All computer systems will eventually be obsolete. That is a fact. However the nMP is so full of compromises that a computer made in 2010 can still beat it soundly. That was a misstep in apple's strategy.
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.
New MacPro 8c D700 getting annihilated in benchmarks :-(
- Thread starter HumpYourWayUp
- Start date
- Sort by reaction score
All computer systems will eventually be obsolete. That is a fact. However the nMP is so full of compromises that a computer made in 2010 can still beat it soundly. That was a misstep in apple's strategy.
And thus the sadness.
At that point the fastest Mac will be a neutered mid-level machine with a short future.
As I have pointed out before, imagine if our 2009s still held either a GT120 or an AMD 4870. How useful would they be? How much value would they have held?
The 2006 machines that are still productively running Yosemite are doing it while all of the other 2006 machines have slid into "old & unuseful" category. Proprietary, and non-upgradeable GPUs were a bad idea when announced in June of 2013. These benchmarks just prove everything that has been said about them since.
Has there ever been another time when a 6 year old Mac could be upgraded to be faster than anything currently available from Cupertino? I don't think so.
I am saddened too.
I awaited the NMP announcement with great anticipation.
I was ready to jump in with credit card in hand.
Then I saw the closed system with very little upgrade potential in the GPU or CPU.
Very high prices for something that was not really a substantial improvement over my 5,1.
With all of Apple's money, they could have done something truly amazing.
Obviously this is not a priority for them.
Maybe they can make an Apple Watch Mac Pro?
I'm hoping the nMP will be a phase, and apple will realize it was a mistake. Just like the old G4 Cube.
I'm disappointed in the nMP. I know its very high tech, and from an engineering standpoint its pretty damn impressive. But, its really just an 'appliance' computer, similar to the new sealed up mini and imac, with soldered on everything. Use it until its too obsolete, then recycle it and buy a new one from Apple. Thats why when I decided to drop a chunk of cash on a mac to replace my older main PC, I got a 2012 tower, and not a nMP.
I'm disappointed in the nMP. I know its very high tech, and from an engineering standpoint its pretty damn impressive. But, its really just an 'appliance' computer, similar to the new sealed up mini and imac, with soldered on everything. Use it until its too obsolete, then recycle it and buy a new one from Apple. Thats why when I decided to drop a chunk of cash on a mac to replace my older main PC, I got a 2012 tower, and not a nMP.
I've had an 8 core nMP D700 1Tb I've setup Windows 8.1 for a client only last week side by side with my 12 core 3.46 with Sintech/PCIe 1Tb Apple blade. I have ran CS6, Solidworks and 3DSMax on both setups in bootcamp though my 12 core is running 7 x64 and not 8.1.
With mulitcore the old girl isn't that far off in terms of performance, feels just slightly slower than the 6,1. I would imagine that a 980 would close that gap even more, though single core the 6,1 always shows a clean pair of heels as it should do. I have no benchmarks nor do I care for them really it's real life use I care about.
But the 6,1 isn't a replacement for the 4-5,1 it's a totally different product and like comparing a Granny Smith with a Washington Red. It's a 440w single box Xeon that makes no noise and has very different uses I've found to it's predecessor...
With mulitcore the old girl isn't that far off in terms of performance, feels just slightly slower than the 6,1. I would imagine that a 980 would close that gap even more, though single core the 6,1 always shows a clean pair of heels as it should do. I have no benchmarks nor do I care for them really it's real life use I care about.
But the 6,1 isn't a replacement for the 4-5,1 it's a totally different product and like comparing a Granny Smith with a Washington Red. It's a 440w single box Xeon that makes no noise and has very different uses I've found to it's predecessor...
I have a 6 core nMP with D700's and a PC, core i7 3770k + GTX 780 - the PC is a much better gaming experience and yes it was much cheaper than the nMP.
So while the D700's aren't going to win any awards for their gaming performance, it's pretty much in a class of it's own when it comes to packing up to 12 cores and 128GB or RAM into a tiny, whisper quiet package. For me I would rather use this as my every-day workstation over a PC or an iMac. It makes a brilliant VMware lab when you are short on space. It's pretty good for Photoshop too.
Yes the D700's will get their asses kicked every time in gaming benchmarks, but if you are a gamer the nMP isn't aimed at you. Neither is the iMac for that matter. If you want good gaming performance build a PC, then either run it as a Hackintosh, or put Windows (or Linux) on it.
