Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

JesterJJZ

macrumors 68020
Original poster
Jul 21, 2004
2,481
847
So yeah, it's at least a year away, but what are the top candidates other than speed increases?

Personally, I'm rooting for a second processor and SSD option. What else? USB3.1 maybe, I hear that's coming a long soon, also very Apple-like, being that it takes a lot of cues from the Lightning connector.
 
Haswell EP Processor
DDR 4.
(May be) 6-Core as standart configuration. (As 8 core Haswell E will lower price)
(May be) Dual processor model.
 
Same as this years model with upgrades to CPU/gpu + higher storage.

No second CPU, the current design does not allow for it, PSU is too small, and the heat generated would require the size of the Mac pro to increase.

The Mac Pro will remain a video editing solution of 1x Xeon CPU and 2x GPU.

Look at the 2006-2012 Mac pro, change is minimum, and in-line with Intel Roadmap
 
it will have the Gold Satin option "c3P0" and upcharge it 1,000 and fingertip security.

YAY!!!!!

All kidding aside, no idea. could anyone predict the changes from the 1,1 to 5,1?

My best guess some way to get rid of that hideous rat nest from the wires bunched around the trashcan. In all honesty it is one of the best looking Macs before you add all the components.
 
Late deliveries probably.

Other than that the latest Xeon CPU and more GPU options are most likely. I'd like to see a single GPU option, with a second CPU socket and more RAM slots. Dual 10Gig-E would also be useful.
 
Second SSD slot
Faster processor
Improved TB slots to hold better
New processor options

Now that I think of it, there's not much in there so not much to speculate on, especially for incremental updates, which the next one will most likely be.

The second SSD slot is a definite need though.
 
I wouldn't expect too much from the next revision. Faster CPUs, GPUs, and memory.

What would be my dream machine Mac Pro:

Love the new circular design, just think Apple went too small for the power needed from a WORKstation with their first version. Probably not likely to get any bigger but I can of course dream...

Let's turn that tri-hull into a hex-hull with a 8-10" diameter. With six sides,
1 - CPU 1
2 - CPU 2 (optional)
3 - GPU 1 (Nvidia or AMD)
4 - GPU 2 (optional)
5 - 8 Memory slots (DDR4)
6 - Raid Card w/ 4 SSD slots

Will obviously need a much bigger power supply, enough to run everything at full clock. A workstation shouldn't be overly concerned about being Green. Mine as well add another TB 2+ controller as well for those with lots of monitors, DASs, and external TB enclosures.

Of course, probably won't be able to afford it and engineering-wise may not be able to build it, but again, it's my dream. :)
 
I wouldn't expect too much from the next revision. Faster CPUs, GPUs, and memory.

What would be my dream machine Mac Pro:

Love the new circular design, just think Apple went too small for the power needed from a WORKstation with their first version. Probably not likely to get any bigger but I can of course dream...

Let's turn that tri-hull into a hex-hull with a 8-10" diameter. With six sides,
1 - CPU 1
2 - CPU 2 (optional)

3 - GPU 1 (Nvidia or AMD)
4 - GPU 2 (optional)
5 - 8 Memory slots (DDR4)
6 - Raid Card w/ 4 SSD slots

Will obviously need a much bigger power supply, enough to run everything at full clock. A workstation shouldn't be overly concerned about being Green. Mine as well add another TB 2+ controller as well for those with lots of monitors, DASs, and external TB enclosures.

Of course, probably won't be able to afford it and engineering-wise may not be able to build it, but again, it's my dream. :)

One tweak:
  1. CPU 1 + 4 or 8 DIMM slots
  2. CPU 2 + 4 or 8 DIMM slots (optional)
  3. GPU 1 (Nvidia or AMD)
  4. GPU 2 (optional)
  5. Raid Card w/ 4 SSD slots

The memory controller is on the processor die, and the DIMMs must be very close to the processor.

I've only seen them on opposite sides of the processor, like they are on the nMP. This photo of a ProLiant motherboard shows 24 slots, 6 on each side of each processor.

it_photo_181420_52.jpg

(click to enlarge)
 
There's a rather lengthy wish list features thread already here...
https://forums.macrumors.com/threads/1683404/

As for what's realistic...
- Haswell CPU (entry level 6-core?)
- DDR4 RAM
- New GPUs

You're probably looking at a 6-8% CPU performance boost clock-per-clock. I wonder if Intel will offer a low priced 6-core around the $350 mark that would allow Apple to offer an entry level 6-core for $2999.

As for I/O:

SSD: I use to think that a second SSD might be a no-brainer in a refresh, but someone pointed out that there may be a limitation on the pin count of the GPU to Motherboard connectors. Given the GPU without the SSD currently has to offer connectivity for 6 displays, I'm not sure there are enough pins to offer another x4 PCIe connection on that interface.

USB 3.1: Highly unlikely to see this in the near future in any Intel chipset and thus even less likely in any Apple product.

TB3: This is a possibility with support for PCIe 3.0 over Thunderbolt, but I'm skeptical of it being ready in 12 months.

HDMI 2.0: Very likely as this will likely be natively supported by the next gen GPUs.
 
As for what's realistic...
- Haswell CPU (entry level 6-core?)
- DDR4 RAM
- New GPUs

I think Haswell will shift the cores up and 6 will be the base.

I think the 6 core should be the base model now. It's only a $500 CPU and its performance is not that awesome as is.
 
Last edited:
There's a rather lengthy wish list features thread already here...
https://forums.macrumors.com/threads/1683404/

As for what's realistic...
- Haswell CPU (entry level 6-core?)
- DDR4 RAM
- New GPUs

You're probably looking at a 6-8% CPU performance boost clock-per-clock. I wonder if Intel will offer a low priced 6-core around the $350 mark that would allow Apple to offer an entry level 6-core for $2999.

As for I/O:

SSD: I use to think that a second SSD might be a no-brainer in a refresh, but someone pointed out that there may be a limitation on the pin count of the GPU to Motherboard connectors. Given the GPU without the SSD currently has to offer connectivity for 6 displays, I'm not sure there are enough pins to offer another x4 PCIe connection on that interface.

USB 3.1: Highly unlikely to see this in the near future in any Intel chipset and thus even less likely in any Apple product.

TB3: This is a possibility with support for PCIe 3.0 over Thunderbolt, but I'm skeptical of it being ready in 12 months.

HDMI 2.0: Very likely as this will likely be natively supported by the next gen GPUs.

Out of interest... what's all the fuzz about with HDMI 2.0, if thunderbolt delivers 4k at 60hz? Also was wondering what next GPU's would be like? Seems like the PSU will be a limiting factor and I have a feeling nvidia will still not be involved :)
 
SSD: I use to think that a second SSD might be a no-brainer in a refresh, but someone pointed out that there may be a limitation on the pin count of the GPU to Motherboard connectors. Given the GPU without the SSD currently has to offer connectivity for 6 displays, I'm not sure there are enough pins to offer another x4 PCIe connection on that interface.

No-brainer - use a higher pin count connector. ;)
 
Out of interest... what's all the fuzz about with HDMI 2.0, if thunderbolt delivers 4k at 60hz? Also was wondering what next GPU's would be like? Seems like the PSU will be a limiting factor and I have a feeling nvidia will still not be involved :)

Personally, I agree... DP1.2 is all I need. I guess HDMI 2.0 may be important of you want to use it to connect a third 4K display or use a TV as a monitor.

As for next-gen GPU's, they hinge on the availability of 22nm which is expected to become available from TSMC in volume later this year. Both Nvidia and AMD have done as much as they can with 28nm. However, from what I've read 22nm and even smaller process nodes are going to become an increasing challenge for fabs leading to potentially longer generational updates. As you'd expect, transitioning from 28nm to 22nm will provide for a significant increase in transistor count for a given TDP, enabling more cores (for added parallel performance) and/or more features (such as TrueAudio).

Whether Nvidia wants to get involved in the nMP, is likely up to Nvidia - whether they want to meet Apple's terms. I suspect Apple would love to have them, but only if they offer compelling OpenCL performance at a price point that supports Apple's desired margins (and yours and mine price expectations). Nvidia doesn't seem interested in tangential products like consoles or the nMP... I guess they must feel they are doing the best thing by focusing strictly on mainstream products. Like you, I can't see anything changing substantially in the next 12 months for them to reverse course in that regard. It may be for the best actually, because a constant flip-flop of vendors supplying GPUs for each generation of Mac Pro would only hurt driver development (especially for those of the previous generation). I personally hope Apple and AMD's partnership continues for the foreseeable future.
 
