Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

As we get closer Mac Pro - peoples thoughts

TB2 channel bonding is 20Gbits of bandwidth per device, a PCIe 16x 3.0 slot is 126Gbits of bandwidth per device

new MacPro has 6 ports but only 3x TB2 controllers aggregated off only 12 lanes of Pcie 2.0 which is a TOTAL bandwidth of 48Gbits

The total bandwidth of ALL the new MacPros TB2 ports is less than a single 16x 2.0 slot from the original 2006 MacPro

When you consider your sharing that bandwidth just to drive your monitors, and the additional overhead for thunderbolt, it it not very appealing.

The bandwidth for dual e5-2600 workstation board is 80lanes pcie3.0, most limited to 72 lanes for space, which is bandwidth of 567Gbits/sec

That is not "most"

In a workstation environment, I'll take 567Gigabits/sec bandwidth with cards that "suck" over 48Gbits any day.

and cards don't suck, Cards that are poorly designed oems and cost to "repair" them is the same as what you purchased them for - they suck.

MacPro will be fine for photoshop users, designers, etc. Those who process large amounts of data, need fast I/O - no, you were forgotten about long ago on infinite loop.

Its vaguely interesting for on location work, its easier to fly with a few small boxes than one giant one, but a small pc is just a good and you can add SAS for LTO and Raid cards which wont have bottlenecks.




M Most said:
I've said it before, but I'll say it again: Cards suck. Thunderbolt 2 is the first one to come along that provides most of that speed and none of the fragility of slots. Cards are a legacy design, born in an era when there was no alternative. That era is ending, and I for one am happy to see that happening.
Click to expand...
 
Christopher*Seguine said:
TB2 channel bonding is 20Gbits of bandwidth per device, a PCIe 16x 3.0 slot is 126Gbits of bandwidth per device

new MacPro has 6 ports but only 3x TB2 controllers aggregated off only 12 lanes of Pcie 2.0 which is a TOTAL bandwidth of 48Gbits

The total bandwidth of ALL the new MacPros TB2 ports is less than a single 16x 2.0 slot from the original 2006 MacPro

When you consider your sharing that bandwidth just to drive your monitors, and the additional overhead for thunderbolt, it it not very appealing.

The bandwidth for dual e5-2600 workstation board is 80lanes pcie3.0, most limited to 72 lanes for space, which is bandwidth of 567Gbits/sec

That is not "most"

In a workstation environment, I'll take 567Gigabits/sec bandwidth with cards that "suck" over 48Gbits any day.

and cards don't suck, Cards that are poorly designed oems and cost to "repair" them is the same as what you purchased them for - they suck.

MacPro will be fine for photoshop users, designers, etc. Those who process large amounts of data, need fast I/O - no, you were forgotten about long ago on infinite loop.

Its vaguely interesting for on location work, its easier to fly with a few small boxes than one giant one, but a small pc is just a good and you can add SAS for LTO and Raid cards which wont have bottlenecks.
Click to expand...
+1 it just seems like were going backwards...I mean if you put your cards in and leave them in the tower in the proper optimized slots, why ever touch them at all? M Most you have proved me wrong many times before, perhaps you will come out on top in this discussion as well...but it just doesn't seem very appealing to me....the new mac pro is like the mac mini pro...I think the whole point people bought mac pros/towers was for the expandability to begin with...otherwise why not use a mobile machine?
 
I'm not sure how expandable the old Mac tower really was, with essentially a mATX board that had at best a Gen2 x16 / x8 / x4 / x4 PCIe slots, and very limited bus power available . I think they just wanted a more powerful super Mac mini that could share accessories with the iMac and MBPR.

Apple basically makes the same profit on each device, 10 million iPhones is better than 10,000 Mac Pros in that regard, but hopefully apple realized it's the 10,000 Mac Pros that help sell 10 million iPhones.
 
Last edited:
Brian Merlen said:
+1 it just seems like were going backwards...I mean if you put your cards in and leave them in the tower in the proper optimized slots, why ever touch them at all? M Most you have proved me wrong many times before, perhaps you will come out on top in this discussion as well...but it just doesn't seem very appealing to me....the new mac pro is like the mac mini pro...I think the whole point people bought mac pros/towers was for the expandability to begin with...otherwise why not use a mobile machine?
Click to expand...

This isn't a contest. I'm just expressing an opinion, not trying to start, finish, or "win" an argument. I don't care about "coming out on top," coming out on bottom, or any variation thereof. But for the sake of conversation, the new mac pro as currently described (I would reiterate that it doesn't exist quite yet...) is a hell of a lot more than a "mac mini pro," just as Final Cut Pro X is a lot more than "iMovie Pro." There is a completely new internal architecture for memory, storage, GPU addressing, and a number of other things (including a new OS) that all have the potential of delivering throughput that is pretty impressive. Sticking with legacy designs would not permit those things, which is at least one reason they didn't do that. Expandability was always only one reason for the towers. Pure power was always the primary one. For me, expandability was never as important as flexibility, and no, they're not the same thing. It was always more about adding functionality than having "more" of what was already in the box. Doing that with modules is no different than doing it with cards as long as the connectivity bandwidth required for that particular function is not an issue. For many things, like video, I/O, and others, it isn't an issue at Thunderbolt 2 speeds. For other things, like high speed processing of large image formats at accelerated speeds, it might be, and it might not be. Time and experience will reveal those answers, not guessing based on past designs.
 
@ Wixom right but as pointed out above that single 16x provides more throughput then all the TB2 ports combined in the new system... which is why us on set DITs might be a little worried about the viability of this workflow. Granted I appreciate the savings on weight in the form factor, but not if it means I need a ton of external solutions with cords and in the end I still am trying to work around the less throughput situation... and Mike no argument here, simply a discussion of preferences and I have since came around on the "IMovie pro" debate. I don't use it and am Avid guy, but it certainly has improved a lot! I am just a little bummed out on what I see as a possible issue with the new machine. I just wish we had options...like lose the mac mini, keep this mac pro, and keep a tower for guys who need one...
 
Last edited:
Brian Merlen said:
right but as pointed out above that single 16x provides more throughput then all the TB2 ports combined in the new system... which is why us on set DITs might be a little worried about the viability of this workflow. Granted I appreciate the savings on weight in the form factor, but not if it means I need a ton of external solutions with cords and in the end I still am trying to work around the less throughput situation...
Click to expand...

You keep guising this as a "ton of external solutions." I would suggest asking the question, exactly what cards does your typical system currently carry, and how many of those are still needed with the new machine and upcoming software? And also ask the question, how many machines do you currently carry and how much space and weight do they consume? If I'm guessing correctly, the cards are probably one or two GPUs, an SDI card, a storage HBA, and possibly a Red Rocket or two. Out of those, the additional GPU's may not be necessary at all, the SDI card gets replaced by a rather small box, the storage theoretically won't need an HBA, and the Red Rockets may or may not be necessary going forward, depending on what you do and how fast you have to do it. But even if you do need them, the overall space taken up by even two of the new machines with an expansion chassis on each is likely to be considerably less than two current towers, and you still need cables coming out of the towers for the video I/O and the external storage. I just don't see where this amounts to "a ton" of external boxes. It amounts to one, and possibly two. And a host with a fraction of the form factor of the current one, and with far faster memory and internal storage.
 
This is going to be way interesting when it actually happens. I wonder how many people will be surprised, pleasantly or otherwise. I, for one, have pretty much no idea about the specifics, but - and I know there are many APple detractors who'll vigorously disagree with this - it's hard to imagine the guys at Apple have designed this machine in a total vacuum with no idea about what other technology is being developed by Intel, nVidia, etc. And, Apple being the dog rather than the tail, other software and hardware companies are usually pretty eager to adapt to whatever it is Apple is doing.

I am, of course, optimistic, but also I have recently enhanced my 2010 Pro Tower to be as relevant as possible even after the new cylinder comes out.

Exciting times, these! :-)
 
well all the drives now need to live externally, the capture card, the LTO, any GPU(s) if needed, rocket(s)....but the main issue is letting this plus monitoring work together without causing issues. granted it might work, but it just seems like based on the math above there might be issues. I frequently need 3 drives, plus LTO, it really depends what I am doing. I work on a variety of systems and workflows as well as cameras. I love OSX and I really hope that the Mac Pro works for my needs, believe me I want it to! I just worry that it'll be like there other systems where they compromise on performance for form factor...as you mentioned tho its all just conjecture right now. but heck its fun to talk on RU, better then sitting here bored between takes lol! I enjoy our discussions immensely, and I hope for both our sakes your correct and I am wrong! If this small thing can really replace my sore back problems then I am 100 percent for it!
 
A 2160p or 1600p main monitor, 2 1080p side monitors, and a 6g hdsdi output - you've maxed your thrunderbolt throughput.

What about storage? Single Internal PCIe ssd, all mass storage is external. Your bandwidth limited already.

What about GPU compute? 2x proprietary format - claimed 7 teraflops at least on the high end config. sounds impressive on paper compared to a single 4.5 teraflop titan, if your app uses opencl well. Cuda, your out of luck.

Ram? 4 slots. Pretty limiting or at the least very expensive.

Interconnectivity - 2 x 1gbe ethernet ports? 10Gbt and I would have said ok, now they are pushing the market forward.

This was a design exercise: Make a "workstation" that people want to put on their desk, not under it.
Yes, make a mac mini pro.
Builds apple brand recognition in business,education, and higher end home markets to drive more customers to their consumer products.

Its a nice design, but for compute or bandwidth intensive applications - film post - doubt it.

M Most said:
You keep guising this as a "ton of external solutions." I would suggest asking the question, exactly what cards does your typical system currently carry, and how many of those are still needed with the new machine and upcoming software? And also ask the question, how many machines do you currently carry and how much space and weight do they consume? If I'm guessing correctly, the cards are probably one or two GPUs, an SDI card, a storage HBA, and possibly a Red Rocket or two. Out of those, the additional GPU's may not be necessary at all, the SDI card gets replaced by a rather small box, the storage theoretically won't need an HBA, and the Red Rockets may or may not be necessary going forward, depending on what you do and how fast you have to do it. But even if you do need them, the overall space taken up by even two of the new machines with an expansion chassis on each is likely to be considerably less than two current towers, and you still need cables coming out of the towers for the video I/O and the external storage. I just don't see where this amounts to "a ton" of external boxes. It amounts to one, and possibly two. And a host with a fraction of the form factor of the current one, and with far faster memory and internal storage.
Click to expand...
 
Christopher*Seguine said:
A 2160p or 1600p main monitor, 2 1080p side monitors, and a 6g hdsdi output - you've maxed your thrunderbolt throughput.

What about storage? Single Internal PCIe ssd, all mass storage is external. Your bandwidth limited already.

What about GPU compute? 2x proprietary format - claimed 7 teraflops at least on the high end config. sounds impressive on paper compared to a single 4.5 teraflop titan, if your app uses opencl well. Cuda, your out of luck.

Ram? 4 slots. Pretty limiting or at the least very expensive.

Interconnectivity - 2 x 1gbe ethernet ports? 10Gbt and I would have said ok, now they are pushing the market forward.
Click to expand...


Theory: Everything you said.

Reality: I don't run specs. I don't run teraflops. I don't run slots. And frankly, I don't run bandwidth. I run applications, and until I can run them and test them, I'm making no guesses or prognostications on how they will run on Apple's new hardware. Once I can, I can make a sensible, informed judgement as to whether these machines are useful for what I have to do.
 
Brian Merlen said:
+1 it just seems like were going backwards...I mean if you put your cards in and leave them in the tower in the proper optimized slots, why ever touch them at all? M Most you have proved me wrong many times before, perhaps you will come out on top in this discussion as well...but it just doesn't seem very appealing to me....the new mac pro is like the mac mini pro...I think the whole point people bought mac pros/towers was for the expandability to begin with...otherwise why not use a mobile machine?
Click to expand...

Look at it this way... With the old/current Mac Pro tower design, you had 4 slots. If you added dual GPUs, eSATA/ SAS card and a Rocket, you were full. One of those GPUs had to be single-width if you wanted to use all 4 slots. Or you went with single GPU and then had an extra slot free and most people ended up adding in a DeckLink or AJA I/O card or a USB3 host. There are PC's out there that offer more slots, but none have more PCIe lanes to allocate unless you went with kludgy server boards with cascaded PCIe controllers that added more compromises than additional slots and lanes.

With the Xeon E5, the PCIe controllers have been moved into the CPU package, so there is no more cascaded chipsets for additional PCIe lanes. In systems like the Z820 or T7600, you have to install both CPUs if you want to use all the slots in the system. We have PCIe 3.0 now and it sure is fast, however there are no cards out there that can actually saturate a PCIe v3 X8 slot, let alone an X16 slot.

The new Mac Pro gives us high-speed FLASH storage connected direct to PCIe, no ~580MB/s per link speed limit of SAS/SATA. It gives us dual PCIe v3 GPUs, it gives us multiple USB3.0 ports (dual headers if I interpret it correctly). And then there are 6 TB2 ports at 20Gbps each. Roughly the equivalent of 6 x PCIe v2 X4 slots.

I'm not going to do the Apple fanboy thing and tell you all that the Mac Pro is the end all, be all of workstations. It's not. There are plenty of PC configurations that have more to offer in terms of cores and even expandability to some extent. The big thing to take away from this here is that we're seeing a paradigm shift of sorts. PCI cards are no longer the only way to expand a system and in many cases are not going to be the most desirable way to do it as we move into the future. A DeckLink or AJA I/O card doesn't need to be a card at all. It can be a Thunderbolt connector to a small box with a variety of I/O ports and the necessary electronics that would usually be on the card can be incorporated inside this box in a way that is more applicable and ergonomic. PCIe expansion boxes are a kludge solution at best and are not the proper way to handle peripheral expansion as this design progresses forward. Furthermore, the GPUs inside the new Mac Pro are not a proprietary design. This is an open spec that has been jointly agreed upon by AMD/ATI and Intel. No word yet as to whether or not nVidia is onboard. The hope here is that the smaller GPU form factor will be adopted by more PC makers in the future as it offloads the display interface from the GPU cards and allows them to be smaller and for more of them to be added into systems for computational purposes.

I mostly agree with M Most on his view of PCIe cards. Cards suck. Sure, they're plenty reliable if you stick them in a system and never touch them again. Or at least they are until they fail. Not so common for them to fail, OTOH, look at all the reports of dead Rocket cards on here.

As for locking connectors, there are times when you want them and times you don't. DisplayPort does offer locking connectors, however, Mike the Fireman Ross is correct, mini displayport is indeed a consumer-grade connection and non-locking. In most situations and installations, the active components in the ends of TB cables are going to fail vs. a physical problem with the connector itself. Locking connectors are generally unwanted unless you're tying together components in a rack. I've lost track of all the damaged equipment I've seen over the years due to cable snags and "professional" locking connectors. Most of the time the equipment damage can be avoided if the cables are secured properly, as it tied down to other support past the point of connection, but even then it doesn't always save you.

All things considered, I think the new Mac Pro is a very intriguing and even a nice design for a desktop workstation. If the price is right, it will make a nice small server or render "appliance" system. The promise of TB2 used as a 20Gbps interconnect fabric to build a lattice network for computational grids is huge and every system has that right out of the box.

I agree with others here that the new Mac Pro could really use an "expansion port". Would be great if they piped a PCIe v3 X16 slot out the rear of the unit that could be connected to an expansion chassis. PCIe cards are not going away anytime soon... OTOH, there really are no free PCIe lanes to give unless they add another CPU.


I'm hoping Apple brings back the XServe. A 1U blade style compute node and server could do real well here. Would love to see it with dual CPUs and quad GPUs (of the new small form factor), onboard FLASH storage, a few TB2 ports and dual 10G-E ports built in. Give it one PCIe slot for adding a RAID controller.
 
