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

Thunderbolt Vs Upgradeable GPU + PCIe slots?

  • Thunderbolt ports + Proprietary, non-upgradeable GPUs, NO free PCIe slots [new Mac Pro]

    Votes: 61 32.4%
  • Four PCIe 3.0 slots sharing 40 lanes with NO thunderbolt at all

    Votes: 127 67.6%

  • Total voters
    188
@wonderspark,

USB3 should be able to deliver the needed speeds for your 8-drive array tho. Two 7200 drives (in RAID0) on each of the four USB3 ports again RAID0'ed is 8 drives and no bottleneck. No?

TB2 is the same I would imagine. It's going to be very close to 2GB/s (either per controller or per port - I'm not clear on that). So if you sold your RAID card for the $800 I think it goes for used then you could probably buy two 4-Drive RAID enclosures for TB2 with that money. So there's your 8-drives again and nowhere near using up all the bandwidth of even one controller. If you put SSDs in all of those 8 bays I suppose you would at that point have to place one enclosure on separate controllers.

Back to USB3 for a moment, my tests show I can get extremely close to the full 500MB/s over each dedicated port and three 7200 HDDs come in at just a little under 500MB/s (with the drives I tested) sustained over the first 65% of the platter. So with the 4 nMP USB3 ports you could probably put 12 (3x4) drives on and get around 1.8 or 1.9 GB/s out of it.

Either way... I think USB3 is the cheapest way. 4-drive enclosures are like $150 or less IIRC. And doing that you wouldn't need that expensive RAID Card.
I could do that, but I bought this card so I could expand to 16 disks, and using up all the USB3 ports would still force me into Thunderbolt whenever I needed to attach a client drive full of raw video. I could use a wireless keyboard, but they're not full length, and I love my numeric keypad.

I know I could force myself into a nMP, but it's not elegant for me at all, and totally unnecessary. Aggregating TB or TB being v2 x8 lane fast is far more elegant for the next couple of years. You could say I'm too slow, and the nMP is a puck that's been flicked way, way too far ahead of my stick, or a long bomb into the endzone while I'm still on the 50 yard line. I'm happy about that, because it means I'm riding a very convenient wave of timing, whereby I can skip this version and pick up the next wave when it's better, and save a lot of money in the present.
 
WD Thunderbolt dual-bay enclosures are relatively inexpensive now as well...

Amazon is selling their 8GB TB setups for about $150 plus the cost of the included drives ($650 total). RAID a couple of those for 16TB at 400MB/s! :D

If you don't need that much storage, the WD TB enclosure with dual Velociraptors is on for $580 and a couple of those would give you 4TB at 800MB/s! :eek:

Yeah, I'm thinking smaller drives and more of them is better than bigger drives. Mainly because we're talking about RAID0 for the speed and that really really needs real-time (every 2 to 4 hrs scheduled) backup. With eight 1TB drives the BU drive could be two 3TB drives - and with twelve 1TB drives it could be two 4TB drives - concat or RAID as you like.

But now as soon as we go to 2TB drives 8 of them is 16TB and you'll need about 12TB of BU space. 12 of them would be 24TB with about 18TB needed for backup. And it of course gets more ridiculous with 3TB or 4TB drives. :eek:

It's really good news to hear the TB RAID enclosures are coming down in price! This will make for a much more robustly configured nMP if I end up getting one.
 
I could do that, but I bought this card so I could expand to 16 disks, and using up all the USB3 ports would still force me into Thunderbolt whenever I needed to attach a client drive full of raw video. I could use a wireless keyboard, but they're not full length, and I love my numeric keypad.

You could hub the RAW video drive in and just deal with the bottleneck while transfering the files over. It'll probably still be fast enough not to make you wanna break things. :)

I know I could force myself into a nMP, but it's not elegant for me at all, and totally unnecessary.

Well, if that's the case, for sure, skip this version and wait for the MP 7,1 or 8,1 whatever grabs you by then. Maybe even a 6,1 used/refurbished.

Aggregating TB or TB being v2 x8 lane fast is far more elegant for the next couple of years. You could say I'm too slow, and the nMP is a puck that's been flicked way, way too far ahead of my stick, or a long bomb into the endzone while I'm still on the 50 yard line. I'm happy about that, because it means I'm riding a very convenient wave of timing, whereby I can skip this version and pick up the next wave when it's better, and save a lot of money in the present.

