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
If there is no Skylake Xeon available to test, and see what is capable of, we cannot state that ...
🙂
It's better to wait and see first...
In general Xeon family is targeted at the non-consumer workstations, they have higher core counts, and support for ECC memory, better support for virtualization etc. I also think that they also have more pci-e lanes (?).
Also there have been reports that Xeon processors are actually the better quality samples and have been through better testing, for stability reasons.
The higher core counts only is enough reason for using them in nMP.
🙂
It's better to wait and see first...
In general Xeon family is targeted at the non-consumer workstations, they have higher core counts, and support for ECC memory, better support for virtualization etc. I also think that they also have more pci-e lanes (?).
Also there have been reports that Xeon processors are actually the better quality samples and have been through better testing, for stability reasons.
The higher core counts only is enough reason for using them in nMP.
Last edited:
Hi filmak
I agree but I wouldn't expect a larger (more powerful that is) PSU on the nMP. They might improve it's efficiency though, and even take that improvement at up a little bit, but no major overhaul should be expected.
That would mess with the overall design, which I don't will happen soon.
Still, I believe it's already well designed and efficient, at least for what Apple wants the machine to be.
What they will do is adapt the components to the PSU, or the available power output. CPUs will almost always have the same TDP (not necessarily the same power consumption though), and GPUs will always be clocked according to the available envelope, downclocked and/or downvolted accordingly. Other components need to be accounted for as well (PCH, SSD, etc). And if Apple decides to go TB3 and uses the power delivery option that will need to be considered too.
If the PSU should be upgraded in terms of power delivery, probably the cooling solution would also need a revamping to be able to dissipate the extra power. Don't think that will happen.
However, CPUs on 14nm should consume less power, but if you check TDP will still be as high. Newer GPUs should have better power management but the available ones are still on 28nm, only next year we'll get 14/16nm GPUs with possible lower power draw, although with improved specs (more CUs, more transistors) which will again take power consumption to the same levels.
Capitan, forget about DT CPUs in the MacPro, will not happen. I understand you want a great gaming machine, with loads of OC potential, but that is not the MacPro nor it will ever be. This is not a PC nor will it ever be like one, they will keep it Pro with Xeon processors, because that's their focus with the nMP.
Apple keeps a very tightly controlled ecosystem and having people Ocing and upgrading their machines would be a complete support nightmare I guess, they won't go for it.
You need to go the Hack... way for that 🙂
Xeons will usually have more cores and additional features enabled, like ECC mem support. They also have more mem channels, and more PCIe lanes. Imagine having an i7 CPU on the nMP? What would you do with the 2 GPUs? OK, now you will tell me you don't need to GPUs, right? Or that you can have each on only x8 lanes that it won't make a difference? What about when they do, when things are optimized to the point that it will matter?
I'm not saying I know how Apple thinks, but it seems clear to me that they won't go that route. They don't want you to have control over your machine, they want you to use it as it is. And they want you to use it within specs.
I understand your wish/need but I can also see their point.
This is still a Pro machine, like it or not. Some will argue right away 🙂
I agree but I wouldn't expect a larger (more powerful that is) PSU on the nMP. They might improve it's efficiency though, and even take that improvement at up a little bit, but no major overhaul should be expected.
That would mess with the overall design, which I don't will happen soon.
Still, I believe it's already well designed and efficient, at least for what Apple wants the machine to be.
What they will do is adapt the components to the PSU, or the available power output. CPUs will almost always have the same TDP (not necessarily the same power consumption though), and GPUs will always be clocked according to the available envelope, downclocked and/or downvolted accordingly. Other components need to be accounted for as well (PCH, SSD, etc). And if Apple decides to go TB3 and uses the power delivery option that will need to be considered too.
If the PSU should be upgraded in terms of power delivery, probably the cooling solution would also need a revamping to be able to dissipate the extra power. Don't think that will happen.
However, CPUs on 14nm should consume less power, but if you check TDP will still be as high. Newer GPUs should have better power management but the available ones are still on 28nm, only next year we'll get 14/16nm GPUs with possible lower power draw, although with improved specs (more CUs, more transistors) which will again take power consumption to the same levels.
Capitan, forget about DT CPUs in the MacPro, will not happen. I understand you want a great gaming machine, with loads of OC potential, but that is not the MacPro nor it will ever be. This is not a PC nor will it ever be like one, they will keep it Pro with Xeon processors, because that's their focus with the nMP.
