips ARE different. They are not interchangeable, require different motherboards, support different amounts of and speeds of memory, support different numbers of PCIe lanes, differ in whether they support hyperthreading, have different numbers of quicklink interconnects, are tuned differently with respect to the increase in speed and length of time spent in Turbo mode and have different wattage envelopes. If this doesn't make them different chips, then I don't know what does.

The i5 and i7 Lynnfield chips are consumer chips as opposed to the Nehelem chips which are directed at the enthusiast section of the market. The Lynnfields have significant advantages for use in the iMac over the Nehelem chips, in particulr the fact that they draw less power and run cooler, clock cycle per clock cycle. They are limited compared to Nehelem in several ways, including memory bandwidth, amount of memory supported, the number of total lanes supported in PCIe connections and the number of Quick Connect interlinks, Intels answer to AMD's Hypertransport, This makes Lynnfields slower and less optimal for gaming. Of course, Intel decided to use the i7 label on Lynnfiewld chips to confuse consumers, just as you are apparently confused. Get your facts straight before you pull out your can of whip ass.

AS for hyperthreading, It most certainly does draw significantly more power. This ios why reviewers turn it off when they measure power consumtion at idle and at load. In a normal PC, activating hhyperthreading isn't much of a problem, but in the space confines of an iMac, I worry a bit because this is a brand new architecture for Apple, and Apple hasn't always had the best of luck in product design on thier 1.0 releases. Are you old enough to remember the Mirror Door Power Macs? They sounded like vacuum cleaners and still had a tendency to over heat. A number of Apple's recent laptop offerings have also run so hot you can't use them on your lap if you intend to procreate. And I stand by my comment that hyperyhreading can degrade system performance. I may not have correctly explained why it happens, but the fact that it does can be seen in any number of reviews on line. I don't have the time to hunt them down right now to satisfy your puerile inability to check for yourself, but if you call me out again, I will.

Love and kisses.
 
Actually it is not

Well not everything that people write here in lenghty postings and that sounds technical is in the end correct.

Of course, the Apple Store is correct in calling the i7 a Nehalem processor.

These are Lynnfield processors. Period. They are not Nehelems. Visit the Intel website and see for yourself.

By all means, ignore what I have to say. Go ahead and buy an i7 system. I guarantee that you won't see a 35% boost in performance on real world apps compared to the i5. And since no one has had much experience with the systems yet, no one can say what temps and fan noise will be l;ike with hyperthreading always on.

I was merely trying to sound a note of caution.
 
SLI systems might suffer from the lack of PCI express lanes, but that will concern only a tiny minority of people. AFAIK, SLI is not even possible on a Mac.
There's a lot of interesting issues occurring in regards to SLI/CrossFire and the P55.

Ideally, if you use SLI or CrossFire on a Lynnfield system, then the two cards will default to using 8x PCIe lanes for each, so on most graphics cards before the current Radeon 58** series (and the upcoming GT300 cards), the impact should be negligible in terms of bandwidth concerns.

However, in game testing, there seems to be an issue where going with a SLI or CrossFire setup on P55 results in poorer performance scaling in games compared to the X58 chipset. It's especially evident in quad GPU configurations (albeit, this is a small % of the market). As single-GPU cards though continue to push increased bandwidth thorough put, you'll begin to see even two card SLI/CrossFire setups more heavily impacted.

The reality is though, that most enthusiasts going for a CrossFire or SLI setup, would go with the X58 chipset anyway, as it allows drop-in upgrades for Gulftown when it debuts. There's also the fact that Intel themselves have stated that socket 1156 will likely remain solely in the quad core market (you'll likely never see six-core or more 1156 processors).

As you mentioned though, SLI/CrossFire is currently not available with Macs anyway.

The power consumption of the new model both idle and under load is MUCH lower. That will outweigh the tiny performance disadvantage for the vast majority of people.
For something like the iMac, yeah, the power consumption difference is huge. P55 is the much better option, lol.

For desktop systems, it's not really a big deal.
 
These are Lynnfield processors. Period. They are not Nehelems. Visit the Intel website and see for yourself.

By all means, ignore what I have to say. Go ahead and buy an i7 system. I guarantee that you won't see a 35% boost in performance on real world apps compared to the i5. And since no one has had much experience with the systems yet, no one can say what temps and fan noise will be l;ike with hyperthreading always on.

I was merely trying to sound a note of caution.
Hi. So before you try and open your mouth and look even more like an uneducated troll, maybe you should do a little bit of research yourself first.

Here is an Intel whitepaper on the Lynnfield-based i5 and i7: http://download.intel.com/products/processor/corei7/319724.pdf. Notice that the URL is from Intel's website itself. Here's the URL from the page you can access it from: http://www.intel.com/products/processor/corei7/index.htm

Now, if you actually open the article and read it, you'll see this blatantly obvious line right away: "Intel Core i7-800 Processor Series and the Intel Core i5-700 Processor Series Based on Intel Microarchitecture (Nehalem)"

Notice that last word: Nehalem.

Exactly.

Before you go off spouting like some expert, the chips ARE different. They are not interchangeable, require different motherboards, support different amounts of and speeds of memory, support different numbers of PCIe lanes, differ in whether they support hyperthreading, have different numbers of quicklink interconnects, are tuned differently with respect to the increase in speed and length of time spent in Turbo mode and have different wattage envelopes. If this doesn't make them different chips, then I don't know what does.
No one is debating that Lynnfield is a different implementation of Nehalem than Bloomfield, etc. However, they are *all* Nehalem. Nehalem is the underlying microarchitecture that all of them are based on. When Intel first released Nehalem-based products, they essentially offered two variants: Gainestown, for the server/high-end workstation markets, and Bloomfield, a "consumer" variant that is targeted towards the high-end enthusiast crowd. Bloomfield was never intended to be a mainstream consumer processor line. In reality, Bloomfield was simply Gainestown with a few features disabled or unsupported (the most important two being one QPI link being disabled and the lack of ECC support in the supporting X58 chipset).

Intel wanted to release a Nehalem-based consumer product, but there was no way they could affordably do so with Bloomfield. Also, as others have mentioned, the Bloomfield/X58 combination consumes a lot of power and has considerable thermal output, so it just wasn't ideal for the mainstream market.

Thus, Intel took the Nehalem microarchitecture. They eliminated the QPI links altogether, eliminated a memory channel (which is why Lynnfield only has two), and moved many of the features of a traditional northbridge chipset onto the processor die itself (such as the PCIe links). They also enhanced the Turbomode and reduced its power consumption (partly due to the elimination of some features). Thus, it is definitely a different implementation of Nehalem than the likes of Bloomfield, but it *is* Nehalem, simply in a form more conducive to the mainstream market.
 
I agree

Nehalem is not the name of Intel's first generation i7 900 CPU and its Xeon derivative (they were called Bloomfield and Gainestown respectively) but the name for the entire new architecture. So, Lynnfield is merely another member of the Nehalem family.

Read it up here.

Furthermore, the i7 9xx is not much faster in most applications than the i7 8xx. Only in very specialised benchmarks like specInt or specFP you'll see the benefits of the triple-channel memory controller. And even then it's not more than 10-12%. In real world application the i7 900 is about 0-2% faster than the i7 800 at the same clock speed.
The faster interconnect of the i7 900 is only relevant for multi-CPU systems and since neither i7 900 nor i7 800 are available in multi-CPU configurations, this is simply a "virtual" disadvantage.
SLI systems might suffer from the lack of PCI express lanes, but that will concern only a tiny minority of people. AFAIK, SLI is not even possible on a Mac.

The power consumption of the new model both idle and under load is MUCH lower. That will outweigh the tiny performance disadvantage for the vast majority of people.

I don't disagree with many of your points, except to say that the people should understand exactly what they are getting. Lynfields are significantly from different from the original Nehelems, which were intended for the enthusiast segment of the market. The faster interconnect actually does make a difference in some applications that are memory bandwidth starved, even in single core nehelem i7's. The issue of the PCIe bus likely won't affect end users for the reasons that you've pointed, mostly because Apple offerings have never been strong on video performance in games. And yes, SLI is still not possible, even with Mac Pros which is rather a tragedy because it forces a significant market segment to shun macs.