There ya go, same conclusion. And why do I think TB3 will be out before MP7,1 ??? ;)

Also of interest is this quote from Wikipedia:

"From mid-2012, manufacturers including LaCie and Drobo have started to include USB 3.0 connections in addition to the Thunderbolt one(s) on some of their devices."
- http://en.wikipedia.org/wiki/Thunderbolt_(interface)​
I guess meaning a potential solution to the problem you just outlined above?
 
yeah, that's a bummer if you're stuck with what you have when you buy it.. i mean, not really a bummer for me because i have less demanding peripheral needs that as far as i can gather, TB2 will handle perfectly well.. but i can at least sympathize with people needing faster transfer not being able to upgrade as the technology becomes available.

But will it be enough in two years or three?
It may be a costly oversight if you don't take this into account.
 
But will it be enough in two years or three?
It may be a costly oversight if you don't take this into account.
for me personally, I don't see a problem. basically, with the kind of applications I use, the better you get at drawing, the smaller your file sizes end up being..
but in the case of say a photographer whose cameras keep getting more&more megabeams, I can understand how their storage/transfer requirements are going to grow
 
for me personally, I don't see a problem. basically, with the kind of applications I use, the better you get at drawing, the smaller your file sizes end up being..
but in the case of say a photographer whose cameras keep getting more&more megabeams, I can understand how their storage/transfer requirements are going to grow

Then you're begining to understand why for some of us, the nMP just doesn't fit the bill as presented here. Some of us already have needs that go beyond what TB2 can serve right now. We can expand the old model to fit our need, but not with the nMP. We're getting short changed by Apple.
 
for me personally, I don't see a problem. basically, with the kind of applications I use, the better you get at drawing, the smaller your file sizes end up being..
but in the case of say a photographer whose cameras keep getting more&more megabeams, I can understand how their storage/transfer requirements are going to grow

Doubtful. Most prosumer formats are very very happy with USB3 and SATA3. TB2 is way overkill for anyone doing 1080 stuff for the most part. And for sure any still photogs...

The only folks who might outgrow TB2 are professional filmmakers layering 4K. And I guess they won't be on a MacPro for the heavy lifting.. ;)

All typically speaking. More than likely no one buying the nMP will in 2 or 3 years be saying; Damn this thing is too slow for me - I wish I had more I/O bandwidth. :p
 
Then you're begining to understand why for some of us, the nMP just doesn't fit the bill as presented here. Some of us already have needs that go beyond what TB2 can serve right now. We can expand the old model to fit our need, but not with the nMP. We're getting short changed by Apple.

nah.. I understand it for the most part and I don't think I've ever implied (or at least I hope I haven't) the new mac is more than enough computer for everyone. even someone like me who is not the most demanding of users will be maxing it out to it's fullest capacity for maybe 6 weeks per year.
what I do think tho is that at least something like 80% of typical mp users (probably more) are going to see enhanced performance in many ways (hd, ram, gpu, cpu.. and yes, even tbolt) compared to what they'll be upgrading from..
so in that regard, it's sorta a tough call as to whether or not apple is shortchanging it's users.

----------

Doubtful. Most prosumer formats are very very happy with USB3 and SATA3. TB2 is way overkill for anyone doing 1080 stuff for the most part. And for sure any still photogs...

The only folks who might outgrow TB2 are professional filmmakers layering 4K. And I guess they won't be on a MacPro for the heavy lifting.. ;)

All typically speaking. More than likely no one buying the nMP will in 2 or 3 years be saying; Damn this thing is too slow for me - I wish I had more I/O bandwidth. :p

I guess my main misunderstanding regarding video editors is whether or not they typically import/edit on the boot drive then move projects to storage afterwards. because if that's the case, it seems to me that even they will benefit from the new mac .. but if it's more typical to edit off externals then yeah, as wonderspark outlined, I can see where tbolt may induce lag as opposed to today's pcie in mac pros
 
nah.. I understand it for the most part and I don't think I've ever implied (or at least I hope I haven't) the new mac is more than enough computer for everyone. even someone like me who is not the most demanding of users will be maxing it out to it's fullest capacity for maybe 6 weeks per year.
what I do think tho is that at least something like 80% of typical mp users (probably more) are going to see enhanced performance in many ways (hd, ram, gpu, cpu.. and yes, even tbolt) compared to what they'll be upgrading from..
so in that regard, it's sorta a tough call as to whether or not apple is shortchanging it's users.

----------



I guess my main misunderstanding regarding video editors is whether or not they typically import/edit on the boot drive then move projects to storage afterwards. because if that's the case, it seems to me that even they will benefit from the new mac .. but if it's more typical to edit off externals then yeah, as wonderspark outlined, I can see where tbolt may induce lag as opposed to today's pcie in mac pros

But at what price... Those who would benefit the most, as you say, are those in need of something more than an imac but less than a full, expandable, workstation. Anandtech and others have already priced the CPU in the ballpark of $1.8 to 2.2k for volume buy and presently the only firepro card with 6gb or ram is the w9000 which details at $3.2k each. Even if AMD slashes the price by half your are up to $5k just for those 3 components.

