The new Mac Pro is the xMac

[deconstruct60]

That one's on DDR4's upcoming architecture of a single DIMM per channel. Quad-Channel Memory -> four DIMMs. DDR4 also has provisions for highly increased density, so the Mac Pro with 128GB RAM using four DIMMs is already around the corner (or at least 64GB get more affordable).

It isn't just DDR4 it is also users.

https://forums.macrumors.com/threads/1598539/

Not claiming this has optimal sampling demographics but even on geeky macrumors only < 10% of Mac Pro users crack the 64GB barrier. Four DIMM slots is not a huge market barrier. That is one reason why DDR 4 goes there. It isn't particularly to gratuitously kick users in the shins. It is where the mainstream workload is.

There probably are going to be more expensive rank "cheats" for DDR4 where something pretends to be a single range while juggling two DDR4 dimms. It will be a low density for latency trade off that some subsets of servers make. Probably will not be mainstream though.

As opposed to wireless FibreChannel for the Pros?

802.11ac is not so much Prosumer but normalizing across the Mac (and Airport ) line up. It is a normal technology upgrade. If had 11n it is about time for 11ac. Just like it is time for USB 3.0 for systems that used to have USB 2.0.

Thunderbolt = External PCIe.

No external PCIe is external PCIe. There already is a external PCIe standard. Thunderbolt is something different. Yes, part of that is external PCIe but it is more. That has upsides and downsides.

SonnetTech will gracefully supply you with an Expansion Chassis for your Pro cards, which was stated by Phil Schiller in the keynote as a use-case for all those Thunderbolt 2 ports.

Those work fine for the x1-x2 PCI-e Audio card and the x1 USB 3.0 or FW800 port expander PCI-e card. There is a x8 and up cards where that doesn't work so well.

A) Xeon-E5 v2 starts with eight cores,

No it doesn't. Xeon E5 will have an 4 core model. The minimum number of cores likely targeted at a Mac Pro is likely 4 cores ( E5 1620 v2 ) since that actually results in the lost CPU cost which will in turn result in the lowest user system cost.

E5 is not the 2600 series and even the 2600 series doesn't start with 8 in core count. Minimum 2600 is also 4.

and the only part that would fit into the 85W per CPU TDP of the old Mac Pro is a 10-Core that runs at 1.7GHz, B)

There is no "per CPU TDP" in this new Mac Pro. There is only one. So it is only the CPU TDP. 2 * 85 => 170W is higher than 130W.

C) One 12-core is probably less expensive than two hexa-cores because you remove on-chip redundancies that not only cost money but also draw energy,

Err no. Only the same core performance on same mirco-archecture design (which hard to do with v2 options since all move up 2 cores) . If go to lower clock sped on 6 cores two of those older ones will likely be cheaper than the $1,800-2,200 12 core is going to be. But the 12 core is going to be at the top of the Xeon E5 v2 price spectrum. Within the E5 2600 v2 series, you'll be able to get two entry level 8 cores for less money than one 12 core.

the 12 core costs more money because the die is bigger. So intel gets less chips per wafer (and more likely a defect pops up in chip). So each core pays more of wafer..... plus ... they can get away with it. It isn't just purely wafer costs.

there isn't alot of redundancies with two packages. You actually get more. More PCI-e lanes. More memory controllers. More cores. More L3 cache. It is really a question of how much is enough for most of the targeted users.

Frankly above the E5 1600 and 2600 series there is the E5 4600 and E7 series. If count core is the primary object and space and price is no object then can go higher still.

and one less CPU means less components on the logic board that draw power, produce heat and cost money, and ultimately a lower cost of the system.

It don't think the new Mac Pro is trying to reduce system costs as much as reassign them. Some cost reductions have been traded for a standard 2nd GPU.

The lower power maximum is more so from throwing out requirements to power and cool arbitrary cards with arbitrary requirements.

Do you have a projection of "Prosumers" using a triple 4K setup?

The disconnect is more so an implication that the display limit is just 3. Three 4K display doesn't necessarily mean limited to 3 displays.

Pros love gray plastic?

Well it actually isn't plastic being stabbed at here but the fact a certain class of "Pros" don't particularly care what the box looks like because they are going to just stuff it in a rack/closet/cabinet/etc. It could have purple flowers on it .. no one is going to see it most of the time.

I don't even understand your argument on 802.11ac because there isn't anything better.

Again it goes back to the mindset of it is going in a rack/closet/cabinet/etc in which 802.11ac wouldn't work well in anyway. Not that is where most Mac Pros are going. Just the presumption that a subset of the Mac Pro market is actually the Mac Pro's primary market.

Also, a Prosumer is a normal guy who buys Pro equipment

Typically the argument isn't even that the other buyers are normal. It is that they are "pro" enough because you have to have bought x, y , z, and w and have a , b, c, and d legacy integration requirements to be a pro.

It is also the notion that have to cover every single possible sub niche to be pro. Can't be a pro swiss army knife if only have 4 function.... keep doubling that and will get to pro knife.

Most of this is a knee-jerk reaction to Apple's adjustments on targeted subsets of broader markets. If their subset is left out somehow the whole pro connotation evaporates.

 

[toughboy]

Let us not forget this thing is being assembled in the good old USA. I highly doubt the price point will be that low.

It will be "Made in USA" because Tim talked with Obama about bringing some of the revenues back to US and got a tax-cut by promising an American assembly line, therefore I dont think it will have huge impact on the production cost of the new Mac Pro.

 

[ssgbryan]

That's an insane comment. Clearly, the new Mac Pro is far more powerful than the current Mac Pro.

I have a really difficult time seeing your problem here. When I need a new computer, I buy the highest-performing computer I can afford. I don't spend any energy worrying about computers that don't exist. And then I keep that computer until I can afford a computer that performs better (a cycle that usually takes me about 5 years as well).

Ok, I'll dumb it down some more. I am not an Apple fanboi. I use what works best for my workflow.

I am not wedded to OSX. My software is feature compatible on both the OSX and Windows front. I don't really care whether the OS is Apple or Windows - quality control for OSX has dropped to the level of Windows in my opinion.

And having Sir Idiot Boy handling the software end of it doesn't give me confidence in Apple's software direction - he is a 1 trick pony - make it less capable, make it smaller.

My situation is simple. I am at the end of cycle for my current MacPro.

TODAY, for around $4800 I can purchase a 16 core Dell Workstation.

Whatever the cost of the nMP, I will need to add $1500 to the price to handle the storage requirements I have TODAY - that does not take into account future growth over the next 5 years.

The child-like belief that some people have around here that the nMP will force TB products to drop in price apparently weren't around for Firewire, or SCSI for that matter.

If the top end nMP is over $3000, it won't compete with the Dell/HP line up.

And that nMP will still have fewer cores than I can get today on the Dell/HP front.

It will also have less capabilities on the RAM front. This may change, but not in the near future.

Then there is the issue of the SSD size in that nMP. I don't have 2 240Gb in a Raid0 for the speed (although that is nice), it is because I filled the first 1 up, and it was cheaper to add another as opposed to getting a 480Gb one.

 

[drsox]

Also, a Prosumer is a normal guy who buys Pro equipment (the Mac dictionary is just a triple-tap away and defines it as "someone who uses a $10,000 camera to make home movies of his dog"). If you look at Pro equipment and differentiate features on the level of Prosumer vs. Pro, then your metrics are a bit off.

Nope. Prosumer may include these people but also includes enthusiasts who don't use their Tech equipment to earn their living but just want to use interesting gear. Might also be gamers, but also might not.

Me, for example. I want a powerful "workstation" that I can run headless and the fastest I can find is only 50% faster than the one I built last year. If I can't run this headless then it also won't work for me.

Apart from the CPU. I'll have to look at GPGPU, but there's a limit to what I can run easily.

 

[Giuly]

It isn't just DDR4 it is also users.

https://forums.macrumors.com/threads/1598539/

Not claiming this has optimal sampling demographics but even on geeky macrumors only < 10% of Mac Pro users crack the 64GB barrier. Four DIMM slots is not a huge market barrier. That is one reason why DDR 4 goes there. It isn't particularly to gratuitously kick users in the shins. It is where the mainstream workload is.

There probably are going to be more expensive rank "cheats" for DDR4 where something pretends to be a single range while juggling two DDR4 dimms. It will be a low density for latency trade off that some subsets of servers make. Probably will not be mainstream though.

DDR4 has provisions for 3D stacking, up to eight dies are layered on top of each other, interconnected and then enclosed in a single package, so there is no need for more DIMMs as you just increase the capacity via the layers in the chips instead.

Those work fine for the x1-x2 PCI-e Audio card and the x1 USB 3.0 or FW800 port expander PCI-e card. There is a x8 and up cards where that doesn't work so well.

If you're running two graphics cards, you're left with two PCIe 2.0 x8 slots which happen to have a total bandwidth of 80GBit/s. Six Thunderbolt 2 ports with 20GBit/s each offer 50% more bandwidth at 120GBit/s. It's just a matter of splitting the x8 PCIe cards into something akin to PCIe x4. For your professional video capture card with four inputs, that means two inputs and a single Thunderbolt 2 port per device.
Most cards already work fine in the Thunderbolt 1 expansion chassis, though, and most cards work with PCIe 1.0 on the darn old Mac Pro, of which eight lanes offer the same 20GBit/s as Thunderbolt 2.

Not to mention that you need new gear for 4K anyways so there's little incentive of bringing them over, but that only goes for video.

No it doesn't. Xeon E5 will have an 4 core model. The minimum number of cores likely targeted at a Mac Pro is likely 4 cores ( E5 1620 v2 ) since that actually results in the lost CPU cost which will in turn result in the lowest user system cost.

E5 is not the 2600 series and even the 2600 series doesn't start with 8 in core count. Minimum 2600 is also 4.

According to news from yesterday, this is the E5-2600 v2 lineup:

The highlighted models are those with known part numbers.

Then there's E3 v3 which are single-CPU models and basically rebranded Haswell i3-i7s (= same as the top-of-the-line 27" iMac).
Other than that there's E5 2400 v2 which has those 4-6 core models and E7, of which one CPU costs more than a complete Mac Pro. Both aren't due for September, but for Q1'14.

There goes the E7, and also the E5 2400 v2 as those are intended for Entry-Level DP servers. E3 v3 should also be safe to dismiss as that's a completely different platform. E5 1600 v2 would technically be a fit, but there hasn't been any news on those lately.

the 12 core costs more money because the die is bigger. So intel gets less chips per wafer (and more likely a defect pops up in chip). So each core pays more of wafer..... plus ... they can get away with it. It isn't just purely wafer costs.

It's not twelve vs. six, it's twelve vs. six times two. They've the same twelve cores, minus one memory controller, queue, uncore and IO, so a 12-core die has a smaller area than two hexa-cores.

Frankly above the E5 1600 and 2600 series there is the E5 4600 and E7 series. If count core is the primary object and space and price is no object then can go higher still.

E5 4600 is not only for quad-processor systems (even more expensive), but also still Sandy Bridge-EX, and E7 isn't due until Q1'14.

Well it actually isn't plastic being stabbed at here but the fact a certain class of "Pros" don't particularly care what the box looks like because they are going to just stuff it in a rack/closet/cabinet/etc. It could have purple flowers on it .. no one is going to see it most of the time.

The creative pro who puts it on his desk. There's the Pro who prefers a great design right there.

Again it goes back to the mindset of it is going in a rack/closet/cabinet/etc in which 802.11ac wouldn't work well in anyway. Not that is where most Mac Pros are going. Just the presumption that a subset of the Mac Pro market is actually the Mac Pro's primary market.

Immagine what happens to the Mac Pro subforums if they'd ship a $3000+ workstation without 802.11ac and no third-party accessory to add it. It's just an essential feature for any computer, and 802.11ac is currently the state-of-art technology.

It is also the notion that have to cover every single possible sub niche to be pro. Can't be a pro swiss army knife if only have 4 function.... keep doubling that and will get to pro knife.

Most of this is a knee-jerk reaction to Apple's adjustments on targeted subsets of broader markets. If their subset is left out somehow the whole pro connotation evaporates.

Nah.

 

[Macsonic]

There is options. From "sneak peek" till new Mac Pro launch folks can choose to buy a current Mac Pro. At least there is a "change is coming" window.

The Mac Pro was on track to being canceled. That there is anything is an option. Spliting the volume between dual tower and new format would only both at greater risk of failure. Likewise cranking up the fraticide is tried overlapping single package model.... both entire line ups would be at risk.

The Mac Pro was not succeeding in current format. If it had been doing well ( as good or better than its mac cohorts ) from 2008-2010 it never would have gone into EU market withdrawal in 2013. There may have been a more gradual transition with 2012-2014 being the transitional years and 2014-2015 the transition to the new form factor, but it probably still would have come.

Hi Deconstruct60. Hope you had a good weekend. To clarify, what I meant of Apple giving us an option to still offer the tower Mac Pro, I was thinking a revised or updated version with USB 3.0 and PCIe 3.0 but not necessarily having Thunderbolt. Maybe the tower design is also different. Sorry this was not clear in my post. But I know this won't happen so I called it my "Wish List"

A couple of months before the 2013 WWDC I decided to buy the current 5.1 Mac Pro as I had a feeling Apple will embed or built in the CPU and GPU so it is possible to have Thunderbolt, less expansion. Could you confirm and verify if it is true that Thunderbolt is only possible if the GPU and CPU are built in to the logicboard? Not sure if this was mentioned in another thread.

Another question and would appreciate your elaborating. A Mac tech told me, with the new 2013 MPro, to hook up an external PCI box using Thunderbolt, you would just get the speed of PCIe 2.0 ? Thanks

 

[thekev]

Not claiming this has optimal sampling demographics but even on geeky macrumors only < 10% of Mac Pro users crack the 64GB barrier. Four DIMM slots is not a huge market barrier. That is one reason why DDR 4 goes there. It isn't particularly to gratuitously kick users in the shins. It is where the mainstream workload is.

It's like I've said before, 64 gigabytes of ram is the geek equivalent of a fur coat, jeweled cane, and spinning rims.

Admittedly that was funnier a couple years ago when 8GB dimms were still ~$500 each.

 

[Giuly]

A couple of months before the 2013 WWDC I decided to buy the current 5.1 Mac Pro as I had a feeling Apple will embed or built in the CPU and GPU so it is possible to have Thunderbolt, less expansion. Could you confirm and verify if it is true that Thunderbolt is only possible if the GPU and CPU are built in to the logicboard? Not sure if this was mentioned in another thread.

If you look closely, you see that the CPU and GPU(s) are on two separate logic boards. It should be safe to assume that they're interconnected via PCIe x16 or x32 plus something to bring either DisplayPort or complete Thunderbolt 2 data back to the CPU board.
Here's the kicker: You can replace the GPU board, and you run into the same issues as you'd run on the old Mac Pro for the first couple of years: Nobody manufactures a compatible third-party card and dual-GPU replacements will be expensive even after they become obsolete. They took away the opportunity of using flashed PC cards, other than that everything is still the same.

Another question and would appreciate your elaborating. A Mac tech told me, with the new 2013 MPro, to hook up an external PCI box using Thunderbolt, you would just get the speed of PCIe 2.0 ? Thanks

The SonnetTech enclosures are PCIe 2.0 x16 per slot, but the maximum speed is 10GBit/s regardless of that.

 

[monaarts]

In our discussions here I've become more and more convinced that the new Mac Pro is what was called the xMac; a small configurable Mac Pro. A mid tower in the parlance of back then.

Thesis: The new Mac Pro is designed more for the prosumer than the Pro, here are the Prosumer friendly choices

• Removed all PCI card options (Prosumer)
• Removed all internal disks (Prosumer)
• Removed 2nd CPU option (Prosumer)
• Removed 8 DIMM slot option (Prosumer)
• Removed all legacy ports (Prosumer)
• Made Small (Prosumer)
• Made Quiet (Prosumer)
• 802.11ac standard (Prosumer)
• Added extensive external ThunderBolt 2 ports (Prosumer)
• Three monitors standard out of the box (Prosumer)
• Zippy lights that turn on when you rotate it around (Prosumer)
• Polished Darth Vader case (Prosumer)
• Low end CPU (TBD)
• Low end Graphics (TBD)

Some may be double counting but it spells out all the Prosumer features added. The last two are the only two pieces missing for this to be a prosumer machine.

What is left for the Pro?

• Dual high end FirePro Graphics (Pro)
• 12 core processor (Pro)

Again my observation is that they had to tip their hand on the Pro at WWDC, so all they did was show off how powerful this computer to keep the Pros from leaving. Left completely unsaid is how low end this can go.

On pricing

• If it starts at $3,000 it's DOA.
• If it's $2,500 it will do OK.
• If it's $2,000 then it will do really well.

I can be completely wrong but I believe the starting price will be $1,999.

I think everything you say makes sense. I think it makes sense for Pro's though and that is going to mainly be because of pricing. At a low starting price, Pro shops will be able to purchase these PC's and upgrade them more often than they have in the past. Especially once you add all the external devices, add-ons, what have you... It will create it's own eco-system that helps influence Mac Pro owners to stay within the Mac family. Increased purchasing with a decrease in profit margin is the way Apple is moving. They'd rather make less money per device and sell you 3 times as many devices to ultimately wind up with a higher profit from you. My prediction is a starting price of $1499 on the new Mac Pro.

 

[deconstruct60]

If you look closely, you see that the CPU and GPU(s) are on two separate logic boards. ... plus something to bring either DisplayPort or complete Thunderbolt 2 data back to the CPU board.

It isn't Thunderbolt. The TB controller has to be 1-2 inches from the physical ports. Routing DisplayPort and PCI-e don't have those limitations.

The SonnetTech enclosures are PCIe 2.0 x16 per slot, but the maximum speed is 10GBit/s regardless of that.

x16 physical. Not electrical. Physical just means the pins go in the socket. It doesn't mean they are connected to anything. The connection is what matters.

"The PCIe slot will physically accommodate up to a x16 PCIe card, however the actual electrical bandwidth of Thunderbolt is x4..."

http://www.sonnettech.com/product/echoexpresschassis.html

 

[Cubemmal]

My prediction is a starting price of $1499 on the new Mac Pro.

While I think $1999 is likely $1499 seems impossible. I think they'd have a hard time getting the BOM to hit that price target (think of it, a cell phone costs $600-$700 when not subsidized) and additionally that would cannibalize their other sales. Why buy a Mac Mini when you can get a Mac Pro for only $500 more? It intrudes too much into their other lines.

Starting at $2k nicely interleaves with their other products.

 

[deconstruct60]

DDR4 has provisions for 3D stacking, up to eight dies are layered on top of each other, interconnected and then enclosed in a single package, so there is no need for more DIMMs as you just increase the capacity via the layers in the chips instead.

Which is a "cheat". Physically that isn't going to work in the new Mac Pro since it changes the widith/height, depending upon your perspective, of the DIMM itself. It will require physically workarounds since the 4 DIMMs have to be close to the CPU but now don't fit next to each other neatly.

Also going to take a latency hit but then multiple ranks always take a latency hit.

If you're running two graphics cards, you're left with two PCIe 2.0 x8 slots which happen to have a total bandwidth of 80GBit/s.

Not sure if this is the "New math" but 40 - ( 2 * 16 ) = 8 ; not 16. And they are actually v3.0 if provisioned off an Xeon E5 (v1 or v2 ).

If you attach two TB controllers which only talk v2 onto two x4 v3 links they will drop down to v2.

The only way to leave 16 lanes is if one of the GPU cards is choked off down to just x8.

Six Thunderbolt 2 ports with 20GBit/s each offer 50% more bandwidth at 120GBit/s.

TB ports aren't additive. You don't keep getting more individual connect bandwith on TB any more an a 1GbE switch gets individual port bandwidth if add more sockets. A port on a 4 or 8 socket 1GbE switch is going to give you 1GbE .

The 20Gb/s in is largely there so can transport 20Gb/s out. Otherwise the daisy chain network isn't going to work so well.

The on/off ramp from the Thunderbolt backbone network for PCI-e data is much smaller than than. Pragmatically it is around x3 PCI-e v2 (theorectically x4 ) which is around 12Gb/s . All this bluster about adding up ports won't see real work performance. You can add controllers not ports. With three controllers have around 36Gb/s total spread out over three separate networks. 120Gb/s is a fantasy from the remote PCI-e controller perspective.

It's just a matter of splitting the x8 PCIe cards into something akin to PCIe x4.

LOL. And if you have isochronous data delivery requirements what then? If the x8 card was hooked to a single device (e.g., 12 RAID drive bays that is single set. ) .... what need two downstream devices on the two x4's ? You also not going to get the full x4 ( probably closer to x3). So each TB controller is x1 drop off as you try to "glue" the bandwidth back together. It doesn't.

TB is not a solution targeted at x8 and up cards. It just isn't. If the cards happen to work at slower bandwidth for some (maybe most ) fine but will be a performance hit.

According to news from yesterday, this is the E5-2600 v2 lineup:

I guess you missed the 2609 v2 and 2643 there that are lower than 8, huh? The minimal 2600 is not 8. The current 2609 and 2643 are both four cores now. In the 2600 v2 line they got a +2 cores ( just like about every else). In the current line up there is a 2637 that has 2 cores. Same formula ...... surprise 4 cores. If there is no 2637 v2 the min probably is 6 not 4 but if there is one it will be what I said 4 cores.

E5 1600 v2 would technically be a fit, but there hasn't been any news on those lately.

Most rumors chases are going to be more interested in the Core i7 variants ( Ivy Bridge E ).

http://www.cpu-world.com/news_2013/...ge-E_extreme_CPUs_to_launch_in_September.html

Just replace E5 1620 1650 1660 for the first column for what the Ivy Bridge EP details.

It's not twelve vs. six, it's twelve vs. six times two. They've the same twelve cores, minus one memory controller, queue, uncore and IO, so a 12-core die has a smaller area than two hexa-cores.

The un-core portion of the die is not redundant. It is also not particularly not that relatively big. L3 cache soaks up a far bigger proportional share than the un-core portion does.

The marginally bigger sure, but it isn't a power / expense thing. Base clock rate and upper end turbo have more impact on power and expense inside the same implementation when hold core count constant (with one or two dies to get to core count).

The creative pro who puts it on his desk. There's the Pro who prefers a great design right there.

The point was there are different groups of folks. Some folks don't need to gaze longingly at their Mac Pro.

Immagine what happens to the Mac Pro subforums if they'd ship a $3000+ workstation without 802.11ac

Again different sets of current users have different demands they put on the Mac Pro.

The presence or not doesn't make it "Prosumer" or not.

Nah.

You say that isn't the definition of "Pro", but for better or worse that is the definition that some use. When different groups decided that words means different things then communication breaks down.

 

[macines]

Consumer Computer?? Prosumer computer ?? very much stuck in the past.

Computers are converging more and more. Theses days you can do work on a macbook air that was called "pro" on a high end computer 5 years ago.

Most people dont even use the power they have in their notebooks. So assuming that pros know how to take advantage of the computing power thats at their disposal - who could possibly need more then 12 cores in one computer these days?(except maybe for NASA computers otl.)

Just because computers don't take a room's space anymore like back in the 80s people didnt stop calling them computers, right?

It's called progress.

 

[deconstruct60]

Could you confirm and verify if it is true that Thunderbolt is only possible if the GPU and CPU are built in to the logicboard?

Same logical/virtual mainboard. There isn't a restriction against a single conceptual mainboard being split into daughter-cards ( multiple pieces), but pragmatically there is a requirement for a GPU that isn't going to disappear later. If there is a GPU in the CPU package then done ( since computer isn't going to work so well if no CPU. ). Otherwise it can be embedded onto the same board in a fashion where it isn't going to be causally removed.

Thunderbolt isn't just "external PCI-e" . There already a standard for external PCI-e. Thunderbolt transports two things: DisplayPort and PCI-e data. If it is not doing both then it isn't Thunderbolt. That means the host PC has requirements that the rest of the devices don't necessarily have. It must be a source of video out. Dual port peripherals must also be able to deliver DisplayPort in backwards compatibility mode ( if not consuming that single to use internall, e.g., Apple TB Display).

to hook up an external PCI box using Thunderbolt, you would just get the speed of PCIe 2.0 ? Thanks

Yes. most TB controllers are hooked to v2 devices. Inside of most laptops/desktops they are hooked to the Intel IOHub/Southbridge chipset which only supplies v2 lanes. The vast majority of FW , SATA , USB 2.0 & 3.0 , 1Gb/s Ethernet , etc. controllers are all PCIe v2 also.

The notion that Thunderbolt was primary designed so that folks could build external PCI-e card enclosures that folks would cards into is deeply flawed. It works in many cases, but that wasn't the primary design objective.

It tends to work better where the external box was already there in the old set up ( e.g., PCIe card + external drive box ) and now become an integrated unit. Likewise where the external box was already there and now shipping more formerly internal PCI controllers to that external box (e.g, ship FW , 1GbE , and USB controllers out to external display ) .

Things that were full size , full height , large bandwidth cards ( x8 very high end RAID card, x8 or x16 Inifinband , x16 GPGPU card ) never were a primary target.

 

[deconstruct60]

Consumer Computer?? Prosumer computer ?? very much stuck in the past.

Computers are converging more and more. Theses days you can do work on a macbook air that was called "pro" on a high end computer 5 years ago.

.....- who could possibly need more then 12 cores in one computer these days?(except maybe for NASA computers otl.)

It is for more whether there has been progress in users workload to match the progress in the computer ability to cover that workload.

That is a completely and utterly different thing from professional or not. If your computational workload is not increasing but still getting paid to do work then..... it is still professional.

There are some folks whose computation workload is going up. If it goes up about at the same rate the computers improve then they will stay in the same class of machines. If had old workstation workload 5 years ago and that increased as fast (or faster) than workstations have then require another workstation.

For example only handled standard def TV 5 years ago and now only handle 4K . Likewise if was doing standard def TV and now only do 720p HDTV then maybe not.

Generally, over time more users "move down" because doing the same things as they were 5-10 years ago. That isn't quite as widespread in the workstation space, but it is a factor. One trick is to take folks that were higher up and catch them as they move down. Another is to find new high value added apps/computations that require more workload.

 

[Jakimo]

• Removed all PCI card options (Prosumer)
  
  Disagreed. Many pro media devices are switching to Thunderbolt, most Thunderbolt expansion products include PCI for those who need it.
  
• Removed all internal disks (Prosumer)
  
  I've met very few professionals who even know they can put more than one hard drive in their current Mac Pro. Internal disks are pointless today. Media companies store everything on networked drives and servers. The only people that want lots of internal space are prosumers that don't need to invest in systems that enable working in groups.
  
• Removed 2nd CPU option (Prosumer)
  
  First, we don't know this 100%. Second, the previous model used two CPU's to get to 12 cores. 12 cores are now available on a single CPU.
  
• Removed 8 DIMM slot option (Prosumer)
  
  This is a somewhat dated option anyway. Those that need more RAM will buy it on single DIMM's. Additionally, the speed of the SSD will relieve the system when it does page out.
  
• Removed all legacy ports (Prosumer)
  
  Get over it. Legacy options are still supported with adapters. Why would you buy this machine if you're planning on using ancient peripherals with it anyway?
  
• Made Small (Prosumer)
  
  So what? If you can accomplish the same thing in a smaller package why wouldn't you? There are lots of advantages to being smaller and lighter, even in a pro machine. We do move things around sometimes.
  
• Made Quiet (Prosumer)
  
  This is an exceptional feature, especially for audio pros. I don't know what kind of pros you think use this machine, but I can tell you that the ones it's designed for will all appreciate a noise reduction.
  
• 802.11ac standard (Prosumer)
  
  So you're suggesting that wireless connectivity that's as fast as ethernet doesn't benefit me as an actual professional? First, I can move my machine wherever I want it. Second, it eliminates the need for yet another cable in the rats nest that already exists. Third, 802.11ac is faster, that's always good--for everybody. Are you suggesting that because I'm a pro I would prefer slower wireless connections?
  
• Added extensive external ThunderBolt 2 ports (Prosumer)
  
  Have you seen Thunderbolt peripherals? Prosumers simply don't buy these. Pros want expandability and modularity. Thunderbolt gives us that.
  
• Three monitors standard out of the box (Prosumer)
  
  Additional monitors increase efficiency. Especially for media related tasks. Again I ask what exactly do you think people do with these machines?
  
• Zippy lights that turn on when you rotate it around (Prosumer)
  
  Being able to see what's what without getting out a flashlight. You make no sense.
  
• Polished Darth Vader case (Prosumer)
  
  Pros like stuff that looks cool too.
  
• Low end CPU (TBD)
  
  We'll see about that.
  
• Low end Graphics (TBD)
  
  We'll see about that too.

As an actual professional, that uses these machines, I'm getting really sick and tired of seeing people assume what it is that I need or want to implement in my business. Particularly when they clearly have no idea what a professional workflow is like.

Excellent response. I'm stunned at how slow people are to accept change that's forward thinking.

Outstanding machine, my ONLY concern is where in the world do I put my coffee with no CD tray?!

 

[monaarts]

While I think $1999 is likely $1499 seems impossible. I think they'd have a hard time getting the BOM to hit that price target (think of it, a cell phone costs $600-$700 when not subsidized) and additionally that would cannibalize their other sales. Why buy a Mac Mini when you can get a Mac Pro for only $500 more? It intrudes too much into their other lines.

Starting at $2k nicely interleaves with their other products.

An extra $500 is a 50% increase - that is the point. People who don't need the power will save $500 while the people who want power will spend the money for extra power and more ability to add to it.

 

[macines]

It is for more whether there has been progress in users workload to match the progress in the computer ability to cover that workload.

That is a completely and utterly different thing from professional or not. If your computational workload is not increasing but still getting paid to do work then..... it is still professional.

There are some folks whose computation workload is going up. If it goes up about at the same rate the computers improve then they will stay in the same class of machines. If had old workstation workload 5 years ago and that increased as fast (or faster) than workstations have then require another workstation.

For example only handled standard def TV 5 years ago and now only handle 4K . Likewise if was doing standard def TV and now only do 720p HDTV then maybe not.

Generally, over time more users "move down" because doing the same things as they were 5-10 years ago. That isn't quite as widespread in the workstation space, but it is a factor. One trick is to take folks that were higher up and catch them as they move down. Another is to find new high value added apps/computations that require more workload.

Yes, I understand that.

And thats exactly what is happening.

Hardware is developing way faster than software is. (just look at the smart phone market)

And then within the next two years there'll be 24 core processors for desktop machines.

If a computer is not working properly its rather the lack of software optimisation than the lack of computing power.

Arent macs known for their usability and exactly that OS every "professional" is looking for?

So why not take a leap in a new direction, a direction that will define the desktop computer branch for the next 5-10 years.

I can see that many are scared because this maybe comes a little early, but in two years or so, all of us will love the risk apple is taking with this - which always comes with innovation.

 

[Cubemmal]

An extra $500 is a 50% increase - that is the point. People who don't need the power will save $500 while the people who want power will spend the money for extra power and more ability to add to it.

But that's not what Apple is thinking. They have enough products at $1500 already, the MBP and the iMac. Why try to put the Mac Pro there? Marketeers try to create a spread of products and prices to capture as much market share as there is. In this case they need a premium product and that is the Mac Pro. At $2500 base model it is outside the comfort zone of most Prosumers (except the well heeled). It rather sticks out like a sore thumb and is a niche product.

If they can simply build a mac with a BOM that allows a $2k base price, as I believe the new Mac Pro does, then they can bring in much more of the higher end mac buyers and turn the product into a mainstream computer. Look at the 17" MBP, they killed it because there were few buyers at $2400+. They have graphs of this and I bet that $2k captures 25%+ of the market, where $2.5k catpures < 1%.

If they push it down to $1.5k they simply cannibalize the iMac and MBP market instead of capturing more of the $2k market, which is only served by the Retina MBP (presently) and iMac. I won't buy either so they aren't getting my dollars, but a $2k MP would.

Just Marketing and pricing 101 ...

 

[monaarts]

But that's not what Apple is thinking. They have enough products at $1500 already, the MBP and the iMac. Why try to put the Mac Pro there? Marketeers try to create a spread of products and prices to capture as much market share as there is. In this case they need a premium product and that is the Mac Pro. At $2500 base model it is outside the comfort zone of most Prosumers (except the well heeled). It rather sticks out like a sore thumb and is a niche product.

If they can simply build a mac with a BOM that allows a $2k base price, as I believe the new Mac Pro does, then they can bring in much more of the higher end mac buyers and turn the product into a mainstream computer. Look at the 17" MBP, they killed it because there were few buyers at $2400+. They have graphs of this and I bet that $2k captures 25%+ of the market, where $2.5k catpures < 1%.

If they push it down to $1.5k they simply cannibalize the iMac and MBP market instead of capturing more of the $2k market, which is only served by the Retina MBP (presently) and iMac. I won't buy either so they aren't getting my dollars, but a $2k MP would.

Just Marketing and pricing 101 ...

But what you are saying isn't marketing and pricing 101. First off, the MBP has an entirely different purpose and target market. Regarding the iMac, it isn't the same price. I am talking about a tower being priced at $1500 without a display whereas the iMac is roughly the same price with one. I am also talking about a low spec MacPro vs. a high spec iMac. At the end of the day neither of us know what will happen. We are both speculating from different sides.

 

[deconstruct60]

Hardware is developing way faster than software is. (just look at the smart phone market)

Somewhat but how times has Apple got up at WWDC and said that OpenCL , GCD , etc were the future. (not CUDA , no just a couple of cores, etc) ? What Three or four ? The first couple of time yeah sure that gets put in the "yeah we'll do that later" bucket. 3-4 later ... it is a heck of lag.

Usually the lag is far more motivated by lack of hardware infrastructure penetration. For 2013 all Macs can do OpenCL . that is 14M boxes a year your software is ignoring. Go ahead. Maybe competitors will do exactly the same thing.

And then within the next two years there'll be 24 core processors for desktop machines.

It isn't about complex x86 cores. That is one of the major points. To some extent it also isn't about complex ARM cores either. That is a point that will develop to a bigger gap in a couple of years.

Arent macs known for their usability and exactly that OS every "professional" is looking for?

I think more known for getting out of the way so don't have strange nobs and buttons to push due to quirky behavior.

So why not take a leap in a new direction, a direction that will define the desktop computer branch for the next 5-10 years.

The iMac probably far more so defines desktop computing over next 5-10 years. The new Mac Pro is carve out a subset of the workstation market that is going to grow. It is the subset where more change and less "sunk cost" focus (at least inside the box itself) is going to be.

I can see that many are scared because this maybe comes a little early, but in two years or so, all of us will love the risk apple is taking with this - which always comes with innovation.

Folks probably would have liked a transitional progression. Sort of like the log jam of 13" Mac laptops right now. That is much tougher to pull off in the Mac Pro space.

If the Mac Pro has been going full speed then 2012 would have been last new tower design and then 2013 could have been the overlapping transition year. Or perhaps do the major shift on the E5 v3 transitional move.

Well see if this is too early or not. If the prices spike .... it is way too early to make this move. If about the same it is aggressive but about right.

 

[Macsonic]

Same logical/virtual mainboard. There isn't a restriction against a single conceptual mainboard being split into daughter-cards ( multiple pieces), but pragmatically there is a requirement for a GPU that isn't going to disappear later. If there is a GPU in the CPU package then done ( since computer isn't going to work so well if no CPU. ). Otherwise it can be embedded onto the same board in a fashion where it isn't going to be causally removed.

Thunderbolt isn't just "external PCI-e" . There already a standard for external PCI-e. Thunderbolt transports two things: DisplayPort and PCI-e data. If it is not doing both then it isn't Thunderbolt. That means the host PC has requirements that the rest of the devices don't necessarily have. It must be a source of video out. Dual port peripherals must also be able to deliver DisplayPort in backwards compatibility mode ( if not consuming that single to use internall, e.g., Apple TB Display).


Yes. most TB controllers are hooked to v2 devices. Inside of most laptops/desktops they are hooked to the Intel IOHub/Southbridge chipset which only supplies v2 lanes. The vast majority of FW , SATA , USB 2.0 & 3.0 , 1Gb/s Ethernet , etc. controllers are all PCIe v2 also.

The notion that Thunderbolt was primary designed so that folks could build external PCI-e card enclosures that folks would cards into is deeply flawed. It works in many cases, but that wasn't the primary design objective.

It tends to work better where the external box was already there in the old set up ( e.g., PCIe card + external drive box ) and now become an integrated unit. Likewise where the external box was already there and now shipping more formerly internal PCI controllers to that external box (e.g, ship FW , 1GbE , and USB controllers out to external display ) .

Things that were full size , full height , large bandwidth cards ( x8 very high end RAID card, x8 or x16 Inifinband , x16 GPGPU card ) never were a primary target.

Thanks Deconstruct60. It's clearer now.

 

[Giuly]

The notion that Thunderbolt was primary designed so that folks could build external PCI-e card enclosures that folks would cards into is deeply flawed. It works in many cases, but that wasn't the primary design objective.

It tends to work better where the external box was already there in the old set up ( e.g., PCIe card + external drive box ) and now become an integrated unit. Likewise where the external box was already there and now shipping more formerly internal PCI controllers to that external box (e.g, ship FW , 1GbE , and USB controllers out to external display ) .

Things that were full size , full height , large bandwidth cards ( x8 very high end RAID card, x8 or x16 Inifinband , x16 GPGPU card ) never were a primary target.

Nevertheless, the list of professional cards that work with the Thunderbolt 1 enclosure is impressive. Thunderbolt 2 should add even more cards to that list.

 

[-hh]

I normally sit out on these types of threads... But... ...
2) Only 4 DIMMS? OMG so what? 64GB (4x16GB) is only $580. 128GB (4x32GB) is $1300. If you need that much memory, then is price really an issue? No, not in my field.

If I could match your prices, I'd tend to agree. However, I'm finding that 4x16 is presently $880 and 4x32 is $3500.

3) "Made Quite" is not a Pro feature? LOL okay.. I've been in no less than 12 studio "sound booths" that have the hardware in the room with them.

I had some concerns with this claim too, but more because this is a "less is always better" and because part of the debate with the Tube is its potential for higher noise levels because the now-external expansion devices will each require their own subsystem cooling.

FYI, I recently pulled out one of our lab's sound meters to quickly gage how much the ambient noise level increased on one of my MP configurations (w/4 internal HDDs spinning) when a pair of external drives were also turned on: it went from 58dB to 60dB at ~0.5m (location of operator's head). However, I did note that the meter had an effective floor of 60dB and getting readings was a bit erratic...I want to rerun this with another meter (probably a non-portable one), as I really expected the ambient floor to have been a good 5dB lower than reported.

7) Single CPU only.. I'm not sure on this one yet. In general I don't think this is a big deal. 12 core is pretty nice. 24 even better for rendering. But again, if you are doing work that needs more, wouldn't you have the money for a nice little Mac Pro farm?

True, but this presumes a workflow which can utilize a render farm...please provide guidance as to how this would be set up in Adobe Photoshop CS6. In any case, which really is illustrating that not all users are the same: there will be some Mac Pro use cases where several TB of fast local storage is more beneficial to the workflow productivity than having more cores.

Originally Posted by ssgbryan
Which matters not one bit if the software I use doesn't take advantage of them.

Which won't matter much if the OS libraries those apps leverage them transparently. If Apple dumps more of the standard graphics workload off the CPU onto the GPUs there is no major rewrites required.

An interesting assertion, but since GCD has been baked into the OS X Golden Master since 2009 (Snow Leopard), would not the counterpoint be that this source of performance gains should have had already happened?

One reason software has slow roll out is that most deployed systems( and users ) don't have the underlying foundation. If vast majority of folks have 2 cores not going to get 6-8 core optimized from a wide spectrum of software vendors. Similarly if 50% of all Macs have less than 2GB memory in them and running 32 bit kernel mode then most won't be in big hurry to get to 64 bit ( to soak up more space to address less than 2GB ? not a big winner).

True, but this appears to be disregarding the customer-based self-selections which occur and which also very strongly apply to the "Power" niche.

For a specific example, let's take your notional example of an observation that ~50% of all deployed Macs have <2GB RAM.

For the developer of Power Niche Software N, is this metric really the one that's most appropriate for them to plan to? Not necessarily: what's significantly better is for them to know is what their paid customers have (plus where they see as growth markets), because if 98% of their customers have >4GB, then they may make an informed decision to ignore the fact that 50% of the overall population because it represents only 2% of their customer base.

If one enjoys an automotive analogy, the manufacturer of high performance tires is going to make sure that he makes the sizes that his fifty thousand Porsche, BMW & {etc} customers have been buying .. he simply does not need to care that the automotive marketplace also includes 20 million Buicks because that customer demographic has zero history or interest in ever buying his specialty product.



-hh

 

[Cubemmal]

First off, the MBP has an entirely different purpose and target market.

No it doesn't. A customer wants a Mac, that is the target market. Like all successful companies Apple has created a product line. Look inside a Mini, iMac or MBP and what do you see? The same thing! Same chips, same capabilities, just minor differences and a different board layout. They just take the same ideas and repackage them. A different purpose and target market would be between the iPhone and the Mac. Within the Mac line they're essentially the same, with variations to fit whatever specific the customer wants. The MBP can certainly act as either a desktop or mobile regardless.

Regarding the iMac, it isn't the same price. I am talking about a tower being priced at $1500 without a display whereas the iMac is roughly the same price with one. I am also talking about a low spec MacPro vs. a high spec iMac.

You're comparing a Xeon with dual GPU's against a monitor with a consumer CPU inside which is the largest difference between the Macs. Any way you cut it given the BOM on the new Mac Pro the chances of it coming in at $1.5k is practically zero IMO.