Nehalem Mac Pros Arrive: Unboxing and Benchmarks

[Anonymous Freak]

Don't feed the trolls.

I know, don't feed the trolls. I normally don't.

PowerPC was noticeably better than x86 up until about the Pentium III vs. G4. Then, the massive improvements in clock speed on the P3 side more than made up for the G4 being faster per-clock.

G5 thunderously stole the crown back to PPC. But then PPC stagnated, and Intel released the Core 2 line, which was ever-so-slightly faster than an equally clocked G5.

Now i7 is absolutely blowing Core 2 out of the water; and PPC hasn't improved one bit since then. (Indeed, the 2.5 GHz dual-core G5 isn't even available any more.) Yes, IBM has improved PPC-derived systems (the chips in the PS3 and Xbox 360, for example,) but not 'desktop' PPC chips; because Apple was realistically the only customer for them. (Yes, IBM uses them in their own products, but they don't put nearly as much effort in maximizing CPU speed on those.)

Their much higher-end POWER series chips are still very competitive, but they never put that much effort into the lower-end PowerPC. The G5 looked like it was going to be IBM's big push; but then it just petered out. Heck, third-party PowerPC-licensed chip designers had better stuff at the end than IBM.

 

[subaqua]

Looks nice, but not $5000+ nice

Agreed. You can probably buy two Jan 2008 8-core/2.8GhZ systems for that money now which will give you 16-cores in a cluster ... hehe.

Gone are the days of the MacPro being a deal I guess; very sad. I was hoping for maybe just a slight increase in price for the 8-core version but $3,300 vs. $2,800 is too much considering the speed of the new low-end 8-core model.

I'm crossing my fingers for better value in 2010.

 

[iMacmatician]

Snow Leopard will probably NOT be able to recover enough ground to make the difference if it makes any at all.

Snow Leopard would also improve the Harpertowns, although it may help the Gainestowns more.

This really IS the year to skip upgrades IMO. Maybe next year if they leave the prices and speeds the same but offer 6 or 8 cores per chip.

I'm crossing my fingers for better value in 2010.

In Q2 2010 Intel will release Gulftown, which will have 6 cores at presumably similar clock speeds to Gainestown. Hopefully Apple will update the Mac Pro with 6 and 12 cores…or they might just add it as an option like with the Clovertown.

 

[eeboarder]

I think you and eeboarder are either confused by bar-graphs, are working for Apple, or have a very different idea of what "significant improvement" means.

Between the new 2009 MP 2.66 and my ancient MacPro v1.1 at 2.66 x8 there's a 35% to 40% increase when the cores are all (16 and 8 respectively) hitting at 100% load! That's dismal! The rule of most professionals has been to upgrade when it's 100% increase (AKA: Twice as fast) for about the same price.

There's only a 20% to 25% increase for something like photoshop or pretty much anything that isn't maxing out your cores.

The new 2009 2.66 is $4,700 in it's wimpiest configuration. My v1.1 MacPro was I think, $2,500 and $2,600 all together after upgrading. That's a 80% price increase for a 25% speed differential. Sorry, that totally sucks!

What was the 2008 Mac Pro v3.1 at 2.8GHz? Oh yeah, $2,800.

The New MacPro 2009 2.66 is between 10% slower and 10% faster in everyday use than the 2008 2.8. No question about it and Snow Leopard will probably NOT be able to recover enough ground to make the difference if it makes any at all. When all cores are at 100% the new 2.66 is just 21% faster than the older 2.8 octad.

The price difference is 68%. You really think a 68% price increase is justified for a machine that is sometimes slower and only between 10% and 21% faster in some cases???

Me? Nope!!! No way I just did real numbers and real percentages based on real benchmarks and real prices and the results are in. This years' Mac Pros are a total rip-off compared to last years. I really mean a total screaming in your face, raising hell, boycotting kind of rip-off too. So I guess we can expect there to be lots and lots of noise on the forums about all this. I won't but I bet others will! PS: And it's even worse if we go outside Apple and look at machines from other vendors.

This really IS the year to skip upgrades IMO. Maybe next year if they leave the prices and speeds the same but offer 6 or 8 cores per chip.
Last year you could upgrade to a machine that was MORE than twice as fast for just $200 above the previous year's. This year we're bending over and taking an almost $2,000 ramming and getting a box which is sometimes even slower on top of that!