That price is more inline for professional/business than prosumer, except for the fact that this machine is too locked down by proprietary standard for the pro/business. And what would be the point of buying the entry level 6 core & single gpu if you can't upgrade it yourself later on without paying the apple tax...
This is especially true if your applications exist on another plateform and you can get a better specc'd machine for less money than the base nMP. It just doesn't make sense...
 
...it's sorta a tough call as to whether or not apple is shortchanging it's users.

I guess my main misunderstanding regarding video editors is whether or not they typically import/edit on the boot drive then move projects to storage afterwards. because if that's the case, it seems to me that even they will benefit from the new mac .. but if it's more typical to edit off externals then yeah, as wonderspark outlined, I can see where tbolt may induce lag as opposed to today's pcie in mac pros

Well, they offered a boot drive that's 2.5 times faster than anything they're ever previously offered and 15 times faster than what they typically offer in the MP line. The I/O bandwidth is 6.7 times. The expandability is ten times that of the previous machine. The computing potential if we consider 1GPU = 1CPU (6 core) (just generally) is over 30 times the previous machine, and so on. This is the 10 times jump over the 10% increase Jobs said was needed to capture the imaginations (and pocketbooks) of users.

There's no (bottleneck) lag unless misconfigured. Unless I'm mistaken, the card he has is not even close to TB2 speeds. Totally I guess we have 6GB/s throughput on TB2 and 2GB/s throughput on USB3. The cards he has (if I'm not mistaken) is about half of the TB2 throughput and it costs close to $1k. The reason he's thinking "lag" (if he is?) is because he's thinking about placing that card on one TB2 connection and then adding 16 drives to that. I guess he loves his card - but that's a misconfiguration as should be obvious. With the TB2 configured to maximize storage I/O he should be able to get full speed from around 40 of the fastest 7200 HDDs (all in RAID0) instead of the 16 he's talking about. Add another 12 to (almost) max out the USB3 and we're talking sufficient throughput to support around a 55-drive RAID0 array comprised of the fastest 7200 drives you can buy. (mathematically speaking - real-world would be more drives yet because you can't get the full speed from each drive in such an array - typical it tapers off till you're only getting about 100MB/s from each drive added after the first 4 or 5.)

Lag? Mmm-kay... :rolleyes:

And actually (I'm still not clear on this) according to this it's "a data rate of 20 Gbps in both directions (on each of the two channels)" which means that it's not 2GB/s per controller (of the three controllers) but rather per port (channel) of which there are six. So that's actually all of my TB2 numbers times two. :D This means 12GB/s and not 6GB/s or 110 HDDs not 55 - as compared to his 16. :D

This seems to suggest the same thing: "Thunderbolt 2 enables two 20Gbps bi-direction channels". The each of the 6 total channels in the nMP can do 20Gb/s - or about 2GB/s. And reading further they say we won't actually get 2GB/s but more like 1.5GB/s is expected. If so that's a total of 1.5x6 or 9GB/s total - although I would think the final spec closer to 1.6 or 1.7 GB/s per port myself - if TB1 is any indication.


.
 
Last edited:
...
And actually (I'm still not clear on this) according to this it's "a data rate of 20 Gbps in both directions (on each of the two channels)" which means that it's not 2GB/s per controller (of the three controllers) but rather per port (channel) of which there are six. So that's actually all of my TB2 numbers times two.

The engadget article is being sloppy with "channel".

here is Intel's explanation of what they did:

"... It is achieved by combining the two previously independent 10Gbs channels into one 20Gbs bi-directional channel that supports data and/or display. Current versions of Thunderbolt, although faster than other PC I/O technologies on the market today, are limited to an individual 10Gbs channel each for both data and display, less than the required bandwidth for 4K video transfer. Also, the addition of DisplayPort 1.2 support in Thunderbolt 2 enables video streaming to a single 4K video monitor or dual QHD monitors. All of this is made possible with full backward compatibility to the same cables and connectors used with today’s Thunderbolt. ..."
http://blogs.intel.com/technology/2...ndwidth-enabling-4k-video-transfer-display-2/

The fact that it 100% backwards compatible should be a huge clue that the physical channels didn't change at all. That "combining" ( bonding/aggregating) is logical, not physical channel. Very similar to how the 10 up/down was a bonding of two sets of wires ( one up and one down). Now have two up and two down... so 20 both ways.

This seems to suggest the same thing: "Thunderbolt 2 enables two 20Gbps bi-direction channels".

