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.
Mac Pro Rumors
- Thread starter Serban
- Start date
- Sort by reaction score
Given Apple's direction towards consumer and not Pro user, I expect the next Mac Pro to be a much smaller form factor and possibly have limited expandability. Apple doesn't want us to expand and upgrade hence why they force us to buy new as thats where they make their money..
Me personally I will always remember the Mac Pro as a cheese grater type of workstation, and expandability is worth it to me.
Me personally I will always remember the Mac Pro as a cheese grater type of workstation, and expandability is worth it to me.
Not for me. If there were no drive bays at all and just a wad of dedicated thunderbolt ports I'd be happy. I totally wouldn't mind two matching or semi-matching cases sitting adjacent each other. One with drives in and the other the CPU. For boot the new-ish mSATA units rock! They're really tiny, very low power, and my current one profiles very well:
With two of those even better - but I don't need rotational drive bays nor full-sized SSD bays at all. Eight or so TB ports would do me fine in their stead.
Same for the PCI slots. If there was only one PCIe v4 slot at x32 that would be fine. Two would be better - especially if both were x32 and room for double-height cards. If I needed more slots; again, just supply us with expansion options that look good sitting next to (or on top of) the main (half-sized) box.
Last edited:
Seriously?
You REALLY want all of your storage to be external? Forget any speed differential. Forget the mass of boxes on your desk. Forget the huge problems if one were to come undone at the wrong moment. Forget that not many drives yet even support TB (unless you count firewire).
Sorry, if there were no drive bays you would probably be just as well off getting a couple iMacs and slaving them together. You would always know they would be regularly updated.
^^ This
Damn right. I like the fact that most of my stuff is contained in the one air ventilated box (admission: I do have a couple of external clone drives for back-up, but they sit on top of the MP case, no balancing act required.)
There are already too many cables - I want even more stuff in the box!
I get fed up with the 'oo - bright, shiney' kids who want change cuz they're 'bored' with the same old. Unless it's demonstrably better, please don't change the pro case, Apple.
Damn right. I like the fact that most of my stuff is contained in the one air ventilated box (admission: I do have a couple of external clone drives for back-up, but they sit on top of the MP case, no balancing act required.)
There are already too many cables - I want even more stuff in the box!
I get fed up with the 'oo - bright, shiney' kids who want change cuz they're 'bored' with the same old. Unless it's demonstrably better, please don't change the pro case, Apple.
Yes, my #1 worry (well #2 after no 2013 MP) is that they will reduce things somehow. Unless they make things better with any change I just want the old MP but with modern specs (faster, USB3, etc). The discussions about what they might get rid of strike me as defeatist.
Apple did this with the 2012 iMac but that is for non-techy consumers. They are the 'ooh, new shiny!' types. Pros and computer geeks care about capabilities more than looks. They can improve the design (rack-mountable is the top request) but that is just external stuff.
Keep a couple of things in mind:
- TB attaches via a 4x PCIe lanes, and adds additional latency on top.
- You won't have 4x lanes per port (TB chips provide multiple TB ports, and switch it).
- If each TB chip doesn't get it's own dedicated PCIe lanes (also switched), you only end up with a max throughput of ~ 4x PCIe lanes. Thus, parallelism of storage devices for increased throughput will be limited vs. a PCIe RAID card, all other things being equal (member count & configuration).
Hey. This is apple we're talking about. You will have no evidence until they announce it.
Rgds,
Interesting. But we're talking about a MacPro design which hopefully would implement the standard with the considerations of a workstation grade system. Implemented on a PCIe v4.0 x8 lane, we could embed 8 buffered fully dedicated 20 GB/s link 16GT/s transfer ports - each I would assume could then be hubbed as you describe their desktop implementation of it. That's like 10 or 12 GB/s per port throughput after considering 8/10 protocol, latency, and etc. Hehe, we could put a 8 or 10-Drive SSD RAID0 sets on each connection and get full speeds from all of them.