.

Believe me, I would be the last person to drink Apple's cool-aid. I really don't like how Apple does business, but the alternative is much worse. I don't want to deal with Linux, either.

Overall, you're jerkin' yourself off to bars and numbers. If you sat down and worked on a 2.8GHz previous-gen MP and then worked on a 2.26GHz Nahelam MP without knowing it, you wouldn't notice a single difference. You're just set on benchmark numbers. Get real.

Regarding the price, I believe people are paying a premium for these Nahelam processors. The DDR3-8500 RAM makes up for it. I just got 3x2GB on newegg for $71. Not too shabby. I actually purchased a dual 2.26GHz with the ATI upgrade and wireless keyboard and mouse for $2,818 on a developer discount. Sounds worth it to me.

You may want to run benchmarks all day and get your rocks off, but for people like me who are going to use these machines for real work and production, the benchmarks don't matter. I'll be running several heavy applications simultaneously, and I'm looking forward to having all the power available for use for the next 5-7 years.

I don't know what other vendors you're looking at. These processors aren't even released yet.

 

[Tesselator]

Believe me, I would be the last person to drink Apple's cool-aid. I really don't like how Apple does business, but the alternative is much worse. I don't want to deal with Linux, either.

Overall, you're jerkin' yourself off to bars and numbers. If you sat down and worked on a 2.8GHz previous-gen MP and then worked on a 2.26GHz Nahelam MP without knowing it, you wouldn't notice a single difference. You're just set on benchmark numbers. Get real.

Regarding the price, I believe people are paying a premium for these Nahelam processors. The DDR3-8500 RAM makes up for it. I just got 3x2GB on newegg for $71. Not too shabby. I actually purchased a dual 2.26GHz with the ATI upgrade and wireless keyboard and mouse for $2,818 on a developer discount. Sounds worth it to me.

You may want to run benchmarks all day and get your rocks off, but for people like me who are going to use these machines for real work and production, the benchmarks don't matter. I'll be running several heavy applications simultaneously, and I'm looking forward to having all the power available for use for the next 5-7 years.

I don't know what other vendors you're looking at. These processors aren't even released yet.

Mmmm, I do know what you're saying. But I've been doing this so long now that I know within reason how those benchmarks translate into usage of all kinds. Not bragging or anything but I've been doing this since Commodore Pet and even before. That's years before there was an Apple or a PC-XT.

Again though I know what you mean - some people place too much relevance on benchmarks and may not know how they translate into diverse performance across actual applications. I'm pretty careful not to trip over that one though.


PS: Corei7 has been out for awhile now already.

.

 

[iBug2]

PS: Corei7 has been out for awhile now already.

.

Only the single CPU ones. Dual CPU Nehalems will be out end of march to everyone else.

 

[Bye Bye Baby]

How much faster is the new mac pro than my iMac?

Do I need to be jealous? And if so, how much?

 

[drsmithy]

but, when 2 threads of virtual machine runs on 1 physical core - you will get 60-70% performance hit in VM(under full load)

and You cant control which virtual threads it will use.

Not nice imho.

This will only happen if your OS's CPU scheduler is stupid. Given how long SMT CPUs have been around, one would hope OSX's is not.

 

[drsmithy]

Mac OS X may / may not be laying out memory in the best way for a NUMA box.

Given this is the first NUMA hardware Apple has released, it's probably a safe bet that OS X isn't well optimised for NUMA, if it is NUMA-aware at all.

(Maybe NUMA support is some of the pixie dust in Snow Leopard ?)

 

[eeboarder]

I believe the Nahelem Xeons aren't released yet other than with Apple.

 

[AidenShaw]

This will only happen if your OS's CPU scheduler is stupid. Given how long SMT CPUs have been around, one would hope OSX's is not.

Early versions of OSX were rather weak on SMP systems, even though those have been around for decades.

I would not expect that Apple invested a lot of effort in making hyper-threading aware schedulers before now - when hyperthreaded chips are actually here.

Given this is the first NUMA hardware Apple has released, it's probably a safe bet that OS X isn't well optimised for NUMA, if it is NUMA-aware at all.