Two 20Gb/s channels is a presumption talking about a 2 physical port controller. If you look at previous TB controllers there are 4 channels for those that handle two physical ports.

http://arstechnica.com/gadgets/2011...rbolt-controller-could-broaden-reach-of-spec/

Intel may or may not do a "chain ender" controller ( one with one physical port) for TB v2. A Ethernet or Firewire dongle doesn't need anything close to 20Gb/s ( or 10Gb/s even). Likewise any chain end dock like device is likely trying to save costs and hit a lower price point rather than blazing speed ( 2 or 3 USB 3.0 and/or Firewire ports don't need 20Gb/s either. ).


The each of the 6 total channels in the nMP can do 20Gb/s - or about 2GB/s.

two channels on the same switch can of course transfer from incoming to outgoing port at 20Gb/s. That doesn't make them additive bandwidth. Nor does it say much for the bandwidth out of the switch which is via a different interface.

In short, the PCI-e and DisplayPort data are transported at TB speeds on the TB network. However, if talking about what shows up on the other side after transport that is necessarily TB speed. 20Gb/s isn't going to make your SATA 3Gb/s controller go any faster. The data will still show up on the host side at some small multiple of 3Gb/s even if there are a couple of drives.


And reading further they say we won't actually get 2GB/s but more like 1.5GB/s is expected. If so that's a total of 1.5x6 or 9GB/s total - although I would think the final spec closer to 1.6 or 1.7 GB/s per port myself - if TB1 is any indication.

Ports aren't additive. One of the primary reason the PCI-e bandwidth to host is lower is because using a shared resource in the switch. Both ports on a single controller share the same PCI-e switch gateway to the host.

lightridge_thunderbold_inside_600px.png


http://www.tomshardware.com/reviews/thunderbolt-performance-z77a-gd80,3205-4.html

TB v2 that diagram changes a bit because the TB switch is going to get more complicated by being able to feed and route more DisplayPort v1.2 traffic and also interleave the traffic without introducing ischoronous delivery problems. However, the x4 PCI-e v2 interface to the host/peripheral remains exactly the same.
 
But at what price...
[...]
It just doesn't make sense...

you're right.. especially on paper.. you're absolutely right.

but i think that type of discussion goes well beyond apple vs the rest of computerland..

we see the same phenomenon over and over again in pretty much anything that's on the market.

i mean, there are a few boutiques (and even [sorta] mainstream stores) around here selling $10,000 shoes.. WTF!! that's absolutely ridiculous to me.. much more than something like spending 5g on a computer when theres a 4000 model with better specs..

point is, i get your point.. but i don't really have a good explanation for it other than humans are retarded when it comes to money :) ..all but the most frugal of people will knowingly spend more on something in their lives.. that's just how we roll and i'm sure if i knew of everything you bought in the last year or so, i'd certainly find something that i feel you way overpaid for.. and if we were on friendly terms-- i'd call you a ***** idiot for spending so much on it :D

as far as i'm concerned when looking at consumerism/capitalism as a whole, i could be buying a new computer every year and still save money compared to all you fools spending mega bucks on things like, you know, cars.

but if someone wants to spend an extra grand on something, it's not really that big of a deal.. for whatever reason, they generally feel happy about their buy.. maybe that's why apple always wins those customer awards or whatever?-- people buy something they consider fancy and/or more unique than most stuff out there and that alone gives some sort of happiness (even if the happiness isn't truly genuine etc).. likewise, i bet bmw buyers are some of the happiest car owners in the world.. not because of how much better/functional/powerful their cars are. rather how much they paid for their cars..

but i guess that's going way off topic

----------

[but all that said-- i'm pretty sure once i'm actually using the computer and not thinking about money or anything like that.. i'm going to be super hyped using it.. it's a luxury buy, sure.. but at least it's still awesome to use.. just like how a bmw is awesome to drive]

relatively speaking, apple tax isn't that much.. prada tax anyone? evian?

.
 
Last edited:
besides cars, i'll make another sweeping remark that's probably aimed at 75% of you all..

lottery tickets?

that's not apple tax.. that's straight up stupid tax.
you're literally giving money to a corporation with nothing in return (other than maybe a little thrill and the perceived joy you get during your weekly 'if i won the lottery' conversation)..

but hey.. i'm not going to sit around chastising people for that.. i mean- do you thing.. have fun.. it's completely fine.
but i'm not going to buy those stupid things ;)


[EDIT] and who do you have to talk to around here to get that wink emoticon changed? sometimes i feel like what i'm saying deserves a wink but that blue guy doesn't sum up the right feel..
 
