Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

phong

macrumors newbie
Original poster
Nov 30, 2002
22
0
Guys,

I know barefeats have not posted their benchmarks on the new Nahelem Mac Pros yet. I have also seen some preliminary cinebench scores.

Anyways, my college computer store is selling the 2008 2.8 Octo-Core for $1999.

Or I can use a ADC Discount and get the 2009 2.66 Nahelem Quad for $1999. The 8GB limit pisses me off.

I do Photoshop work (just dealing with RAW photographs, and basic work, nothing huge) and some Final Cut Pro work. Probably won't use every core efficiently but that depends on any new projects, which my involve scientific work for my academics. But according to the cinebench graphs from another thread, even in a single thread, the 2008 2.8 compares favorably with the 2.66 Quad.

I am thinking about picking up the 2008 2.8 Octo-Core Mac Pro sometime this week or next while the store still have it.

I am leaning towards the 2008 model.

Any thoughts? Same price, what would YOU choose?
 
Personally, I would always wait for real-world tests - without that I wouldn't make a move. That said… time does seem to be of the essence and you do know the kind of performance you would get from 2.8 machine, so you do know what to expect from this machine, but you’re not in this position with the 2.6 model yet.

With regards to RAM, again until we’ve seen some testing, I wouldn’t like to give a definite opinion. As has been pointed out on other threads, with certain machines Apple has stated a maximum amount of RAM in their specs, which was lower than the actual possible amount.

Hopefully, we’re going to see sites like Barefeats test the machines very soon but whether that’s within the timeframe you have, is another thing.

As long as you’re happy the performance that the older machine offers (which you seem to be) and prepared to accept that the new machine might well offer more than the initial benchmarking software and 8GB RAM suggests, go for the 2.8GHz one.
 
For photoshop work, the 2.66GHz Quad will be faster...but for FCP, the 2.8GHz Octo will be faster...

so it depends what you'll do most

also....the new one is more future-proofed as far as snow leopard, multi-threading etc...
 
Sorry we haven't got benchmark results up yet. We're still waiting for our test units to arrive even though we ordered them on the 3rd. We're getting some help from "remote mad scientists" who are running some real world tests including with Photoshop.

Meantime, since Geekbench includes memory transfer tests and since Photoshop heavily uses memory for caching, take a look at these Geekbench scores (higher is faster):

2009 2.66GHz 4-core = 8355
2008 2.8GHz 8-core = 8098
2009 2.26GHz 8-core = 11397

Based on that, it could be argued that the 2008 2.8GHz is almost as fast as the 2.66GHz 4-core but with more memory expansion potential. Yet if you can afford the 2.26GHz 8-core with 8 memory slots, you might be better off with it for editing in Photoshop using multi-layered RAW files.
 
I'm in a similar quandary re: the 2008 2.8 Octo and the 2009 2.26 Octo. I do both 2D and 3D graphic work. The unknown factor is Snow Leopard. Will it make a big difference in performance on the Nehalem chips, or will it be a negligible speed up. I'm sort of leaning toward a refurbished 2008 2.8 Octo myself.
 
In Cinebench R10 The 2.66 quad benchmarks slightly faster than 2008 8 core 3.0 in single threaded tests and considerably slower than 2006 2.66 8 core Mac Pro in multithreaded tests. While there are many benchmarks I would personally like to see before forking out $3000, Cinebench is heavens closer to real life raw utilization than synthetic geekbench will ever be, so this is probably what we are going to see throughout other tests. 30% improvement in single threads, clock for clock and slightly mixed bag in terms of multicore performance at least until key applications learn to utilize hyperthreading well.

My money at the moment is still on 8 core 2008 octo, because they are excellent value for money and don't have 8Gb memory limitation like 4 core 2009 Mac Pros do. 4 cores will also never be popular second hand choice, so the values will suffer the same way single CPU 2008 MPs values do.
 
