Anandtech's Mac Pro Review Is Out!

[kazmac]

Yes this made me giggle too

lol..awesome

and brought Gollum to mind.

Actually, this review knocked all my delusions of owning a Mac Pro out of the water. I do not need that kind of computing power and now my little delirium has been promptly kicked to the curb (a retina MBA or Mac Mini would be fine for me. )

And I love Pixelmator. A joy to use.

 

[nanofrog]

By the way, what are you doing that makes ECC GPU so important?

I can't speak for beaker7, but there are instances where this would be important.

For example, scientific/engineering type of usage patterns that rely heavily on recursive algorithms that leverage GPGPU processing to speed up floating point performance past what a CPU would be capable of.

And Apple has targeted the scientific and engineering markets in the past, albeit half-heartedly IMHO, so there is a possibility such an issue could arise for some users.

 

[theSeb]

I can't speak for beaker7, but there are instances where this would be important.

For example, scientific/engineering type of usage patterns that rely heavily on recursive algorithms that leverage GPGPU processing to speed up floating point performance past what a CPU would be capable of.

And Apple has targeted the scientific and engineering markets in the past, albeit half-heartedly IMHO, so there is a possibility such an issue could arise for some users.

Thanks nano. I am aware of those applications, but I was genuinely curious why it was important to beaker7 specifically.

 

[derbothaus]

All of the iMac gaming tests I've seen turn the resolution down to 1920 x 1080 to get acceptable performance but boast about gaming at 'full HD'.

1080p is 'full HD'. It is not 'native' though and doesn't look as good scaled on a 27" screen.

 

[Spinland]

Thanks nano. I am aware of those applications, but I was genuinely curious why it was important to beaker7 specifically.

I'm specifically interested in the load-throttling behavior of the nMP, especially of the hex core that I'm awaiting, hence my curiosity about what Beaker7 might have encountered during work loads. One of my more popular original 3D character designs is that of a cartoonish black bear (we do live in the Adirondacks region, after all) and I endowed it with a full, lustrous coat of dynamic fur that interacts with inertia, gravity and wind effects.

Needless to say rendering scenes where he's a feature, using Lightwave that is totally CPU bound for rendering, makes my CPU cores beg for mercy. My best render node right now is a quad core i7 and I see wait times of 18 hours or more for involved scenes, with my machine's fans spinning like jet engine turbines. I am anxious, eager and nervous to see how my new investment is going to fare under these loads, and any advance intel I can glean while I wait is welcome.

 

[Cubemmal]

My stock 6-core does not appear to have this problem. I use USB audio anyway - the startup chime is the only time the built-in speaker is used. So far I have had no audio glitches whatsoever, and I am able to drive my old FW audio interface through Tbolt without any issues.

That's good to hear, er, not hear. I ordered a BTC hex core.

I have driven all cores for hours on end, and the nMP is without question the quietest computer I have ever owned. It is much quieter under load than the i7 mini powering my home theater or the i7 iMac on my desk at the university.

Quieter than the iMac, with all that horsepower - that says a lot.

 

[FredT2]

I have driven all cores for hours on end, and the nMP is without question the quietest computer I have ever owned. It is much quieter under load than the i7 mini powering my home theater or the i7 iMac on my desk at the university.

My Mini gets very noisy with anything that pushes all of the cores (2012 i7). Does the iMac behave similarly? I've always heard that it is very quiet.

 

[flat five]

Interesting, thanks for the tip. In my case I actually mostly use the Photoshop Extended 3D stuff for painting textures directly onto 3D model files, as well as quick and dirty video rendering, but this is certainly worth looking at when I get some play time.

yeah, pixelmator isn't going to help out there.. it's just image creation/editing for now.

 

[grouch]

My Mini gets very noisy with anything that pushes all of the cores (2012 i7). Does the iMac behave similarly? I've always heard that it is very quiet.

most imacs are pretty quiet, rarely gets as loud as the mini. it also depends which mini as well.. some tend to get louder than others.

 

[Spinland]

it also depends which mini as well.. some tend to get louder than others.

I have the mid-2011 Mini Server 5,3. It's primarily my file server, but also works as a distributed render node and, when I have all four cores cranking out render frames, it sounds like a jet engine turbine whining at full throttle.