The engadget article is being sloppy with "channel".

here is Intel's explanation of what they did:

"... It is achieved by combining the two previously independent 10Gbs channels into one 20Gbs bi-directional channel that supports data and/or display."

The "and/or" qualifier suggests that the 20Gb/s can be used solely for data, there is no mention at all about the controllers, for example how many ports a host controller will have. It will be interesting to see when it is actually available, but Apple puts it like this: "up to 20Gb/s data transfer speed".
 
The "and/or" qualifier suggests that the 20Gb/s can be used solely for data, there is no mention at all about the controllers, for example how many ports a host controller will have. It will be interesting to see when it is actually available, but Apple puts it like this: "up to 20Gb/s data transfer speed".

Apple says:
"Thunderbolt 2 delivers twice the throughput, providing up to 20Gb/s of bandwidth to each external device. So you’re more than ready for the next generation of high-performance peripherals. You can connect massive amounts of storage, add a PCI expansion chassis, and work with the latest external displays — including 4K desktop displays and peripheral devices capable of broadcast monitoring in 4K. And since each Thunderbolt 2 port allows you to daisy-chain up to six peripherals, you can go all out by plugging in up to 36 external devices via Thunderbolt alone."​
Which also seems to reaffirm that it's 20Gb/s per port - 6 ports. I didn't see anything official in deconstruct60's post which would make me think otherwise either. I'm betting it is indeed per port and the finished product will deliver 1.7GB/s like TB1 delivers 850MB/s (real world). So that's a nice total of 10.2 GB/s. And we'll probably get very close to all of that if we connect nothing but drives (SSD better than HDD) to the 6 ports. What would that be in SSDs anyway, about 16 or 18 drives. I guess we'd have to use 18 drives - 3 per port to really saturate for the maximum throughput.

And we could also connect 18 GPGPUs to the sucker as well, and get full speed out of them when the 18 drive RAID00 wasn't being accessed too heavily. Probably we could have the entire thing (18 drives in 6 enclosures plus 3 more GPGPUs daisy-chained off each of the 6 ports) powered up and rendering sequential 3D 4K animation frames without ever hitting a bottleneck - or if so not for more than a few milliseconds per frame.

That would be some power draw though! If each GPGPU draws 150W while computing and each drive averages 15W that's real close to 3,000W all total. :rolleyes:

----------

[EDIT] and who do you have to talk to around here to get that wink emoticon changed? sometimes i feel like what i'm saying deserves a wink but that blue guy doesn't sum up the right feel..

Yeah, the rolls eyes guy is a better tongue-in-cheek wink-wink bit-of-the-nudge character IMO. Heh :rolleyes:
 
Last edited:
The "and/or" qualifier suggests that the 20Gb/s can be used solely for data, there is no mention at all about the controllers, for example how many ports a host controller will have.

To a specific host or specific peripheral the data has to travel through a x4 PCIe v2 link. That link theotrically maxes out at 2GB/s ( 4 * 500MB/s ). That's 16Gb/s if convert to bytes. Data can only theoretically saturate a Thunderbolt network if there is device to device PCI-e data traffic that bypasses the host. As long as there is just one target ( the host) for in/out going traffic the PCIe switch ( not the Thunderbolt one) becomes the bisection bandwidth limiter.

Doing additive counting of ports of a switch is wrong. For a switch it is the bisection bandwidth that will indicate max throughput.

Thunderbolt is a relatively cheap switch ( go price 10 or 40 GbE switches per port solutions ) because it keeps the switching limited. 1 or 2 ports keeps things very simple and affordable. It is highly unlikely going to get TB controllers with very wide fanout. Similarly, it is highly unlikely going to have individual controllers for each port. Each TB controller consumes x4 PCI-e v2 lanes. Six controllers would be 24 lanes. The Mac Pro doesn't have that kind of budget to blow on Thunderbolt. Since Intel said the sockets, cables, and PCIe interface are exactly the same.... it is unlikely the number of physical TB channels has changed either.
 
Apple says:
"Thunderbolt 2 delivers twice the throughput, providing up to 20Gb/s of bandwidth to each external device.... you can go all out by plugging in up to 36 external devices via Thunderbolt alone."​

Apple also says that:
" Thunderbolt is the fastest, most versatile I/O technology there is. "

Which is hyperbole and cow manure. It isn't the fastest. Similarly, you are skipping over the "up to 20Gb/s" above. There is a depending upon configuration and conditions involved.



Which also seems to reaffirm that it's 20Gb/s per port - 6 ports.