Of course I don't mean "like on their laptops".
Although... knowing Apple... Or do I still have my wires crossed?
Last edited:
biggest benefit would be an additional connectivity option for most users, that's it. best case scenario would be having the option to share a peripheral with a laptop or iMac, and maybe a RAID'ed storage array (like the CalDigit HDPro2) that doesn't rely on PCIe connectivity.
if the option is thunderbolt or more available PCIe slots inside a tower, I'll take the PCIe slots... only time will tell I guess, just hope it's not LESS available PCIe slots
It depends.
If the MacPro went (nearly) half size then it plus the external storage wouldn't take up any extra room and short cables of about 40cm whouldn't be a clutter.
If the ports were implemented as I outlined above, there would be no lose of speed. In fact it would be MUCH faster than SATA III - which is already a bottleneck as it is.
If Apple sold the complimentary peripherals in the right form-factors (looks like MacPro matching style, available in 2, 4, and 8 bay configurations, etc.) Then they could make more profit and maybe keep MacPro going a little longer. Folks on a budget could buy their plain vanilla cases from whoever - thus saving money and fueling an aftermarket.
Then yes, I wouldn't mind all the storage being external. Except, like I mentioned a dual mSATA or mPCI SSD tucked out of the way somewhere (Many people may opt for only using those when 2TB units are around $500 or so...)
And if similar things were true of the PCIe expansion dock then I wouldn't mind having yet another [mini-mac-looking] case sitting next to those with room for 4, 6, or 8 PCIe expansion bays.
IMO, the only advantage to internal drives is lugability and speed. The speed i already addressed and have you ever actually tried to lug a stuffed MacPro? Grr, it's pretty much a permanent fixture wherever it gets planted. So IMO, all things considered, it wouldn't really matter if storage and cards were internal or external. As long as the external configuration were designed with that in mind.
The advantage to my envisioned system of course is massive amounts of super high-speed storage. Given current storage device speeds (maybe mixed with Apple's new Fusion system), we're talking about up to eight $1k (vanilla) boxes which are each 32.5TB in size and deliver about 1.5GB/s streams. And plenty of bandwidth to spare when faster device interfaces like SATA IV come along and (hopefully) double everything up. No other workstation system that I know of offers anything similar - although you can come close in a DIY if your wallet size is extra-large.
Also in no case would the drives actually need to support TB. TB to SATA III already exists and TB to SATA IV will exist when that comes along. It's just like now with the Fiber storage arrays... None of those drives actually have an optical interface ya know... They're SATA or whatever...
And finally when was the last time something connected to the back of your MacPro "came undone" at any time let alone the "wrong time"? For me, between 6 and 24 machines in use simultaneously pretty much 24/7 from sometime in the early/mid 90"s till about a year ago the answer is: never.
Last edited:
All TB chips get are 4x lanes, and at the speed the TB chips can operate (IIRC, PCIe 2.0, so 500MB/s per, which is more than enough to handle 10Gb/s <1.25GB/s>).Interesting. But we're talking about a MacPro design which hopefully would implement the standard with the considerations of a workstation grade system. Implemented on a PCIe v4.0 x8 lane, we could embed 8 buffered fully dedicated 20 GB/s link 16GT/s transfer ports - each I would assume could then be hubbed as you describe their desktop implementation of it. That's like 10 or 12 GB/s per port throughput after considering 8/10 protocol, latency, and etc. Hehe, we could put a 8 or 10-Drive SSD RAID0 sets on each connection and get full speeds from all of them.
Of course I don't mean "like on their laptops".
Or do I still have my wires crossed?
Now most consumer boards right now offer 20x lanes, so that would allow for 16x for a GPU (card or soldered to the main board), and 4x for the TB chip. This is essentially what you'd see in a laptop or iMac (i.e. LGA1155 socket based).
Assuming, and this may be a BIG IF at this point, the next MP (or replacement product), has 40x lanes, that would leave 3 * 4x lane configurations. Under these conditions, it would be possible to strap 3x TB chips to the board, but given the cost involved (larger socket = more money, and the TB chips are also $49 per last I checked), would most likely be deemed too expensive to run this way.