 

[ha1o2surfer]

It does not throttle under heavy load. It throttles under a power virus. That's not heavy load, that's unexpected load. Have you actually read the review? He said that he couldn't get throttling under any regular process loads.

----------



Exactly. Power viruses are, well, power viruses. Every computer will eventually throttle since you are throwing at it a load it isn't designed to take.

You should read the review.
He could only get it to throttle using what he calls a "power virus", not any kind of real workflow. You would be able to game all day with no throttling.

Nah, A computer should not throttle no matter what load is put on it, if it's working properly. I honestly couldn't accept that as an answer lol go ahead, bash me

 

[Spinland]

Nah, A computer should not throttle no matter what load is put on it, if it's working properly. I honestly couldn't accept that as an answer lol go ahead, bash me

No intent to bash here, but I believe that if you want there to exist a computer--any computer--that doesn't eventually start throttling as you ramp the load arbitrarily high...you are doomed to disappointment.

ETA: iBug2 has posted below a delightful use case that torpedoes my objection. I have to laugh and withdraw it with due compliments.

 

[iBug2]

No intent to bash here, but I believe that if you want there to exist a computer--any computer--that doesn't eventually start throttling as you ramp the load arbitrarily high...you are doomed to disappointment.

No I don't think he is. There always is an option to use a computer inside an industrial freezer. Then you can push it to the max and the fans won't even spin.

 

[Beta Particle]

I'm a bit annoyed that the light graphics tests were PS 5 and LR 3, a bit old there, and leaves me wondering if performance is still the same, or as he indicated, there are improvements in CS6.

Doesn't give me the info I need.

I agree that it would have been more useful to have benchmark results from the most recent versions of each application, but if there's no reason to upgrade - or good reasons to avoid upgrading (like the subscription model) then it's somewhat understandable.

I'm not sure that any of the professionals I know are using the latest versions of Adobe's programs now.
Adobe's costs have always been extremely high, but it seems that they are only interested in large scale customers these days with the changes they have made to how they license their products.

1. If I recall correctly CS5 was the last time you had the opportunity to upgrade from more than one version ago.
The release of CS5 was when they announced that the licensing model had changed, and you would only be able to purchase an upgrade from the previous version.
So even though I did not really need it (the content-aware tools are useful I guess) I went from CS2 to CS5. People that I knew who were running CS3 also went to CS5, as they would not have had the opportunity to upgrade to CS6. (Adobe used to let you go back three versions with an upgrade)

2. While there may have been improvements to multithreading in CS6, this is not really something that Adobe made a lot of noise about. CS5 handles four cores just fine, which is all that most systems built to run Photoshop have. (I suppose that's a bit of a chicken-and-egg situation)

3. CS6 did not really add anything relevant to my work. On the PC side at least, it just felt like a minor UI refresh.

4. After CS6, Adobe moved to a subscription model, which I have no interest in being a part of.
I would be really interested in seeing how that has affected Adobe's business, as most of the professionals I know are holding onto their CS licenses and would rather seek alternatives than subscribe to CC.


While I have kept Lightroom up to date, there's also been very little reason to upgrade from Lightroom 3 to version 4 or 5 unless you need support for a specific camera. Lightroom 3 was the last time there were significant improvements to Camera Raw's image quality.

It's great that crossfire is working on Windows. It should put everyone's worries to rest whether or not they could game with this machine. And on OS X it'll perform slightly worse than a single 280X, which is still quite decent gaming performance.

Honestly, it's rather poor performance if you plan on gaming with it.
While I understand there are people that want a machine to use for work and for games, you could do a lot better with your money if your main focus was gaming.

The D700's are a $1000 upgrade, and a 780Ti is only ~$700 which offers the same or better gaming performance on a single card. It would be even cheaper if you went with SLI or Crossfire, but I would not recommend that, as using a single card gives you a smoother experience.

It does not throttle under heavy load. It throttles under a power virus. That's not heavy load, that's unexpected load. Have you actually read the review? He said that he couldn't get throttling under any regular process loads.

I think calling that a "power virus" is making an excuse for the system. Running FurMark and Prime95 or IntelBurnTest are standard operating procedure for testing custom PC builds.

I think you should read the article again. By the way, what are you doing that makes ECC GPU so important? I am sure that this isn't the first time you have mentioned it.