The misdirection that you seem to keep hyping is the difference between Thunderbolt data bandwidth and the bandwidth of the other two data protocols that Thunderbolt transports. To a single device they are not the same thing. One if the bandwidth/through put on the TB network. The other is the time it takes to "get on" , ride TB to remote designation , and "get off". Those "get on" and "get off" portions for the invididual transported protocols to get back to their native state can be the bisection bandwidth bottleneck.

The other issue that Apple's hyperbole sweeps under the table that the Thunderbolt is basically a switching network. Even just a Thunderbolt network of just two devices involves two switches talking to another. It is *NOT* like mapping ethernet endpoints in the task of assesing bandwidth. That methodology is just wrong.

Technically Apple is correct. The bandwidth across only the TB switch is 20Gb/s. That says little directly to what the host is going to see because the TB switch has to deal with the PCI-e switch to get data to the host. You can claim how that is "unofficial" till the cows come home if you wish, but that is how TB works.
 
To a specific host or specific peripheral the data has to travel through a x4 PCIe v2 link. That link theotrically maxes out at 2GB/s ( 4 * 500MB/s ). That's 16Gb/s if convert to bytes.

I don't follow. 20 gigabits/s equals 2.5 gigabytes/s because 20 / 8 = 2.5. That is a theoretical max, but PCIe uses 8b/10b encoding, that is 10 bits are used to transfer 8 bits of data, so the maximum data rate is 2.5 * 0.8 = 2 gigabytes/s or 2GB/s.

Data can only theoretically saturate a Thunderbolt network if there is device to device PCI-e data traffic that bypasses the host. As long as there is just one target ( the host) for in/out going traffic the PCIe switch ( not the Thunderbolt one) becomes the bisection bandwidth limiter.

Doing additive counting of ports of a switch is wrong. For a switch it is the bisection bandwidth that will indicate max throughput.

You keep pretending that you know something about this new controller which no-one has seen. I'm not doing additive counting of ports.



Imagine the following where both devices gets data from the host:

[host] <---> [device 1] <---> [device 2]

If device 1 consumes 1GB/s and device 2 consumes 2GB/s, then device 2 can never get 2GB/s because the link between the host and device 1 needs to carry 3GB/s of data for that to happen. That is unrelated to what we are talking about, but it's one reason to use the "up to" wording.​



Thunderbolt is a relatively cheap switch ( go price 10 or 40 GbE switches per port solutions ) because it keeps the switching limited. 1 or 2 ports keeps things very simple and affordable. It is highly unlikely going to get TB controllers with very wide fanout. Similarly, it is highly unlikely going to have individual controllers for each port.

You obviously do not know any of this.
 
Last edited:
The core issue though is that the 1/3 of all Mac users is still several multiples larger than the Mac Pro PCI-e card market. 14M Macs per year. 33% is about 4.6M. That's likley at least 10x bigger than the Mac Pro market, let alone the add-on card market to of that bigger subset.

WTF does that have to do with the Mac Pro? I already said the new Mac Pro would make a good gaming Mac or Mac Mini replacement, etc. if it were at reasonable price. The Pro market is what it is. If Apple is going to support it then they should support it, not screw it.

The fact that Thunerbolt products that are variants of PCI-e cards are just the primarily just the same board in a new wrapped along with some updated drivers, actually would help support the PCI-e card market to last longer than it would have if limited to just the Mac Pro demand.

Who is talking about PCI-e? This is about PCI vs Thunderbolt instead of it. I don't know anyone who thinks trying to get an adapter for PCI-e for Thunderbolt is a better idea than a dedicated Thunderbolt port.

It actually was necessary. To get placed on all Mac it needed to be a dual use socket. There is limited edge space for sockets on Mac products. You can create an alternative universe where that isn't true, but in this one that is a very real design requirement.

You seem to be talking about notebooks that have very very limited space. This is a freaking TOWER for god sakes. There's plenty of room for multiple ports. Design requirement? Who's?

Second, adaption of Thunderbolt would be even slower if there was no "backwards compatible" mode to build inertia off of. USB 3.0 got quick

Yeah, but who else is doing it Apple's way? Sony is going to use a USB port for backwards compatibility. Once again, Apple is trying to build support for something no one else seems to want (i.e. Firewire compared to USB).

uptake because there was USB 2.0. USB 3.1 will much slower because USB 3.0 is no where near as large ( USB 2.0 had almost a decade to build a deployed user base). Thunderbolt with a clearly proprietary, Intel only, socket would have even more problems than TB has now. There is a reason why Lightpeak was in USB form factor and why USB folks didn't want their port hijacked.

USB folks? I think you mean the USB consortium. I think a lot of US would just assume they double with the USB ports so we would get MORE OF THEM instead of wasting space on the cases for Thunderbolt ports NO ONE WANTS.

