Some people dont understand that.
Apple's Processor Options for Early 2012 Mac Pro Begin to Firm Up
- Thread starter MacRumors
- Start date
- Sort by reaction score
Some people dont understand that.
nickeditor
macrumors newbie
Perhaps Apple will surprise us and wait until march to use new Ivy Bridge processors instead of this Sandy Bridge-E 
2011 is not finished ¿an update for Mac Pro like Macbook Pro in 2011 and a renovation in 2012? I think it is smart idea
2011 is not finished ¿an update for Mac Pro like Macbook Pro in 2011 and a renovation in 2012? I think it is smart idea
KnightWRX
macrumors Pentium
Get a 599$ Mini every year, and in 7 years, you'll have paid almost the same price as a base Mac Pro. Except by that 7th year, the Mac Mini you'll be buying will be much more powerful than that initial 3500$ Mac Pro you bought. Not to mention with inflation, in 7 years, those 7 Minis will have ended being cheaper to buy than dropping 3500$ on a Mac Pro today.
If your needs are served by a 7 year old machine, here's a hint : You don't need a Mac Pro. Sometimes, you learn to just buy "enough" instead of always going for "top of the line". Go with your needs rather than your wants.
something3153
macrumors 6502
I've never found that building a quality desktop PC was cheaper than buying one. Dell will happily sell you a cheap POS or a quality machine, but anything I'm willing to build is more expensive than the POS by quite a bit. It's even sometimes more than their good machines. The finished product is incomparably better, of course, and precisely configured, but not cheaper.
Then I buy HP Elitebooks for laptops, which are no cheaper than comparable Apples, but very high quality, and more configuration options.
deconstruct60
macrumors G5
That is a strange hypothesis given the rash of MHz junky posts that litter this and other Mac Pro threads. (e.g., "I want fewer, faster cores" ). The old and new single package line up (in "entry to top end" order )
Old 2.80 , 3.20 , 3.33
New 3.60 , 3.20 , 3.33
In the old line up, MHz junkies needed to buy the most expensive option to maximize clock speed. In the new line up, MHz junkies only need to buy the entry level model to maximize clock speed. There is probably a $1000 reduction in price for them. I would say that is a "feature" that many will buy.
In fact, if a user wants the fastest base MHz clock rate of the single and dual package models, then the entry level one, 3.6GHz, will be the fastest.
Comparing Hz is only transparent when the architecture is the same. Once change arch then it is increasingly an ineffective "apples to apples" metric.
The other fundamental problem is that people confuse instruction sets with architecture. The same instruction set can be implemented by two different architectures. Even multiple architectures implemented by the same company.
If you bother to drop down into the actual architecture and are concerned with real throughput on modern applications then the L3 cache is larger and ring bus connected to support speeds like 3.33GHz better for 6 cores. Likewise, there are 4 instead of 3 memory channels so can stream data faster (if don't pick a bonehead memory layout). The branch prediction and instruction fetching (new decode cache) are all substantially improved. While it may look like the mid and top tier clocks are the same the actual throughput through the core is likely 10-15% higher on average and much higher on more than a few applications. That is the problem with looking at single somewhat superficial characteristics to make large purchasing decisions.
And even for the MHz junkies, the E5 1660 will likely "Turbo" up to match the 1620 in speed. So if get your thrills running Mac Paint in Sheepshaver (or whatever last century software addicted to ) it will be just as fast.
Similar to the dual package models across old and new models (if drop down to low number of cores of previous generation the clock frequency goes up). MHz junkies should be looking in the "Turbo" column, not the base rate column, if primarily buying a machine just on MHz.
The dual package models will obviously be better because end up with 4 more cores. Even if the architecture was the same (which it isn't) that is just more "horsepower" to do multiple things at the same time. The myopic focus on single applications is humorous also. Two ( or more) applications where one (or more) are in batch processing mode and one is in interaction mode can consume multiple cores if. Anybody who "walks and chews gum at the same time" can soak up multiple cores in a box. [ For example, this rumor about CS 6 (http://www.appleinsider.com/article...aperture_like_theme_new_3d_functionality.html ) states that Photoshop will have an
feature. There are numerous non esoteric actions that can happen in parallel if bother to implement them.
If get past myopically looking at cores and MHz then PCI-e v3.0 is substantially better than PCI-e v2.0. Even a limited set of SATA III connectors would be an substantive improvement.
Only by the uninformed. The design parameters for the upcoming Mac Pro were finalized probably over a year ago. "Post Steve Jobs" is likely 1 - 2 years from now for the fundamental design constraints.
You need to understand computer architecture and EE a bit more.
CPUs have thermal limitations: you can't just increase clock for free. There are diminishing returns as you increase the speed not just if you can cool: you have leakage across transistors with higher voltage (required to drive clock speeds higher).
Tick/Tock is how Intel does this. A pure die shrink is your tick, and clock speeds *do* increase within the same TDP.
I'm going to avoid going too deep into concept as well of clock cycle efficiency: 1 Million clock cycles on a 2008 Xeon is not going to be the same as 1 Million clock cycles on a SB Xeon: micro-architecture has changed, and for just about any task the SB chip will do more in 1MM cycles then before: often a lot more. There are many different reasons for this and you can explore them on your own. New instruction sets, different cache/latency speeds, more I/O speed, re-architected pipelines, better branch predictors. It goes on and on. Engineers spend years in order to make things just one clock cycle more efficient when they know that op is used billions of times a minute.
Speed simply is how long the task takes. How often does any company introduce a new product as faster and it runs things *slower*? Yeah, exactly
I can think of one recently: AMD Bulldozer will be using less FPUs, so if you are FPU heavy with your app, the Phenoms will currently still be 'faster'.Most folks are integer bound, not FP.
- b
deconstruct60
macrumors G5
1. The base Mac Pro is around $2,400-2.500 .
2. 7 * $599 = $4,193 ( about $1,600 more than a base Mac Pro ).
every other year... so 4 (rounded up) instead of 7 ... you'd be close: $2,396.
But yes, the mini 6 years from now will likely dominate the single package Mac Pros being sold now.
When Apple puts a box on the "vintage computer" list .... it has left the performance realm. Past 3-4 years, most likely the computer is "fast enough" as opposed to "fast". Especially in this era where increased transistor budgets allow the designs to subsume more classically external CPU functionality into the "CPU" package. ( As opposed to making the basic CPU functions more complex).
The mini will become a miniaturized early 2000 era "Mac Pro like" machine and the Mac Pro will become a miniaturized cluster of 2000 era "Mac Pro like" machines.
However, with Intel's split of the E5 into the 1600 and 2600 configs that may not work out so well going forward. As long as they significantly suppress the core count on the entry level single package model the base Mac Pro will do better over a longer period of time. However, 6 years is 3 "tick-tock" cycles so the mini will catch up in performance with that time window. Way too many 100's of millions of R&D being poured into R&D to make that happen will get traction over that amount of time.
MacSince1990
macrumors 65816
Hmm. Hopefully they do something with the quad-core for the base machine. Or overclock the hex-core, because 2.0 GHz is a bit silly... and there's an enormous gap in pricing as you climb up the ladder..
Oh, Lord... do you use the phrase "the good ol' megahertz myth" in every post you make?
Obviously he was speaking tongue-in-cheek.
The phrase "The MHz Myth" was coined by Apple when they NEEDED it for PR reasons. Yes, technically it was true; a 500 MHz G4 was not the same speed as a 500 MHz PIII, and was generally faster than a 1 GHz P3 (always faster than a 1 GHz P4, and generally faster than a 1.3 GHz P4..), but seriously, there is no "MHz myth" anymore.
Yes, a 1.6 GHz C2D would be moderately faster than a HyperThreaded 3.06 GHz Northwood, although it's got two cores, which isn't exactly a fair comparison, especially since HT was in its infancy (literally.. the 3.06 was the first CPU to use HT). Of course, a 1.6 GHz C2D would *not* be faster than a DC 2.3 GHz G5. At least not in G5-optimized situations.
[/quote]
Moore's law speaks on the relative cost of transistors as time passes and transistor density increases.
And for the record... no, the G5's aren't "ancient technology," they're just not current. If you outfitted a Quad G5 the way it ought to have been in the first place-- on a 65 or 45nm fabrication process, with an integrated memory controller, with 2MB of cache per core (8 MB total, shared), and 1.33 GHz DDR3 (not available at the time of course), and gave it some form of threading, it would actually perform quite well. Especially if you cut down on the number of in-flight instructions and the relatively large register count that often tied up the machine. These are all relatively basic changes that would have translated into massive performance increases.
Oh, Lord... do you use the phrase "the good ol' megahertz myth" in every post you make?
Obviously he was speaking tongue-in-cheek.
The phrase "The MHz Myth" was coined by Apple when they NEEDED it for PR reasons. Yes, technically it was true; a 500 MHz G4 was not the same speed as a 500 MHz PIII, and was generally faster than a 1 GHz P3 (always faster than a 1 GHz P4, and generally faster than a 1.3 GHz P4..), but seriously, there is no "MHz myth" anymore.
Yes, a 1.6 GHz C2D would be moderately faster than a HyperThreaded 3.06 GHz Northwood, although it's got two cores, which isn't exactly a fair comparison, especially since HT was in its infancy (literally.. the 3.06 was the first CPU to use HT). Of course, a 1.6 GHz C2D would *not* be faster than a DC 2.3 GHz G5. At least not in G5-optimized situations.
[/quote]
Moore's law speaks on the relative cost of transistors as time passes and transistor density increases.
And for the record... no, the G5's aren't "ancient technology," they're just not current. If you outfitted a Quad G5 the way it ought to have been in the first place-- on a 65 or 45nm fabrication process, with an integrated memory controller, with 2MB of cache per core (8 MB total, shared), and 1.33 GHz DDR3 (not available at the time of course), and gave it some form of threading, it would actually perform quite well. Especially if you cut down on the number of in-flight instructions and the relatively large register count that often tied up the machine. These are all relatively basic changes that would have translated into massive performance increases.
Battlefield Fan
macrumors 65816
You're so wrong sir. Please never advise yourself or others on what computer to buy.