(Maybe NUMA support is some of the pixie dust in Snow Leopard ?)

I won't bet that OSX is NUMA aware, for sure.

Same issue as SMT - only if you have lots of idle engineers do you prepare for possible future architectures.

I would also expect that a non-NUMA aware system would be just as fast as a NUMA-aware system for many applications. It would feel just as fast - you'd have to run careful experiments to see the effects of NUMA - especially for the simple dual-socket systems that Apple is building. (Quad/Octo/Hex socket systems with multiple QPI hops are more of a problem - for dual socket systems where memory is either local or one hop away the difference is not huge.)

I believe the Nahelem Xeons aren't released yet other than with Apple.

I've already ordered 32 systems with DP Nehalems from a major vendor - the order is in the queue waiting for announcement/ship day. (I already have a few proto systems for evaluation - hot stuff.)

I expect that Apple's early announcement won't translate into a signifcant lead in early shipments - when it's all over you may find Mac Pros arriving a handful of days before the other systems. A few fanbois who are insecure about their other equipment might claim that this is significant, but it really isn't in the bigger picture.

 

[im2fast]

I'm new to all this macrumors thing, but I have read nearly all the replies on the new MacPro and got a lot of good info. I was planning to move on from my G4 to a new Macpro (tough to upgrade an imac when you want). All this dual quad core stuff for you serious folk earning a living from your machines- impressive, hope you're making tons of money. I just don't want to get bogged down with a machine I can't upgrade and the single 2.66 machine may do me good for five years or so. Your thoughts can help. thanks

 

[mjteix]

I've already ordered 32 systems with DP Nehalems from a major vendor - the order is in the queue waiting for announcement/ship day. (I already have a few proto systems for evaluation - hot stuff.)

Without naming the manufacturer, are some of those systems equivalent to the new Mac Pros and if so, are the prices similar too?

 

[NeoSage]

I'm loving my Mac Pro 2.26

This is a great machine, and I am a power user. There are 16 virtual cores, and this nifty grid with 20 bars, making a 20x16 grid of cells, which are green when CPU is active. For the vast majority of apps and daily usage, you will be disk or memory or graphics card bound, not cpu bound.

So the way I look at it, if I wanted serious rendering or supercomputing time, I would rent capacity with a service that offered it. Otherwise, I will buy much faster drives, much more RAM, and always stay current with the best video cards - and if I max those out, I can always throw in a couple of the latest greatest chips and void my warranty. But buy the time I max out the non-CPU resources, there will be a new motherboard out, new set of chips, and I will be able to pick up the best of the previous generation chips for a fraction of todays chips - and they will be faster and probably have more than 4 cores per chip. SO 2.26 made the most amount of sense for me, especially when I realized that the couple grand difference with the 2.93 would be best spent on high speed disks, memory, and graphics cards. It's a simple equation. Oh, and as for quad vs octo, and old vs new, all SW dev trends point to the more cores the better, and the better interconnects/memory mgt the better, so for future proofing, its the way to go.

 

[Anonymous Freak]

I believe the Nahelem Xeons aren't released yet other than with Apple.

You are very, very correct. As I've commented previously, I find it hilarious that there are people in this thread who have them in their hands; yet I'm still under NDA and can't talk about it.

 

[Tesselator]

How much faster is the new mac pro than my iMac?

Do I need to be jealous? And if so, how much?

Depends on which two you compare. But no, for what iMac is designed for the $5,900 Mac Pro does not beat it by any amount a human being can typically detect. Your user experience will typically be as good or better on the iMac.

Of course when you start getting into multi-core computing then... well...

 

[Tesselator]

Only the single CPU ones. Dual CPU Nehalems will be out end of march to everyone else.

Right, yes, of course. Just like Apple has two Mac Pro models with only a single processor.

As far as the Xeon line we already know the prices. And that puts Apple's offerings almost consistently close to $2000 over most everything else.

So we're back to the same percentages that I for one predict will have two affects: One being to cut off Apple's sales at the knees or even below! And the other being to increase the troll and bitchfest frequencies on sites like this.


BTW, for those of us that think these guys here talking about NUMA are referencing an old Pink Floyd album or something have a read here:
http://en.wikipedia.org/wiki/Non-Uniform_Memory_Access

.