And that includes wasting a Mini-DP port (i.e. it's a royal PITA if you are using Thunderfart to have to keep unplugging the freaking monitor every time you want to change a TB device in line with it (since it's all daisy-chained). I haven't seen ANY hubs for it yet except Apple's monitor and it's still using USB2 ports so NO ONE WANTS IT. Apple needs to get their crap together. They made a deal to have Light Peak (in Thunderbolt form) for the first year EXCLUSIVELY and that only SLOWED the adoption of it by "everyone" that much more. For a company that would supposedly like to see Thunderbolt be the next big thing, they gave away that much more ground to USB3 in the mean time. And as it is, the #1 reason hardly anyone is using TB right now is that most of the devices that ARE available are hard drives and they cost like 50%-300% more than the equivalent USB3 or eSata version with the exact same transfer speeds. One would have to be a flipping idiot to buy a 3TB Thunderbolt drive for $300 when the same one for USB3 costs like $110 with the same throughput (since neither interface is saturated). Thunderbolt makes a good "high end" drive connector, but then "true" high-end uses PCI 3.0 16x ports for high speed, not an external cable.... that runs at 4x (for ONE device since it's shared).


Third, frankly Display Port can also use the help since DVI and HDMI seem to holding their ground. It need another demand push and while Thunderbolt won't help sweep those other two from the field of play it isn't going to hurt places where holding ground gained either.

How does Thunderbolt "help" (mini) Display Port? By making you unplug all your monitors to change your Thunderbolt drives???? :confused:


1. Sharing bandwidth with GPU. High throughput GPU vendors aren't going to be happen with that. There are GPU cards that uses PCI-e switches. Apple works with none of them. It is a solution with overhead which folks want to shovel under the rug in these alternative universe option enumerations.

All the more reason the GPU should never have been tied to Thunderbolt in the first place, except perhaps as an OPTION. Why would ANYONE want to waste Thunderbolt throughput on graphics output? It makes no sense except perhaps to drive 3 monitors (that have thunderbolt pass-through) from one port on a notebook where ports are a premium. That's not enough reason to tie Thunderbolt to the GPU IMO.

3. Exactly why would the GPU vendors want to create basic designs which puts more money into Intel's pockets. Intel is already replaced them in terms of deployed GPU units. Clearly on its way to becoming #1 in the Graphics market. So Nvidia and AMD want to speed that up? Not. If think the PC system vendors are a bit skittish about Intel only solutions.... the GPU vendors are in another zipcode.

You're just giving more reasons why the new Mac Pro should have used all PCI 3.0 expansion. In fact, they could have used the video cards they did use and STILL offer PCI 3.0 in a conventional case. The ONLY reason they can't use PCI 3.0 expansion in the new Mac Pro is that they chose to shove the thing in that cylindrical trash can instead of a normal tower case.

Given they moved away from rack mount servers they haven't particularly been trying to fit into square rack holes. The gratuitous handle height on the current model isn't friendly to horizontal rectangular holes either.

Yeah and that just proves how unfriendly Apple is to professionals. They don't update the software or hardware in a timely fashion and don't listen to their customers. It's going to cost them the market in the end. I hope spoiled rich kids is enough of a market to keep Apple going indefinitely.

Large shops seem more likely target. One-man-band shows aren't. Past video ingest into a network storage for a team solution custom video cards for what? Transform? Done. Output to reference monitors... not that huge of a gap.

And yet one-man-bands is who this new model is targeted to. Ask any of the REAL professional on here how much they like having no PCI expansion in the new Mac Pro rather than the spoiled rich kids who want the thing to play Doom 4.0.

What pro audio gear company doesn't have new USB 2.0 offerings?

"Pro" and USB 2.0 don't go together. USB 2.0 tends to be laggy due to overhead (not bandwidth which is why most devices are still FW400 since they don't need more bandwidth, just low overhead and since FW800 is backwards compatible with FW400, it keeps the largest market possible). USB 3.0 would solve that dilemma, but they don't seem to be embracing USB 3.0 very quickly either.

What pro audio gear company couldn't sell more product to more Mac users if enclosed their PCI-e card in a Thunderbolt box. Their "card" would work not just with Mac Pro but with millions more Macs. Why wouldn't they be interested in trying to sell those million more users?