For photoshop work, the 2.66GHz Quad will be faster...but for FCP, the 2.8GHz Octo will be faster...

so it depends what you'll do most

also....the new one is more future-proofed as far as snow leopard, multi-threading etc...

future proofed my ass
4x2.66= 4 physical cores = 8 (virtual) cores and 8GB ram wall
8x2.88= 8 physical cores = 8 ... cores and 32GB ram wall

they are on par as far as snow leopard is concerned, what penryn lacks is memory throughput
as far as ram wall goes: its still MAX 16gb atm (very expensive actually with 4gigs sticks) and its not been tested.
uMBP 15" could go 8gigs but only goes 6 because its crippled too.
 
Meantime, since Geekbench includes memory transfer tests and since Photoshop heavily uses memory for caching, take a look at these Geekbench scores (higher is faster):

2009 2.66GHz 4-core = 8355
2008 2.8GHz 8-core = 8098
2009 2.26GHz 8-core = 11397

Based on that, it could be argued that the 2008 2.8GHz is almost as fast as the 2.66GHz 4-core but with more memory expansion potential. Yet if you can afford the 2.26GHz 8-core with 8 memory slots, you might be better off with it for editing in Photoshop using multi-layered RAW files.

Are you sure about that last statement? The 2.26GHz machine wouldn't do
nearly so well on GeekBench if it weren't for its multi-threaded performance.
But Photoshop isn't heavily multi-threaded, as I understand it. In addition,
we already have some results showing that the new 2.26GHz machine is
slower on the Retouch Artists benchmark than the older machine, even
when that older machine had a slower stock hard drive and less memory.
 
Infrared,

I think his comments were "memory throughput specific". With other benchmarks YMMV.

With those comments in mind check out this blog:

http://blogs.adobe.com/jnack/2006/12/photoshop_and_multicore.html

Here's the quote that caught my eye:

"What may not be obvious to a non-engineer like me, however, is that not all operations can or should be split among multiple cores, as doing so can actually make them slower. Because memory bandwidth hasn't kept pace with CPU speed (see Scott Byer's 64-bit article for more info), the cost of moving data to and from each CPU can be significant. To borrow a factory metaphor from Photoshop co-architect Russell Williams, "The workers run out of materials & end up standing around." The memory bottleneck means that multi-core can't make everything faster, and we'll need to think about doing new kinds of processing specifically geared towards heavy computing/low memory usage."

Nahalem's game changing architecture is the key to getting around the bottle neck.

-fate

Sorry for the propaganda link :)

http://www.apple.com/macpro/features/processor.html
 
Infrared,

I think his comments were "memory throughput specific". With other benchmarks YMMV.

With those comments in mind check out this blog:

http://blogs.adobe.com/jnack/2006/12/photoshop_and_multicore.html

Here's the quote that caught my eye:

"What may not be obvious to a non-engineer like me, however, is that not all operations can or should be split among multiple cores, as doing so can actually make them slower. Because memory bandwidth hasn't kept pace with CPU speed (see Scott Byer's 64-bit article for more info), the cost of moving data to and from each CPU can be significant. To borrow a factory metaphor from Photoshop co-architect Russell Williams, "The workers run out of materials & end up standing around." The memory bottleneck means that multi-core can't make everything faster, and we'll need to think about doing new kinds of processing specifically geared towards heavy computing/low memory usage."

Nahalem's game changing architecture is the key to getting around the bottle neck.

-fate

Sorry for the propaganda link :)

http://www.apple.com/macpro/features/processor.html

That's an interesting quote and it does seem to make some sense. But, you
know, when I hear people say multi-threaded stuff is too hard, I wonder if
they're making excuses. I.e., yeah, it is hard, but then again, we can't be
bothered doing the recoding. The path of least resistance is very often the
most attractive.

And I have doubts as to the competence of Adobe's programmers. I mean,
PS CS3 is an application that uses 99% of 1 core if (with no document open)
if you just hold down the space bar. I mean, how well coded is that? :)

ps.png
 
Actually the comment about resale value may end up being more important than the rather unnoticeable (in the real world) difference between the quad and octo choices.
 
That's an interesting quote and it does seem to make some sense. But, you
know, when I hear people say multi-threaded stuff is too hard, I wonder if
they're making excuses. I.e., yeah, it is hard, but then again, we can't be
bothered doing the recoding. The path of least resistance is very often the
most attractive.