 

[Bye Bye Baby]

Depends on which two you compare. But no, for what iMac is designed for the $5,900 Mac Pro does not beat it by any amount a human being can typically detect. Your user experience will typically be as good or better on the iMac.

Of course when you start getting into multi-core computing then... well...

Here's hoping you are right. So for the moment I will be content with a small amount of jealousy.

 

[RUAerospace]

I think you and eeboarder are either confused by bar-graphs, are working for Apple, or have a very different idea of what "significant improvement" means.

No, you simply have no clue about chip design and how difficult it is to produce sizeable improvements in performance within the given span of time, especially given all hurdles faced by the engineers and designers today and the strong performance of the previous architecture. Very few designs have achieved the boost that Nehalem has gained given these conditions.

Between the new 2009 MP 2.66 and my ancient MacPro v1.1 at 2.66 x8 there's a 35% to 40% increase when the cores are all (16 and 8 respectively) hitting at 100% load! That's dismal! The rule of most professionals has been to upgrade when it's 100% increase (AKA: Twice as fast) for about the same price.

Refer to what I said above first of all. Secondly, you have no say over when someone might consider buying a new computer. If you do heavy rendering, or design or anything work-related, than the time saved with Nehalem over the course of the life of using the machine over the previous could well make up the difference in price. You simply failed to see that point whereas other professionals who have such needs could well buy one for that exact reason.

[

There's only a 20% to 25% increase for something like photoshop or pretty much anything that isn't maxing out your cores.

Barefeats disagrees with you, the new 2.93 GHz Nehalem is about 3 times as fast as the previous 3.2 GHz in MP actions. You act as if most professional apps don't max out their cores when really, software like 3D rendering, design, and technical software have been employed in machines with 8+ cores from years before and are considered among the most parallel and well-threaded of programs. If you had any idea of what your talking about, maybe you'd realize that this is why the largest performance boosts in Nehalem took place in exactly these applications and they typically range from 35-60+% with some even greater.

Me? Nope!!! No way I just did real numbers and real percentages based on real benchmarks and real prices and the results are in. This years' Mac Pros are a total rip-off compared to last years. I really mean a total screaming in your face, raising hell, boycotting kind of rip-off too. So I guess we can expect there to be lots and lots of noise on the forums about all this. I won't but I bet others will! PS: And it's even worse if we go outside Apple and look at machines from other vendors.

No you simply made generalized conclusions without any real knowledge of what to expect in performance gains or any context of how this boost compares with the gains seen in other new architectures. If you actually looked at the reviews of Nehalem, you'd see quite unanimonously that people who DO KNOW these things have generally praised the Nehalem for its gains. Don't waste your time with comments like doing "real numbers" and "real percentages" on "real benchmarks" when you don't even have a clue what any of those things mean.

 

[Tesselator]

Here's hoping you are right. So for the moment I will be content with a small amount of jealousy.

Expansion envy? See a shrink abut that.

 

[Tesselator]

No, you simply have no clue about chip design and how difficult it is to produce sizeable improvements in performance within the given span of time, especially given all hurdles faced by the engineers and designers today and the strong performance of the previous architecture. Very few designs have achieved the boost that Nehalem has gained given these conditions.



Refer to what I said above first of all. Secondly, you have no say over when someone might consider buying a new computer. If you do heavy rendering, or design or anything work-related, than the time saved with Nehalem over the course of the life of using the machine over the previous could well make up the difference in price. You simply failed to see that point whereas other professionals who have such needs could well buy one for that exact reason.

You're right. I should be banished into the abyss for ever and ever. You know so much more than I. You should be a god so i can worship your every wise word.

Barefeats disagrees with you,

I don't see how they could. They're using my compiled benchmarks. And I ran the numbers twice. There's no mistake.

No you simply made generalized conclusions...

Poppycock! I made extremely accurate and specific conclusions based on real benchmarks, real experience, real prices, and actual facts - that I personally compiled myself! If some company or persons such as yourself wishes to extort, exaggerate, or falsify the data because of some personal problem that's on them - if indeed anyone did do that besides yourself. I'll not be a party to it. I am sorry that you personally don't like the results but I won't apologize for that.