I don't see TB as a primary interface for desktops. As per sharing peripherals with laptops or other systems that can run faster than other interfaces, absolutely (biggest aspect of the PC side of the TB coin).
Regarding TB based RAID systems, they're toys, not professional products. Even if the designer put the effort into creating a proper RAID controller, it's still not a proper professional product due to the fact these products are using consumer grade HDD's.
Even the Promise Pegasus R4 or R6 are using consumer grade drives.
So fabricate a better TB chip.
The mods to get PCIe version 4 going probably would not create much of an engineering challenge. And if the New Mac Pro only has PCIe v2 slots you can count me out anyway. Version two... heh... Sounds like desktop secs. I want a classy workstation.
The kind of bump we had with the introduction of MacPro1,1 slash 2,1... It's time for another bump. Already MacPro4,1 and MacPro5,1 are too closely spec'd. As are 1,1 through 3,1. Give us two full spec PCIe v4 busses. And we don't even need connectors for more than two slots of one buss internally. Or maybe one would be enough... a single Version4 lane is four times the speed of a Version 2 lane...Yeah, of course I know I'm dreaming here. But this is kind of a dreamers thread anyway. So might as well dream a nice dream. And this is actually realistic and doable too. Sure MacPro would be base priced kinda high but decent spec'd MacPro have always been a little out there in relationship to other systems and DIY price-points. What I'm talking about does require some engineering/reengineering efforts but Apple has had plenty of time for that as of about now... So it's actually feasible as well. Well, within the dream of course.
Last edited:
The limitations are in the TB chip, and not with how it interfaces with the PCIe bus. That's quite an engineering challenge, and they're going to take their time solving it in order to recover as much $$$ as possible before improving the specs.So fabricate a better TB chip.
Thing is, there's a real possibility that whatever replaces the MP will be more of a consumer model so it can be cross marketed in order to generate a growth product.Sounds like desktop secs. I want a classy workstation.
A very real possibility!
= Quite sad for folks looking for something significant to upgrade to tho. =You are ignoring the factor that TB controller chips in a personal computer device need two inputs source types. x4 PCIe and 1-2 DisplayPort inputs. 3 TB controllers would mean 3-6 DisplayPort outputs. Where getting those from?
Not a very effective use of the logic board space even if wanted to do it.
It is also a bit of a waste to hook them to the PCI-e v3 lanes of the CPU package. The four of the eight lanes of the chipset are a more natural fit, which limits to just one controller. Besides it would be interesting if the firmware actually could deal with 3 different switches with potentially 36 other PCI-e switches in external TB controllers dangling off of them.
the TB spec bump in 2014 might pick up PCI-e v3 updates although the remote controllers are likely still primarily stuck on v2 so perhaps not. They may just use speed increase largely to decrease latency in large and/or long TB networks. In short, using Fat-tree ( http://en.wikipedia.org/wiki/Fat_tree) setup with a faster backbone. It is already somewhat Fat Tree anyway but longer distances and more switches only makes it necessary to get "fatter" on the backbone haul branches.
And the dream continues to crumble...
Same place as these Motherboards?
But I'm not sure it HAS TO include display signals or if just CAN if so implemented. Wiki doesn't say specifically and I haven't really looked into the interface engineering spec - just the userish parameters. <shrug> I suppose 2 or so ports could contain raster and the other not.
Sounds disappointing except for the 2014 bump you speculated on. Anyway <sigh> if Apple did do something like this, I wouldn't complain.
If it's the same-old same-old with PCIe v2 (4 internal slots), four SATA III drive bay, and usual round of minor enhancements & speedups, I'll be skipping yet another MacPro release. I wanna be wowed! I know this might sound a little strict but even the difference between MacPro1,1 and 5,1 isn't all that much of a wow IMO. Maybe I could hang if Apple went with PCIe v4, included 2 mSATA MB connections, added front loading hot-swap (x4?) of some kind on top of the 4 internal bays, and got a deal from Intel on some of the sweeter Xeons or something like that. <Shrug>
.