WTF are you talking about? WTF would an audio company want to put their Pro Audio box on a PCI-e card instead of just using FW400/800 and/or even Thunderbolt directly? PCI boxes cost a lot of money (hell the hubs we are promised cost hundreds of dollars). My FW audio box cost $400. It would cost me like $1000 to get a PCI card + Thunderbolt box and it would take up a lot more room. It's a FW400 box. My current MBP can have it and a FW800 drive plugged in at the same time since my MBP has BOTH FW400 and F800 ports (now that's convenience). It also has an expansion port with a USB 3.0 card plugged into it (sorry newer Macbook Pros before USB 3.0 came out; you can't have USB 3.0, but ironically my 2008 model CAN and DOES have USB 3.0 on it now). 2008 was the pinnacle of Macbook Pro design. It has gone all aesthetic since then with fewer expansion options and ports available. If my battery dies on a plane, I can plug a backup in. If your battery on newer Macbooks dies on a plane, you're SOL.

A FW400 cable with a FW800 shaped socket at the end is no more an adapter as a USB socket A to microUSB cable is.

??? An adapter is an adapter dude. It's still a PITA compared to having a dedicated port since you have to carry the things around and NOT LOSE THEM on the go.

Frankly the whole stuck in the FW400 ghetto is in what part what killed FW. No need for speed increased and standard dies off on devices tracking higher workloads.

I don't know WTF you're talking about. If a device doesn't NEED the extra bandwidth, why would you want to force it to be FW800 when FW800 can run FW400 devices also. They're only mistake was not making a backwards compatible port that doesn't need an adapter (but that would have probably required some foresight).

A non fact. As big as the USB market? No. No one? Also no. Your whole post started off alluding to millions of users.

Again, I don't WTF you're going on about. You seemed determined to argue for arguments sake rather than actual relevance.

Just got through harping on only really need a FW400 speed connection. That's PCI-e v2 x1 territory. Thunderbolt is significantly faster than that. There is hardly anything audo that pushes Thunderbolt except for small corner cases of extremism.

I was talking about audio adapters and convenience (i.e. you have to waste one of those precious Thunderbolt ports with an adapter that won't have a pass-through on it in order to use a Firewire device of any kind with a Thunderbolt port PLUS pay $50 for the ability to do so) whereas a dedicated port costs nothing extra and doesn't tie up a port. The Mac Pro has plenty of room on its case. The fact my tiny little Mac Mini 2012 model has a DEDICATED FW800 port and the new Mac Pro does NOT is a freaking JOKE.
 
i thought thunderbolt IS pcie?

He of course means card-edge slots when he says "PCIe". So far the language we're all using is:

TB : PCIe 4x over cable. 10Gb/s 4 channels unidirectional (2x up, 2x down)
TB2 : PCIe 4x over cable. 20Gb/s 2 channels bidirectional.
PCIe : PCIe vX x16 - x4 card edge connection.

and:
It's becoming extremely clear to me that each connector (such as seen on the MP) is capable of 20Gb/s bidirectional DATA - of which we'll probably get 17 to 17.5Gb/s. One controller as seen in the nMP offers two full speed independent connections for Data and/or Display (either or both). Both Data and Display can be on the same single connection chain - as demonstrated live by Intel themselves. They showed a 2K LCD monitor and two individual SSDs configured in a software RAID0 where the LCD was displaying an I/O speedometer and the drives were getting about 1160MB/s (which is about max for two SSD drives in RAID0) all simultaneously on a single TB2 chain (one connector - daisy-chained).
 
Last edited:
He of course means card-edge slots when he says "PCIe". So far the language we're all using is:

TB : PCIe 4x over cable. 10Gb/s 4 channels unidirectional (2x up, 2x down)
TB2 : PCIe 4x over cable. 20Gb/s 2 channels bidirectional.
PCIe : PCIe vX x16 - x4 card edge connection.

and:
It's becoming extremely clear to me that each connector (such as seen on the MP) is capable of 20Gb/s bidirectional DATA - of which we'll probably get 17 to 17.5Gb/s. One controller as seen in the nMP offers two full speed independent connections for Data and/or Display (either or both). Both Data and Display can be on the same single connection chain - as demonstrated live by Intel themselves. They showed a 2K LCD monitor and two individual SSDs configured in a software RAID0 where the LCD was displaying an I/O speedometer and the drives were getting about 1160MB/s (which is about max for two SSD drives in RAID0) all simultaneously on a single TB2 chain (one connector - daisy-chained).

Right, but lets say you've got a 4K display which needs about 16Gbps bandwidth, and a 1000MB/s (8Gbps) drive array. I assume if you run those daisy chained on the same connector, the max bandwidth of 20Gbps means your drives will be bottlenecked to maintain the display signal integrity. However, if you plug the drive array into the second connector instead (still the same TB2 controller) will your drive array then run at full speed? Or, do you need to plug it into a whole separate TB2 controller to get full speed? (Is there any documentation that makes this clear or are people assuming stuff?)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.