The numbers speak very clearly and I'm not biased either way. Ooops, actually I am biased! I hate Microsoft! Maybe if I could fudge a tad without selling my soul completely in order to make Windoze and Windoze hardware look worse I might stoop to it. So, when I say:

Between the new 2009 MP 2.66 and my ancient MacPro v1.1 at 2.66 x8 there's a 35% to 40% increase when the cores are all (16 and 8 respectively) hitting at 100% load! That's dismal! The rule of most professionals has been to upgrade when it's 100% increase (AKA: Twice as fast) for about the same price.

There's only a 20% to 25% increase for something like photoshop or pretty much anything that isn't maxing out your cores.

The new 2009 2.66 is $4,700 in it's wimpiest configuration. My v1.1 MacPro was I think, $2,500 and $2,600 all together after upgrading. That's a 80% price increase for a 25% speed differential. Sorry, that totally sucks!

What was the 2008 Mac Pro v3.1 at 2.8GHz? Oh yeah, $2,800.

The New MacPro 2009 2.66 is between 10% slower and 10% faster in everyday use than the 2008 2.8. No question about it and Snow Leopard will probably NOT be able to recover enough ground to make the difference if it makes any at all. When all cores are at 100% the new 2.66 is just 21% faster than the older 2.8 octad.

The price difference is 68%. You really think a 68% price increase is justified for a machine that is sometimes slower and only between 10% and 21% faster in some cases???

Me? Nope!!! No way I just did real numbers and real percentages based on real benchmarks and real prices and the results are in. This years' Mac Pros are a total rip-off compared to last years. I really mean a total screaming in your face, raising hell, boycotting kind of rip-off too. So I guess we can expect there to be lots and lots of noise on the forums about all this. I won't but I bet others will! PS: And it's even worse if we go outside Apple and look at machines from other vendors.

This really IS the year to skip upgrades IMO. Maybe next year if they leave the prices and speeds the same but offer 6 or 8 cores per chip.
Last year you could upgrade to a machine that was MORE than twice as fast for just $200 above the previous year's. This year we're bending over and taking an almost $2,000 ramming and getting a box which is sometimes even slower on top of that!
​

I pretty much mean exactly that! Nothing general! No assumptions!

.

 

[Jeffrosproto]

This Mac Pro really looks great, the past models were clean, but this one!!!

 

[YahonMaizosz]

Anyone knows where I can get a 2008 Mac Pro with the 8 Core 2.8??

The new 2009 Mac Pro is rubbish in my opinion... Its Price/Performance does not match..

 

[nontroppo]

What really annoys me, is based on published 1000-unit prices, the 2X2.26ghz Nehalem chips are actually $848 cheaper than the 2.8ghz harpertowns, yet Apple have hitched a $500 price hike on the MP over the that generation. We can't know exactly how Apple's margins have changed (but cheap'n'cheerful graphics cards (reheated NVidia cards from 2007) and DDR3 memory and a potentially simpler motherboard), means for a questionably "better" machine (depending on whether you are largely single or multithreaded user), Apple are pumping their margins to the extreme. Apple should at least be trying to be competitive with their own previous generation, the $ premiums Apple are putting on Nehalem, considering the better (guesstimated for the 2.26 vs 2.8) margins and questionable performance for a large number of MP users (photoshop and company) leaves a sour taste in the mouth.

 

[gnasher729]

I would not expect that Apple invested a lot of effort in making hyper-threading aware schedulers before now - when hyperthreaded chips are actually here.

There is very little that needs to be done to make a scheduler hyper-threading aware. You have eight cores, 1 to 8, and each one can run two threads that need scheduling, 1a and 1b, 2a and 2b, ... 8a and 8b. There is a very simple rule: Don't schedule threads on the same core (like 1a and 1b) unless there is at least one thread running on each other core.

For the VM example: With this simple scheduling method, a VM using two virtual cores would usually get two real cores running a single thread. That is, until your software makes real use of more than eight threads. So yes, if I have fourteen threads running encoding videos, and the VM is given two threads, it suddenly will get only one eighth of the whole machine instead of two eights, but that is because the video encoding takes 7/8ths instead of 6/8ths.

And the scheduler on Leopard is already aware of complex multi-processor systems, so you can ask it to schedule threads on different physical processors or on the same physical processor, whatever makes it run faster.