Same place as these Motherboards?
- ASRock Z77 Extreme6/TB4
- ASUS Maximus V Extreme
- ASUS P8Z77-V Premium
- ASUS P8Z77-V PRO/THUNDERBOLT
- Gigabyte GA-Z77MX-D3H-TH
- Gigabyte GA-Z77X-UP4-TH
- Gigabyte GA-Z77X-UP5-TH
- Intel DZ77RE-75K
- MSI Z77A-GD45 Thunderbolt
- MSI Z77A-GD80
But I'm not sure it HAS TO include display signals or if just CAN if so implemented. Wiki doesn't say specifically and I haven't really looked into the interface engineering spec - just the userish parameters. <shrug> I suppose 2 or so ports could contain raster and the other not.
Sounds disappointing except for the 2014 bump you speculated on.
Anyway <sigh> if Apple did do something like this, I wouldn't complain. If it's the same-old same-old with PCIe v2 (4 internal slots), four SATA III drive bay, and usual round of minor enhancements & speedups, I'll be skipping yet another MacPro release. I wanna be wowed! I know this might sound a little strict but even the difference between MacPro1,1 and 5,1 isn't all that much of a wow IMO. Maybe I could hang if Apple went with PCIe v4, included 2 mSATA MB connections, added front loading hot-swap (x4?) of some kind on top of the 4 internal bays, and got a deal from Intel on some of the sweeter Xeons or something like that. <Shrug>
.
Last edited:
So? The MP is not designed to be a laptop (of which Apple has plenty of different ones for people who care about portability).
And finally when was the last time something connected to the back of your MacPro "came undone" at any time let alone the "wrong time"? For me, between 6 and 24 machines in use simultaneously pretty much 24/7 from sometime in the early/mid 90"s till about a year ago the answer is: never.
Are you familiar with Murphy's Law? I have had all sorts of 'it will never happen' things happen with computers (macs, PCs, mainframes, whatever). An external drive is an unneeded point of failure. Maybe it is less of a problem in a work office but when you have to worry about other people/animals bumping into cords ...
And the dream continues to crumble...
Same place as these Motherboards?
No. To my knowledge Intel's iGPUs (via the associated IOHUB chipset) can't drive more than two DisplayPort outputs. What those boards are hooked is the iGPU output.
Peripherals? no. Personal computers that nominally generation video? Yes. There isn't a "data only" Thunderbolt system. Two major issues:
1. if going to re-use a video port standard then nominal user expectation is going to be that if they plug in a DisplayPort cable it will work. Agreeing to that is likey the only way they could easily get the Display folks to get on board with hijacking their port. Thunderbolt contributes to more folks adopting DisplayPort devices. If it doesn't why let them hijack the port?
the USB folks had already said no because Thunderbolt would hijack a future USB 3.5 or 4.0
2. If end up with a Thunderbolt network where there are no producers of video ( PC doesn't. The peripherals don't ) then it is quite suboptimal to have reserved channel bandwidth minimums for that traffic on the wires.
Could users choose to use it in a data only fashion? Sure. That's OK. However, to set it up so that the system vendors on both sides can block the choice is only going to cause fragmentation of the new Thunderbolt ecosystem. In the early stages of a new ecosystem unnecesary fragmentation is bad thing. It typically leads to failure. Especially when not particularly compete on costs.
Thunderbolt is far more like SCSI and firewire ( nominally daisy chaining for expansion) than USB a wide tree from a central device. However, it is like USB in that this is a hierarchial network. The root PC is special. The switching network once everything is plugged in is a hierarchy. It shows up as a nested set of switches to the 'root' PCI-e controller on the device.