No-brainer - use a higher pin count connector. ;)

There's three possibilities...
1. Apple designed the interface and connectors between the GPU and mainboard with more than enough spare pins to make the addition of a 2nd SSD an easy proposition in the future.
2. Apple designed the interface and connectors between the GPU and mainboard with enough pins to accommodate the displays off one card and an SSD off the other and they have no intentions of redesigning that with every refresh.
3. As in #2, but Apple has no problems redesigning the interface and PCBs with every refresh.

My money is on #2. But only time will tell for sure.

There's also the issue of PCIe lane budget to consider. Unless Haswell Xeons offer more PCIe lanes, that will also add to the complexity of offering another SSD as you'll need a bridge/switch.

I thought I read USB 3.1 is almost ready, but you're saying Intel is far from ready to use it?

History is what makes me skeptical. USB 3 was ratified in 2008 and the first Intel chipset that shipped with USB 3 was in 2012, which incidentally was the first time Apple computers offered USB 3. On top of that, the C600 series Xeon chipsets as used in the nMP, still don't even offer USB 3. So, again, I'll be very surprised if Intel has implemented USB 3.1 within 1 year of standards ratification across their entire chipset line.

Regardless of potential delays, what can we really expect from USB 3.1? Is a ten-fold increase in USB transfer rates really the cure-all for that interconnect? I don't think so. Keyboards, printers, and even flash card readers won't benefit at all. Even most external drive enclosures don't or can't come close to saturating USB 3. The only real benefit is for external RAID arrays, and from what I've seen, most USB 3 external storage solutions today suffer from poor performance, latency, and/or sleep/wake issues that have nothing to do with the rated bus speed of USB 3. And this is now 6 years after the standard was released. I think Thunderbolt offers a vastly superior interconnect solution for high-speed peripherals than USB and I don't expect USB 3.1 to change this at all.
 
Last edited:
More importantly since everybody is calling the 2013 MacPro a "nMP" what will we call the nMP rev 2.0?

:)
 
SSD: I use to think that a second SSD might be a no-brainer in a refresh, but someone pointed out that there may be a limitation on the pin count of the GPU to Motherboard connectors.

The other issue is even if have the pins to run a x4 PCIe v2 connection. A connection to where? Unless Intel comes up with a chipset with more PCIe lanes, the 2nd SSD is going to do is take away bandwidth from the 1st SSD . Can get large total capacity but speed is going down. It wouldn't go down if add the 2nd SSD though TB and have same total capacity.

The real question is whether Intel going to deliver a chipset more closely matched with the design objectives this iteration or in some future one.


USB 3.1: Highly unlikely to see this in the near future in any Intel chipset and thus even less likely in any Apple product.

There is no Intel support for USB 3.0 the current Mac Pro. So lack of Intel support is much of an impediement. USB 3.1 is unlikely because Apple might dump the discrete controller to bring down costs. The other big problem with USB 3.1 is once again the PCIe v2 lanes to where? USB 3.1 is going to have to increase from the USB 3.0 controller's x1. In competition with the SSD or TB controllers where is the bandwidth to support this coming from without engaging in "Rob Peter to pay Paul" ?



TB3: This is a possibility with support for PCIe 3.0 over Thunderbolt, but I'm skeptical of it being ready in 12 months.

TB v1 -> v2 2011 -> 2014 TB v3 is likely about equally as long. Until Intel's chipsets move to x8 PCI-e v3 there isn't much of a match for some large bandwidth boost out of TB.



HDMI 2.0: Very likely as this will likely be natively supported by the next gen GPUs.[/QUOTE]

I think Haswell will shift the cores up and 6 will be the base.

There probably will be specialized 4 core options somewhere in the E5 range of products , but a bit out of the mainstream that Apple selects.

I think the 6 core should be the base model now. It's only a $500 CPU and its performance is not that awesome as is.

The base model needs a $300 CPU given Apple's other constraints they have on pricing and parts included. That isn't going to happen till next iteration. Dragging the core count higher isn't going to help the base clock rate. Sticking with 4 cores allows the E5 1620 v2 to have one of the highest E5 base clock rates and certainly the most affordable high clock rate.
 
I thought I read USB 3.1 is almost ready, but you're saying Intel is far from ready to use it?

To problems with USB 3.1

A. It is new. Which makes it a bad candidate to put into a core chipset. Any bug that pops up with this new standard's implementation not just delays the chipset it will effectively delay all the system designs. ( since Intel is the only chipset vendor if they screw up, all the new designs are screwed up).

Intel is going to wait until the discrete controller vendors have flushed out a decent portion of the bugs. Buggy discrete chips can be tossed from a system design much easier than a CPU or its essential campanion chipset.


B. There little to no bandwidth for USB 3.1 With all of the multilane SATA 6Gb/s and PCIe SSDs being attached to the core chipset in many new system designs adding USB 3.1 probably means trading down on one of those two. The bandwidth from chipset to CPU ( DMI ) is relatively limited. Two interfaces around the same bandwidth as DMI might juggle. Three not so much.

Intel doesn't have to incorporate USB 3.1. They can just throw x8 PCI-e lanes out there from the chipset. If some folks want USB 3.1 fine consume 3-4 lanes. If some folks want PCI-SSD then fine consume 2-4 lanes.


[ sure there are lots of PCI-e lanes in E5 designs; especially dual CPU package ones. But those aren't the Mac Pro or the vast majority of x86 PC system designs. There are x16 CPU lanes and x8 on the chipset and now there are multilane consumers coming out of the woodwork TB (x4) , USB 3.1( probably x2-3) , SSDs ( x2-4 ) , etc. ]
 
Thanks for the explanation!

So the huge advantage of building a computer from E-series chips (Xeon) vs. i-series (i7, i5) is that Xeon has many more "lanes" of bandwidth which are called PCI? Another is that the Xeon can take heat better so it wouldn't have to throttle down or throttle later compared to the i-series. And finally, Xeon use ECC RAM whereas i-series can't.

Do you think the next nMP will have ECC RAM for the video cards?

Is there such a feature of ECC when writing to storage?



To problems with USB 3.1

A. It is new. Which makes it a bad candidate to put into a core chipset. Any bug that pops up with this new standard's implementation not just delays the chipset it will effectively delay all the system designs. ( since Intel is the only chipset vendor if they screw up, all the new designs are screwed up).

Intel is going to wait until the discrete controller vendors have flushed out a decent portion of the bugs. Buggy discrete chips can be tossed from a system design much easier than a CPU or its essential campanion chipset.


B. There little to no bandwidth for USB 3.1 With all of the multilane SATA 6Gb/s and PCIe SSDs being attached to the core chipset in many new system designs adding USB 3.1 probably means trading down on one of those two. The bandwidth from chipset to CPU ( DMI ) is relatively limited. Two interfaces around the same bandwidth as DMI might juggle. Three not so much.

Intel doesn't have to incorporate USB 3.1. They can just throw x8 PCI-e lanes out there from the chipset. If some folks want USB 3.1 fine consume 3-4 lanes. If some folks want PCI-SSD then fine consume 2-4 lanes.


[ sure there are lots of PCI-e lanes in E5 designs; especially dual CPU package ones. But those aren't the Mac Pro or the vast majority of x86 PC system designs. There are x16 CPU lanes and x8 on the chipset and now there are multilane consumers coming out of the woodwork TB (x4) , USB 3.1( probably x2-3) , SSDs ( x2-4 ) , etc. ]
 
There's also the issue of PCIe lane budget to consider. Unless Haswell Xeons offer more PCIe lanes, that will also add to the complexity of offering another SSD as you'll need a bridge/switch.

There are four unused PCIe 3.0 lanes on the IO board. Just route them to GPU2.
 
Second SSD slot
Faster processor
Improved TB slots to hold better
New processor options

Now that I think of it, there's not much in there so not much to speculate on, especially for incremental updates, which the next one will most likely be.