----------
I agree. I think Apple should have introduced what is now the nMP as a product that sits between the Mini and the Pro. They should have made a revised version of the Pro, keeping a large case with plenty of expansion slots. I would have put more drive bays in there - some 2.5" for SSD's and some 3.5" for capacity. Add thunderbolt and the ability to have up to 512GB RAM (like the Z810 and Precision 7910).
While I like my nMP, I can see that I will out grow it in the next couple of years.
So while the D700's aren't going to win any awards for their gaming performance, it's pretty much in a class of it's own when it comes to packing up to 12 cores and 128GB or RAM into a tiny, whisper quiet package. For me I would rather use this as my every-day workstation over a PC or an iMac. It makes a brilliant VMware lab when you are short on space. It's pretty good for Photoshop too.
Yes the D700's will get their asses kicked every time in gaming benchmarks, but if you are a gamer the nMP isn't aimed at you. Neither is the iMac for that matter. If you want good gaming performance build a PC, then either run it as a Hackintosh, or put Windows (or Linux) on it.
----------
I agree. I think Apple should have introduced what is now the nMP as a product that sits between the Mini and the Pro. They should have made a revised version of the Pro, keeping a large case with plenty of expansion slots. I would have put more drive bays in there - some 2.5" for SSD's and some 3.5" for capacity. Add thunderbolt and the ability to have up to 512GB RAM (like the Z810 and Precision 7910).
While I like my nMP, I can see that I will out grow it in the next couple of years.
I'm very happy with my nMP and that's the only thing that matters to me.
Ditto!! Just got my 8 core and I'm very happy!
I'm very happy with my nMP and that's the only thing that matters to me.
If it does what you do well, then that's actually great to hear. Me and my classes are trying to find that sweet spot. As I said, so far, Cinema 4D has at least been pretty darn solid.
Given the amount of time and effort Apple spent redesigning the nMP, I wouldn't be surprised if they allowed GPU upgrades on the 6,1s after the 7,1s come out. Upgrades would all be in house, and (likely) expensive, keeping things beneficial for Apple, and it would quash most of the complaints about the anemic and rapidly aging GPUs in this line up.
No, a different strategy. Similarly what they did with non-removable batteries and no microSD slots. The face of technology is constantly changing. The trouble is many don't want to change with it. Its also about shifting to a different platform and different way of doing things.
Last edited:
I think it's safe to say that Apple is not going 180 on this design - they will either launch the same machine with bumped specs or discontinue the MP altogether. The absence of a 5k Cinema Display is a worrying sign, especially considering that they actually have a 5k panel designed and manufactured already (for the iMac).
Hackintosh is a more attractive option than ever considering the limitations in the nMP. Still going strong after 6 months with my latest Hackintosh build:
Asus X99-S
Intel Haswell-E 5960x (OC to 4.0GHz)
32GB Kingston DDR4 3000Mhz
Nvidia Gigabyte GTX 980 G1
Kraken X61 Water Cooler
NZXT S340
Total about ~$2299
Would be cool to compare benchmarks with the ones posted by OP, but don't have 4k display so I guess it would be useless. Should probably come close or even out-perform the nMP in some tasks, which is pretty awesome considering it's almost 1/3rd the price.
I should note however that the 5960x is pretty tricky to Hackintosh, not for the faint-hearted.
The only thing that's kept me from going Hackintosh is the finicky nature of iMessage/FaceTime on Hacks. It worked, it didn't, it worked with a clone of certain low level info, it didn't work with that clone, it knocked real Macs off of Messages if you recycled a serial number, and then it didn't. Now it works again with cloned info. It's just too weird and unpredictable. It sounds stupid, but I don't want to give up SMS from my computer. There's a great thread on tonymacx86 that covers it.
I feel badly for you guys with demanding work loads. The very best machines Apple sells fall behind simple hand-built desktops that cost 1/4 as much. You're the people Apple seems to be giving up on. One orphaned machine (MP5,1) replaced with, apparently, another one (MP6,1). Hopefully they have at least a footnote worth of stuff for you guys today. I don't have incredibly demanding requirements, though. I used a 2010 Mac Mini (with VMs and everything, thanks to a big SSD) for a long time before getting this cMP, and I'm sure it'll serve me for a long time too.
I feel badly for you guys with demanding work loads. The very best machines Apple sells fall behind simple hand-built desktops that cost 1/4 as much. You're the people Apple seems to be giving up on. One orphaned machine (MP5,1) replaced with, apparently, another one (MP6,1). Hopefully they have at least a footnote worth of stuff for you guys today. I don't have incredibly demanding requirements, though. I used a 2010 Mac Mini (with VMs and everything, thanks to a big SSD) for a long time before getting this cMP, and I'm sure it'll serve me for a long time too.
