The PSU in the nMP is a strange choice. Surly they did not compromise on the 450 PSU just cause it fit in the sexy new body.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Anandtech's Mac Pro Review Is Out!
- Thread starter Brandon263
- Start date
- Sort by reaction score
The PSU in the nMP is a strange choice. Surly they did not compromise on the 450 PSU just cause it fit in the sexy new body.
The i7 mini gets very loud under load. The i7 iMac is not nearly as loud, but it is still much louder than the new Mac Pro (the i5 iMac, btw, is dead silent). The new Pro is beautifully designed for airflow, whereas the mini and iMac place the processor in a cramped space.
It uses them, just not very effectively. I've watched Activity Monitor with various activities, imports, exports, etc, and on my Mini it's lucky to ever hit 400%, more often between 2-300%. HandBrake on the other hand will push 800%. If it scales to six cores, that will be a nice bonus.
BTW, did you notice the CS 6 bench that Anand did on the nMP? 6 seconds versus 13.5 with CS 5.
Yeah, just wish we had the iMac comparison for that.. would help a lot.
The more I think about it though, given the very few other choices (iMac, hackintosh), the hex/d500 is probably going to be the all-around performer that I want. Not ideal at everything, but solid, stable, and a huge leap from my current 1,1.
Anandtech was using Core Temp app, which gets data from thermal sensors on CPU.
Clocks dropping to 2 Ghz might've been Intel's Thermal Monitor kicking in, to prevent CPU melting.
It really looks like there is no adequate heat dissipation for 100% CPU+GPU load.
Well, it depends on whether CS6 or CC actually bring meaningful improvements to multithreaded performance.
It certainly seems that if you're still using CS5 or earlier, you're better off with a consumer quad-core CPU which is a generation ahead of the Mac Pro.
Honestly though, even if there is better multithreaded performance in newer versions, it's rare than I'm actually waiting for Photoshop or Lightroom to process an image when I'm editing. The only thing which actually takes any amount of time is batch exporting images.
I agree that it would be a lot more useful to have up-to-date benchmarks, and they could probably have used a trial for that even if they don't have a license for the latest versions.
But it doesn't surprise me that CS5 and LR3 is where they stopped upgrading - I've seen the same from just about all the professionals I know that work with these applications.
If you need CC to benefit from the 6-12 cores of the Mac Pro, that really bumps up the price of the machine.
I wonder if that's due to optimizations or better use of the GPUs in CS6 than better multithreading. We would have to see the same test run on an iMac as well.
Even if you assume the best case scenario for the Mac Pro where the iMac time stays the same as CS5, it's not scaling particularly well - you move from 4 to 12 cores and only halve the time.
Last edited:
It's was either heat or the PSU struggling.
It's a slick design, but there is going to be a lot of heat produced from a CPU and 2xgpu. I believe your right, if pushed the system will struggle. I would highly recommend AppleCare if one is going to be pushing their systems .
Well, for my case, it's really LR where I want those improvements. Imports, exports, and especially the preview redraws... Going through a lot of D800 images is painful on my 1,1.
I'm disappointed that Anan made up this silly term "power virus" its like he is trying to be diplomatic with Apple. He knows much better.
For anyone that builds their own PCs, it's call stress testing.
And the Mac Pro failed.
At this point a PC builder would be looking at updating the cooling in their system, and the PSU. They would re run the stress tests and see if it passes.
This.
Part of my job includes designing and/or purchasing workstations for about 30 users and the ~300 node cluster. It's a mixed bunch of users, some engineering, some scientific, and a lot of DCC. Odd mix but it's an internal group in a large multinational.
In any case, the cluster runs 24/7 unless doing maintenance. Workstation cycles, both CPU and GPU, are put to use whenever idle.
We've recently gone to custom built because none of the major OEMs seem to be able to build machines anymore with the thermal performance and reliability we need, although BOXX's stuff is getting better. Our new builds are dual Ivy Xeon, air cooled, and will run any major stress test for 48 hours with a max CPU temp of 72C (22 ambient). Max allowable dBA is 20 from 3 feet away.
We do have to endure the absolute horror of a mid-tower size, though, this is a trade we will make.
The 2008-2010 Mac Pro was actually very good in this regard and we deployed many for users who preferred Mac.
I've got an 8 and 12 core nMP coming to test but if Anand's numbers are correct it is not deployable in current form. Typical v1 stuff for Apple. I imagine v2 or 3 will see a larger PSU and improved thermals, though the size will likely have to increase for that to happen.
I agree that the only talk of throttling was under the "power virus" situation.
However, Anand didn't really benchmark it at standard loads over time, as far as we know, these were all cold benchmarks. In short, neither of you know for sure (or if you do, please point me to the part I missed).
Many gaming sites measure FPS over many minutes to see how the coolers handle it. It'd be nice to see a pretty graph showing how it performs over the course of an hour or so at gaming and other tasks.
