I am a little confused as to why these cost so much because I remember reading that an ad hoc 10G connection was possible in the thunderbolt 3 spec if you connected two thunderbolt 3 computers together. I guess the circuitry to make thunderbolt 3 into an normal Ethernet jack is expensive
The expense doesn't come from being able to communicate at 10Gb/s, it comes from being able to do so over long distances. That's why Thunderbolt adapters and cables are more expensive than a passive PCIe riser card or cable. The Thunder3 takes the hit twice: first for the Thunderbolt interface, then for the 10G Ethernet interface that can push those speeds over much worse cabling.
Maybe it's just me, but I caught myself nodding along like it made sense, but then I took step back and thought, "Wait, what?"
>The expense doesn't come from being able to communicate at 10Gb/s, it comes from being able to do so over long distances.
It seems here that you're insinuating that 10GbE is more expensive due to cabling, rather than the actual 10GbE integrated circuit at both ends of the cable enabling that throughput. As a reminder, 10GbE over twisted pair copper will work even with CAT 5e cabling, assuming the run is like a few feet. CAT6A quality cabling is what's "approved" for 10GbE.
Given that this product assumes the user already has a 10GbE network, that assumes the user also has good ethernet cabling that enables that throughput, so then there shouldn't really be anything that the product itself (the 10GbE thunderbolt 3 adapter) cares about long distances since the included short thunderbolt 3 cable is what is expected to push that data through to the PC, and given the short run, there really shouldn't be any issues.
The expense has to fundamentally be from the adapter itself, not the cabling, as better/worse cabling is ultimately the only thing enabling the distance that the data can be streamed without losing data due to noise. So I don't understand where the "being able to do so over long distances" has anything to do with the cost of the product.
It's the PHY, PCB design work and actual physical port that really drives up the cost. The MAC itself is in the order of about $30. If this were a n*1Gbit box, it'd be somewhere around the $70-120 mark, depending on how many 1G ports were involved.
This is amazing price difference. With TB2 people paid 400 for simple adapter. Where TB3 version is around 200 (https://www.span.com/compare/52589-62948). This will start new 10GbE era. It has been knocking on a door for while now.
So what are "cheap passive cables"? Well intuitively, active cables are optical fiber SFP+ cables where active electronics at both ends of the cable have to decode the light signals into electrical signals for the port. "Cheap passive cables" are essentially 10GbE-BaseT, in other words, traditional twisted pair copper ethernet cables.
Again, there's this insinuation that "it's expensive because we need to push a signal over long distances over copper", when the distance that ethernet frames/packets can be pushed over a cable is completely up to the quality/length of the cable run itself, this NIC just pushes the data over the cable.
There's literally nothing extra in terms of electronics that needs to happen for this adapter to work with a raggedy CAT5 ethernet cable as opposed to a high quality 100ft CAT6A ethernet cable run. It's literally part of the ethernet transmission protocol to have CRC bits so that the receiving end can do a frame check sequence to determine if the received ethernet frame was corrupted in transit.
Passive Thunderbolt cables more than three feet long lose signal. Active thunderbolt cables are very expensive. Using the adapter to leverage existing network infrastructure is cheaper than replacing the Ethernet cabling with expensive active thunderbolt cables, just so you can use the ad hoc networking.
and then there's Intel royalties for anything TB related right? A theoretical USB3.1 to 10G Ethernet would have cost less. Might be interesting if 10 x 1GBit USB adapaters + a switch still cost less than this ;)
I'm pretty sure Intel gave up all royalties on TB when they open sourced the standard. It's my understanding that the USB forum is even going to integrate some of the TB standard into the next USB standard.
Intel tried to handicap USB3.1 with TB and failed miserably and in the end open sourced the spec royalty free so it didn't just die. Even without royalties it's still a mostly dead standard precisely because USB is better.
Same. I hope a similar one comes out using Aqauntia's controller. Their single port 10G cards are MUCH cheaper. Hell, they had a deal for their own for $69 now that long ago. So bill of material costs can be reduced on this kind of device with the same, if not better performance, at least on the network controller part. Most 10G singe port NICs are $200+, some $300+. I bet with Aquantia with a partnership could make a similar device for $100ish. I think it's a pretty good add-on or tool for folks without the port already onboard and are mobile.
The prices will come down as the technology becomes more commonplace, but it may not reach the same level of ubiquity as something like a 10/100 NIC because wireless technologies are now so prevalent. Wired Ethernet has many uses still, but it's a rare animal in residential computing and lots of businesses now put WiFi to use on company networks.
All 10Gbe are expensive, the most expensive part is the PHY, TB is expensive and also the most expensive part is the PHY...
The PHY is the chip that is responsible of converting the data signalling from the regular data to the corresponded signalling in the bus cable ( being a cat 6 cable for Ethernet or a Type-C cable for TB3 )...
10GbE now has a price around $100 for just a basic single card... I don't know how much a basic single input TB3 cost but you can imagine...