Your educational institution could have first purchased a couple of units of the new Mac Pro and tested it out before making a bigger purchase. I've learned that before buying anything new specially involving huge expenses, best to have hands on actual usage of the computer before deciding aside from looking at benchmarks. There's a discussion in the Adobe forum on users reporting having problems with the New Mac Pro with After Effects and Premiere Pro. https://forums.adobe.com/thread/1412328?start=120&tstart=0
It's a long discussion and an Adobe staff was doing his best to come up with solutions.
A client of mine offered to me for sale an 8 core 3.0ghz new Mac Pro that was used for a few months. I told the client that I need to actually test out the computer hands on before making a decision. Was able to try out the new Mac Pro. It's fast and quiet though certain tasks was also fast with the classic Mac Pro. The actual hands on usage greatly help with my decision to just continue using the classic Mac Pro.
If it runs well on an old Mac, there's no reason it would run worse on a modern Mac Pro except bad software.
Nothing in AE requires CUDA except the raytrace renderer, which is EOL anyhow. There are valid reasons to need CUDA but AE and PP aren't them.
"Requiring" CUDA and "running better" with CUDA may be 2 different things. Imagine what a drag it is to have to keep updating it to run on newer & better hardware in the free computing world while also trying to keep it running on the fixed and unchanging old tech in nMP.
We get grants to purchase software and can't purchase individual machines that wouldn't possibly be used. This is where that system bites us on in the ass. We have been using the top available Macs since the early 90s and this situation is a first. We gambled and lost.
But...but...there's two of them! Two is better than one. And the machine is silent! Guys...don't you get it. Two video cards and silent tubes boost productivity by ten fold.
???
----------
Premiere is amazing with my CUDA setup. Not so much on the 2013 Mac Pro.
I've done plenty of OpenCL versus CUDA renders out of Premiere, and haven't seen any significant performance deltas when the machine specs were controlled for. The only realistic possibility beyond Adobe being Adobe is some sort of throttling issue, which I haven't seen observed as an issue with the single unit we have access to, nor any evidence from Anandtech.
The bench marks given have insufficient information
The bench marks given have insufficient information - they do not tell you what mode was used - that can make a huge difference. All of the timings below are for 1004653 color interpolated triangles, but the speed goes from 5 frames/sec to 438.60 frames/sec depending upon the mode. On Mavericks, the GPU will not go faster than about 60 frames/sec unless the disablebeamsync program is run first. The Macintosh always blows away the PC away using compiled mode. On the other hand, the vertex array mode is much slower that the Vbo modes on the Macintosh. These numbers all shift as the number of triangles changes - so picking mode and problem size can slant the results. Desktop PCs scale up the speed numbers, but the weakness in the compiled mode seems to persist on Windows 7 machines - so if I want to show than the Macintosh is faster - I simply test using compiled mode.
runMode 0 - immediate
runMode 1 - Compilied
runMode 2 - VertexArrayMode
runMode 3 - VboMode - one buffer
runMode 4 - VboMode - separate buffers
Cylinder Mac Pro using 1 AMD FirePro D500
buildtriangles Count 1004653 vender ATI Technologies Inc.
19.51 Seconds 100 Frames 5.13 (Frames/sec) 5150481.89 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.65 Seconds 100 Frames 60.42 (Frames/sec) 60704109.81 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
2.41 Seconds 100 Frames 41.53 (Frames/sec) 41721470.23 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.66 Seconds 100 Frames 60.10 (Frames/sec) 60375780.00 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.84 Seconds 100 Frames 54.35 (Frames/sec) 54600709.07 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
After running disablebeamsync
Cylinder Mac Pro using 1 AMD FirePro D500
buildtriangles Count 1004653 vender ATI Technologies Inc.
