Apple is currently outfitting its new computer with Thunderbolt 3 and USB-C (3.1). Apple wants to put Kaby Lake Xeon's in them but Intel hasn't come out with a 6, 8, or 12 core Kaby Lake Xeon. Plus, Kaby Lake is just incremental from Skylake. We may be waiting for Cannonlake.
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.
Apple waiting on Intel [Kaby Lake or Cannonlake]
- Thread starter cohen777
- Start date
- Sort by reaction score
I don't wait anymore.
Last edited:
Over the last 25 years I've been watching CPU advancements it's the same old story; nobody is satisfied with the current generation of CPU and everyone is looking forward to the next, next iteration.
Well let me pass along some words of wisdom: Whatever new thing Intel pushes out ain't gunna satisfy any of us for long.
More more more. Faster faster faster.
Well let me pass along some words of wisdom: Whatever new thing Intel pushes out ain't gunna satisfy any of us for long.
More more more. Faster faster faster.
If Apple really want to update the Mac Pro line, they won't wait for anything. If they always use / wait for the next tech / standard / etc. Why the 6,1 come with HDMI 1.4 only? Why not wait for the DDR4? Why 5,1 didn't wait for the Thunderbolt / SATA 3 / PCIe 3.0?
This theory just doesn't fit the history. Apple is more then happy to sell the old / cheap stuff with next gen / expensive price tag.
This theory just doesn't fit the history. Apple is more then happy to sell the old / cheap stuff with next gen / expensive price tag.
Intel haven't even released SkyLake Xeons yet.. I think that's more likely.
That's kind of what they're doing now. I genuinely think there is some merit to the theory that Intel hardware has stagnated in terms of speed progress that Apple is wrestling with moving to longer update cycles on their desktop hardware. I built an i5-2500K system back in 2011 and it still holds its own against modern equivalents.
Skylake was a massive increment but with Moore's Law no longer applicable and Intel struggling to shrink their dies, there's no massive advantage to moving to a newer model of Xeon chips every two years. Admittedly, the GPU situation isn't great but dual D700s are pretty adequate for most usage cases. Professionals are still getting by comfortably with 4,1/5,1 Mac Pros after all.
1. Apple uses Xeon E5 in Mac Pros, not Xeon E3. This is possibly no Kaby Lake Xeon E5 coming any time soon. I can't find the article at the moment but there was a comment/leak by Intel somewhere that the "Skylake-X" (and implicitly Skylake-W would already incorporate the 14nm+ process technology that was used in the mainstream "Kaby Lake" models. There isn't much to bump to. There may be a bump to 14nm++ ( Coffee Lake) is 10nm has major hiccup, but Coffee Lake could be passed over too (as it is primarily a stop gap solution due to the somewhat slow 10nm rollout.)
In some roadmaps it looks like Intel may drop some Kaby lake -X ( and possible Kaby Lake -W which Apple might use) this year but that is really taking the 4 core mainstream E3 die, stripping off the GPU, and making a few tweaks to put into the larger 2066 socket (and PCI-e and memory uncore subsystems). It would be labeled Kaby Lake but would be capped at 4 cores. Good enough for "single core drag racing", but probably not a good match for overall Mac Pro market. 6 or more cores, there is nothing paritculary "Kaby" coming any time this year so single CPU package set ups.
"... The Skylake-X lineup will include six-, eight-core and 10-core SKUs while the single Kaby Lake-X CPU will be a quad-core SKU. ... "
http://www.fudzilla.com/news/processors/42692-more-skylake-x-and-kaby-lake-x-details-show-up
2. Cannon Lake single socket -X/-W class will probably come first some time in very late 2018 or early 2019. Would be ludicrous to wait on those. [ More likely to get a 4 core max Coffee lake stop gap filler if 10nm is messy an anything that goes 6+ cores. ]
The 10nm process is going to be aimed at the much higher priced data center server processors first ( not the workstation stuff that borrows from that technology stream. ). Intel is probably looking to throw up bigger blockage to the server versions of AMD's Zen based offerings. The workstation stuff is going to be a bit thinner blockage for a while. ( Optane probably would be help more than tech porn benchmark scores. )
3. it really isn't Intel. TB v3 would work fine with Xeon E5 v4 options. Apple couldn't do 6 "full speed" TB v3 sockets, but that isn't technically necessary. If Apple had OCD and wants to dump all of the Type A sockets . then yes then the Aug-Oct 17 rollout for Skylake-W is "necessary". But the mental disorder is of Apple's making not Intel's.
Last edited:
Change "not any time soon" to "never".
Because server Xeons don't have a GPU, I presume they also lack the video encode/decode blocks. Further, as you mentioned, Skylake Xeons are likely to be on the 14nm+ process. Thus, a Skylake Xeon and a Kaby Lake Xeon would be one and the same. Hence, there won't be a Kaby Lake named server Xeon.
Do you know any of this, or are you just making assumptions? The most important upgrade for the Mac Pro is the GPU(s) anyway. Pro users need the power provided by cards like NVIDIA's Pascal line. Also, Intel has been releasing chips regularly the past few years, so we shouldn't try to shift the blame to Intel. Apple are definitely the ones behind.