The second SSD slot is a definite need though.

they don't really have the pci-e lanes for it.

16*2 for video takes 32 from the cpus 40. 8 goes to PLX switch (now in theory you may be able to get 16 lanes at 2.0 out of 8 at 3.0 speed if you are not useing all buses at the same time) But right they useing 13 lanes (usb 3.0 is useing only X1 that is one the X4 port) But 16 2.0 lanes needs more then what X8 3.0 can do.

http://imgur.com/ItIqxDY

they should put USB 3.0 on the open PCH X1 lane and flash slot 2 on the PLX switch. maybe they ran out off io room on the ribbon cables or that that the load was to high on the PLX switch with SSD and TB. Or routing was in the way pulling the X4 back to video card 2.

But then they may need to have both slots on the PLX switch other then 1 on PLX and on one PCH.

also PCH is feed from the X4 2.0 DMI link (intel needs to boost that to 3.0)

----------

There's a rather lengthy wish list features thread already here...
https://forums.macrumors.com/threads/1683404/

As for what's realistic...
- Haswell CPU (entry level 6-core?)
- DDR4 RAM
- New GPUs

You're probably looking at a 6-8% CPU performance boost clock-per-clock. I wonder if Intel will offer a low priced 6-core around the $350 mark that would allow Apple to offer an entry level 6-core for $2999.

As for I/O:

SSD: I use to think that a second SSD might be a no-brainer in a refresh, but someone pointed out that there may be a limitation on the pin count of the GPU to Motherboard connectors. Given the GPU without the SSD currently has to offer connectivity for 6 displays, I'm not sure there are enough pins to offer another x4 PCIe connection on that interface.

USB 3.1: Highly unlikely to see this in the near future in any Intel chipset and thus even less likely in any Apple product.

TB3: This is a possibility with support for PCIe 3.0 over Thunderbolt, but I'm skeptical of it being ready in 12 months.

HDMI 2.0: Very likely as this will likely be natively supported by the next gen GPUs.

PCIe 3.0 on Thunderbolt?? the video cards eat up 32 lanes. Moving usb 3.0 to PCH may free up x1 or the X4 slot on the PLX lanes but still 8 3.0 lanes may only be 2 TB buses. But you also will have USB + SSD and other stuff all on the 2.0 X4 DMI bus.
 
Last edited:
There are four unused PCIe 3.0 lanes on the IO board. Just route them to GPU2.

According to Anand, the nMP is actually short 4 PCIe lanes (hence the switch for the TB controllers)...
http://www.anandtech.com/show/7603/mac-pro-review-late-2013/8

they don't really have the pci-e lanes for it.

16X2 for video gives you only 12 from the cpus 40. 8 goes to PLX switch (now in theory you may be able to get 16 lanes at 2.0 out of 8 at 3.0 speed if you are not useing all buses at the same time) But right they useing 13 lanes (usb 3.0 is useing only X1 that is one the X4 port) But 16 2.0 lanes needs more then what X8 3.0 can do.

You're not making much sense to me here but maybe I'm reading it wrong. At any rate (according to Anand) lanes are allocated as follows...
CPU:
- x32 for GPUs
- x8 for TB Switch (shared across x12 2.0 lanes - x4 for each TB controller)
Total: 40 3.0

PCH:
- x4 for SSD
- x3 for networking (x1 for each of WiFi, GigE1, GigE2)
- x1 for USB 3
Total: 8 2.0

So There are no free lanes.

We may all agree that this allocation is a bit nonsensical (esp. x3 for networking) but it is what it is. Regardless, there's not enough lane budget with the current architecture for another x4 SSD without potentially bottle-necking something.


PCIe 3.0 on Thunderbolt??

Yes, agreed... For this to be feasible, we'd need to see more 3.0 lanes.

If Intels next gen chipset were to offer x8 3.0 lanes (instead of 2.0 lanes), you could do something like this...

CPU:
- x32 for GPUs
- x8 for TB1 and TB2
Total: 40 3.0

PCH:
- x2 for SSD
- x1 for networking (x1 shared by WiFi, GigE1, GigE2)
- x1 for USB 3
- x4 for TB3
Total: 8 3.0
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.