18.79 Seconds 100 Frames 5.32 (Frames/sec) 5345889.43 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.23 Seconds 100 Frames 438.60 (Frames/sec) 440636963.69 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
2.39 Seconds 100 Frames 41.86 (Frames/sec) 42053282.77 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.26 Seconds 100 Frames 386.10 (Frames/sec) 387897174.00 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.32 Seconds 100 Frames 314.47 (Frames/sec) 315928775.53 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
Asus Replublic of Gamers - NVDIA GeForce GTX 670M (vbm-1903)
-----------------------------------------
buildtriangles Count 1004653 vender NVIDIA Corporation
9.11 Seconds 100 Frames 10.98 (Frames/sec) 11028024.28 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
9.03 Seconds 100 Frames 11.07 (Frames/sec) 11122030.34 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
2.43 Seconds 100 Frames 41.10 (Frames/sec) 41292764.64 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
2.09 Seconds 100 Frames 47.85 (Frames/sec) 48069523.51 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
3.09 Seconds 100 Frames 32.37 (Frames/sec) 32523567.64 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
2.31 Seconds 100 Frames 43.33 (Frames/sec) 43529157.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
0.81 Seconds 100 Frames 123.15 (Frames/sec) 123725733.39 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878281.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878318.21 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878281.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
The bench marks given have insufficient information - they do not tell you what mode was used - that can make a huge difference. All of the timings below are for 1004653 color interpolated triangles, but the speed goes from 5 frames/sec to 438.60 frames/sec depending upon the mode. On Mavericks, the GPU will not go faster than about 60 frames/sec unless the disablebeamsync program is run first. The Macintosh always blows away the PC away using compiled mode. On the other hand, the vertex array mode is much slower that the Vbo modes on the Macintosh. These numbers all shift as the number of triangles changes - so picking mode and problem size can slant the results. Desktop PCs scale up the speed numbers, but the weakness in the compiled mode seems to persist on Windows 7 machines - so if I want to show than the Macintosh is faster - I simply test using compiled mode.
runMode 0 - immediate
runMode 1 - Compilied
runMode 2 - VertexArrayMode
runMode 3 - VboMode - one buffer
runMode 4 - VboMode - separate buffers
Cylinder Mac Pro using 1 AMD FirePro D500
buildtriangles Count 1004653 vender ATI Technologies Inc.
19.51 Seconds 100 Frames 5.13 (Frames/sec) 5150481.89 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.65 Seconds 100 Frames 60.42 (Frames/sec) 60704109.81 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
2.41 Seconds 100 Frames 41.53 (Frames/sec) 41721470.23 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.66 Seconds 100 Frames 60.10 (Frames/sec) 60375780.00 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
1.84 Seconds 100 Frames 54.35 (Frames/sec) 54600709.07 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
After running disablebeamsync
Cylinder Mac Pro using 1 AMD FirePro D500
buildtriangles Count 1004653 vender ATI Technologies Inc.
18.79 Seconds 100 Frames 5.32 (Frames/sec) 5345889.43 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.23 Seconds 100 Frames 438.60 (Frames/sec) 440636963.69 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
2.39 Seconds 100 Frames 41.86 (Frames/sec) 42053282.77 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.26 Seconds 100 Frames 386.10 (Frames/sec) 387897174.00 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender ATI Technologies Inc.
0.32 Seconds 100 Frames 314.47 (Frames/sec) 315928775.53 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
Asus Replublic of Gamers - NVDIA GeForce GTX 670M (vbm-1903)
-----------------------------------------
buildtriangles Count 1004653 vender NVIDIA Corporation
9.11 Seconds 100 Frames 10.98 (Frames/sec) 11028024.28 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
9.03 Seconds 100 Frames 11.07 (Frames/sec) 11122030.34 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 0 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
2.43 Seconds 100 Frames 41.10 (Frames/sec) 41292764.64 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
2.09 Seconds 100 Frames 47.85 (Frames/sec) 48069523.51 (Cells/sec) xsize 800 ysize 400 PolyList 1 runMode 1 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
3.09 Seconds 100 Frames 32.37 (Frames/sec) 32523567.64 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
2.31 Seconds 100 Frames 43.33 (Frames/sec) 43529157.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 2 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
0.81 Seconds 100 Frames 123.15 (Frames/sec) 123725733.39 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878281.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 3 stereoType 1
buildtriangles Count 1004653 vender NVIDIA Corporation
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878318.21 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
0.81 Seconds 100 Frames 123.30 (Frames/sec) 123878281.79 (Cells/sec) xsize 800 ysize 400 PolyList 0 runMode 4 stereoType 1
I'm sorry, I don't understand what you are referencing.
Is this one of the tests that was used by Barefeats?
Are you suspicious/implying that he "changed modes" to slant the results?
Otherwise, we have to assume he wasn't "switching modes" due to boredom or incompetence. Any reputable bench site like his would have very short life if easily proven to be running inaccurate tests.
I am able to replicate many of his tests and get exactly the same numbers.(As could many other people, hence why cheaters get caught)
Your post has many technical details but for the life of me I have no idea what you are trying to get at.