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.
Will there be a new Mac Pro in 2015?
- Thread starter Anto38x
- Start date
- Sort by reaction score
- Status
koyoot
macrumors 603
Upgrade along the D300 -> D500 > D700 path? Relatively low likelihood. I don't think folks have or will ask for it in a coherent way that presents a creditable market opportunity and Apple isn't particularly looking to do it. (e.g., folks yelling for "go back to box with slots" is not going to motivate Apple to add a install service. )
Upgrade to HBM solution ? Relatively substantially less likely than than the previous option. Mainly because the fitting and connectivity is custom. From iFixit teardown
(https://www.ifixit.com/Teardown/Mac+Pro+Late+2013+Teardown/20778 )
The cards are thermal glued to the those plates (which are in turn bolted to the Thermal core). The patterns of where the "hot spots" are is driven by the GPU board layout. HBM cards are not going to have the same layout. Nor are they likely the same thickness as previous implementations. So minimally new thermal plates needed (worse case need a different thermal core + mounting config ). Similar with the data pin sockets. Possibly can use the same connectors but the pin mappings are not necessarily the same. The power transfer mechanism looks to be reusable though.
If buying it just to hold for 3-9 months before selling it to buy another "even newer" one then the price you are paying isn't as critical as the resale price you'd get when sold. If pay $3,000 now and get $2,500 five months later then you have paid $100/month to defacto "lease" a Mac Pro. Used Mac Pro prices are unlikely to implode in a relatively short 6-10 month span of time.
If buying for giggles then probably shouldn't buy anything; just wait the one you actually appear to want. . If short term Mac Pro can produce incrementally higher revenue far exceeding $100/month then it is not a difficult decision to make.
If the GPUs are upgradable, I suspect it will be an instore Upgrade only. Not a terrible thing as they would likely be then covered by a warrenty
AdamSeen
macrumors 6502
The shape of things to come in the next gen of graphics cards: http://wccftech.com/nvidia-pascal-gpu-17-billion-transistors-32-gb-hbm2-vram-arrives-in-2016/
Well which one is it? You are excited about OVERCLOCKING a power hungry chip or UNDERCLOCKING & UNDERVOLTING?
They are polar opposites, you know that, yes? This is referred to as being "all over the map".
Have a look at those Fury Reviews. Look at the charts about power consumption. Find one that includes 7970/R9 280X. Those are same thing as D700. Apple had to drop the clocks on 7970 to make it run as D700. Since Fury uses quite a bit more power then 7970, guess what they would have to do to fit 2 of them in 7,1? Yep, under clock and undervolt it like crazy. The GPU rendering power will wilt like licorice in the sun.
I have no doubt that Apple is working on these. I also have no doubt that they will end up neutered shadows of the real thing and very sad compared to what Dell & HP will have in their workstations. And certainly nothing more then a weak whisper compared to what could have been in a machine with real PSU and cooling options.
koyoot
macrumors 603
Undercooking the core, and overclocking the HBM memory. Is that so hard to read it from my post, where I specifically pointed it, in the post you quoted? Power consumption goes down very much from down clocking the core on AMD GPUs, however, oc on Memory does not affect the power consumption in such way, as core does.
The numbers are promising, because the bigger gains are from Overclocking the Memory, than from overclocking the core. The only problem is this - from Memory bandwidth Pro apps are not benefiting as much as games. Thats what makes me slightly apprehensive for it, because Im not sure in this case Apple would consider running HBM on 640 GB/s even if it would not change the thermal envelope at all.
But that is only theory, so far.
The numbers are promising, because the bigger gains are from Overclocking the Memory, than from overclocking the core. The only problem is this - from Memory bandwidth Pro apps are not benefiting as much as games. Thats what makes me slightly apprehensive for it, because Im not sure in this case Apple would consider running HBM on 640 GB/s even if it would not change the thermal envelope at all.
But that is only theory, so far.
Exceptionally good advice, thank you.
The work I do can be done with a 2012 machine, so I'll likely soldier on with it hoping that refreshes come sooner rather than later and then buy at the beginning of the curve.
I suppose it's partly a sense of the "grass is always greener" and you always feel things would be much more comfortable with a more powerful machine, whether you strictly need it or not!
I just wish Apple would help the dilemma a bit by shuffling their refurbs between countries. Many nice Mac Pro configs appear in the US store at acceptable prices, but virtually none in my country.
MVC, long time you didn't come up with one of those typical replies of yours.
Maybe it's because English is not our native language (koyoot and meas well) but I perfectly understood what he wrote and it seems to me clear as water that he was mentioning underclocking the core (which ends up not being a big performance killer) and overclocking the mem, which provides some, but no much, benefits in some apps. But that might be just me as well... or not. koyoot, good thing you seem a calm person, no shooting back like crazy ;-)
Still, I believe there's plenty of bandwidth already with HBM.
Artic Islands and Pascal are still far off, but are very promising.
Maybe it's because English is not our native language (koyoot and meas well) but I perfectly understood what he wrote and it seems to me clear as water that he was mentioning underclocking the core (which ends up not being a big performance killer) and overclocking the mem, which provides some, but no much, benefits in some apps. But that might be just me as well... or not. koyoot, good thing you seem a calm person, no shooting back like crazy ;-)
Still, I believe there's plenty of bandwidth already with HBM.
Artic Islands and Pascal are still far off, but are very promising.
koyoot
macrumors 603
Also keep in mind that if Apple will stay with 125W TDP power envelope for their GPUs, and Full Fiji will be able to be squeezed to it at that level to 850 MHz - that will mean from 250W TDP you will get 14 TFlops of performance on two GPUs.
Thats, according to Wikipedia stats, over twice the performance of single Titan X in similar power envelope.
https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#GeForce_900_Series
And to add to that, as we have seen on few pages before, AMD finally cracked the OpenCL drivers for OS X. So there can be a lot to look for. All of this still can simply be diminished by Apple decisions, tho.
However, it looks very promising.
Thats, according to Wikipedia stats, over twice the performance of single Titan X in similar power envelope.
https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#GeForce_900_Series
And to add to that, as we have seen on few pages before, AMD finally cracked the OpenCL drivers for OS X. So there can be a lot to look for. All of this still can simply be diminished by Apple decisions, tho.
However, it looks very promising.
Pascal taped out in June ('source' that link is leveraging off of)
http://www.fudzilla.com/news/graphics/37937-nvidia-pascal-gp100-gpu-reportedly-tapes-out
AMD taped out in June-July ( same 'source'. )
http://www.fudzilla.com/news/processors/38291-amd-tapes-out-finfet
They don't seem far apart on the development cycle, but both certainly won't make/enable any 2015 Mac Pro product. Nothing likely in Q1 and vast majority of Q2 2016 either. The very high density die option for HBM2 probably won't make mainstream GPU space due to expense. 16GB would still be quite generous in terms of capacity at more acceptable price points. In terms of making the next Mac Pro hardware update cycle window, these look like they will probably be a miss ( 2nd half 2016 is likely "too late").
Doubtful that AMD will be shooting for 17B transistors range though (at least as a just a GPU). Most likely a smaller budget that is easier to get higher wafer yields on a brand new process. Likewise it would be bit surprising if Nvidia first to market version was not the "max budget" option.
What GPU implementers do with those billions of transistors and double digit GBs is probably will be the key differentiating factor more so than just "big budget".
Refurbs likely come from at least two major sources ( 'abandoned' repairs and expired leases ). If there is no large Apple repair/lease business in your country then the supply of refurbs is likely highly limited. Who does most of the local repairing and/or leasing may not be Apple depending upon how skewed the local distribution and service market is.
Swapping between countries probably gets in the way of whatever double-Irish with a triple Bermuda twist with a Isle of Man flip their tax shelter. Once send the machines down some tax shelter rabbit hole they can't be moved anymore.