Secondary for Intel is that in most situations this connects their iGPUs to the outside world. Thunderbolt "happens" to forward Intel's CPU+GPU direction on their mainstream CPU packages. All those boards being hooked up is indicative of what Intel would like every board to be like. (e.g., use displayPort/Thunderbolt to push legacy VGA off standard backplanes.)
That is going to cause user expectations mismatch. If both ports are labeled with the same symbol. They are physically identical to a mini DisplayPort cable. How come monitor works in one and not the other? That is just "mystery plugging results" for no good reason.
Sounds disappointing except for the 2014 bump you speculated on.
It is called being informed.
http://www.anandtech.com/show/5405/the-first-thunderbolt-speed-bump-likely-in-2014
Intel basically said as much. Again it is also rational as I outlined against churning specs and fragmentation in early standards evolution.
Most folks buy workstations to work. Not because it is 'cool' or the 'thrill'.
If it doesn't do work well then it will fail. There is not enough folks with more money than sense to float a product like the Mac Pro.
PCIe v4 probably isn't going to make it down to the Mac Pro for a long time. Even the PCI-e v4 folks know they are working on a niche update that is primarily oriented toward a smaller subset of computers. The spin is that it will be generally useful. But it won't even finish standards till 2014-15 ( http://www.pcisig.com/news_room/Press_Releases/November_29_2011_Press_Release_/ ). Like v3, the actually deployment is not likely to even start until a year or so until the testing standards have solidified. More likely it will be used as a more standard processor interconnect than for generic expansion cards.
Frankly, besides those who are pushing 100Gb Infiniband and Intel's Aries Interconnect ( http://www.theregister.co.uk/2013/02/15/cray_xc30_plans_q4_2012_numbers/ ) there aren't many drivers to move past PCI-e v3 Frack, totally behind the curve AMD hasn't even implemented v3 yet in any of their offerings. ( Although they seem to be willingly dropping out of the high performance market. )
v4's adoption curve is gong to be quite long unless more affordable usages show up in the next year or so.
Like the XServer there is hardly need for more than one mSATA like device.
Similarily hot swapping anOS drive (or possibly mirrored OS drive) is all that is really required in a racked environment. If want front load access to lots of disks there are storage arrays designed just for that. the "compute" box can just be attached.
Apple doesn't buy enough Xeons to get super sweet deals. Folks make far to much of that one shot event of Apple jumping the gun on a Intel Xeon release window. that was likely more indicative of how small of orders Apple for that class than anything else.
Point taken...
--
Man, you guys just aren't gonna let me dream are ya...
Wait... I know!!! I got it!!! A new port - and we'll call it "Lightningbolt". It'll be 32GT/s on each line (no display funny stuff), and the new MacPro will have 8 of them! Ha!
--
What? Heh... You ain't no Apple user...!
.
Last edited:
Wait... I know!!! I got it!!! A new port - and we'll call it "Lightningbolt". It'll be 32GT/s on each line (no display funny stuff), and the new MacPro will have 8 of them! Ha!
http://www.theverge.com/2012/1/13/2706097/amd-lightning-bolt
Damn! OK, "SPARC-Z" then...
It's an acronym for SPeedy Apple Resurrection Connection and the "Z" is a secret!
Though it's there, it's possible to leave it unconnected (data only configuration), only connect DP up to a single TB chip (& leave the remaining TB chips' DP signals unconnected), or share the same DP signal between all TB chips (switched DP signal).
The last option of course, allowing a video signal all the way around to prevent confusion over what TB ports can/can't do. But in a desktop setting, the primary use would be for data transfer with shared peripherals than handling a video signal IMHO.
Never said this was efficient use of PCB real estate, or that there wouldn't be additional challenges created by a multi-TB chip implementation.
Just showing limitations in regard to what Tesselator would primarily use the TB ports for, which is storage (not just the chips, lane consumption and bandwidth, but current TB storage products as well). It's just not suited for the type of storage use he needs at this time (proper RAID configuration on enterprise grade media).
Given how new TB is, I don't see this happening that quickly in order to maximize profits.