My reason for the original post was informational. It is important for people to understand that a 35% boost in performance on a synthetic benchmark simply doesn't reflect actual results on real world software. (It's a shame, but no one has developed the kinds of real world performance benchmarks that are available for the PC.) Intel is being disingenious in calling the hyperthreading Lynnfields i7's, given the numerous architectural differences. Every review site that I've read, and I've looked at about a dozen, agrees. Finally, because this is a brand new iMac design from a number of standpoints, I sincerely believe it is too early to tell if hyperthreading will be more trouble than it's worth. I still remember the mirror door G5 Mac Pros. They were an abomination to the ears and the energy bills. They also acted like space heaters. It seems at least possible to me that the Lynnfield i7 could cause problems.

I can't waste anymore time on this. I've responded three times. Each of you will decide for yourselves in the end. I simply, and humbly, suggest you educate yourself first before pulling the trigger on the top of the line i7. If I'm a jerk for doing, so be it.
 
Thus, a Core 2 Duo at 3.33 GHz is the fastest you can get for games.
I'm actually fairly certain that the "fastest you can get for games" would be the Intel Core i7 975 processor. Admittedly you'd be dumping $1k for it to do so.
 
Ughh a Celeron is just a P$ by any other name

Hi. So before you try and open your mouth and look even more like an uneducated troll, maybe you should do a little bit of research yourself first.

Here is an Intel whitepaper on the Lynnfield-based i5 and i7: http://download.intel.com/products/processor/corei7/319724.pdf. Notice that the URL is from Intel's website itself. Here's the URL from the page you can access it from: http://www.intel.com/products/processor/corei7/index.htm

Now, if you actually open the article and read it, you'll see this blatantly obvious line right away: "Intel Core i7-800 Processor Series and the Intel Core i5-700 Processor Series Based on Intel Microarchitecture (Nehalem)"

Notice that last word: Nehalem.

Exactly.

Based on does not mean identical to. Intel Celeron chips were always based on the faster running, more powerful cousins, such as Core Duops, P4s and P3s. It was in Intel's commercial interest to confuse the public on thier capabilities. Do you always regurgitate what the vendor posts on its websites? Or do you ever read reviews from independent hardware reviewers. Before you post again, and embaras yourself, read a little more broadly and stop being an Intel shill. Perhaps you own stock?

You will not, repeat, will not see a 35% difference in performance between the Lynnfield core i5 2.6 GHZ and the Lynnfield core i7 2.8 GHZ chip. Maybe 5-7 % in a few, highly parallelized applications, such as video encoding, but note that you'll only see this if the chip remains within its wattage envelope. If its temperature exceeds the maximum prescribed at anytime, the chip first shuts off hyperthreading and then the cores, one by one. The maximum boost is contingent on all cores, real and virtual, firing on full throttle. It really depends on how well Apple has designed its cooling in the latest iMacs. And given how many different directions Apple is going now, I'm not sure that the 1.0 release of the Lynnfield core i7 quad core will necessarily hold up under heavy continued use. But by all means, order one and let us all know. I;m more than happy to let early adopters suffer the pain so that I can make a rational and economically viable choice. Myself, I'll be getting the i5, doubling my RAM and adding a 1.5 TB backup drive for the price of the so called i7. But that's just me.
 
No disagreement

Hi. So before you try and open your mouth and look even more like an uneducated troll, maybe you should do a little bit of research yourself first.

Here is an Intel whitepaper on the Lynnfield-based i5 and i7: http://download.intel.com/products/processor/corei7/319724.pdf. Notice that the URL is from Intel's website itself. Here's the URL from the page you can access it from: http://www.intel.com/products/processor/corei7/index.htm

Now, if you actually open the article and read it, you'll see this blatantly obvious line right away: "Intel Core i7-800 Processor Series and the Intel Core i5-700 Processor Series Based on Intel Microarchitecture (Nehalem)"

Notice that last word: Nehalem.