There's got to be a reason Apple lowered the base clock of the D700 30%. That's a heck of an assumption that they did so to simply have it run at boost all the time.
As the metal tube and heat sink warms up, performance will suffer, Anand did not explore this sufficiently -- this design in particular and especially the target buyers of this machine would benefit from this.
Do you have any evidence at all of this? These are some of the benchmarks I've been waiting for.
You and seb seem to be making a lot of assumptions. As far as we know these are downclocked 7970's (a $300 card) with extra RAM. I've seen no indication they have any of the performance advantages of the other FirePro cards over Radeon in Windows or that 7970's don't perform as well as they do in OS X.
----------
This. Though Anand was happy to equate the D700 to the 7970--at least he tried to bring some sanity to this madness.
I suspect it's a 6gb 7970. Those were released .
I also suspect it's been underclocked to cater for the small PSU .
To be honest the 7970 is a decent card, just a generation behind.
Anand didn't make up that term though. Power virus is a known term used precisely for these situations.
----------
Ah, have you missed the fact that under Windows the D700's are recognised as Fire Pro's and are using FirePro drivers? There's your answer.
----------
He actually equated to R9 280X, which is basically the same as HD7970 but only when comparing it to other gaming hardware.
----------
Gaming shouldn't really stress the machine since it won't be using more than couple cores on the CPU. It seems like the machine only throttles when you push the CPU and the GPU's to the max, like we assumed.
From your friends MVC:
"It would seem logical that 7970 running at higher clocks would win most graphics tests but it regularly doesn't."
http://forum.netkas.org/index.php/topic,8154.0.html
Last edited:
guess he hasn't seen the ifixit teardown and/or how they arrived at the 8/10
either way.. thanks for linking.. pretty much the best thing i've seen online this year..
(he's right about the jesus part though.. sat next to him last night on the subway)
Last edited by a moderator:
Seconded on my 6 core.
No odd audio bits and SUPER QUIET under a heavy Vray rendering load.
Exactly.
Workstation/server computers and components cost such a hefty premium exactly because they are able to run 100% loads for extended periods of time. Hell, one of my DB servers been running a very busy Oracle instance for 695 days - this is what you pay for in professional computing.
I'm pretty sure there are real-world compute tasks that would fully utilise both CPU and GPUs, and Macpro will underperform in these.
Sadly, in this regard Macpro is just a glorified desktop PC, not a workstaion.
----------
Or lowering base/boost clocks for chips.
Changing chassis size would be very expensive, with all the redesign and factory retooling and training.
Last edited:
Exactly. They boast about gaming at full hd on a screen that is capable of a much higher resolution. In a nutshell the iMac graphics aren't powerful enough for 1440p gaming despite having 1440p resolution.
That's not an answer. Just because Windows calls it something doesn't mean it is something or will behave that way. Several of Anand's testing apps specifically called it a 7970.
Waiting for SpecPerf.
.... Regardless, not a FirePro card. He went on to say that the DIY is a "better deal"
Really? Again, you have some hot benchmarks to prove this?
I'm just asking to see some benchmarks before making assumptions -- as we have no hot benchmarks at this point and to do otherwise is baseless conjecture. Crossfire generates a ridiculous amount of heat. When I put two 6870's in CF on my 1,1, I had to manually boost the PCIe bay fan or it wouldn't even boot.
This is a special design that consolidates 3 heat-sinks into one. It's also a workstation. It is definitely not a lot to ask to see how it behaves after it's warmed up.
The ONLY data we have is that when running Luxmark it nearly burns out the CPU -- not entirely relevant to every day tasks, but this is in fact the only data.
Did not see that (it was posted yesterday and I don't browse netkas)! Most of the scores (apart from that one) were very similar but in general the D700 does a little better. Looks like the D700 has some optimizations which give it a little edge even at a lower clock in OS X.
.... Though the D700 were running over PCIe 3.0 and a much faster processor, in addition to having 6GB each.
The real test will be in Windows with relatively equivalent hardware, comparing a 3 year old machine to the new one is a little off.
I play BF4 on 1440p with Medium settings. Not stellar frame rates, but definitely playable. GTX 680MX.
iMac graphics are barely powerful enough for 1440p.
heh.. yeah.. a little too funny for my tastes.
tell your boy mvc that everything i've been interested in talking about over the past few months has climaxed.. the design has been revealed (for the most part) and there's not much else i find exciting enough to argue about.. and i certainly don't want to get caught up in this reality show drama.. so i'm out of here and he should come back to his old stomping grounds..
(ie- i'm the reason he got his recent extendo time out.. well, he's the reason but i was the catalyst.. he went on a 6or7 post tirade about how stupid i am and i guess the mods didn't like that too much)
so i'll see you guys in about 5years.. (or maybe in the mbp forum next year at some point)..
adios
##
Last edited: