Thunderbolt 3 Motherboard?

[Eggz]

Hi Tom's People,

I am looking into a mITX build with USB 3.1 Type-C and have found a couple good options. So far the Gigabyte GA-170N-Gaming5 is the top contender. It lacks Thunderbolt 3, though, even after Gigabyte announced its Gaming series boards with USB 3.1 Type-C would get an upgrade for the Type-C connector to support Thunderbolt 3. Don't understand why they'd exclude the mITX board but include all others in the series.

As for the board that do support Thunderbolt 3, there's an apparent discrepancy in the power deliver that I hope someone on here can help me understand. The Thunderbolt 3 standard says it supports 100 watts of charging. But the only motherboards I've seen with Thunderbolt 3 say that the port supports only 36 watts of charging (e.g. Asus board, Gigabyte boards). What's that about?

So the questions given this landacape are (1) do you expect Thunderbolt 3 to be on mITX boards, and (2) will they deliver the full power spec of 100 watts?

If you could please chime in, I would appreciate it. I haven't been able to find much on this.

Thanks

-Eggz

 

[jimmysmitty]

1. Eventually, yes. TB3 is great as it finally has UCB Type-C to jump on and have one connector for everything while also being a protocol that will allow for faster, smoother and better data transfer rates (USB was not designed with external storage in mind). It will take time though as it is new, Skylake is the first with official support.

2. The spec is up to 100w and I would assume that is more for laptops than for external devices. Most cell phones with fast chargers are around the 10w range so 36w is already a lot. I assume they will but will require a lot of good solid work on the motherboard to support that amount of power.

 

[Eggz]

I guess I'll hold off. Connectivity is very important for compact workstations, which is what I'm building.

Since the spec says up to 100 watts, the strong implication is that the certification process should require 100 watts. After all, Type-C USB 3.1 already handles up to 100 watts, though it doesn't have to. The whole point of a set standard like Thunderbolt is to put an icon there so that you know all of the features contained in the set are present. Providing less than the actual spec degrades confidence. Very strange to me because Intel is usually pretty good about that, and "up to" means a maximum, so you'd expect the same maximum. Perhaps there was an exception because the boards didn't have the capacitors to support that much power delivery. It would be nice to charge a laptop directly from a desktop, which would allow for better utilizing these high wattage PSUs. There's always next year

 

[jimmysmitty]

My assumption is that the boards are not able to support that amount or even the port they are using might have some limitations.

As I said since TB3 is new to Skylake it will probably be the next generation that fully adopts it. It would have been faster but older versions were much more expensive due to the proprietary connector, Intel wanted something like USB Type-C but was stuck waiting on the USB consortium, which made adoption much slower. With the USB Type-C connector this will allow TB3 to be cheaper and to be adopted much faster.

I am pretty excited for the potential uses such a external PCIe or a single cable that charges your laptop, powers the monitor and provides a display out.

 

[Eggz]

I am too! I wonder whether 10 Gbps networking would work. I know that TB has network protocols in it. If so, it could be a much more elegant solution for small businesses trying to put together fast networks in small spaces. The main limitation I can think of would be maximum cable length. But dropping in a small PCI-e card for passthrough would be much cheaper and easier to install than a 10 Gbps switch as well as cards on each computer. TB may not even appear on the computers as a network drive, just as a storage device. That would be cool, and it would probably perform better.

 

[jimmysmitty]

The problem on the network side would be length limitations as you said. Cat5/6 is capable of up to 100 meters (333FT) while most USB is limited to around 15 feet. Of course TB might allow for longer lines but I have yet to see if it will allow for USB to go beyond its current limits.

A switch is actually very negligible in performance loss. I have connected my work system directly to the firewall and to a switch and it performs the same. Switches are technically just pass throughs that allow you to split a single line into multiple. Of course it has hardware to manage each port.

 

[Eggz]

Yes, harder to manage, and they also add a lot of cost.

I actually just looked into the Thunderbolt networking, and there has been support for al ittle over a year: https://thunderbolttechnology.net/sites/default/files/Thunderbolt%E2%84%A2%20Networking%20Bridging%20and%20Routing%20Instructional%20White%20Paper.pdf

Thunderbolt 2 did 10 Gbps networking, and version 3 will be faster. It seems simple enough to setup, and there are even optical versions of Thunderbolt 2 that allot VERY long cables. Here is a 100 ft optical TB 2 cable: http://eshop.macsales.com/item/OWC/CBLOPTTB30M/

I imagine that TB 3 optical cables will come out soon too. That would be sweet!

 

[jimmysmitty]

Optical is a different beast. Great though as they have yet to hit any sort of bandwidth cap (I think last I read they were hitting Exobytes in transfer rates).

Optical does have some downsides though. One is no PoE (Power over Ethernet) and second is less flexibility vs copper. The more you curve it, the higher chance of loss with optical.

As I said each has their ups and downs. I would love to design a fiber based backbone for my home and have GB/s worth of bandwidth available. Of course I also need a 10Gbe NIC and router, neither of which are easy to get yet.

 

[Eggz]

Yeah, I spent some time looking into it yesterday. Going optical on Thunderbolt 3 would be great for a small office, where only high priority components get on it for local connections. You can make a main workstation with a beefy PSU the router. It would have the 100 watts per port for laptop charging, but one port would just be data for the optical.

Then other computers would have a pass-through for data, and where power was needed, you'd just have either a powered hub or a computer with sufficient power. The powered hub thing is great because you can make stations with a monitor and an external GPU for a laptop to plug into, which would help offset the overall hardware cost by allowing laptops without dedicated GPUs. Then at the end of the chain can be a relatively low powered machine with a crap-ton of SSD storage on it, and it would have permissions set to act as a file server. Optical cables can be put through the walls and then break out into copper for durability at the stations.

That would be really cool because the connection essentially blurs, if not eliminates, the distinction between "internal" and "external" components. All of the machines would have access to very fast storage, such that it may be even faster than the device's internal SSD. That essentially turns a digital studio into one large computer. Something like that would be great for a post-processing shop.

There would be no need to buy a special 10 Gbps router, and the total bandwidth would be quadruple that which you could attain with a 10 Gbps anyway.

If you were editing a lot of photos, or putting together a movie, you could just walk over to the beefy workstation, and then have it render the files with output set to the SSD NAS. There'd be no need to even transfer the working files from a station to the main tower because the remote connection would be fast enough. Then the editor would walk back to any available station and start the next project. This could be going on for multiple editors simultaneously. Each would just rinse, wash, and repeat.