[doublepost=1490310730][/doublepost]I think the issue is that Apple is selling a 4 year old Mac with 5 year old hardware. "Waiting" for a new CPU shouldn't be a problem, because if Apple would just use current CPUs then there wouldn't be a problem at all. Everyone would be happy.
Apple just needs to start giving the Mac Pro a speed bump each year. Intel's doing their job. Apple's not.
The apologists will disagree, and claim that newer CPUs are no faster than the Ivy Bridge, and Apple is correct in sitting on their collective innovative ass.
In the real world, however, while Apple is stuck at 12 cores - Intel is shipping E5-2600 processors with 22 cores and improved instruction sets. Other vendors are selling dual-socket systems with 44 cores / 88 threads. (In the server market, quad-socket systems with many more.) HPE dual socket servers support 3 TiB of RAM - and Apple is still stuck at support for 64 GiB in the MP6,1.
If the people making your softwares are optimizing for quad-core Nehalem processors from 2009, you might feel that the E5-2600v4 CPUs aren't a big step up.
On the other hand - if they're optimizing for higher thread counts and the newer instruction set additions you might see a big jump.
Intel has made some big advances, don't let the people with "single-thread 2008 instruction set" blinders mislead you.
That's the problem with playing the waiting game. You keep waiting and the rest of the world has passed you by. I've know friends that keep waiting around to find the perfect house. Years go by and now they are near 50 and don't have a house and the market is too expensive for them.
No sympathies for Apple. Even the current Xeon's they use in the nMP would be satisfactory if they just stuck a pair of them in instead, and improved the GPU's and I/O. There are solutions. They would have had a roadmap off Intel and AMD. They would have known what was coming down the pipeline. They went for a total form over function enclosure for the nMP which, whilst it looked cool, didn't allow it to contain the kind of horsepower and expansion capabilities of the previous generation, which if they'd stuck with, could easily have accommodated more CPU's and better GPU's.
I completely agree with joebclash with regards to the waiting game.
Apple may release an updated Mac Pro someday. On the other hand, maybe they intend to get rid of the platform (just like XServe) in the very near future. No one really knows. In the meantime, life goes on and the PC world moves forward. I view the purchase of consumer technology as a way to obtain tools (for work and play) and not necessarily as an investment. Do you purchase a car as an investment?
Viewed in this manner, I finally got a refurbished Mac Pro 2013 from the Apple Store. I'm very happy with it and I hope that it will give me 10+ years of service.
richmlow
Apple may release an updated Mac Pro someday. On the other hand, maybe they intend to get rid of the platform (just like XServe) in the very near future. No one really knows. In the meantime, life goes on and the PC world moves forward. I view the purchase of consumer technology as a way to obtain tools (for work and play) and not necessarily as an investment. Do you purchase a car as an investment?
Viewed in this manner, I finally got a refurbished Mac Pro 2013 from the Apple Store. I'm very happy with it and I hope that it will give me 10+ years of service.
richmlow
Sky Lake? Kaby Lake? Cannon Lake? More like Swan Lake? Leave it up to Timmy and it's Swan Song! 
![Swan Lake.png]( "Swan Lake.png")
I've lost all faith in Apple and the Mac Pro. I needed something faster than the 6-core 2013 MP that I had, but what should I get an 8-core or 12-core version of the same thing for a lot more money but only a small performance increase? Okay I could rip it apart and put a bigger Xeon in there, but these are v3 Xeons not the newer v4 models and besides it's not a very straightforward process. I ended up selling it and buying a PC instead from the proceeds of the Mac Pro. It's around 30% quicker, but I didn't have to spend any more and if I need to upgrade I can just swap out the CPU to a 22-core Xeon if I like, and the process for this is very easy. I also now have a much faster GPU, something that I couldn't do with the Mac.
Apple could have updated the Mac Pro. There has been an update to Xeon since the 2013 model was released. On the face of it I can't see much of a technical reason why they haven't. Behind the scenes perhaps there is a technical reason, but the way Apple do business they just give the impression they don't care about the product or the customers who want to buy this product. This wall of silence doesn't work Apple. Customers need to either see a road map, or have some indication there will be an update in the pipeline.
Apple could have updated the Mac Pro. There has been an update to Xeon since the 2013 model was released. On the face of it I can't see much of a technical reason why they haven't. Behind the scenes perhaps there is a technical reason, but the way Apple do business they just give the impression they don't care about the product or the customers who want to buy this product. This wall of silence doesn't work Apple. Customers need to either see a road map, or have some indication there will be an update in the pipeline.
The MP6,1 uses E5-x6xx v2 CPUs, there have been two updates since it was released.
Apple's new moto:
"Apple: Yesterday's Technology, Tomorrow's Prices"
Last edited:
If we were talking about cars, then this would be like Honda saying that there's been no significant advancement in automotive technology, so we've decided to stop producing new designs and instead will only sell the 1986 Honda from this day forward.
If we're now in 2017, and you have a choice of buying a brand new Toyota for $20,000 or purchasing a brand new 1986 Honda for $20,000, which choice do you think most buyers would make?