Laco Zamba said:
One more new thing in Mavericks - IP over Thunderbolt
Don't now much about it, but sounds promissing
Click to expand...

It works awesome. TB to TB is a bit faster than 10G ethernet with lower latency and overhead. And the speed per connection doubles with TB2 coming up here as each connection can have access to the full 20Gbps on the TB port.

There are Thunderbolt hubs/ switches forthcoming. A system with multiple Thunderbolt ports can act as a switch or router. And Thunderbolt hubs and switches can connect multiple types of Thunderbolt devices and peripherals, not just serve as a network device.
 
Fair enough, except hardware is not really a grey area. Its has 48gbits of throughput, thats it.

There have been no technical design advancements, same lga2011 socket, gpus on same pcie interconnect, same memory bus. They can be proprietary, they dont have to worry about compatibility. Where is the innovation of the past? The improvements are purely physical and cosmetic, and that is the really sad part. Your apps will run the same.

I cant believe you said you don't care about bandwidth. Your shooting 4 epics and a phantom, you know red ssds max out at x and phantom 10gbe maxs out at y, you need to turn around all cards as soon as possible because they never rent enough, download to a raid-6 storage, make lto6s, do a color pass and deliver a dnx to editorial . Its all about bandwidth, either you have enough or you have very unhappy people who cant run their apps.

By calling it pro and making it single socket, Apple is impeding the needed further development in parallel computing of professional applications, which is definitely taking us backwards.

Great if a small stylish computer with limited bandwidth works for your apps, it will look cute as a carry on.


M Most said:
Theory: Everything you said.

Reality: I don't run specs. I don't run teraflops. I don't run slots. And frankly, I don't run bandwidth. I run applications, and until I can run them and test them, I'm making no guesses or prognostications on how they will run on Apple's new hardware. Once I can, I can make a sensible, informed judgement as to whether these machines are useful for what I have to do.
Click to expand...
 
Christopher*Seguine said:
Fair enough, except hardware is not really a grey area. Its has 48gbits of throughput, thats it.

There have been no technical design advancements, same lga2011 socket, gpus on same pcie interconnect, same memory bus. They can be proprietary, they dont have to worry about compatibility. Where is the innovation of the past? The improvements are purely physical and cosmetic, and that is the really sad part. Your apps will run the same.

I cant believe you said you don't care about bandwidth. Your shooting 4 epics and a phantom, you know red ssds max out at x and phantom 10gbe maxs out at y, you need to turn around all cards as soon as possible because they never rent enough, download to a raid-6 storage, make lto6s, do a color pass and deliver a dnx to editorial . Its all about bandwidth, either you have enough or you have very unhappy people who cant run their apps.

By calling it pro and making it single socket, Apple is impeding the needed further development in parallel computing of professional applications, which is definitely taking us backwards.

Great if a small stylish computer with limited bandwidth works for your apps, it will look cute as a carry on.
Click to expand...

I didn't say I don't care about bandwidth. I said I don't "run" bandwidth. What I was pointing out is that numbers based on legacy designs don't always apply identically when it's not a legacy design.

As for your proposed scenario, I would argue that you should probably be using multiple computers and shared storage if you have do everything that you listed. I don't happen to be a DIT or a data wrangler, so I don't try to multitask to that level with one CPU, and quite frankly, I don't think others should, either.
 
One other cool thing about thunderbolt... 100m runs at full bandwidth (using fiber), though you might need to sell a kid to be able to afford the cables.
 
I did notice the new Asus Z87 Deluxe/Quad was released today, it has TB2.
 
In terms of technology, it is a legacy design, just in a different physical shape.

I tried out 10.9. It still only supports 8bit per channel color.

Ridiculous that they add firepro graphics to the new macpro and still do not support this basic feature of workstation level graphics cards.



M Most said:
What I was pointing out is that numbers based on legacy designs don't always apply identically when it's not a legacy design.
Click to expand...
 
You must log in or register to reply here.
Back
Top