Supersonic
macrumors regular
Apple I = 1.023 MHz
MacPro = 8 core 2.6 GHz
Can someone please do a YouTube video to show the difference between both?
Good, interesting geeky question...
How do you benchmark stuff that old, against current technology?
something3153
macrumors 6502
It's tricky. Particularly, do you run benchmarks that allow the new CPU to execute multiple threads and the countless new instruction set extensions? Even without those, the new ones would crush the old ones, of course, but allowing those would make it even more ridiculous. Since the old CPU won't run a modern benchmark suite at all, I'd approach it with a custom-written calculation script to loop through a certain number of times and record the length of time taken.
Interesting, but I don't believe that your friend built a 16 core machine for that, although he may have been able to use AMD opterons or whatever. Probably an eight core 16 thread machine using Xeon.
Intel hasn't had any Xeon processors beyond 6 core for a dual processor configuration to this point, so that the most that you could get is 12 core.
reallynotnick
macrumors 65816
What are you talking about?
PowerMac Dual Core 2.3Ghz G5 - 2163
Macbook Air C2D 1.6ghz - 2358
Mac Mini C2D 2.26ghz - 3204
MacBook Pro Core i5 2.3ghz - 5910
http://www.primatelabs.ca/geekbench/mac-benchmarks/
You can't compare mhz/ghz across different architectures. It wasn't just an PowerPC vs x86 thing.
Lacking Rosetta - preventing upgrade for us
Sorry Apple, but as long as the new Mac Pro will not have Rosetta or are able to run 10.6, they are not interesting to us in a long time.
We have to many older things like scanners, chemical analyzers and other external equipment which needs Rosetta and where new drivers won't come.
Apple, please reconsider to make Rosetta for 10.7 and many professionals will become happy campers!
Sorry Apple, but as long as the new Mac Pro will not have Rosetta or are able to run 10.6, they are not interesting to us in a long time.
We have to many older things like scanners, chemical analyzers and other external equipment which needs Rosetta and where new drivers won't come.
Apple, please reconsider to make Rosetta for 10.7 and many professionals will become happy campers!
KnightWRX
macrumors Pentium
Well, considering you could build a 32 core machine :
http://www.amazon.com/dp/B003EM6B0O...e=asn&creative=395105&creativeASIN=B003EM6B0O
1000$ - motherboard
1000$ - 4 processors (AMD Opteron 6128 Magny-Cours 2.0GHz)
400$ - RAM 50$ for 2x2GB kits, need 8 of those.
...
Hum, you have 2600$ left over for a case, some drives and a PS... (so his claim doesn't sound so far fetched).
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Chris Tilton
macrumors newbie
For me to do my job, a Mac Mini still won't do. I need at least 16GB of RAM (I have 24GB). I have 4 SSDs which run thousands of audio samples at once. 2 video cards to support 4 monitors (3 on my desk, and one in my recording booth). Doing this without an expandable desktop is impossible. With things like thunderbolt, we're getting closer to where I can do my work on an iMac, but we're not there yet.
I used to have to have 6 PCs running sampler software that all fed into my main mac. Computers got powerful enough about 4 years ago that I could consolidate everything into one computer, but we are a long way off where I can do my job on an iMac, and even further away from doing it on a Mac Mini.
firestarter
macrumors 603
That horse is dead. Buy a new scanner and migrate your data to newer apps.
Contact your analyser manufacturer and ask for newer drivers.
Last edited:
christian_k
macrumors 6502
You can easily sell a $600 MacMini that is just a year old for $300. So each of them will only cost $300 in the end. Will be $2,100 after 7 years.
A seven year old Mac Pro will be pretty worthless in the end. Probably it won't be able to run a new OSX version after 5 years or so.
Christian
typecase
macrumors 6502
Well said. Few people realize the charm of a truly expandable PC. For video, audio or photography, it's the only way to go.
That's not true. I use a 2003 PowerMac G4 running OS X 10.4, and it's fine for my purposes [just dog slow].
firestarter
macrumors 603
The 'replace Mac every year' strategy is an interesting (and good) one.
You get a fresh year's warranty, and the latest version of the OS thrown in too.
Geekbench is not a real test to test how fast computers are. MBP are beating iMacs in geekbench, but in real test the MBP loose to the iMac's.
steadysignal
macrumors 6502a
You're doing orchestral composing I see if you need a machine like that for audio production.
Is disk streaming with SSD's in raid not fast enough instead of pushing all those samples into your RAM? The iMac has 4 memory slots so they can have 32 gb of RAM in total? I know the MBP can use 8gb Ram modules so the iMac's can do it too probably.
A single ATI 6xxx card can support up to 3 monitors I believe so we're close. With Thunderbult you can add an other monitor and then we are at 4.
reallynotnick
macrumors 65816
Geekbench is not a real test to test how fast computers are. MBP are beating iMacs in geekbench, but in real test the MBP loose to the iMac's.
Correct, Geekbench just measures CPU and memory performance, doesn't take into account for HDD or GPU performance.
Thunderbolt doesn't magically make your graphics card able to support another monitor, it is still limited by the GPU.