Exactly.


No one is debating that Lynnfield is a different implementation of Nehalem than Bloomfield, etc. However, they are *all* Nehalem. Nehalem is the underlying microarchitecture that all of them are based on. When Intel first released Nehalem-based products, they essentially offered two variants: Gainestown, for the server/high-end workstation markets, and Bloomfield, a "consumer" variant that is targeted towards the high-end enthusiast crowd. Bloomfield was never intended to be a mainstream consumer processor line. In reality, Bloomfield was simply Gainestown with a few features disabled or unsupported (the most important two being one QPI link being disabled and the lack of ECC support in the supporting X58 chipset).

Intel wanted to release a Nehalem-based consumer product, but there was no way they could affordably do so with Bloomfield. Also, as others have mentioned, the Bloomfield/X58 combination consumes a lot of power and has considerable thermal output, so it just wasn't ideal for the mainstream market.

Thus, Intel took the Nehalem microarchitecture. They eliminated the QPI links altogether, eliminated a memory channel (which is why Lynnfield only has two), and moved many of the features of a traditional northbridge chipset onto the processor die itself (such as the PCIe links). They also enhanced the Turbomode and reduced its power consumption (partly due to the elimination of some features). Thus, it is definitely a different implementation of Nehalem than the likes of Bloomfield, but it *is* Nehalem, simply in a form more conducive to the mainstream market.

You are entirely correct. But Intel has a history of confusing consumers with its naming conventions in order to push product. For much of the last twenty years, Intel; has sold Celeron chips as P3's, P4's and Core Duos. Of course, the Celeron was based on the same microarchitecture, but had so many things pulled out, so many compromises, that the performance of the chip suffered considerably. Nevertheless, these chips were advertised as Celeron P4s and Core Duos with no mention of the fact that they were hobbled. Now I'm not suggesting that the Lynnfields are hobbled like the Celerons, merely that they include compromises that people may not be aware of if we simply call all of these chips Nehelems, and don't distinguish between subclasses which behave differently. My real intent behind the post was not to get into a pissing match over nomenclature. But rather to call attention to two issues that troubled me. First, the implication of the post that the 2.8 GHZ i7 would perform 35% faster than the 2.6 GHZ i5. It won't. And all the reviews that have included these chips with the Bloomfields along with the faster Core 2 Duos show this. The i5 IS significantly faster than the Core2Duos, even the 3.3 GHZ variety, because of the additional cores and most likely because of the TurboBoost function that will crank the 2.6 chip up to 3 GHZ when not at load. Second, I felt it importnat to piont out that hyperthreading is still a marginal proposition under most conditions and actually can result in slower performance on some software. Simply do a Google search of recent reviews on hyperthreading. Given Apple's full plate, and given its failures in the past with regard to system thermal output (I'm thinking of the mirror door G4s and the hot running G5s that ultimately forced Apple to institute water cooling of its systems), isn't prudent to at least question whether the core i7 iMac on offer is necessarily the best option for best performance in the iMac line?
 
willdenow said:
I don't disagree with many of your points, except to say that the people should understand exactly what they are getting.
I agree, they should. And Apple isn't being misleading about what the i5 750 or i7 860 are. They *are* Nehalem-based processors.

willdenow said:
Lynfields are significantly from different from the original Nehelems, which were intended for the enthusiast segment of the market.
You're mis-using the term Nehalem, as if it is a product itself. Nehalem is nothing more than the underlying microarchitecture, which Intel then readily adapts for various markets. They've been doing this pretty much since the first time where the product lines diverged to provide consumer and server/workstation/etc variants.

The faster interconnect actually does make a difference in some applications that are memory bandwidth starved, even in single core nehelem i7's.
No, what has made the difference in memory bandwidth-intensive applications is the third memory channel. Don't confuse the two. For most performance applications, there's been little shown to support that Lynnfield has suffered because Intel went with a DMI link instead of the QPI link. The QPI links are more critical for multi-socket systems, which is why Intel even *disabled* one of the lanes for Bloomfield - consumer X58-based systems were not going to have multiple sockets, and so the additional QPI link was redundant/not needed.

The issue of the PCIe bus likely won't affect end users for the reasons that you've pointed, mostly because Apple offerings have never been strong on video performance in games.
That's generally been more to Apple's slow support/lack of desire to provide the latest GPU hardware (and yes, I realize it's partly due to GPU manufacturers seeing a limited market for top-end hardware also, so the blame goes to both).

For pretty much every GPU Apple currently offers through its store or via BTO options, the P55 chipset would be more than adequate for SLI/CrossFire if it were available.

And yes, SLI is still not possible, even with Mac Pros which is rather a tragedy because it forces a significant market segment to shun macs.
You realize that one of the biggest criticisms until recently about SLI was its lack of multi-monitor support? How then would a Mac Pro user, many of whom likely utilize dual monitors at least for what they're using it for, benefit from an SLI setup?

SLI and CrossFire combined make up a decent percentage of the gaming market, yes. But the gaming market is largely confined to the Windows PC market, and that's unlikely to change any time soon. Given the other reasons listed above, even if Apple started to support SLI and/or CrossFire, I don't think you'd see it make much of a difference.

My reason for the original post was informational. It is important for people to understand that a 35% boost in performance on a synthetic benchmark simply doesn't reflect actual results on real world software.
That's often the case for *all* new processor releases, however. It has more to do with the fact that benchmarks are designed to stress specific aspects of a hardware's performance under extreme circumstances. The PC enthusiast community often use Prime95 to stress-test an overclocked system's stability for an extended period of time (usually 12-24 hours, sometimes up to 48 hours). Do you really think the standard iMac, Mac Pro or even PC user does anywhere near anything that hardware-intensive? I doubt it.

(It's a shame, but no one has developed the kinds of real world performance benchmarks that are available for the PC.)
Most enthusiasts laugh/scoff at "real world performance benchmarks" though. They're mainly used by hardware review sites to try and provide a visual comparison between products that a user can semi-understand. It's still generally a closed environment, testing specific features, and no where near realistic in terms of what the general user will be doing. Think of all of the applications usually found running in the background on a system, and that alone should give you an idea at just how unrealistic synthetic benchmarks are.

Intel is being disingenious in calling the hyperthreading Lynnfields i7's, given the numerous architectural differences. Every review site that I've read, and I've looked at about a dozen, agrees.
Show me a credible review site that claims it's a dubious distinction by Intel. Most of the reputable hardware review sites that reviewed Lynnfield, never seemed to have an issue with Intel stating that Lynnfield is a Nehalem-based derivative, nor that they used the i7 moniker.


Finally, because this is a brand new iMac design from a number of standpoints, I sincerely believe it is too early to tell if hyperthreading will be more trouble than it's worth. I still remember the mirror door G5 Mac Pros. They were an abomination to the ears and the energy bills. They also acted like space heaters. It seems at least possible to me that the Lynnfield i7 could cause problems.
Why would hyperthreading cause problems? In its original implementation with the Netburst-based P4 architecture, it ended up not being a major issue because there was very little consumer software that made use of multiple cores, and thus the benefits of hyperthreading were generally lost.

Times have changed though, and a lot more software can take advantage of it. Sure, it's still not going to see its full potential *right now*, but each year features more and more multi-core software being released, and so it doesn't hurt for Intel to go ahead and include it.

And I think you lack an understanding of how TDP design works, but the 95W TDP that Lynnfield is rated for, is the *maximum* power consumption you will see it utilize. Thus, even with all four physical cores active and all four hyperthreading logical cores active, you will not see it draw more than 95W TDP.

Along those same lines, an intensive single-threaded application will have TurboMode running at a faster processor speed, and at that time, the processor will still be drawing its maximum 95W TDP. Do you think Apple has not adjusted the 27" iMac's thermal design to account for this?

