Why the 2009 Mac Pro is a better buy than the 2008 Mac Pro

[MadisonTate]

I hate repeating myself. Worse, I hate thinking that some new person is going to be talked into buying yesteryear technology because misery loves company.

1. No computer will be "the fastest" for more than a month or two. Faster processors are about to be released right now. 3.2GHz quad's, soon to be followed by 6 or 8 core Nehalem's, with mobile ones right on their heels. Don't get ticked off when the machine you bought yesterday is no longer the best.

2. Buy a machine that makes sense for you. If you're doing Photoshop, I doubt you really need an octo-core anything, anyhow. Overcapacity just wastes energy. It's like having a Lamborghini that you use to commute back and forth to work during rush hour. It's nice to have, but you're never gonna get over 80 mph. And if you get caught doing 100, you'd better have an incredible lawyer.

3. Case in point, 6 years ago, I bought a 1.5GHz Pentium M tablet PC as my primary computer. I do serious software development, and beat the snot out of that thing every day. If the machine hadn't started failing several months ago, I would still be using it right now. Things didn't "stop" working. It still did what I needed it to do. My five year old car still has 335hp, 270 of which only get to wake up once or twice a year when I take it to the track. Am I ticked off that the manufacturer now makes a 360hp version? Am I somehow no longer getting my money's worth out of my car?

4. Don't buy RAM or disk drives from Apple. Seriously. You can get it MUCH cheaper aftermarket.

5. DO buy the latest, greatest processors you can get your hands on, within your budget. You can always add RAM, drives or whatever. The best you can do with processors is get another 20-25% clock rate bump, and even that can be hard unless it's designed in.

6. Don't pay attention to synthetic benchmarks. The only benchmarks that count are ones from the applications YOU will use. Some benchmarks measure how fast you can move things in memory. Who cares! A Nehalem can execute nearly twice as many instructions per clock as its predecessor, but the memory speeds aren't that different. If you're running a real application, it's the processing that you care about.

I don't want to waste too much time here, but here's the Mad Tater's view on things...

The second most dangerous person in the world is someone who thinks they know everything, but doesn't know anything. The most dangerous is the person who follows him.

Oh yeah, and even MORE dangerous is a moron like Jim Cramer who leads thousands off a cliff.

 

[More]

 

[Grimace]

This thread can go directly to wasteland or by merged with the 900 other comments on this matter.

We don't have enough benchmarks to mark concrete conclusions about anything right now.

Clockspeed differences affect some applications -- for others, multi-core and single-core applications may behave differently -- to say nothing of the RAM advantages which at an early stage, heavily favor Nehalem.

Right now, and for the foreseeable future, the answer is "it depends".

 

[More]

^agree to a certain extent.

 

[jjahshik32]

 

[MadisonTate]

This thread can go directly to wasteland or by merged with the 900 other comments on this matter.

We don't have enough benchmarks to mark concrete conclusions about anything right now.

Clockspeed differences affect some applications -- for others, multi-core and single-core applications may behave differently -- to say nothing of the RAM advantages which at an early stage, heavily favor Nehalem.

Right now, and for the foreseeable future, the answer is "it depends".

Grimace,

I'm trying to warn those who hope to future-proof themselves. The 2008 machines really can't be compared to the 2009 machines yet.

You need (a) a fully 64-bit OS like Snow Leopard, (b) a fully 64-bit Cocoa application (or a non-UI 64-bit app), (c) an OS that can take advantage of hyperthreading, which doesn't help pure memory benchmarks but can do wonders in the real world. Pull all that together and THEN do the comparison.

 

[surflordca]

Don't read it. It will save you a big brown mess that you won't have to clean up

 

[MCHR]

Grimace,

I'm trying to warn those who hope to future-proof themselves. The 2008 machines really can't be compared to the 2009 machines yet.