And I have doubts as to the competence of Adobe's programmers. I mean,
PS CS3 is an application that uses 99% of 1 core if (with no document open)
if you just hold down the space bar. I mean, how well coded is that? :)

View attachment 161800

Multi-threaded programming isn't hard, managing the threads is the hard part. Like the quote says, If only 1 guy is doing the assembly and 9 guys are carrying the parts, to him, that 1 guy is the bottle neck.

In building metaphor, you can have electricians, plumbers, etc... all ready to go, but they can't do their thing until the walls have been built.

Another good adage is that one guy can make a baby in nine months, but nine guys can't make a baby in one month. Some things just can't be multi-tasked.

Maybe the software is fast as hell, properly multi-threaded, but is still slow. It could be limitations in transferring data to/from the disk. Every time you solve something, there's always something else to be solved.

As far as your spacebar comment, you are doing something and the software is obligated to respond to it. The software has to wake up, look at what your doing in the context of where it currently is, and respond appropriately. If you repeat rate is high enough, your sucking up CPU cycles.

Which is pretty much like the old Doctor joke... "Doctor it hurts when I do this! To which he replies... Don't do that"
 
As far as your spacebar comment, you are doing something and the software is obligated to respond to it. The software has to wake up, look at what your doing in the context of where it currently is, and respond appropriately. If you repeat rate is high enough, your sucking up CPU cycles.

Which is pretty much like the old Doctor joke... "Doctor it hurts when I do this! To which he replies... Don't do that"

Tons of other applications manage not to max out a core when a user
holds down a key. It's crappy coding that causes that. It does not take
Crysis-level CPU usage to handle keyboard input.
 
I was in the same boat just the other day... But I ended up going with the 2009 Quad-Core 2.66 over the 2008 2X2.88... I just have a feeling that it will be slightly faster... Plus, I'm always about the latest technology... Like last time, I'm sure it will be at least another year before an update to the Mac Pro line... I can only assume that Apple built this machine around SL!

Another reason for my decision was based on the fact that I don't use any Pro applications... I think the craziest I get is iPhoto! Okay, maybe running Parallels with Windows XP, but other than that, I just needed the extra hard drives, superdrive, PCI USB ports, and 8GB of RAM is more than enough for me... I already have a 30"ACD on my desk, so that was a none issue.

It should be at my doorstep soon! I can't wait!
 
dual core is overkill for iPhoto :p

quadcore 2.66 is not faster than octo 2.8 in multithreaded applications.
 
Maybe not

dual core is overkill for iPhoto :p

quadcore 2.66 is not faster than octo 2.8 in multithreaded applications.

Maybe not, but it's not much slower, and it's at least 10% faster on single threaded apps. Add that it comes stock with a faster video card, more RAM and twice the HD and at the same price the Quadcore is compelling.

What will be the kicker is if Intel produces chips in a year that can drop into the Nehalem boards, but the 5400 Xeon boards are left in the lurch.
 
Where does this idea that Snow Leopard is geared toward Nehalem
come from? Does SL even run properly on the new machines yet?

Apple keeps new hardware quite secret. Obviously a few people had to make sure that Leopard runs properly on the new machines, but the vast majority of developers who work on Snow Leopard had no reason to try their software on the new machines. So they probably saw their first 2009 MacPro's at about the same time as we did.
 
also....the new one is more future-proofed as far as snow leopard, multi-threading etc...


There is no such thing as a "future proof" computer... or any high tech device really. The '06 Quad 2.66 looked pretty expandable just 2.5 years ago. Today it looks like an ordinary computer. What ever the "big" feature is today there will be something bigger in two years, and "today's" computer will look lackluster.
 
Maybe not, but it's not much slower, and it's at least 10% faster on single threaded apps. Add that it comes stock with a faster video card, more RAM and twice the HD and at the same price the Quadcore is compelling.

What will be the kicker is if Intel produces chips in a year that can drop into the Nehalem boards, but the 5400 Xeon boards are left in the lurch.

What makes the Quadcore less compelling is the maximum RAM you can put in there....
 
picked up the 2008 2.8 Today

thanks for all the replies everyone. after heavy consideration, I went ahead and went to my college computer store and picked up a new 2008 2.8 8 core Mac Pro for $1999. I think the price is a steal. :)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.