In relation to being docking stations, I think Apple should've included a PCIe slot in the ATDs to facilitate further expansion, such as a GPU.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Acer and Asus to Provide Momentum to Thunderbolt Adoption
- Thread starter MacRumors
- Start date
- Sort by reaction score
In relation to being docking stations, I think Apple should've included a PCIe slot in the ATDs to facilitate further expansion, such as a GPU.
We need to simply agree to disagree on this one.
In my view, a "TBolt peripheral" has two TBolt (mDP) connections and can be inserted into a TBolt daisy chain. (god, how I hate daisy chains - but that's another topic)
A "USB 3.0 disk connected to a USB 3.0 hub connected to a USB 3.0 PCIe controller connected to a TBolt->PCIe bridge" is not a TBolt device in my parlance. It's a USB 3.0 disk.
And, mostly a moot point, since no TBolt -> USB 3.0 devices exist on the market.
Last edited:
Well it sort of all comes down to semantics really...
What are you calling native?
If TB is what seems to be a special purpose interconnect between two PCIe switches and your saying that attaching any sort PCIe interfaced function controller to that switch means the overall device isn't native TB. Then really yes there will never be a native TB device. The point wasn't to have native TB devices like SSD modules, it was to have PCIe device makers be able to install outside the box.
Well it sort of all comes down to semantics really...
What are you calling native?
If TB is what seems to be a special purpose interconnect between two PCIe switches and your saying that attaching any sort PCIe interfaced function controller to that switch means the overall device isn't native TB. Then really yes there will never be a native TB device. The point wasn't to have native TB devices like SSD modules, it was to have PCIe device makers be able to install outside the box.
So you agree, there are no native TBolt devices, and probably never will be.
If there's a PCIe driver for the device, it's not native TBolt.
I'm still confused ... what exactly is native TB to you? Your earlier example of two mDP is what the Thunderbolt display has. As for the rest, I thought Thunderbolt was simply a fancy, long-range PCIe connection that also carries video. So what does having a PCIe driver have to do with a device not being native Thunderbolt? What would be the advantage of being "native"? My questions are in earnest. I'll admit this is not an area I know much about and your discussion is somewhat difficult to follow ...
Last edited:
Well yeah but that doesn't mean we agree on the outlook for TBolt "connected" devices in the future. I mean you look at the Next Gen SATA spec and a TBolt/SATA Express RAID has bugger all silicon in it.
So, do TBolt peripherals exist or not given that TBolt is a native PCIe external interface?
There are several examples that fit your parameters of a TBolt peripheral, including the Apple Thunderbolt Display.
So, as occurred in other threads, you are being prone to negating your own arguments.
Obviously, a TBolt peripheral would require a TBolt port to be deemed a TBolt peripheral.
Another article about Thunderbolt, another thread full of facepalm.
Some fun facts:
PCs and motherboards with USB 3.0 host controllers only began shipping in January of 2010. In the first 12 months, 12 million PCs were shipped with USB 3.0, a 3.5% attach rate overall. Estimated shipments of USB 3.0 enabled PCs in 2011 are at 68 million, or an 18% attach rate.
The first PCs with Thunderbolt controllers were shipped in March 2011. In the first 12 months Apple alone will ship more than 12 million Macs with Thunderbolt. With Acer and Asus on board, and if Sony fully embraces the Thunderbolt standard (and Intel can crank out enough silicon), second year shipments of Thunderbolt enabled PCs could rival the 68 million figure expected of USB 3.0.
What were witnessing here is the initial adoption rate of a brand new I/O interface rivaling that of the third major iteration of the most widely deployed I/O interface in history. Granted, Thunderbolt will not be able to maintain the same cadence as USB 3.0 once USB 3.0 is included in every major CPU and chipset by both Intel and AMD. Thunderbolt will be featured on some 2012 Intel platforms, but not integrated into Ivy Bridge chipsets. However, PCIe and DisplayPort, the two interfaces that Thunderbolt is based on, will be present on every major CPU and chipset.
So where are the devices? Although there are more than 10 billion USB devices in the wild, and sales of USB devices in 2011 alone are pegged at well over 3 billion, estimated sales of USB 3.0 devices are only at about 84 million, or a little more than 2% of the total. This has a lot to do with the fact that most USB devices cant really benefit much from SuperSpeed mode, so most companies arent rushing to redesign their products around a controller that currently sells at a premium. Due to this, and the fact that there isnt a heck of a lot of device controller silicon around yet, the only USB 3.0 devices currently on the market at all are HDDs/SSDs/drive enclosures, flash memory thumb drives, hubs, and memory card readers. None of these are terribly innovative or exciting, and even if USB 3.0 were to dominate these device categories entirely, they would still only account for about 15% of USB devices shipped annually.
Thunderbolt is a means of providing access to the PCIe and DisplayPort interfaces of a PC through a compact external connector which can be used with either copper or fiber interconnects. It is designed to sit right alongside existing USB ports. USB works great for more than 10 billion devices, its not going anywhere. Aside from the fact that both interfaces offer gobs of bandwidth (a single Thunderbolt channel currently provides 1000 MB/s of real world throughput, and a USB 3.0 host controller can muster 325-370 MB/s on a good day), the protocols used are radically different and have hugely different applications. This is evident in the Thunderbolt products that have shipped or been announced in the 6 months since Thunderbolt made its debut. Some of which can be seen here: http://www.anandtech.com/show/4779/tons-of-thunderbolt-peripherals-at-idf
So wait, whats that? Its been six months and we havent seen 82 million Thunderbolt mass storage devices shipped to equal the sum total of boring USB 3.0 devices to hit the market thus far in 20 months? Thunderbolt is clearly DOA... Either that or device manufacturers are having a little trouble getting their mitts on sufficient quantities of Thunderbolt controller chips. But even if that werent an issue, just as SuperSpeed mode is only useful for a small percentage of all USB devices, Thunderbolt is also best suited for only certain types of applications. Im sure we will see some drive enclosures for SSDs or disk arrays as well as pro audio and video gear eschewing USB 3.0 for Thunderbolt, but in general there wont be a ton of overlap with what is typically USB territory. This means that the device shipment volume numbers between the two interfaces will never be comparable (sort of like the situation with FireWire). Im sure if everyone tries hard enough they can conjure in their mind an analogy that illustrates that a lower volume shipment figure does not equate to a total fail. Anyway, Thunderbolt devices are actually shipping. People have actually purchased and are using Promise Pegasus RAID products. Most products are not shipping yet because all the host controllers are belong to Apple right now.
What of the relative costs? Chips, like any commodity, have an associated production cost which when combined with supply and demand factors results in a certain market price. With most new technologies, initial supply is often lower than demand, and production costs are significantly higher than later on in the production cycle. Both USB 3.0 and Thunderbolt host controllers are still in the early stages of production. As die sizes are reduced, yields increase, further production capacity is brought online, and supply pressure is eased, chips become much cheaper. Renesas (NEC) just released their 3rd gen USB 3.0 host controller which reduced die size by 51% over the previous generation. They also more than doubled monthly production volume, and consequently prices have dropped by about 70% from 2010 levels to about $1 per controller. Similar economies of scale will apply to Thunderbolt, although the die sizes might necessarily remain a bit larger, and unfortunately it would appear that Thunderbolt devices will always carry an additional price premium. First of all, its Intels baby, and for the most part they can keep their margins where they like. Secondly, Thunderbolt devices require the same host controllers that PCs do. The architecture being essentially peer to peer doesnt lend itself to cheap and dumb device controller silicon the way USB does. This was something FireWire also suffered from. In the future, though, this price premium for devices should be in the neighborhood of $15 at the retail level, give or take. Right now you also have the $49 cable issue, but once again, its only because Apple is the only game in town (and the cheapest cable they sell lists for $19). There is no reason why we cant expect lesser priced Thunderbolt cables to hit the market in the next 12 months.
The semantics over what constitutes a Thunderbolt device are ridiculous. If it contains a Thunderbolt controller and is connected to a PC using a Thunderbolt cable, its a Thunderbolt device, pure and simple.
Nobody has announced a Thunderbolt to USB 3.0 host adapter yet, despite the number of people who believe such a thing would be worth overpaying for at this stage in the game. Im guessing that not many developers feel like investing the time and effort into coding their own USB 3.0 drivers for Mac OS X when Apples drivers will be freely available in less than six months. The other factor here is that the current crop of Sandy Bridge Macs will likely be the only machines to ever ship with Thunderbolt and not USB 3.0... So the total audience for such a product isnt exactly a massive one. In the end though, Im sure such a creature will come into existence, as the construction of it should be relatively simple.
Some fun facts:
PCs and motherboards with USB 3.0 host controllers only began shipping in January of 2010. In the first 12 months, 12 million PCs were shipped with USB 3.0, a 3.5% attach rate overall. Estimated shipments of USB 3.0 enabled PCs in 2011 are at 68 million, or an 18% attach rate.
The first PCs with Thunderbolt controllers were shipped in March 2011. In the first 12 months Apple alone will ship more than 12 million Macs with Thunderbolt. With Acer and Asus on board, and if Sony fully embraces the Thunderbolt standard (and Intel can crank out enough silicon), second year shipments of Thunderbolt enabled PCs could rival the 68 million figure expected of USB 3.0.
What were witnessing here is the initial adoption rate of a brand new I/O interface rivaling that of the third major iteration of the most widely deployed I/O interface in history. Granted, Thunderbolt will not be able to maintain the same cadence as USB 3.0 once USB 3.0 is included in every major CPU and chipset by both Intel and AMD. Thunderbolt will be featured on some 2012 Intel platforms, but not integrated into Ivy Bridge chipsets. However, PCIe and DisplayPort, the two interfaces that Thunderbolt is based on, will be present on every major CPU and chipset.
So where are the devices? Although there are more than 10 billion USB devices in the wild, and sales of USB devices in 2011 alone are pegged at well over 3 billion, estimated sales of USB 3.0 devices are only at about 84 million, or a little more than 2% of the total. This has a lot to do with the fact that most USB devices cant really benefit much from SuperSpeed mode, so most companies arent rushing to redesign their products around a controller that currently sells at a premium. Due to this, and the fact that there isnt a heck of a lot of device controller silicon around yet, the only USB 3.0 devices currently on the market at all are HDDs/SSDs/drive enclosures, flash memory thumb drives, hubs, and memory card readers. None of these are terribly innovative or exciting, and even if USB 3.0 were to dominate these device categories entirely, they would still only account for about 15% of USB devices shipped annually.
Thunderbolt is a means of providing access to the PCIe and DisplayPort interfaces of a PC through a compact external connector which can be used with either copper or fiber interconnects. It is designed to sit right alongside existing USB ports. USB works great for more than 10 billion devices, its not going anywhere. Aside from the fact that both interfaces offer gobs of bandwidth (a single Thunderbolt channel currently provides 1000 MB/s of real world throughput, and a USB 3.0 host controller can muster 325-370 MB/s on a good day), the protocols used are radically different and have hugely different applications. This is evident in the Thunderbolt products that have shipped or been announced in the 6 months since Thunderbolt made its debut. Some of which can be seen here: http://www.anandtech.com/show/4779/tons-of-thunderbolt-peripherals-at-idf
So wait, whats that? Its been six months and we havent seen 82 million Thunderbolt mass storage devices shipped to equal the sum total of boring USB 3.0 devices to hit the market thus far in 20 months? Thunderbolt is clearly DOA... Either that or device manufacturers are having a little trouble getting their mitts on sufficient quantities of Thunderbolt controller chips. But even if that werent an issue, just as SuperSpeed mode is only useful for a small percentage of all USB devices, Thunderbolt is also best suited for only certain types of applications. Im sure we will see some drive enclosures for SSDs or disk arrays as well as pro audio and video gear eschewing USB 3.0 for Thunderbolt, but in general there wont be a ton of overlap with what is typically USB territory. This means that the device shipment volume numbers between the two interfaces will never be comparable (sort of like the situation with FireWire). Im sure if everyone tries hard enough they can conjure in their mind an analogy that illustrates that a lower volume shipment figure does not equate to a total fail. Anyway, Thunderbolt devices are actually shipping. People have actually purchased and are using Promise Pegasus RAID products. Most products are not shipping yet because all the host controllers are belong to Apple right now.
What of the relative costs? Chips, like any commodity, have an associated production cost which when combined with supply and demand factors results in a certain market price. With most new technologies, initial supply is often lower than demand, and production costs are significantly higher than later on in the production cycle. Both USB 3.0 and Thunderbolt host controllers are still in the early stages of production. As die sizes are reduced, yields increase, further production capacity is brought online, and supply pressure is eased, chips become much cheaper. Renesas (NEC) just released their 3rd gen USB 3.0 host controller which reduced die size by 51% over the previous generation. They also more than doubled monthly production volume, and consequently prices have dropped by about 70% from 2010 levels to about $1 per controller. Similar economies of scale will apply to Thunderbolt, although the die sizes might necessarily remain a bit larger, and unfortunately it would appear that Thunderbolt devices will always carry an additional price premium. First of all, its Intels baby, and for the most part they can keep their margins where they like. Secondly, Thunderbolt devices require the same host controllers that PCs do. The architecture being essentially peer to peer doesnt lend itself to cheap and dumb device controller silicon the way USB does. This was something FireWire also suffered from. In the future, though, this price premium for devices should be in the neighborhood of $15 at the retail level, give or take. Right now you also have the $49 cable issue, but once again, its only because Apple is the only game in town (and the cheapest cable they sell lists for $19). There is no reason why we cant expect lesser priced Thunderbolt cables to hit the market in the next 12 months.
The semantics over what constitutes a Thunderbolt device are ridiculous. If it contains a Thunderbolt controller and is connected to a PC using a Thunderbolt cable, its a Thunderbolt device, pure and simple.
Nobody has announced a Thunderbolt to USB 3.0 host adapter yet, despite the number of people who believe such a thing would be worth overpaying for at this stage in the game. Im guessing that not many developers feel like investing the time and effort into coding their own USB 3.0 drivers for Mac OS X when Apples drivers will be freely available in less than six months. The other factor here is that the current crop of Sandy Bridge Macs will likely be the only machines to ever ship with Thunderbolt and not USB 3.0... So the total audience for such a product isnt exactly a massive one. In the end though, Im sure such a creature will come into existence, as the construction of it should be relatively simple.
Thanks for your post! A lot of people do seem to have completely forgotten about the slow adoption of USB 3.0, especially in the beginning when a lot of long winded tech articles and blog posts were written about how USB 3.0 was DOA for a variety of reasons not the least of which was a lack of USB 3.0 peripherals on the market instantaneously. And I've never understood this idea that Thunderbolt has to "beat" USB 3.0 to have value as a port. Thunderbolt is by design geared towards a different type of use than USB 3.0. Maybe the silliness will stop a couple of years from now: we'll have USB 3.0 and TB side-by-side and devices (hubs, external GPUs, RAIDs, SSDs, etc...) running off the port best suited for their needs and everyone can just be happy with what they use ...
Last edited:
Ya, it's a shame that USB peripherals don't work with TBolt equipped Macs due to these Macs not having USB ports as well. (Disclaimer: this post contains sarcasm.)
There really are no disadvantages to thunderbolt besides lack of physical backwards compatibility with USB.
i just dislike how much room thunderbolt takes up. hopefully in some time that gets smaller too. I would still very much love to have USB 3.0 come to macs soon
You're kidding right?
Have you looked at a Thunderbolt/mDP port lately? It's the same height and 2/3 the width of a USB A port. Cmon!
Let's think about this one... What type of PCIe slot? Half-height, full-height, full-length, single-slot width, dual-slot width? What lane width for the physical connector? How would you carve out space in the chassis to accommodate this? How would the user access this slot? How would the connector be oriented in relation to the display, and how would this be achieved? What type of additional TDP would you design for? How would you provide the power to this slot, and how would you dissipate the heat?
Being sold as a $999 display and all, it should probably still function as one, so it will still need to accept a DisplayPort signal if there is no dGPU present in the PCIe slot. It would also be neat if the other built in features still worked. If we figure 5.8 Gbps for the display signal plus 1Gbps for Gigabit Ethernet, 1.26 Gbps for the two FireWire 800 ports, 480 Mbps for a single USB 2.0 host controller (even though there could well be more than one in there to support the FaceTime HD camera, audio interface and three downstream ports), plus another 25% for protocol overhead on the PCIe stuff, and we've got 775 Mbps left out of our 10 Gbps Thunderbolt channel. Hmmm... Not really enough for a PCIe slot.
But if you did have a dGPU in your PCIe slot, and you could somehow engineer the switching between multiple DisplayPort inputs, you wouldn't need the DP stream on the TB channel, thus freeing up 5.8 Gbps of bandwidth. Now you'd have the legs to support a dGPU with a PCIe 2.0 x1 connection, which would never be bandwidth throttled and most likely perform just as well as that same card in an x16 slot. Well worth the increased cost, added bulk to the display, potential for starting a fire, etc...
I would think the MXM spec would answer many if not all of these questions.
The big standout would be the limited bandwidth one and could it be used efficiently enough to make it worthwhile.
To bad Intel and nVidia aren't in a working together kind of mood.
In terms of hardware implementation, don't iMacs have a GPU?
Isn't TBolt 10Gbps per channel over 2 channels = 20Gbps?
Last edited:
The fact that USBB3 is cheaper than TB is an advantage. It is also backwards compatible to thousands of USB2 devices and will get quite a bit of support. I think there is a market for both to coexist: USB3 for all users, TB for the high end (video capture, gaming, server, etc) market.
Quote:
Originally Posted by DESNOS
Thunderbolt is by far more flexible. You can connect anything to thunderbolt that you could connect to a PCIe port internally. There really are no disadvantages to thunderbolt besides lack of physical backwards compatibility with USB.
Ok, I take this to mean that Thunderbolt is not backwards compatible with USB.
Quote:
Originally Posted by AppleScruff1
And 99% of the peripherals are usb.
I posted this so I'm pretty sure it means that most peripherals are USB and won't plug into a Thunderbolt port. If this is not right, please let me know what it means.
Quote:
Originally Posted by munkery
Ya, it's a shame that USB peripherals don't work with TBolt equipped Macs due to these Macs not having USB ports as well. (Disclaimer: this post contains sarcasm.)
I don't see where I said anything about Macs not having USB ports. Perhaps you would be so kind as to point that out to me? Thanks.
Why not read this string of posts in context?
USB peripherals won't work in Thunderbolt ports because Thunderbolt is not backwards compatible with USB. Most peripherals in use are USB devices, hence they won't work in a Thunderbolt port. However, a USB device will still work in a USB port.
I'm perfectly clear on this. What part is not coming across well?
It sure would, leaving the user with only the nagging question of how you get your hands on an MXM form factor card as a non OEM. (I know they're out there, but not really in the mainstream retail channels at this time.) So then the logical solution would be for Apple to provide the GPU as well. But wait, they already do provide dGPUs as standard or optional in most Mac models... So this would be adding considerable additional expense and complexity to every ATD for the sake of MacBook Air, 13-inch MacBook Pro or quad-core Mac mini owners who want slightly better gaming performance. But maybe a better solution would be for those who want more graphics horsepower to avoid buying a $999 to $1599, ultra-compact Mac with no dGPU. (Or for them to wait for the much improved iGPU that will come with the Ivy Bridge refresh.)
They are MXM form factor with custom designed cooling, so not exactly reasonable for a user installable upgrade.
A device can only use one channel. The other channel is to provide bandwidth for other devices in the chain.
Last edited: