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
Capitan, those were the EVOs released some time back.
With mass production it will be extended to other series, hopefully the Apple drives as well.
The SFF limits the capacity for the custom Apple drives.
I'm betting the 2TB will also be custom made.
dec, I know about the 1200 v4, but do you also want to include the Xeon-D? It's a different league.
The 1200 is in a different socket because they're based on DT counterparts, I'd say almost the same.
1500 could be as well, and since it's a mobile part a different socket seems likely.
What I was saying is that there are a little too many parts right now, it's confusing for most people.
Do you also like to just fight off everything? I believe either v4 or hopefully v5 will come early 2016, which would be around the 18 month usual refresh cycle for Xeons. If v4 was cancelled then v5 might take it's place sooner than was expected, which was 2017 according to the roadmap.
So, now the 1200 lifespan can be shorter than usual, but not the other series?!
5K is still far from mainstream, let alone 8K. No hardware to deal with it. Although LG slipped once with a comment on Apple going to use their 8K panels.
We don't even have DP1.3 on the desktop yet, mobile GPUs won't be able to pump out that many pixels for years.
With mass production it will be extended to other series, hopefully the Apple drives as well.
The SFF limits the capacity for the custom Apple drives.
I'm betting the 2TB will also be custom made.
dec, I know about the 1200 v4, but do you also want to include the Xeon-D? It's a different league.
The 1200 is in a different socket because they're based on DT counterparts, I'd say almost the same.
1500 could be as well, and since it's a mobile part a different socket seems likely.
What I was saying is that there are a little too many parts right now, it's confusing for most people.
Do you also like to just fight off everything? I believe either v4 or hopefully v5 will come early 2016, which would be around the 18 month usual refresh cycle for Xeons. If v4 was cancelled then v5 might take it's place sooner than was expected, which was 2017 according to the roadmap.
So, now the 1200 lifespan can be shorter than usual, but not the other series?!
5K is still far from mainstream, let alone 8K. No hardware to deal with it. Although LG slipped once with a comment on Apple going to use their 8K panels.
We don't even have DP1.3 on the desktop yet, mobile GPUs won't be able to pump out that many pixels for years.
About that 8K screens. Semiaccurate.com has posted an analysis about them, and says, some OEM is releasing them. Whats more, there may be even TWO OEMs with that type of screens in offer.
LG have already stated that Apple will be releasing it, in their blog post.
http://semiaccurate.com/2015/08/04/who-will-have-the-first-8k-consumer-monitor/
http://appleinsider.com/articles/15...ac-8k-later-this-year-display-partner-lg-says
LG have already stated that Apple will be releasing it, in their blog post.
http://semiaccurate.com/2015/08/04/who-will-have-the-first-8k-consumer-monitor/
http://appleinsider.com/articles/15...ac-8k-later-this-year-display-partner-lg-says
Last edited:
Yes I have. Its pointless OC cause it is on LN2. Other question is that it is still stable to finish a benchmark.
I have dug a little about thermals of Fury, and it turns out that reducing the TDP cap in application does not lower the voltages on the core itself. That means, standard voltages, with power cap result in unstable core clocks and increased and inefficient power draw compared to what is possible. Thats why it jumps between 1000 and 950 MHz in 175W TDP. If we would want to lets say get 7.5 TFLOPs from Full Fiji chip in 175W of TDP we would have to clock it at 925 MHz and possibly maintain the voltages. If we would want to get the same amount of TFLOPs in 125W we would need to get the voltages down.
About what happens in mobile space. Turns out that Nvidia obviously reacts to AMDs proposition in the space. That is why they are releasing 990M. The question is whether it will be based on GTX 980 or 980 Ti. I am genuinely excited to see how much power draw can be pushed down in Maxwell cards. The question is now. What will AMD offer here, that Nvidia needs to counter it. If we will look at some evidence, there are signs of 4 GPUs for iMacs. M380, M390, M395 and M395X. Dell already stated that M295X is rebranded to M390. http://www.dell.com/support/home/us/en/04/Drivers/DriversDetails?driverId=VMRGM What are the other two, higher numbered? Of course, it can be OEM only...
I have dug a little about thermals of Fury, and it turns out that reducing the TDP cap in application does not lower the voltages on the core itself. That means, standard voltages, with power cap result in unstable core clocks and increased and inefficient power draw compared to what is possible. Thats why it jumps between 1000 and 950 MHz in 175W TDP. If we would want to lets say get 7.5 TFLOPs from Full Fiji chip in 175W of TDP we would have to clock it at 925 MHz and possibly maintain the voltages. If we would want to get the same amount of TFLOPs in 125W we would need to get the voltages down.
About what happens in mobile space. Turns out that Nvidia obviously reacts to AMDs proposition in the space. That is why they are releasing 990M. The question is whether it will be based on GTX 980 or 980 Ti. I am genuinely excited to see how much power draw can be pushed down in Maxwell cards. The question is now. What will AMD offer here, that Nvidia needs to counter it. If we will look at some evidence, there are signs of 4 GPUs for iMacs. M380, M390, M395 and M395X. Dell already stated that M295X is rebranded to M390. http://www.dell.com/support/home/us/en/04/Drivers/DriversDetails?driverId=VMRGM What are the other two, higher numbered? Of course, it can be OEM only...
Don't put your hopes for Maxwell GPUs. Apple will not use them. If they would want to use Maxwell GPUs in OS X we would already see them in any of the Macs. Especially Macbook Pro. They could have used Nvidia GPU in it, but instead they went with M370X from AMD.
Last edited:
Nothing was suitable at the time. People wishing for 950M but that wasn't shipping to builders in volume and didn't have great Open CL. The new Nvidia drivers improve OpenCL now and are starting to officially support mobile chips so it's clear something is going on behind the scene.
Nothing is going behind the scene. Those are Nvidia WEB drivers. Nothing else. There is completely no sign of Maxwell GPUs in any system Apple provides. Not in El Capitan, not in Yosemite. Nvidia has to provide them because they know how big position they have in Mac Community. Nvidia made those drivers to counter what AMD has done with drivers in El Capitan. That is the only reason. Applications that use OpenCL in El Cap on AMD hardware will have more performance, and more capabilities then they have now.
This is more so "I wish the Mac Pro entry price would go back to $2,200-2,400" than a deep need for a new product design. If you could get an entry level Mac Pro at $2,299 I suspect you wouldn't skip that for for this "Mac Mini Pro" above.
Core i7 x9xx for the parts that overlap the Xeon E5 1600 in baseline implementation don't buy anything on price improvements at all. Xeon E5 1630 v3 $373 ( 1620 v2 $297) is about the same as Core i7 5820K $389
Mainstream Core i7 doesn't either. Early 6th generation option is $350 (http://ark.intel.com/products/family/88392/6th-Generation-Intel-Core-i7-Processors#@Desktop) and 5th generation
isn't much better $348-366 ( http://ark.intel.com/products/family/84979/5th-Generation-Intel-Core-i7-Processors#@Desktop )
Dropping the dual discrete, relatively high VRAM GPU should help, but not necessarily a $800-1000 swing.
What Apple appears unmotivated on is stopping the death spiral pricing progression. Mac Pro going from near $2,000 when introduced ( $2,199 in 2006 ) to $3,000 ( $2,999 in 2013 ). if Apple grows the gap wider still on the next iteration(s) $3,199 Mac Pro and iMac sags to $2,199 then there is a huge gap for a "Mac Pro Lite" project. Color the shell a different color and reuse a fairly high number of parts wouldn't necessarily mean "whole new product".
The Mac Pro and probably whole Mac line up would be healthier if Apple actually moved the upper end iMac down at or below the $2,199 mark and pulled the Mac Pro back down into the gap.
Apple has a fixed number of Mac Products they are willing to pursue at one time. As long as the Mac industrial design shares the same resources as the rest of the company's products that fixed limitation is probably going to remain in place. To get another Mac product either have to send a current one into comatose state or kill it. ( minor exceptions with overlapping retina laptops as wait to kill off the "low res" version, but it is basically a zero-sum game ).
If the Mac Pro gets axed then could possibly open a window for a "Mac Pro Lite" option that basically took iMac board designs and repackaged that with enough differentiation to put in different value/price category. ( e.g., iMac baseline but the GPU run at full instead of "mobile version" speed. ). With a fixed number of products they are very unmotivated to create products that engage in a relatively high amount of fratricide.
Different league but same general ballgame of path to broader mix of more energy efficient server offerings.
New expansion in Intel product categories in that general direction are not surprising. ARM was suppose to come 'kill' Intel with better (more energy friendly) server chips. Well, now there is alot more competition for any that might choose to arrive.
The desktop and the upper performance mobile chips are basically the same baseline design. There is differences in clocking and GPU subsystems attached and in socket ( which doesn't matter to Apple since mini-iMac-Macbook it is all solder down anyway ).
Dropping E5 v4 moving v5 from Q1 2017 up to Q4 2015. Possibly. Dropping v4 bringing v5 up to Q2 or Q1 2016, highly unlikely. Nuke E5 v4 relatively late in the design-release process (which is multi year in length) and result largely going to be same as what happened on desktop. Basically get stuck with the same generation for 2 years. What can remove is the "extra" quarter or so where trying to milk the max profits out of the design while ramping up on new product. E5 v5 is a new socket and chipset. There is going to be a long system builder vendor QA acceptance cycle on it.
The desktop baseline design, of which E3 1200 is a member, has been on a 12 month tick/tock cycle for last 5+ generations. Which the switch to 14nm hit a stall that whole design pipeline flow piled up into a logjam. To uncork the logjam Intel is shortening the Gen 5 Broadwell desktop/laptop related products.
The E5 1600/2600 never was on a 12 month tick/tock cycle. So the 14nm stall did not and is not causing a logjam. So there good reason to flush it the pipeline that isn't clogged down the toilet. If the v4 design team screw up the design and there are major unfixable flaws due to 14nm process mistmatch then there motivation to flush it..... but there is not much really concrete motivating that.
The server/enterprise is on a more conservative schedule with longer QA testing allocations precisely because you do not want to flush product design efforts down the drain.
Part of the disconnect is layering the Core i7 Extreme parts on top of the E5 1600 baseline. It creates marketing tensions more than technical ones. The "Extreme" customers get their underwear on a twist because while "Extreme" they underlying micro-archetecture and chipset aren't the leading edge. Well guess what? In the modern era it is mobile and very low V first. If primarily concerned about being first , bragging rights , and crotch grabbing smack talking then go mobile.
If Intel is disconnecting the E5 1600 class from the E5 2600 one then perhaps they might skip E5 1600 v4. If E5 1600 is going to shift to a minor mutation of the desktop/laptop chipset and substantively lower QA testing, then they maybe shifted to earlier release. So overall for a microarchitecture release, the 2 core models would come first, followed 4 cores , then 4-8 cores , and then finally the double digit models. That would close some of the gap between the 4 and 4-8 core classes.
Gen 6 (Skylake) 4 cores are coming early. That is probably not the "new normal". That is more so how Intel is choosing to uncork the logjam with minimal damage to the balance sheet. [ pull the ones out first that were tooooo close and unwind the rest closer to their normal spacing. ]
More than a year? Yes. Several years though isn't necessary. Once mobile targeted HBM implementations rollout it isn't going to take too long. Early 2016 smartphone class GPUs are doing 4K 60fps
http://anandtech.com/show/9522/qualcomm-adreno-530-510-snapdragon-820
A bigger power and transistor budget that laptop mobile provides and it shouldn't be that hard. DP 1.3 isn't out more so because the standard is new rather than hard to handle those resolutions. The designs started shortly before it went final will emerge from pipeline over next 2 (and a bit less) years.
Apple would be trying to fill the $2000-2400 price point in the future. That is going to be even less mainstream in year from not than it is now ( average selling price of a classic PC desktop is sub $1000 and still falling. )
Apple tries to use the same Airport card in all Macs of that "yearly" design generation. If Intel makes the laptop/iMac updates coming in the Fall then likely a Mac Pro following in the 6-8 month window there after would get the same card.
Looks to be decent Intel offering but Apple has bought alot of Broadcomm stuff. Unless Broadcomm has failed to iterate to next generation well, it is probably going to be hard to displace them (e.g. recent iPod Touch update has broadcomm based Wi-FI. )
On the consumer consumption side? Yes. But like audio "oversampling" for editing it won't be surprising to find the editing/correction/etc graphics folks to use the "extra" and then convert later to another format for wide spread consumption. Pixel peeping a Nikon 810 image of 7,360 x 4,912 could hog an 8K screen . As long as the sensor pixel growth 'war' goes on the the screens will be trailing behind a couple years back.
Besides, the closer sit to the screen the easier it is to see the pixels. Folks sitting 12 inches away from screen "need" 8K to still see Retina display when nose pressed close to the screen.
Maybe Intel will shine some light at IDF soon enough, a lot is still unknown also regarding the DT Skylake parts.
I hope the 1600 will not come down to DT similar parts, although it would certainly ease the qualification and PCH development. Still, I'd leave it where it is.
SKL-EP really depends on when (or if) BDW-EP was cancelled, there is still no confirmation, maybe next week. If it was soon enough maybe SKL development was pushed early enough so that it can be ready sooner.
I want to see SKL in nMP not for bragging rights or anything, I even believe the CPU will not even bring spectacular performance improvements (maybe AVX), but more so because of the PCH, in particular the PCIe connectivity. If, guessing by the little info on Z170, we get the same modular configurable lanes, it would be a cherry on top od the nMP cake. Apple could choose what lanes are allocated to what periphery, doing away with SATA as it not already used. Of the 26 lanes 6 are for USB3, the remaining 20 are configurable in 5 groups of 4 lanes. Imagine 2 SSDs (that's 8 or 4 or none if one or both on the CPU), 3 TB3 Alpine Ridge controllers (again, would some go into the CPU?) take up another 12 lanes. That leaves 8 more for maybe another Alpine Ridge controller and the usual GbE+WiFi+BT and whatever.
I could even see them not using the USB3 ports (or maybe just a couple for legacy) and have USB 3.1 (Type C) off Alpine Ridge altogether.
I mentioned the 8260 cause I like Intel controllers. Switching to Intel would be hard at this point, since they have used Broadcom for a long time, and good controllers too.
But Intel GbE (i219-V) would save them IO, but the second port still would have to be accounted for. Unless they kill GbE ports on nMP too, and stick with WiFi only, like in all other Macs.
You think Pascal or Evergreen will do 8K? Impressive numbers but still that's a lot of pixels to drive, even with HBM2. Even DP1.3 will only do 8K with subsampling. Full RGB 8K support maybe in DP1.4? I'd say 2-3 years at best.
I hope the 1600 will not come down to DT similar parts, although it would certainly ease the qualification and PCH development. Still, I'd leave it where it is.
SKL-EP really depends on when (or if) BDW-EP was cancelled, there is still no confirmation, maybe next week. If it was soon enough maybe SKL development was pushed early enough so that it can be ready sooner.
I want to see SKL in nMP not for bragging rights or anything, I even believe the CPU will not even bring spectacular performance improvements (maybe AVX), but more so because of the PCH, in particular the PCIe connectivity. If, guessing by the little info on Z170, we get the same modular configurable lanes, it would be a cherry on top od the nMP cake. Apple could choose what lanes are allocated to what periphery, doing away with SATA as it not already used. Of the 26 lanes 6 are for USB3, the remaining 20 are configurable in 5 groups of 4 lanes. Imagine 2 SSDs (that's 8 or 4 or none if one or both on the CPU), 3 TB3 Alpine Ridge controllers (again, would some go into the CPU?) take up another 12 lanes. That leaves 8 more for maybe another Alpine Ridge controller and the usual GbE+WiFi+BT and whatever.
I could even see them not using the USB3 ports (or maybe just a couple for legacy) and have USB 3.1 (Type C) off Alpine Ridge altogether.
I mentioned the 8260 cause I like Intel controllers. Switching to Intel would be hard at this point, since they have used Broadcom for a long time, and good controllers too.
But Intel GbE (i219-V) would save them IO, but the second port still would have to be accounted for. Unless they kill GbE ports on nMP too, and stick with WiFi only, like in all other Macs.
You think Pascal or Evergreen will do 8K? Impressive numbers but still that's a lot of pixels to drive, even with HBM2. Even DP1.3 will only do 8K with subsampling. Full RGB 8K support maybe in DP1.4? I'd say 2-3 years at best.
By the way, even if SKL comes, how would you populate the PCIe 3 lanes?
Who would get the "faster" CPU extra 8 lanes? SSD or TB3? Maybe one each?
Would Apple provide for the extra SSD so desired by some? I'd take a single larger one myself.
The remaining lanes could be for the display outputs.
What bottlenecks could be expected in the TB3 controllers hanging off the PCH? Even DMI3 is "only" PCIe 3 4x.
Any thoughts?
Who would get the "faster" CPU extra 8 lanes? SSD or TB3? Maybe one each?
Would Apple provide for the extra SSD so desired by some? I'd take a single larger one myself.
The remaining lanes could be for the display outputs.
What bottlenecks could be expected in the TB3 controllers hanging off the PCH? Even DMI3 is "only" PCIe 3 4x.
Any thoughts?
another point to consider: if nMP goes Haswell, the power delivery system changes (going by the DT counterparts) because of FIVR, and that is an overhaul in the motherboard design - simplified. Skylake goes back to separate power lines to the CPU. More complex but in line with current Ivy scheme, no major changes in board design.
You are right, the entry level Mac Pro is the most expensive it has been. It could be the new Apple product tax, where they charge extra for the first generation (see 15" macbook pro retina, iMac retina, macbook air, etc). At some point the introductory prices decrease. The flip side is that sometimes new designs increase prices permanently, like the retina macbook pros. It would be nice if Apple dropped the entry level price of the Mac Pro down to $2000-$2500, but I somewhat doubt it. When it debuted, it was relatively competitive to similarly equipped workstation PCs.
If Apple wanted to introduce a Mac Pro lite, I predict they would do it by increasing the price of the entry level Mac Pro. Make the entry point a 6 core configuration with good GPUs. This way, the performance of high end iMacs or a high end mac pro lite would not be stepping on the Mac Pro's toes, as the Mac Pro will have a higher core count. Cost could be saved by not having to include support for the workstation class chipset, no ECC memory, only 1 thunderbolt controller, no second GPU and the entry level CPU could be something like a Core i5 around $200.
Sorry, this is just wishful thinking. Even if for some reason broadwell-EP is cancelled (which is unlikely), Skylake is not going to be released immediately to take its place. -EP class processors typically lag about a 12-18 months behind the consumer introduction. This was true for Haswell. For Broadwell, consumer chips were released Q3 2014, and Broadwell-EP will be released Q1 2016. If you want Skylake-EP, you are going to be waiting 12-18 months from now.
Stacc, there were really no DT Broadwell processors. Only a couple of low powered underclocked SKUs with great iGPUs.
That's why it could be possible that Broadwell was cancelled. Not saying it was, just that it's possible, as rumored.
If it was indeed and quite early on, and Skylake development was taken instead, maybe early next year it's a possibility.
Think about it, if Intel did come up with the 1500 without anyone knowing, and although it's more of a DT or mobile sibling than a real Xeon, at least some of it's development must be synched with the Xeon parts, right? OK, it's not the whole story but some work must be done and qualification could start soon as well.
It is indeed wishful thinking, but also considering (un)available info and recent releases.
I'll have fury (not the AMD one) come down on me for this...
That's why it could be possible that Broadwell was cancelled. Not saying it was, just that it's possible, as rumored.
If it was indeed and quite early on, and Skylake development was taken instead, maybe early next year it's a possibility.
Think about it, if Intel did come up with the 1500 without anyone knowing, and although it's more of a DT or mobile sibling than a real Xeon, at least some of it's development must be synched with the Xeon parts, right? OK, it's not the whole story but some work must be done and qualification could start soon as well.
