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.
2 + 2 = 3 How does D700 & 12 Core fit in 450 Watts?
- Thread starter MacVidCards
- Start date
- Sort by reaction score
The German review has measured 438w peak power. Must be underclocked versions of the GPU's surely.
I think so but I'm no stats guru. With the GPU's running full whack and the e5 maxing out at 130w anyway it's pretty astounding stats for the GPUs and the rest to total 308 watts for a physics guru!
Right, if you look back in the thread, Apple has only listed the specs for the 4 core and 6 core versions. It's probable that the 12 core has the same power supply, but they have not listed the 12 core specs, so I was just wondering if it was verified vs. what Apple has said for the 4 and 6 core versions.
There will probably be a lot of questions about the GPUs. They do seem to be under clocked, but Apple has posted stats on the gflops of the D700, so they can't under clock it that much.
The CPU AND GPU stress test will be the interesting one. My other thinking is for an 8 or 6 core this may matter less as the CPU power consumption will be much less.
Intel won't underclock or customise a CPU just for Apple, all E5's are max 130 watts and most likely socketed. The logic board and components such as usb power ports etc are pretty much fixed. It's got to be mostly in the GPU's.
It's true they won't customize a CPU for Apple, but Intel CPUs (and those AMD GPUs) can dynamically clock. Apple isn't stopped from clocking down the CPU when the GPUs spin up, and clocking down the GPUs when the CPUs spin up.
That's why the CPU and GPU stress test would be interesting. It may be that the CPU can run at full speed, and the GPU can run at full speed, but not at the same time.
Macbook Pros do the exact same thing for power consumption. The battery nor the power adaptor alone can actually power the system at full bore, so it will clock down different components dynamically. If you are plugged in, it will draw from the battery and the power adaptor and go back to full speed.
I agree maxing out the thing with both will be ideal but I reckon that's pretty damn close to peak power when you add up the current for the empty data ports like usb. Also comparisons with different mains voltages running the same benchmarks too as the Germans will be 230v 50hz. Universal switch mode power supplies running 110/240 aren't exactly the same at different voltage inputs no matter what the manufacturers say..
It's most certainly binning and under clocking (as you point out). I wrote more about this in post 20.
I'm sure current benchmarks show that the D700 is the W9000/2 so
Both the D700 working together have about the same performance as the W9000. Maybe splitting GPU in half was decision to improve power efficiency and heat distribution?I'm sure current benchmarks show that the D700 is the W9000/2 so
Actually the current benchmarks show that D700 is W9000 / 1.25. So they basically get 80% of the performance of W9000 with half the TDP, which is quite good.
what I don't get about Apple and the MP, is that surely with an increase of say 10mm on the overall diameter and another 20mm on the height, they could have increased the thermal dissipation, increased the PSU rating and not had to underclock anything. The extra size difference wouldn't make a slight bit of difference with the buyers.
I'm no technical person, but when Apple designed the old MP, it was over-engineered, but the new MP seems to go the other way like the MacMini, engineered to the very limits of thermal capacity. (my MM will quickly reach 97C on some jobs)
I'm no technical person, but when Apple designed the old MP, it was over-engineered, but the new MP seems to go the other way like the MacMini, engineered to the very limits of thermal capacity. (my MM will quickly reach 97C on some jobs)
Further to my reply, it almost looks like Apple designed the MP to read nice on the web page...I mean 9.9" tall, 6.6" diam, 11lbs weight. Don't tell me that is some huge coincidence. So having come up with the marketing spiel, the engineers had to go away and make the damn thing fit, TDP or no TDP.
I'm all for new thinking but not at the expense of performance or other practical factors. Sure its a workstation (it has workstation CPU) but as others have mentioned, it's really a big souped up macMini.
I'm all for new thinking but not at the expense of performance or other practical factors. Sure its a workstation (it has workstation CPU) but as others have mentioned, it's really a big souped up macMini.
http://www.anandtech.com/show/7603/mac-pro-review-late-2013/10
W9000 is basically a R9 280X aka HD7970, and a crossfire D700 is getting 85% of crossfire 280X. Single D700 is getting 80% of a single 280X.
I'd say the whole thing is a good compromise. Spend half the juice, get almost the same performance.
Ah well you will have to wait for more benchmarks then. I haven't seen any Windows tests comparing the 3D drawing performance of D700 to W9000.
Actually it isn't. There was 3 revisions of Tahiti XT chip.
XT, reference GPU clocks 925 MHz, no boost
XT2, reference clocks 1000/1050 (boost), bit lower VDCC per clock
XTL, reference clocks 850/1000 (boost),
Each one has some changes under the hood that affect performance per W.
W9000 is based on XT, 7970 on XT, 7970 GHz Edition on XT2, R9 280x on XT2 (first batch) and XTL (current batch), D500 and D700 both are based on XT2.
Thanks for clarification. Does that mean R280X is actually faster than W9000? (Haven't seen any benchmarks comparing those two)? If that's so, then D700 is even closer to W9000's performance.
Obviously whatever cards these really are wouldn't look good on a spec sheet so they came up with a proprietary name.
The GPU's we were led to believe they are don't fit with power consumption constraints in the nMP. That knowledge combined with the fact that Apple is being really hush hush as to the real identity of these cards suggest it's something we're not going to like. If the identity of these GPUs was something worth bragging about Aople wouldn't have needed to create a new name for them.
Are we really surprised after all? This is Apple we're talking about here. Kings of proprietary "revolutionary products", marketing babble and unicorn dust. Once the dust settles you see the unicorn for what it really is, a donkey with a horn super glued to it's forehead.
The GPU's we were led to believe they are don't fit with power consumption constraints in the nMP. That knowledge combined with the fact that Apple is being really hush hush as to the real identity of these cards suggest it's something we're not going to like. If the identity of these GPUs was something worth bragging about Aople wouldn't have needed to create a new name for them.
Are we really surprised after all? This is Apple we're talking about here. Kings of proprietary "revolutionary products", marketing babble and unicorn dust. Once the dust settles you see the unicorn for what it really is, a donkey with a horn super glued to it's forehead.
Last edited:
I'd risk the statement, that as long as these cards will work for purposes they were chosen for, people won't care about what these card really are.
It's more important for "freaks", who are into technical details (as I am), than for user for whom it has "just work".
Agree. The system does 7 Tflops, which was peak supercomputer performance in 2002, not that long ago. That's enough spec for me. I'm willing to trade some bleeding edge performance for a silent computer, because in a few years I'll get much more than that.
----------
It doesn't work that way. The Apple culture is "small as possible", and the engineers get nice stock options for innovating that. Certainly marketing didn't pick some arbitrary size. And calling it a souped up mini is like calling a race car a souped up bicycle. Duh.