Both cars are brand new. But one has continued to be refined, while the other is 100% unchanged since 1986. Most buyers would choose to buy the more current and refined car if they are going to pay the same price anyway.
... and the CUDA architecture, most software need to accelerate processing/encoding/rendering...
Last edited:
I do think Apple will move to all USB-C, even on their dekstop products.
But, I hope Apple leave "classic ports" available one certain models (thus no fuss with mandatory dongles...):
- Mac mini: HDMI and USB
- Mac Pro: (two) Gb or preferably 10Gb ethernet and USB
- iMac: Gb ethernet and USB
But, I hope Apple leave "classic ports" available one certain models (thus no fuss with mandatory dongles...):
- Mac mini: HDMI and USB
- Mac Pro: (two) Gb or preferably 10Gb ethernet and USB
- iMac: Gb ethernet and USB
But those two updates would have required a different motherboard ( back in the context from where this "just stick in a v3 or v4" comments comes from). That couldn't happen. Would have to swap out a major logic board to move from on "Tick-Tock" technology platform to the next.
Apple releasing the MP 2013 on the tail end of "Tick Tock" cycle basically put it into a 'dead end" in terms of CPU upgrades later in life. Two ways Apple could have gone. One, cut the MP a little short and get back onto doing a major update synchronized to "Tock" ( odd version release in Xeon E5 terms : v1 , v3) . Two, take an even extra long hit on recycle and wait until three had passed and jump onto v5 ('Skylake) and still jump onto an odd.
In case two it would be quite helpful to tell folks "we screwed up. this will take longer than usual" message in the between stage. Or at least to some nominal gesture at a speed bump ( e.g., 2012 Mac Pro ). That would serve as a indication that the R&D effort isn't completely dead in the water.
Jumping onto the tail end of the Tick-Tock cycle also couples you to older I/O chipset. Good, if you are being risk adverse to initial bugs. Bad, if you are I/O bound. Again v5 will would open door for opening up some of the PCIe bandwidth limits. ( As Apple as abandoned SATA completely there is a usage mismatch between their goals and the 610 chipset).
It isn't just a CPU availability problem. Apple appears to have turned the Mac Pro into a "hobby project" that is somewhat of technology demo. Some folks have advocated the idea of a "concept car" halo effect would be an overall good thing for the Mac market. It isn't. The problem Apple appears to have is wanting to pile too many technology upgrade dependancies into the upgrade. New Mac Pro means new Thunderbolt tech + new CPU tech + new GPU tech + new SSD tech + whatever else they can pile on top.
The 2013 MP was in part held up waiting for '2014' technology of TBv2 ( released limited run in 2013 but targeted for mainstream in 2014). Similarly E5 v2 got 12 cores to 'match' the top end core count of dual MP 2012 models. If Apple had done a MP with E5 v1 and TBv1.2 there would have been more regression in throughput. But in terms of CPU and GPU timing ( catching GCN 1.0 and not the more capable OpenCL 1.2/2.0 later editions) and the back-end of the CPU-Chipset cycle it was bad.
Last edited:
Except for the MacBook ( which predated 'apple stable' TBv3 ) Type C has pragmatically ment TB v3 more so than "USB". ( USB is pragmatically subsumed in TB v3 controller implementation. ).
Going to pure TBv3 only a Mac Pro probably would mean a delay to the new CPU socket (and more PCIe lanes ) and chipeset ( another addtion of PCIe lanes ).
The TBv3 controllers would basically swamp the E5 v4 + 610 chipset combo when coupled to the expected SSD + other stufff workload.
E5 v5 (Skyalke-X 2066 socket ) and X299 chipset would open up the bandwidth for 3-4 TB v3 controllers ( 6-8 sockets ) along with substantially better SSD ( and possible Optane bump) and even have leftover for 10GbE (if dropped down in affordable enough range on Bill-Of-Material (BOM) price budget. )
I expect the MacBook to stop being the 'oddball' on the next upgrade and shift to TBv3 also. Type-C socket would more so mean Thunderbolt than USB in the Mac ecosystem.
That said ... that is more OCD on Apple's part than good design.
The mini couldn't do more than 4 TBv3 sockets. Going from 2 TB + 4 USB down to just 4 sockets bad. If Apple left two Type-A's that would draw tons less needless rancor.
Ditto for iMac. 4 TBv3 + 2 Type-A USB would actually be a much better match to what the vast majority of people have.
The Mac Pro... more than 6 TB sockets is a bit deranged. TB has utility but getting into extremism for extremism sake is silly. Again there are lots of Type-A devices out there in that user base and the system is absolutely not short on space on the side of the device; even in the slightest. Nuking the TBv2 sockets (and requiring dongles) is already going to piss off folks. Even more for Type-A sockets is just going to pour gasoline on an already big bonfire.
Four TBv3 (Type-C) sockets on a Mac Pro would be useful in driving uptick in Type-C adoption. Pouring more than that probably will hurt as much as it helps over a broad segment of the population. Just drop the number of Type-A sockets... folks will get the 'hint'.