I can't waste anymore time on this. I've responded three times. Each of you will decide for yourselves in the end. I simply, and humbly, suggest you educate yourself first before pulling the trigger on the top of the line i7. If I'm a jerk for doing, so be it.
You're not a jerk, you're simply uneducated when it comes to Lynnfield and its Nehalem roots.

And there's no way someone can deny that the i5 750 and i6 860 will offer performance increases over the previous Core 2 line, especially since every review of Lynnfield has compared it to the Core 2s and shown that it handily beats them.

Now, if your argument was on the graphical performance of the new 27" iMacs, then you could have made your case, as the mobile 4850 has no business powering intensive graphical applications at that resolution. But that's for another debate.
 
These are Lynnfield processors. Period. They are not Nehelems. Visit the Intel website and see for yourself.
Intel says:
Lynnfield was the codename for the four core desktop Intel® Core™ processors for the Intel 5 series chipset based platforms. Lynnfield was based on Intel microarchitecture codename Nehalem and launched in September 2009.
So, Intel claims Lynnfield is Nehalem. If that is useful or truthful information I do not know, but they do claim that.

I think a warning against 35% performance difference is fine (I agree this is something that probably will not end up being a true gain) and informed people will agree. I also think a warning about Macs becoming noisy is fine too. I had a QuickSilver G4 once, which was not as bad as the Mirrored Door version, but still abominable. And I traded it for a 500MHz G4 Cube, slower but really, really quiet. These days, Apple laptops and minis start cranking up the ventilator as soon as you do anything resembling vaguely serious work, so heat is something I am weary about, as it translates in Apple to noise and I am one of those people who really dislike noise. (Jobs does too, but probably he does nothing serious with his machines so he never notices how fast these machines move to vaccuum cleaner mode. I am exaggerating, I know.)
 
Wow! Even Wikipedia can get it right:
Lynnfield is the code name for a quad-core processor from Intel released in September 2009.[1] It is sold in varying configurations as Core i5-7xx, Core i7-8xx or Xeon X34xx. Lynnfield uses the Nehalem microarchitecture and replaces the earlier Core microarchitecture based Yorkfield processor, using the same 45 nm process technology, but a new memory and bus interface.
http://en.wikipedia.org/wiki/Lynnfield_(microprocessor)

People just don't want to do research anymore, do they?
 
willdenow said:
You are entirely correct. But Intel has a history of confusing consumers with its naming conventions in order to push product.
But benchmarks at many different reputable hardware sites show the i7 860 as essentially equal to the i7 920 in everything but memory-intensive applications, and even then the difference isn't *that large*. It mainly comes down to a difference of features that an enthusiast would be concerned with. But enthusiasts are but a small minority of the overall PC market, so there's no reason for consumers to be concerned. They're not being ripped off by Intel.

willdenow said:
For much of the last twenty years, Intel; has sold Celeron chips as P3's, P4's and Core Duos. Of course, the Celeron was based on the same microarchitecture, but had so many things pulled out, so many compromises, that the performance of the chip suffered considerably. Nevertheless, these chips were advertised as Celeron P4s and Core Duos with no mention of the fact that they were hobbled.
Until recently, when Intel publicly stated that the "Pentium" moniker was going to refer to a lower-cost market that exists between that of the Celerons and the Core 2 lineup, I can't recall any time when Intel tried to pawn off Celerons as anything but Celerons.

I would very much be interested in seeing a story where the product stated "Core Duo" and yet it was in fact a Celeron. Of course Intel is going to base the Celeron off of existing microarchitectures: it's far cheaper to disable some components of the existing microarchitecture, than it is to create a whole new microarchitecture solely for the low-cost market.

However, during the P3 days they were clearly "Celerons", during the P4 Netburst era they were still "Celerons", and as far as I know they're still branded as "Celerons" clearly for the Core/Core 2 derivatives.

Now I'm not suggesting that the Lynnfields are hobbled like the Celerons, merely that they include compromises that people may not be aware of if we simply call all of these chips Nehelems, and don't distinguish between subclasses which behave differently.
Then by your argument Bloomfield contains "compromises" since it has its second QPI link disabled, and also doesn't support ECC. Is Bloomfield thus not an "i7" or a "Nehalem-baed" processor?

My real intent behind the post was not to get into a pissing match over nomenclature. But rather to call attention to two issues that troubled me. First, the implication of the post that the 2.8 GHZ i7 would perform 35% faster than the 2.6 GHZ i5.
Remember though that they're not likely to state when TurboMode is enabled and, when it is , change their naming nomenclature to indicate "well, this stock 2.8 Ghz i7 860 was actually running at 3.46 Ghz". I also haven't looked at the benchmarks myself, but if they're multi-threaded, than the i7 860 would have an advantage there as well.

In situations where both Turbomode *and* hyperthreading could play a role, I have no doubt that the i7 860 could show a good performance increase over the 750 (remember, the i5 750 lacks hyperthreading, while the i7 860 has as many HT logical cores as the 9** series).

I will have to look into what the source of the 35% difference exactly was though, according to that article.

And all the reviews that have included these chips with the Bloomfields along with the faster Core 2 Duos show this. The i5 IS significantly faster than the Core2Duos, even the 3.3 GHZ variety, because of the additional cores and most likely because of the TurboBoost function that will crank the 2.6 chip up to 3 GHZ when not at load.
Ugh, what? The 750 is typically faster than the top-end Core 2s, but not by much (and there are even times when the Core 2s are still faster). Here's a good look at them: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634&p=10. It's on the first page of the benchmarks, so that you can scroll through them.

The i5 750 is not "significantly" faster than the Core 2 Duo E8600 or Core 2 Quad EEs. There are times where it falls behind, times where it beats them by a bit, and a few times where it's a decent bit ahead. But it's far from conclusive to say it's a huge jump in performance.

Second, I felt it importnat to piont out that hyperthreading is still a marginal proposition under most conditions and actually can result in slower performance on some software. Simply do a Google search of recent reviews on hyperthreading.
Well, a lot of hyperthreading's problems were also the result of being coupled with operating systems that didn't necessarily have great multi-core support (Windows XP comes to mind). Multi-core and hyperthreading support under the likes of Windows 7 and Linux (and to a lesser extent Mac OS X) are much more improved. And for any applications where multithreading is present, the benefits of Hyperthreading will be available. Photoshop is probably one of the best examples of a widespread, mainstream professional application where hyperthreading will be beneficial.

Given Apple's full plate, and given its failures in the past with regard to system thermal output (I'm thinking of the mirror door G4s and the hot running G5s that ultimately forced Apple to institute water cooling of its systems), isn't prudent to at least question whether the core i7 iMac on offer is necessarily the best option for best performance in the iMac line?
Why would it be? Lynnfield runs fairly cool. At load, sure, it pulls 95W TDP. But how often do you think most users will actually run at full load? I'm guessing, not that often. And I'm quite certain Apple has designed the iMac to allow for those periods of near-load or full load performance.

Remember, in reality, Lynnfield features quite a few enhancements in regards to power management, and thus I'm willing to bet that for the most part, it'll likely stay far below its max 95W ceiling. I'm not too concerned about the issue of heat and Lynnfield when it comes to the iMac (I mean, I'm running a i7 920 CrossFire system in a mATX case, and even my thermal performance isn't that bad :p).
 
I'm glad my Hackintosh stomps the new iMac by a good 3,000 points, did I mention the GTX 285 mine is running as well? ;) Oh and the upgradable to 24gb of ram..and the option to upgrade to Gulftown next year :D
 
I run GeekBench and only get 6514. Al programs closed. Why the difference? I tried searching forums, but no luck.

Quad i5. 2.66.

Thanks.....
 
Until recently, when Intel publicly stated that the "Pentium" moniker was going to refer to a lower-cost market that exists between that of the Celerons and the Core 2 lineup, I can't recall any time when Intel tried to pawn off Celerons as anything but Celerons.

I would very much be interested in seeing a story where the product stated "Core Duo" and yet it was in fact a Celeron. Of course Intel is going to base the Celeron off of existing microarchitectures: it's far cheaper to disable some components of the existing microarchitecture, than it is to create a whole new microarchitecture solely for the low-cost market.