So technically, a similar product with USB 3.1 can cut the cost for 10GbE adapters as USB 3.1 controller is much cheaper than TB3... ofcourse; you'll need USB 3.1g2 (10gbps) which is more expensive than regular USB 3.1g1 (5gbps).. again the most expensive part of USB 3.1g2 is the PHY it self...
mass producing the PHY and open standards (so multi manufacturers can make the Controllers and PHY) drives cost down, that's why TB3 is still expensive as Intel is still the only maker for these parts...
I mean, literally that box is TB3 > PCIe x4 (or x2) bridge, which then goes to the TN4010 PCIe x4 10GbE controller (backside of pcb), which goes to the 10GbE PHY from Marvell. The cost is in the fancy enclosure, the fact that this is an extremely niche device, that they can charge the amount they're charging, and that people who want it (for ~reasons~) are willing to pay this price. Cost of components themselves are insignificant, I would guess the BOM in production qtys is below 100USD, more like closer to 50-60.
I'd say longer than that. Just like blu-ray didn't replace DVD in PCs/laptops. The problem is that gig-E is fast enough for most people, and wired Ethernet is being replaced by WiFi.
A box of matches is small and the above NIC isn't small. As it's 'only' a nic the price really needs to hit $99 before people take notice. Even then it's too expensive.
Will the mainstream ever stop milking people or maybe, just maybe, take a chance and move straight to the next level. 10Gbe should be standard by now. Same for 4k screens.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
23 Comments
Back to Article
Zeratul56 - Tuesday, February 13, 2018 - link
I am a little confused as to why these cost so much because I remember reading that an ad hoc 10G connection was possible in the thunderbolt 3 spec if you connected two thunderbolt 3 computers together. I guess the circuitry to make thunderbolt 3 into an normal Ethernet jack is expensiveBilly Tallis - Tuesday, February 13, 2018 - link
The expense doesn't come from being able to communicate at 10Gb/s, it comes from being able to do so over long distances. That's why Thunderbolt adapters and cables are more expensive than a passive PCIe riser card or cable. The Thunder3 takes the hit twice: first for the Thunderbolt interface, then for the 10G Ethernet interface that can push those speeds over much worse cabling.JoeyJoJo123 - Tuesday, February 13, 2018 - link
Maybe it's just me, but I caught myself nodding along like it made sense, but then I took step back and thought, "Wait, what?">The expense doesn't come from being able to communicate at 10Gb/s, it comes from being able to do so over long distances.
It seems here that you're insinuating that 10GbE is more expensive due to cabling, rather than the actual 10GbE integrated circuit at both ends of the cable enabling that throughput. As a reminder, 10GbE over twisted pair copper will work even with CAT 5e cabling, assuming the run is like a few feet. CAT6A quality cabling is what's "approved" for 10GbE.
Given that this product assumes the user already has a 10GbE network, that assumes the user also has good ethernet cabling that enables that throughput, so then there shouldn't really be anything that the product itself (the 10GbE thunderbolt 3 adapter) cares about long distances since the included short thunderbolt 3 cable is what is expected to push that data through to the PC, and given the short run, there really shouldn't be any issues.
The expense has to fundamentally be from the adapter itself, not the cabling, as better/worse cabling is ultimately the only thing enabling the distance that the data can be streamed without losing data due to noise. So I don't understand where the "being able to do so over long distances" has anything to do with the cost of the product.
ZeDestructor - Tuesday, February 13, 2018 - link
It's the PHY, PCB design work and actual physical port that really drives up the cost. The MAC itself is in the order of about $30. If this were a n*1Gbit box, it'd be somewhere around the $70-120 mark, depending on how many 1G ports were involved.ZeDestructor - Tuesday, February 13, 2018 - link
PHY being the ethernet PHY, not the TB3 PHYDunkurs1987 - Wednesday, February 14, 2018 - link
This is amazing price difference. With TB2 people paid 400 for simple adapter. Where TB3 version is around 200 (https://www.span.com/compare/52589-62948). This will start new 10GbE era. It has been knocking on a door for while now.Alistair - Tuesday, February 13, 2018 - link
Try reading what Billy Tallis said again. Failure of reading comprehension ;)He said the adapter must be expensive so that it works with cheap passive cables. Which is also what you said, while thinking he said the opposite.
yvizel - Tuesday, February 13, 2018 - link
Alistair, I wish I could "like" your post... :)JoeyJoJo123 - Wednesday, February 14, 2018 - link
You're not making a lick of sense, either.So what are "cheap passive cables"? Well intuitively, active cables are optical fiber SFP+ cables where active electronics at both ends of the cable have to decode the light signals into electrical signals for the port. "Cheap passive cables" are essentially 10GbE-BaseT, in other words, traditional twisted pair copper ethernet cables.
Again, there's this insinuation that "it's expensive because we need to push a signal over long distances over copper", when the distance that ethernet frames/packets can be pushed over a cable is completely up to the quality/length of the cable run itself, this NIC just pushes the data over the cable.
There's literally nothing extra in terms of electronics that needs to happen for this adapter to work with a raggedy CAT5 ethernet cable as opposed to a high quality 100ft CAT6A ethernet cable run. It's literally part of the ethernet transmission protocol to have CRC bits so that the receiving end can do a frame check sequence to determine if the received ethernet frame was corrupted in transit.
eh_ch - Wednesday, February 14, 2018 - link
Passive Thunderbolt cables more than three feet long lose signal. Active thunderbolt cables are very expensive. Using the adapter to leverage existing network infrastructure is cheaper than replacing the Ethernet cabling with expensive active thunderbolt cables, just so you can use the ad hoc networking.xchaotic - Tuesday, February 13, 2018 - link
and then there's Intel royalties for anything TB related right? A theoretical USB3.1 to 10G Ethernet would have cost less. Might be interesting if 10 x 1GBit USB adapaters + a switch still cost less than this ;)rahvin - Tuesday, February 13, 2018 - link
I'm pretty sure Intel gave up all royalties on TB when they open sourced the standard. It's my understanding that the USB forum is even going to integrate some of the TB standard into the next USB standard.Intel tried to handicap USB3.1 with TB and failed miserably and in the end open sourced the spec royalty free so it didn't just die. Even without royalties it's still a mostly dead standard precisely because USB is better.
CheapSushi - Tuesday, February 13, 2018 - link
Same. I hope a similar one comes out using Aqauntia's controller. Their single port 10G cards are MUCH cheaper. Hell, they had a deal for their own for $69 now that long ago. So bill of material costs can be reduced on this kind of device with the same, if not better performance, at least on the network controller part. Most 10G singe port NICs are $200+, some $300+. I bet with Aquantia with a partnership could make a similar device for $100ish. I think it's a pretty good add-on or tool for folks without the port already onboard and are mobile.PeachNCream - Tuesday, February 13, 2018 - link
The prices will come down as the technology becomes more commonplace, but it may not reach the same level of ubiquity as something like a 10/100 NIC because wireless technologies are now so prevalent. Wired Ethernet has many uses still, but it's a rare animal in residential computing and lots of businesses now put WiFi to use on company networks.Xajel - Wednesday, February 14, 2018 - link
All 10Gbe are expensive, the most expensive part is the PHY, TB is expensive and also the most expensive part is the PHY...The PHY is the chip that is responsible of converting the data signalling from the regular data to the corresponded signalling in the bus cable ( being a cat 6 cable for Ethernet or a Type-C cable for TB3 )...
10GbE now has a price around $100 for just a basic single card... I don't know how much a basic single input TB3 cost but you can imagine...
So technically, a similar product with USB 3.1 can cut the cost for 10GbE adapters as USB 3.1 controller is much cheaper than TB3... ofcourse; you'll need USB 3.1g2 (10gbps) which is more expensive than regular USB 3.1g1 (5gbps).. again the most expensive part of USB 3.1g2 is the PHY it self...
mass producing the PHY and open standards (so multi manufacturers can make the Controllers and PHY) drives cost down, that's why TB3 is still expensive as Intel is still the only maker for these parts...
timecop1818 - Wednesday, February 14, 2018 - link
I mean, literally that box is TB3 > PCIe x4 (or x2) bridge, which then goes to the TN4010 PCIe x4 10GbE controller (backside of pcb), which goes to the 10GbE PHY from Marvell. The cost is in the fancy enclosure, the fact that this is an extremely niche device, that they can charge the amount they're charging, and that people who want it (for ~reasons~) are willing to pay this price. Cost of components themselves are insignificant, I would guess the BOM in production qtys is below 100USD, more like closer to 50-60.Vidmo - Tuesday, February 13, 2018 - link
"compatible with both Macs and Windows PCs"Sire, as long as you're not trying to use the Thunderbolt device on Windows server, Intel does not allow that.
https://communities.intel.com/thread/117555
Flunk - Tuesday, February 13, 2018 - link
I wonder how long it will be until 10G Ethernet is on every motherboard, 2 years?fred666 - Tuesday, February 13, 2018 - link
I'd say longer than that.Just like blu-ray didn't replace DVD in PCs/laptops.
The problem is that gig-E is fast enough for most people, and wired Ethernet is being replaced by WiFi.
Samus - Tuesday, February 13, 2018 - link
That's a sexy little thang but daaaamn expensive.damianrobertjones - Wednesday, February 14, 2018 - link
A box of matches is small and the above NIC isn't small. As it's 'only' a nic the price really needs to hit $99 before people take notice. Even then it's too expensive.Will the mainstream ever stop milking people or maybe, just maybe, take a chance and move straight to the next level. 10Gbe should be standard by now. Same for 4k screens.
gamingkingx - Monday, June 4, 2018 - link
Can anybody come with a guess of why they are not making a 10G USB 3.1 model? That should be a no-brainer. I would gladly pay for it as a SFFPC user.Or just integrate the damn AQC-107 in most mainstream and upwards boards. Come on! Lets get the 10G revolution going!!! :-D