You need (a) a fully 64-bit OS like Snow Leopard, (b) a fully 64-bit Cocoa application (or a non-UI 64-bit app), (c) an OS that can take advantage of hyperthreading, which doesn't help pure memory benchmarks but can do wonders in the real world. Pull all that together and THEN do the comparison.

That's exactly what's making any comparison somewhat confusing.

Here's a scenario: fast forward six to eight months, when Snow Leopard drops, and some apps are ramping up to take advantage of 'threading.

Where do you want to be THEN? 2009 machine, or 2008 machine?

 

[EricTheRed71]

I don't think it's as simple as the new being better than the old & each person has to look at their own needs before deciding what to do.

Our situation:

We need 2 new Mac Pros for running a 3dsmax renderfarm under Windows x64. We already have 6 of the original dual 2.66's & 2 of the first 3.0 Octos from 2007. We need to render very large, memory intensive scenes & are struggling sometimes with the macs only having 16GB of RAM.

The plan is to get 2 new machines & pack them out to 32GB of RAM. 3dsmax uses all available RAM & all available cores to render with. Benchmarks show that both the new octo 2.23 Nehalem & the 2008 2.8 octo are roughly even & also the roughly the same as our 2007 octos.

A refurb 2.8 octo costs roughly €2300 in Germany & the new octo 2.23 costs about €2850. No great difference then. The problem comes with adding the RAM - about €1000 for the 2008 model & about €3500 for the Nehalem. Sure the price of RAM may come down but when & by how much?

So we are looking at spending either:

€2300 + €1000 x 2 = €6600 for the 2008's

or

€2850 + €3500 x = €12700 for the 2009's

So which is the better buy?

 

[MCHR]

^^

That's an entirely different equation, one unique to you at this purchase point.

IF you get your machines to be on the cutting edge, and saving minutes from renders, etc. is of importance, then the new machines may benefit.

IF your constraint is purchase price, and how that fits into your budget, that's another matter completely.

However, IF you get your machines to last four or five years (as many do), then an equation of amortization of that price difference over several years is important.

The wildcard is how the coming OSX, apps, etc are developed to best benefit this newer architecture. THAT'S what's giving everyone headaches, IMHO.

 

[EricTheRed71]

^^

That's an entirely different equation, one unique to you at this purchase point.

IF you get your machines to be on the cutting edge, and saving minutes from renders, etc. is of importance, then the new machines may benefit.

IF your constraint is purchase price, and how that fits into your budget, that's another matter completely.

However, IF you get your machines to last four or five years (as many do), then an equation of amortization of that price difference over several years is important.

The wildcard is how the coming OSX, apps, etc are developed to best benefit this newer architecture. THAT'S what's giving everyone headaches, IMHO.

To be honest it's the RAM issue that's important rather than saving a few minutes on a render. If I know that the RAM prices wold be coming down soon then I'd probably go for a 2009 model but then again the 4GB chips are not actually available from my normal reseller atm.

These machines are for business use so having to last a certain amount of time isn't a problem. If something better is released at a decent price point then any current machines are just relegated to the render farm. We have up to now avoided going for the "cutting edge" as it just isn't needed & is soon overtaken by something new anyway.

OSX performance is a non-issue for us unfortunately.

 

[Horst]

So we are looking at spending either:

€2300 + €1000 x 2 = €6600 for the 2008's

or

€2850 + €3500 x = €12700 for the 2009's

So which is the better buy?

Well, you could buy 3 or even 4 '08 MPs for the price of the 2x '09 models, sure that would help with your renderfarm.

It's hard to tell right now where RAM prices and availability of larger modules are going, or what advantages the next revision of the Intel chips will bring.

Right now it isn't even clear how powerful the '09 MPs actually are, or what possible bugs will emerge in the next few months.

The '08 are rock solid, that we know, so for business I'd get those if I had to buy now, but that's just me.

One more thing to look into might be how well 3dsmax is doing with hyperthreading; on the Mac side the benchmarks aren't looking too impressive right now, but that's mainly Cinebench.