Apple keeps a very tightly controlled ecosystem and having people Ocing and upgrading their machines would be a complete support nightmare I guess, they won't go for it.
You need to go the Hack... way for that 🙂
Xeons will usually have more cores and additional features enabled, like ECC mem support. They also have more mem channels, and more PCIe lanes. Imagine having an i7 CPU on the nMP? What would you do with the 2 GPUs? OK, now you will tell me you don't need to GPUs, right? Or that you can have each on only x8 lanes that it won't make a difference? What about when they do, when things are optimized to the point that it will matter?
I'm not saying I know how Apple thinks, but it seems clear to me that they won't go that route. They don't want you to have control over your machine, they want you to use it as it is. And they want you to use it within specs.
I understand your wish/need but I can also see their point.
This is still a Pro machine, like it or not. Some will argue right away 🙂
Not necessarily. The X5690/W3690 for example has an i7 990X Extreme Edition equivalent with the same basic specs and performance, but no SMP or ECC support. There's no need for ECC in most cases, and by most I mean a lot.
Beat me to it, sorry.
Yes, validation for Xeons is more rigorous than desktop parts.
There's a reason they usually only come after DT parts are available for some time.
Consider also that Xeons come in different flavors, 1S, 2S, 4S... Multi socket CPUs will not have counterparts in the desktop for sure.
Some people complain about lack of dual processors. For that you need 2 2600 CPUs, which would skyrocket the price. You will say that only 1 GPU would be fine. But the cost of one (AMD) GPU (which I believe we agreed already must have been a bargain for Apple) will not compensate for a 2600 CPU from Intel, that won't do bargains as AMD does, right? So, you can't have a dual CPU with Core i7 nor with Xeon 1600. Would you pay the premium over an already overpriced machine, according to many? Have you seen the premium for an 8 core in the current nMP, which is already a 2600 part? I still don't know why they didn't use the 1680 though.
Yes, validation for Xeons is more rigorous than desktop parts.
There's a reason they usually only come after DT parts are available for some time.
Consider also that Xeons come in different flavors, 1S, 2S, 4S... Multi socket CPUs will not have counterparts in the desktop for sure.
Some people complain about lack of dual processors. For that you need 2 2600 CPUs, which would skyrocket the price. You will say that only 1 GPU would be fine. But the cost of one (AMD) GPU (which I believe we agreed already must have been a bargain for Apple) will not compensate for a 2600 CPU from Intel, that won't do bargains as AMD does, right? So, you can't have a dual CPU with Core i7 nor with Xeon 1600. Would you pay the premium over an already overpriced machine, according to many? Have you seen the premium for an 8 core in the current nMP, which is already a 2600 part? I still don't know why they didn't use the 1680 though.
Capitan, that's because probably you only game on it, but the folks that do scientific, engineering and design work probably will disagree with you.
This makes no rational sense in the context of Apple's normal practices for several reasons.
1. Apple doesn't try to hit exact, fixed in stone dates. The whole entire purpose Apple having their own scheduled dog and pony shows is so that they can pick what date is right for them and not have the date forced upon them. For the last couple of years, iPhone release around September? Yes. Some exact date in September every year? No. WWDC on same date every year? No. The iPads on same date? No. Mac Laptops ? No.
The only thing special about Dec 19 was that it was technically about as late as you could go and still arrive inside the "Fall 2013" window. Going OCD on that date would be bizarre even for Cupertino kool-aid drinkers.
2. Try to mimic the 2013 announcement schedule is even more dubious. Effectively it has fundamentally different agenda. It went like this
June 2013 "We are going to stop selling the current form factor.... if want this one better buy relatively soon" ( i.e., sneak preview of new Mac Pro )
October 2013 "We told you should buy the old form factor because we are stopping selling it now. " (i.e., "soft announce" the new Mac Pro ... still no firm date for shipping, just Fall 2013, but more details now. )
Dec 2013 Fall technically ends in the month so ship something...
There is no likely radical form factor change coming. Any tweak they may make (e.g, add an inch or so in height and width ) isn't going to motivate many to get the "absolutely smallest" container before they disappear. Any physical tweak would probably be to enable a better thermal/power envelope; that's actually what more than a few folks want.
Announce and ship 2-6 weeks later is likely inside the same quarter. Announce and ship 12-16 weeks later is likely to suppress sales for whole quarter. Remember the iMac screw up (2012?) when they couldn't hit dates but had stopped shipping the current iMacs so it drove overall Mac sales down for the Quarter. I'm sure someone at Apple does. If can't get high volume shipments of Xeon E5 and TB v3 out of Intel unitil 2016 there is zero good reason for Apple to announce early and suppress whatever is left of the Mac Pro 2013 sales.
There is leveraging a slightly scarcity effect and then there is being rude. One contributing reason demand ballooned is because they 'announced' so far ahead. Apple's standard practice is to stop shipping he current after the announcement. In 2013, compounded on top of that because has pragmatically turned off the EU switch back in February, 11 months previous.
So a 2015-2016 strategy of 'starve' folks who need a Mac Pro to get some work down for Nov-Jan is extremely likely going to piss off a decent number of folks. It is a tool for jobs that aren't on Apple's dog and pony show schedule. All that primarily does is make the HP, Dell, Lenovo workstation sales force job easy. "See Apple? One day when you need another workstation they will yank the product off the market and you'll be hung to dry for a couple of months."
If Apple simply waits a bit to let the factor to get a couple more weeks of production built up then demand/supply mismatch won't be so high and so long. Too much scarcity invites hoarding and speculators to jump and makes the problem worse. The Mac Pro doesn't need that. Intel will start to ship to System vendors before their announcement date. Apple can use some of that to build up some inventory a bit higher than what they had last time. There still will be wait times but they don't have to be measured in multiple months.
[ Aside: I think the new MacBook is probably an exceptional case. I suspect it has been on a very tight leash because they are going to kill this v1 instance off relatively quickly. So they don't want any inventory in the channel when they do. Skylake Core M and TB v3 is a far better fit for a one port wonder than what is out there now. ]
And that's why I said ECC memory is only required by a minority. There's no justifiable reason why 90% of customers who are creatives and use Final Cut or Creative suite should be paying a premium for features they don't need. And I can also bet that most customers in engineering and science are not using many Mac Pros. They have mixed workflow comprising of servers and laptops from companies such as Dell and HP, Linux Systems with their own proprietary software, and a small amount of Apple computers. I studied the natural sciences at university and our climate model software wasn't even Mac compatible.
Don't be mad at me, I was just trying to see all angles, not playing the devil's advocate. 🙂
But if Apple drops the Xeon and ECC features, what solution would they have for those few that really need it, even as few as they are? The MacPro has been for that market all along, right?
Brad, I also don't think Apple would commit to an exact date just because it would make 2 exact years, but it would be nice though. 🙂
dec, I'm with you on the MacBook, it's just a matter of time.
But if Apple drops the Xeon and ECC features, what solution would they have for those few that really need it, even as few as they are? The MacPro has been for that market all along, right?
Brad, I also don't think Apple would commit to an exact date just because it would make 2 exact years, but it would be nice though. 🙂
dec, I'm with you on the MacBook, it's just a matter of time.
Should consume less power if the clock and core count ( both x86 and, if integrated, graphics) and clock are held steady then yes. If crank up the core count and/or clock then not necessarily so.
Xeon E5 is primarily targeted toward higher throughput. Lower TDP just purely for lower TDP doesn't buy any more throughput at all. What the Xeon E5 have is a power budget and largely what Intel is going to do is pack as much as they can into that budget. There are several other products they have in the Xeon product line up. There are Xeon-D , Xeon E3 , E7 in the line up to to fill other roles. (even Atom server oriented products that don't technically have Xeon product label... yet). E5's don't have to do everything for everybody.
The Xeon E5 v3 1620 and 1630 are a bit kneecapped on clock.
1620 v2 3.7-3.9 GHz http://ark.intel.com/products/75779/Intel-Xeon-Processor-E5-1620-v2-10M-Cache-3_70-GHz
1620 v3 3.5-3.6 GHz http://ark.intel.com/products/82763/Intel-Xeon-Processor-E5-1620-v3-10M-Cache-3_50-GHz
1630 v3 3.7-3.8 Ghz http://ark.intel.com/products/82764/Intel-Xeon-Processor-E5-1630-v3-10M-Cache-3_70-GHz
Intel backslid 0.2Ghz with v3 on clock and narrowed the Turbo range to almost nothing. That is partially offset by improved micro-architecture but is problematical for several kinds of workloads. Part of this is just market segmentation on Intel's part ( probably could clock a bit higher if tweak how the turbo modes trigger a bit) and they are kneecapped because they don't want to sell "too many" $300 processors at the cost of fewer >$500 ones. But AMD probably isn't going to be quite so uncompetitive in 2016.
If the v4 versions were not so kneecapped (or went to 6 cores ) then Intel would have a better product to fill that space than if they just pushed the same performance into a smaller TDP window. If the v4 versions could Turbo up to 4.0 GHz alot of the folks complaining they can't "single core drag race' over 4.0 GHz would evaporate.
Like I said a few pages back, there should be 3-4 models.
Mac Pro Metal/Gamer. i7, mini tower, 2x PCIE slots for SLI gaming, regular 2.5" SSD. Choice of AMD and Nvidia cards.
Mac Pro Creative. i7, medium tower, more RAM slots, M2 SSD, 3xPCIE slots for more storage and video capture cards.
Mac Pro CAD. Single Xeon, workstation GPU, M2 SSD, nMP design. No need for PCIE slots.
Mac Pro Science. Up to dual Xeon option, high tower design, up to dual workstation GPU option, internal RAID.
The chance of this happening? Zero. Apple doesn't want to cannibalise iMac sales and can't compete against the cost of gaming towers from PC companies or home brew.
Last edited:
Not true. The Xeon Family is aimed at a very broad range of systems of which non-consumer workstations is a small subset. The Mac Pro uses Xeon E5 class processors. There are several other elements of the Xeon family. Xeon-D , Xeon E3 , Xeon E7 , Xeon E5 4xxx class that have nothing to do with the the general space the Mac Pro falls into.
There are already Xeon E3 v5 products announced. They aren't going into a Mac Pro though.
That was my point. TDP is not the actual power consumption as you know.
Usually the extra power that is "saved" on a smaller node is used for additional features (being it cores or anything else) or speed bumps. Still, TDP will not vary much as has been always the case.
As you pointed out, the 1620 has the same core count, lower speed and has higher TDP. TDP is not an actual measure of the normal consumed power.
Usually the extra power that is "saved" on a smaller node is used for additional features (being it cores or anything else) or speed bumps. Still, TDP will not vary much as has been always the case.
As you pointed out, the 1620 has the same core count, lower speed and has higher TDP. TDP is not an actual measure of the normal consumed power.
Capitan, Apple would have kept the MacPro tower design (which is awesome by the way) if they thought that was the way to go, right?
I still think Apple doesn't want to get into the "everything and some boots" war with regular PCs. They make top notch personal workstations, beautiful and reliable, and leave the specs and benchmarks market for the PC manufacturers that try so hard to copycat, without much success I'd say.
Likewise on the smartphone business, they can sell iPhones almost at whatever price they want, and others can't. Why?
I still think Apple doesn't want to get into the "everything and some boots" war with regular PCs. They make top notch personal workstations, beautiful and reliable, and leave the specs and benchmarks market for the PC manufacturers that try so hard to copycat, without much success I'd say.
Likewise on the smartphone business, they can sell iPhones almost at whatever price they want, and others can't. Why?
Let's face it. They are about making stylish somewhat efficient products for people with money and making shareholders happy. It's all very bourgeoise. Utility, choice and options are not the concerns of an elitist ideological approach that is common in corporate California. You don't get what you want, you are given what they 'think' you should have. As the Wikipedia article on bourgeoise states:
'... the bourgeoisie is the social class who owns the means of production and whose societal concerns are the value of property and the preservation of capital, to ensure the perpetuation of their economic supremacy in society.'
'... the bourgeoisie is the social class who owns the means of production and whose societal concerns are the value of property and the preservation of capital, to ensure the perpetuation of their economic supremacy in society.'
Correct, but we have to live with it, or not.
Unfortunately, it's business nowadays. Everyone only cares about making money, and shareholders will have the last say.
If you don't go where the money is, you're gone in no time.
It's everybody's fault really, we all always want more...
Unfortunately, it's business nowadays. Everyone only cares about making money, and shareholders will have the last say.
If you don't go where the money is, you're gone in no time.
It's everybody's fault really, we all always want more...
Of course, and for now only the E5-1600 v3 series is suitable for use in nMP.
From intel's page... I would like to have your opinion about the following...
I think that their descriptions are for higher systems than consumer's.
Intel® Xeon® Processor E5-4600 v3, E5-2600 v3, and E5-1600 v3 Product Families
Brilliant versatility at the heart of your data center
With the Intel® Xeon® Processor E5-4600 v3, E5-2600 v3, and E5-1600 v3 product families at the heart of an agile, efficient data center, your organization can respond rapidly to changing business demands, protect data from cybercriminals and streamline data center operations.
A New Way to Think About the Data Center
Now designed for compute, storage, and networking to work better together, the latest Intel® Xeon® processor can help you meet today’s challenges while laying the groundwork for tomorrow’s fully-orchestrated data center.
- Deliver more automation capabilities for higher-performing, more secure clouds
- Solve increasing demands for storage and quickly crunch through complex data analytics problems
- Meet the speed of execution expected today
- Operate a more cost-effective, energy-efficient data center
- Measure and monitor
http://www.intel.com/content/www/us/en/processors/xeon/xeon-e5-solutions.html
The TDP for the E5 16xx for a given generation is the uniform so that one system design can take any one of the products in the line up and fit the same zone.
The TDP for E5 v3 is higher more so because there is more in v3 chip packages than was in v2 packages. Voltage regulation moved off motherboard and into the package. The net system thermal load doesn't really change. It is simply a ballon squeeze ( squeeze off the board and into the package. same total amount of 'air' ) . It is a bit more dense heating, so if have dubious cooling infrastructure that weakness will be more exposed, but not a big issue for the Mac Pro. The pragmatically more limited over-clocking isn't a contextual issue either.
TDP, rather than the specific load for a specific workload, is important if talking about a processor selection for a product line up and not for a single instance of a product line up for a single instance workload.
The Mac Pro line-up would be better if the 1620/1630 were better. If Intel is holding those back because the other models can't differentiate themselves enough then the whole line-up is a bit hobbled. If v4 generally increases the clock rates all around then it will be a better line-up than v3. Throw those increases on top of the small micro-architecture tweaks made and it will preform better on a wider variety of workloads.
There has been more than alot of handwaving in this thread about Apple waiting for later Xeon E5 because the TDP was going to get dramatically better. That is just not likley to at all as long as Apple is primarily looking at keeping the E5. It isn't going to get dramatically lower in 1-2 generations. Better management of the power budget allocated will probably dramatically get better.... waiting on process shrink because that is going to open up huge offload on the power supply is just wishful thinking.
Intel's description is a bit wide. 1600 are not exactly for data centers but more so for single socket workstations.
VR is coming off chip again, but only with SKL.
Maybe with Broadwell the lower end quad cores will be better than Haswell, that move of lowering speed was a bit non sense. There are already a handful of SKUs, opening up even more performance levels is confusing.
VR is coming off chip again, but only with SKL.
Maybe with Broadwell the lower end quad cores will be better than Haswell, that move of lowering speed was a bit non sense. There are already a handful of SKUs, opening up even more performance levels is confusing.
Maybe, when binning, Intel had a lot of under performing parts and called them 1620, while the ones that passed the test were named 1630, still useful and don't have to be disposed of.
Not going to happen. Having multiple models of a single desktop goes against everything in Apple's core philosophy since Steve Jobs returned to the company in 1997.
The difference between the E5-4600 an the E7 line-ups will fade over time. Other material leaked from Intel looks like they will be splitting off the E5 1600 line is some degree over time also.
A more power efficient data center can be accomplished by putting more workload onto fewer computers. That doesn't mean the individual computers consume less power. Just that you can run fewer of them. Similar with substituting a high number of short stroked HDDs (just using outer 50% tracks ) with fewer SSDs. The root cause problem is that the throughput isn't high enough. Get a better throughput solution inside of the system and will need less systems for same amount of workload.
So if can put 10 Virtual machines on a E5 4600 system now when could only do 7 before then perhaps can retire or repurpose a couple out of 10 systems had running before. Not a massive cut in TDP in individual system instances, but turning off whole systems.
E5 1600 similar factors. If 5 employees can do the work of 7 on older systems .... can consolidate down to 5 or if find more work can keep all 7 busy (instead of having to expand into more systems/seats ). Unlike server consolidation there is using only so much can consolidate onto a single workstation driven by a single human. The human becomes the gating factor. Can put bigger data sets on a workstation so that the 'size' of the human trackable set of problems goes up, but that has limits in workload scope. All workloads aren't getting bigger faster than the hardware is improving.
This all has very little to do with "single core drag racing" workloads and synthetic 'hot rod' benchmarks.