There is rather a glut of Mac Pro currently in the US online store. ( 10-13 only 2 with a discount) [ relative to 27" iMac with 20 over a broader product range]
askunk
macrumors 6502a
Well... as much as I have the same feeling on the issue, it baffles me to think that Apple would have spent millions on the new assembly lines, just to show off.
Besides, the nMP comes from a Unibody enclosure. It wouldn't take much to produce a second, taller model with Dual CPU.
So the issue, I guess, it is not much whether Apple can or can't implement it in the actual enclosure, but - as more eloquently said by others in the discussion - a strategic choice.
Making a taller or even wider case is not the issue for sure. IF you know how it's made you understand why.
There is of course the development effort, since you need to redo all the layout and testing, thermal, mechanical, electrical, whatever.
Still, it's doable, but like I said before, this design is perfect and they won't "ruin" it by changing an inch, that's my understanding of it.
Although Apple seems to have a higher budget for spending on R&D now I don't think they'll invest a lot more on nMP anymore, the current setup works for them since the real pros don't use these anymore.
I believe they will update but always within the same concept.
There is of course the development effort, since you need to redo all the layout and testing, thermal, mechanical, electrical, whatever.
Still, it's doable, but like I said before, this design is perfect and they won't "ruin" it by changing an inch, that's my understanding of it.
Although Apple seems to have a higher budget for spending on R&D now I don't think they'll invest a lot more on nMP anymore, the current setup works for them since the real pros don't use these anymore.
I believe they will update but always within the same concept.
fastlanephil
macrumors 65816
Why hasn't a PC hardware company at least attempted to copy the nMP form factor yet? Isn't the Tube the future and the Box a dinosaur. Wasn't that sort of in the cool-aid that Apple was serving up a couple of years ago.
Well?
Well?
Apple is the only company that was willing to do a completely custom form factor for a very low volume PC. Other vendors will only do custom form factors for high-volume PCs, like the iMac lookalikes that Dell and HP are shipping. They'll sell enough of those to warrant the custom motherboards and cases. Apple made the economics of the nMP work by pre-buying a lot of part inventory (particularly the GPUs), because unlike other vendors, Apple doesn't feel a need to match current specs to be competitive, so they can get 2+ years of return out of their investment. Any other vendor that tried to follow suit would lose a lot of money.
AdamSeen
macrumors 6502
flat five
macrumors 603
i thought that was already possible.. (since 2013/mavericks?)
?
apple__ IP over Thunderbolt
----
edit.
also, don't forget about the 802.11ac wifi.. that's faster than fw800 was.
Last edited:
SoyCapitanSoyCapitan
Suspended
A history of well designed failures.
http://regmedia.co.uk/2013/10/09/apple_g4_cube.jpg
http://www.blakespot.com/sgi/images/sgi_front.jpg
http://ismh.s3.amazonaws.com/next_cube_front.jpg
Those three failed even though they were more modular than the nMP.
flat five
macrumors 603
can't really compare the cube to the nmp.. the cube was a sideshow.. nmp is the main event.
if nmp fails, mac pro is done.
---
[edit]
of course you can compare it.. they're very comparable 😉
just in a historical sense, the outcomes will be not so comparable if both are deemed failures.
Last edited:
flat five
macrumors 603
http://www.theregister.co.uk/2015/07/24/review_acer_revo_one_rl85_/
seems at least inspired by nmp and airport extreme.
?
that aside, it's not an easy copy to make when coming from a box mentality.. box mentality is start with a box then cram it with components.
if you tried to do that with a similar sized cylinder as nmp, it wouldn't work.. nothing would fit and things would overheat etc. it's a (much) more involved design which was worked on for a few years.. i don't doubt other manufacturers are experimenting with smaller form factors at the moment.. they're just not ready yet.
koyoot
macrumors 603
Asynchronous Shaders. Imagine that working in Mac Pro with dual GPUs with extreme parallel/asynchronous capabilities.
I have not been more excited for any type of technology execution, than for Async 🙂.
- Status