Everyone was happy to compare the D700s to AMD's W9000 workstation cards when they thought it had ECC memory and was a significant cost savings.

Now that we know performance is quite a bit lower than the W9000 cards, and the D700 cards do not include ECC memory, suddenly it's not a big deal.

No intent to bash here, but I believe that if you want there to exist a computer--any computer--that doesn't eventually start throttling as you ramp the load arbitrarily high...you are doomed to disappointment.

Put one of these on any CPU and I doubt you would ever have to throttle it - it's quiet too: 12.6 dBA. In early 2014, they will also be adding active noise cancellation.

 

[Spinland]

No I don't think he is. There always is an option to use a computer inside an industrial freezer. Then you can push it to the max and the fans won't even spin.

LOL! Okay, I stand corrected, I withdraw my objection.

 

[wildmac]

I agree that it would have been more useful to have benchmark results from the most recent versions of each application, but if there's no reason to upgrade - or good reasons to avoid upgrading (like the subscription model) then it's somewhat understandable.

I'm not sure that any of the professionals I know are using the latest versions of Adobe's programs now.
Adobe's costs have always been extremely high, but it seems that they are only interested in large scale customers these days with the changes they have made to how they license their products.

1. If I recall correctly CS5 was the last time you had the opportunity to upgrade from more than one version ago.
The release of CS5 was when they announced that the licensing model had changed, and you would only be able to purchase an upgrade from the previous version.
So even though I did not really need it (the content-aware tools are useful I guess) I went from CS2 to CS5. People that I knew who were running CS3 also went to CS5, as they would not have had the opportunity to upgrade to CS6. (Adobe used to let you go back three versions with an upgrade)

2. While there may have been improvements to multithreading in CS6, this is not really something that Adobe made a lot of noise about. CS5 handles four cores just fine, which is all that most systems built to run Photoshop have. (I suppose that's a bit of a chicken-and-egg situation)

3. CS6 did not really add anything relevant to my work. On the PC side at least, it just felt like a minor UI refresh.

4. After CS6, Adobe moved to a subscription model, which I have no interest in being a part of.
I would be really interested in seeing how that has affected Adobe's business, as most of the professionals I know are holding onto their CS licenses and would rather seek alternatives than subscribe to CC.


While I have kept Lightroom up to date, there's also been very little reason to upgrade from Lightroom 3 to version 4 or 5 unless you need support for a specific camera. Lightroom 3 was the last time there were significant improvements to Camera Raw's image quality.

So... I should buy a 2010 computer for 2010 software?...

Actually kind of a serious question. If someone is dead set against Adobe CC, then the nMP might be an egregious waste of money.

However, a lot of folks are keeping LR up to date. So no idea if the test is relevant there or not, since it's 2 versions old.

But that part of the Anandtech test really doesn't help, because not only did they not test against a current iMac, they didn't test against 4,1 or 5,1 MacPros either.

 

[iBug2]

Honestly, it's rather poor performance if you plan on gaming with it.
While I understand there are people that want a machine to use for work and for games, you could do a lot better with your money if your main focus was gaming.

Why would you buy a Mac Pro if your main focus is gaming? Hell with that, why would you buy a Mac for that?

nMP has good performance for gaming and that's that. I didn't say it's the fastest gaming machine on earth or is comparable to similarly priced PC's for gaming.

I think calling that a "power virus" is making an excuse for the system. Running FurMark and Prime95 or IntelBurnTest are standard operating procedure for testing custom PC builds.

Then write to Anand and tell him that.

Everyone was happy to compare the D700s to AMD's W9000 workstation cards when they thought it had ECC memory and was a significant cost savings.

Now that we know performance is quite a bit lower than the W9000 cards, and the D700 cards do not include ECC memory, suddenly it's not a big deal.

The difference of FirePro isn't just ECC. You'll be able to get all the benefits of FirePro on Windows so you are getting a great deal on a dual FirePro's.

 

[ha1o2surfer]

No intent to bash here, but I believe that if you want there to exist a computer--any computer--that doesn't eventually start throttling as you ramp the load arbitrarily high...you are doomed to disappointment.

ETA: iBug2 has posted below a delightful use case that torpedoes my objection. I have to laugh and withdraw it with due compliments.

ok ok I could use it in a freezer. I do have such a laptop that doesn't need a freezer though

 

[FredT2]

SHowever, a lot of folks are keeping LR up to date. So no idea if the test is relevant there or not, since it's 2 versions old.

I think Lightroom 3 is still the fastest version Adobe released. Bottom line though is that Lightroom is not very well suited to be able to take advantage of nMP: doesn't make very good use of multiple cores, barely uses hyper threading at all, and from what I've been able to discover has no Open CL support. That's why an i5 iMac was able to beat an i7 rMBP in Anandtech benchmark, and lots of unused cores in the nMP did not help it overcome the slower processor speed. I'll bet that the 6-core nMP will beat the iMac, though not by much. We can only dream that Adobe finds a way to add Open CL support in version 6.[/QUOTE]

 

[DisMyMac]

Everyone was happy to compare the D700s to AMD's W9000 workstation cards when they thought it had ECC memory and was a significant cost savings.

Now that we know performance is quite a bit lower than the W9000 cards, and the D700 cards do not include ECC memory, suddenly it's not a big deal.

What about running the 2 cards as mirrors and checking each other's results? That's the same thing as ECC memory, almost.

 

[MH01]

It does not throttle under heavy load. It throttles under a power virus. That's not heavy load, that's unexpected load. Have you actually read the review? He said that he couldn't get throttling under any regular process loads.

----------



Exactly. Power viruses are, well, power viruses. Every computer will eventually throttle since you are throwing at it a load it isn't designed to take.

If you have an excellent PSU and adequate cooling, I do not see a system throttling . Unless I am mistaken but power and heat resulted in throttling.

The PSU in the nMP concerns me if the system is pushed , and so is the heat.

 

[wildmac]

I think Lightroom 3 is still the fastest version Adobe released. Bottom line though is that Lightroom is not very well suited to be able to take advantage of nMP: doesn't make very good use of multiple cores, barely uses hyper threading at all, and from what I've been able to discover has no Open CL support. That's why an i5 iMac was able to beat an i7 rMBP in Anandtech benchmark, and lots of unused cores in the nMP did not help it overcome the slower processor speed. I'll bet that the 6-core nMP will beat the iMac, though not by much. We can only dream that Adobe finds a way to add Open CL support in version 6.

From what I've seen, LR does use the cores, at least on my old 1,1 cMP. From what I've read, it uses up to 6. Which of course is part of the problem with the Anandtech test with a 12-core as the test bed.

Really need to see PS and LR tests on a hex-core, which is the most likely choice for graphic designers and such.

 

[MH01]

The difference of FirePro isn't just ECC. You'll be able to get all the benefits of FirePro on Windows so you are getting a great deal on a dual FirePro's.

Agreed. Though the point is that a lot of people have been claiming the D700s are W9000, and the nMP a bargain. Without knowing what the D700 or any of the D series is, the jury is out on exactly what deal we are getting.

----------

I know gaming isn't the intended purpose of the Mac Pro, but two things struck me in the gaming tests. Firstly games had to be run in Windows to use both graphics cards. Secondly the performance of 2 D700's is around the same of a single GTX 780 TI.

Even if you do accept that you need to run games in Windows not OS X, if you want to game on a Mac at 1440p with the highest settings, you will need a workstation that costs over £3k (base spec with D700's), more likely over £5k if you want a decent spec in the rest of the system. I'm not going to say a £300 PC can do it because that's nonsense, but there is NOTHING in Apple's product range other than the Mac Pro that can do this. The Mini doesn't have dedicated graphics. All of the iMac gaming tests I've seen turn the resolution down to 1920 x 1080 to get acceptable performance but boast about gaming at 'full HD'.

You just have to accept that Apple does not do a gaming system.

The Mac Pro was the only machine that ever had a dedicated desktop GPU. It's good to see the new nMP have decent gaming performance in such a small package. Anyone wanting to game is crazy to spent that much money on the performance that the nMP offers in gaming, for a secondary function it's fine.

 

[beaker7]

E5-2697 v2 max operating temp is 86 C.

Running it at 95+ under heavy load not exactly smart.

Make sure you all buy your AppleCare.

 

[Derpage]

Sure, throttling due to heat limitations is one thing, throttling caused by voltage drops is pathetic and piss poor engineering.