There is another set of data center workloads that are relatively lightweight and highly variable. Those machines may needed periodically, but are switched on and largely idling a substantial amount of time. Those need different kinds of processors than E5, E7 class. Intel sells those too. That's why there is broad family of Xeon processors.
I realize I'm way out of my league in this thread, so apologies if this seems like a stupid question:
I'm a little confused about the naming scheme and feature set of the upcoming processors, especially the Xeon line. Everything I've read in the past few months says that Skylake is the architecture to wait for. However, on the Xeon front, we're only getting Broadwell-E. I'm most interested in the new features that Skylake offers: support for the latest APIs, eDRAM+, DDR4, Thunderbolt 3, USB 3.1, DisplayPort 1.3, etc. Will the Xeon / Broadwell-E processors support these as well, or will we have to wait a year or two for these features to make it to the Skylake Xeon line?
FWIW, I'm still rocking a late 2008 MBP, mostly because I've been able to upgrade and max out the internals over the years. Now that Apple has made the rMBP is significantly more difficult for users to upgrade, I've decided to make the jump to the nMP line in hopes that I can extend its life a bit more than the rMBP line allows.
I'm a little confused about the naming scheme and feature set of the upcoming processors, especially the Xeon line. Everything I've read in the past few months says that Skylake is the architecture to wait for. However, on the Xeon front, we're only getting Broadwell-E. I'm most interested in the new features that Skylake offers: support for the latest APIs, eDRAM+, DDR4, Thunderbolt 3, USB 3.1, DisplayPort 1.3, etc. Will the Xeon / Broadwell-E processors support these as well, or will we have to wait a year or two for these features to make it to the Skylake Xeon line?
FWIW, I'm still rocking a late 2008 MBP, mostly because I've been able to upgrade and max out the internals over the years. Now that Apple has made the rMBP is significantly more difficult for users to upgrade, I've decided to make the jump to the nMP line in hopes that I can extend its life a bit more than the rMBP line allows.
There are some 1600's in some data centers. Just like Intel has the unmentioned E5-2400 series in addition to these.
"data center" is also a bit of overkill terminology when it is something just running a small company. Also sometimes virtualization/paravirtualization won't work for some folks that don't have two CPU package workloads.
http://www.supermicro.com/products/nfo/Xeon_X10_E5.cfm?pg=MB&show=SELECT&type=UP#List_MBD
Single socket, blade motherboard is a good fit for a E5 1600 if working with just 4 DIMM slots and reasonable RAM capacity demands.
Xeon is Broadwell-EP, not -E.
There is no info on Skylake-EP yet, so it's hard to say what's new really.
Still, you should be able to count on higher IPC, DDR4 is already supported on Broadwell-EP, possibly SKL will even allow for higher speeds, 2666 maybe?
All the staff you see in the desktop CPUs won't probably make it to Xeons, MHz OC, eDRAM, whatever.
The E5 Xeons will probably continue without iGPU, contrary to E3-1500 v5 released recently.
USB 3.1 will come probably with SKL, current C610 PCH does not support it yet. you can have type C though.
DP 1.3 might come with newer GPUs next year only.
TB3 will work with Broadwell-EP, the issue is that each controller needs x4 PCIe lanes, and nMP should go for 3 controllers which is a bit of a problem. Either Apple cuts on the numbers of ports, or uses switches. SKL has a new PCH with 20 configurable PCIe3 lanes and it could prove better suited, but not required.
You can do USb3.1 on TB3 by the way.
There is no info on Skylake-EP yet, so it's hard to say what's new really.
Still, you should be able to count on higher IPC, DDR4 is already supported on Broadwell-EP, possibly SKL will even allow for higher speeds, 2666 maybe?
All the staff you see in the desktop CPUs won't probably make it to Xeons, MHz OC, eDRAM, whatever.
The E5 Xeons will probably continue without iGPU, contrary to E3-1500 v5 released recently.
USB 3.1 will come probably with SKL, current C610 PCH does not support it yet. you can have type C though.
DP 1.3 might come with newer GPUs next year only.
TB3 will work with Broadwell-EP, the issue is that each controller needs x4 PCIe lanes, and nMP should go for 3 controllers which is a bit of a problem. Either Apple cuts on the numbers of ports, or uses switches. SKL has a new PCH with 20 configurable PCIe3 lanes and it could prove better suited, but not required.
You can do USb3.1 on TB3 by the way.
- Status