However, during the P3 days they were clearly "Celerons", during the P4 Netburst era they were still "Celerons", and as far as I know they're still branded as "Celerons" clearly for the Core/Core 2 derivatives.
Today's Wolfdale based dual core Celeron or AMD's Athlon II X2 offer killer performance for the dirt cheap $50-70 they command.

Regretfully only AMD ships with affordable IGP solutions that aren't terrible in their 780G/785G/790GX lines. It costs an arm and a leg to get a 9300 or 9400 for LGA 775.

Remember, in reality, Lynnfield features quite a few enhancements in regards to power management, and thus I'm willing to bet that for the most part, it'll likely stay far below its max 95W ceiling. I'm not too concerned about the issue of heat and Lynnfield when it comes to the iMac (I mean, I'm running a i7 920 CrossFire system in a mATX case, and even my thermal performance isn't that bad :p).
Which X58 motherboard and power supply are you using?
 
According to the tests I've seen on reputable PC websites, the Core i7 model 860 (2.8GHz Lynnfield) with paired 1333MHz desktop RAM, is every bit as fast as the Core i7 model 920 (2.66GHz Bloomfield) with triple channel 1066MHz RAM.

Of course Apple uses 1066MHz notebook RAM instead of 1333MHz desktop RAM so there will be a small performance hit, but essentially the new iMac is equal to its much more expensive Mac Pro cousin.
 
According to the tests I've seen on reputable PC websites, the Core i7 model 860 (2.8GHz Lynnfield) with paired 1333MHz desktop RAM, is every bit as fast as the Core i7 model 920 (2.66GHz Bloomfield) with triple channel 1066MHz RAM.

Of course Apple uses 1066MHz notebook RAM instead of 1333MHz desktop RAM so there will be a small performance hit, but essentially the new iMac is equal to its much more expensive Mac Pro cousin.

No, the iMac models with desktop chips use desktop memory. The changes to the i7 involve making it a best of hybrid of existing models. That does involve using 2 channels instead of three, but only because the advantages of 3 channels were so minimal.

There are a plethora of revues for the i7 860 used in the top of the line iMac which support Apples use of the i7 860.

The i7 860 is Intels best of (based on price/performance) for this generation. Upgrading or downgrading from the i7 860 provides only minimal changes based on cost.
 
Today's Wolfdale based dual core Celeron or AMD's Athlon II X2 offer killer performance for the dirt cheap $50-70 they command.

Regretfully only AMD ships with affordable IGP solutions that aren't terrible in their 780G/785G/790GX lines. It costs an arm and a leg to get a 9300 or 9400 for LGA 775.
Yeah, at the same time though, it's not like the 9400M is that good. :/ One of the threads I was reading had people criticizing Intel's IGP solutions. Apparently they ignore the 9400M's performance. It's better than Intel's solutions, most certainly, but no where near as good as a discrete GPU. :(

Which X58 motherboard and power supply are you using?
I use the ASUS Rampage II Gene mATX X58 board. The power supply is a Corsair HX750W.

My mATX case has 5x 120mm fans: 2x on each side, and one on the rear. I'm also using the Corsair H50 CPU water cooler to cool the 920. How I setup is that all four side fans are pulling air in, and the GPUs essentially segment off the case into halves: the two 120s on the right side to help to cooler the GPUs, while the two 120s on the left blow in air that helps cool the memory slots and the northbridge.

I have the rear 120mm blowing out, and the H50's radiator can hold a 120mm on the front and back, so the rear 120mm slot technically has the radiator plus two 120mms. The HX750W also has a 140mm fan to cool the PSU, and the Radeons expel their air mostly through the rear slot (they also expel a little through the top of the card, which the PSU's fan pulls up and out). Thus, in the rear, there are three exit locations for air. Overall, the setup is working really well. The 920 is currently only running at 3.4Ghz, but it's temps are around 40C at idle and the low 50s peak at load. The Northbridge usually peaks at about 52-54C, which for the X58 in this setup is rather good.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.