It is indeed wishful thinking, but also considering (un)available info and recent releases.
I'll have fury (not the AMD one) come down on me for this...
Glad that you used the wiggle words "relatively" and "similarly".
The MP6,1 was much more expensive than similar power single-socket workstations equipped with a single desktop (i.e. non-ECC) GPU. It also required expensive external storage for many users, and was handicapped by a shortage of memory slots and an absence of PCIe slots.
You had to force possibly unwanted hardware (GPU2) onto the Linux/Windows box and ignore the expansion costs on the MP6,1 (T-Bolt storage vs available internal slots) to come up with a "competitive" comparison. You also had to ignore the advantages of more memory slots and PCIe slots.
The MP6,1 would have been dead-on-arrival except for the value that Apple users put on Apple OSX support.
Glad that you used the wiggle words "relatively" and "similarly".
The MP6,1 was much more expensive than similar power single-socket workstations equipped with a single desktop (i.e. non-ECC) GPU. It also required expensive external storage for many users, and was handicapped by a shortage of memory slots and an absence of PCIe slots.
You had to force possibly unwanted hardware (GPU2) onto the Linux/Windows box and ignore the expansion costs on the MP6,1 (T-Bolt storage vs available internal slots) to come up with a "competitive" comparison. You also had to ignore the advantages of more memory slots and PCIe slots.
The MP6,1 would have been dead-on-arrival except for the value that Apple users put on Apple OSX support.
I read Stacc's comment as referring to the original 1,1 Mac Pro. I bought one of those and at the time, so far as I could tell, its value was unmatched in the PC world. Sure, that changed. But back in 2006 it was one hell of a value. The 6,1 -- not so much.
How much simpler things would be if Apple changed their model names when the hardware changed. Six different systems over the last 10 years with the same name. Even worse in the portable lines, since hardware changes happen a bit faster there.
And to make matters worse, Apple recycles names for radically different products.
If someone says "I have a MacBook" - how many questions do you need to ask to understand which Apple they have?
is that right?
i'm buying a new laptop soon.. wanting to wait for refurbs (or find a 2014 2.8ghz refurb.. if anybody has a source?) but i pretty much need a new one now.. (as my current laptop isn't being replaced.. it's just being reassigned but still in the loop)
current laptop is 2010 mbp but i bought it refurb for maybe 2200.. the new ones are $2700 at the configuration i get.
so they were cheaper at some point? for whatever reason, i always thought the fastest cpu models were 3grand brand new.
?
---
heh.. i think much of the cost of the mac pro is going towards the manufacturing (thermal core and outer casing.. some small run custom shaped boards.. etc.)
i wouldn't be surprised if 1/3 of the cost of an entry nmp is paying for those things.. another thousand for the components.. and 1000 profit..
that said.. i also wouldn't be surprised (and actually expect) to see a price drop on 7.1
(edit).. or, at the very least, quit with the three-legged-dog ram
Last edited:
Just a point of distraction. The Xeon E5 v3 (Haswell) , v4 (Broadwell), v5 (Skylake) all have a different sockets and chipsets from the current design so any one of them mean changes. v5 isn't particularly any less, perhaps marginal only on the power layout front. There are several others you are ignoring (PCIe allocations , USB 3 routing , etc. ) .
FIVR or not is ballon squeeze distraction issue. Where the waste heat of voltage regulator (VR) is located isn't a major deal. What the canonical power input set up is .... there are Intel reference boards that layout the answer to that "mystery". If a system vendor wants to do something different from what Intel recommends then have some design work to do (because want to support running the CPUs out of spec ). That is extremely unlikely an Apple objective. Clever design summersaults on custom VR when overall power supply is very tightly allocated is going to buy what? A whole lot of nothing.
If Apple had done a E5 v3 system, then they could have done a "drop in" upgrade with a minor firmware bump. They didn't. The advantages for leverage the commonly inside of a tick/tock cycle, they basically passed on. So anything coming that is based on a post v2 baseline is going to be new motherboard work.
- Status