Thunderbolt 3 Overview [pdf] | Hacker News Reader

Thunderbolt 3 Overview [pdf]

thunderbolttechnology.net

18 pointsfezz38 comments

WiGig is a more practical technology, in my humble opinion.

I have a WiGig-enabled laptop and a WiGig dock with 3 1080p monitors, a Ethernet cable, and all of my peripherals connected to the dock including a 10 port hub with a USB3 HDD.

I can transfer files to my HDD (80MB/s), to the network (90MB/s), play 1080p videos on 1 monitor, browser on another, and play games on the last one without noticing any latency at all.

And I can move my laptop to the other side of the room without any noticeable connection issues, which is pretty surprising considering they're advertising WiGig as a short-range wireless technology (although having large obstructions between the laptop and the dock will cause problems very quickly).

For a first-gen product, I'm pretty impressed. The room-wide wireless gigabit Ethernet alone is already a killer feature for me, but the multiple wireless displays and peripherals makes for an even more amazing package. Can't wait to see how they iterate on it.

Really impressive that this can handle basically everything that's thrown at it. "Yeah, it'll provide power and is backwards compatible with other protocols. External GPU to game with your laptop? No problem. Oh, and the connector is smaller than before, too."

That was the promise of thunderbird 2 (or 1)...

TB3 being tied into USB-C means more possibilities on the consumer side if they price it right.

TB1/2 were weak in many respects but pricing and adoption were the main ones.

I believe 3 is finally fast enough.

Not quite, eGPU wasn't officially supported for instance. Port compatibility with USB (and thus not having to give up USB ports) might also be a big factor, and TB2 power delivery, while higher than pre Power Delivery USB, was only 10W IIRC.

The problem remains the extra cost of integrating TB. Hopefully the cost will come down with Alpine Ridge.

Can somebody give me a quick overview of how and when to spend my money on this?

The new xeon for laptops will have TB3.

As a video editor that deals with terabytes of footage, in practice thunderbolt is much faster than USB3. It's common that i have to copy 1-4 terrabytes onto several drives for editors, fx artists, archival, etc usually on deadline. I don't care that thunderbolt isn't as ubiquitous as usb, but like firewire before it, I'm glad it's there and common enough.

Except that many companies use large scale centralized network storage and are exporting the files over their network protocol of choice, and only need local storage for caching, temp, and swap.

None of the Macs we use for editing have 10GbE or FiberChannel interfaces onboard, so you can use a Thubderbolt device to plug into high-speed network storage. ATTO Technology make some good ones.

That's nice in theory but every post house I've worked at still needs plenty of thunderbolt and usb drives. The footage isn't shot at the post house, it has to be transported and copied over to the SAN, and everything is on a tight deadline so it needs to be done yesterday. Or rather than using the in house colorist (or fx artist) the client decides to go with someone else, so that often means copying footage for that delivery. And the client wants two copies of it, one for archive and one because the creative director wants to dick around with it. Or the post house is over-booked and the only edit station is an imac not connected to the SAN, so copy it over to a drive so someone can start working. Plus with everything shot on digital it's not unheard of to get 20 or more hours of footage for a campaign. On a RED or ARRI, that takes up a lot of space.

Half my work now is out of my home now, I don't need a SAN, instead clients ship me a thunderbolt raid (typically a G-Raid) that I ship back once the project is finished. And I know more and more editors and motion graphic artists who are doing the same, less work in edit suites, more work at home. Fast, portable drives makes that possible (and efficient high quality codecs like ProRes and h264).

I've been doing this for 20 years, we'll still need fast, portable storage for many years to come and right now, for video editing, Thunderbolt is the best choice.

Wouldn't something like eSATA be a better choice? The hdd's speed would be the limiting factor instead of the bus.

SATA tops out at 6Gb/s. One 7200 rpm spindle might not be able to saturate that, but a RAID array could, and any decent, single SSD would trivially saturate it.

So, for the kind of use case where you need to move lots of data, very quickly, eSATA may not be better :)

Well yeah, but a single ssd wouldn't hold lots of data, and for the use case described, I assume its a spinning rust drive. Also you wouldn't run a large raid array over a single sata bus, at the very least you'd use SAS and get multiple channels in one cable.

Edit: The g-raid mentioned looks like a 2x spinning array, so well within the limits of a single sata III bus.

esata is less ubiquitous than both USB and Thunderbolt (and firewire before it). External ports have never been standard on any macs and I don't ever remember them being standard on any of the HP machines we used (the majority of big budget editorial, i.e. feature films, tv, and commercials, are cut on macs, although fx leans heavily toward Windows and Linux). We use to install esata ports on some of our machine room systems, as the speed difference was definitely useful. I wish esata would have become more common but at the moment Thunderbolt makes it unnecessary (and USB3 seems to have stifled it's standard adoption across non-Apple systems).

Also, aside from the G-Raid, we also use 1 and 2tb SSD's (more for portability than the extra speed).

when you have some special equipment that needs it.

keep in mind that USB2 is already overkill for external HDDs. and USB3 become overkill for SSDs. technically, you won't ever need usb3.1.

USB2 maxes out at 60MB/sec. I cannot remember when I had a drive that was so slow, I think I'd have to go back to single or double digit gigabytes.

In comparison, modern HDDs exceed 150mb/sec on writes, more on reads.

please, do connect a budget external HDD on USB3/firewire/esata/or even sata directly.

the best i got here was 63Mb/s here. extenal HDDs usually have 5400RPM laptop drives. with one plater/head.

Except I have an external drive that is a 5400RPM 3.5" drive, a Samsung Spinpoint (the last generation before Seagate bought the unit from Samsung), and it does at least 120MBps sequential writes.

Intel continues to refuse to deploy Thunderbolt by default on Z series chipsets (Z87, Z97, Zwhatever LGA1151/Broadwell uses).

I'd love to adopt Thunderbolt, Intel literally refuses to sell it to me.

There are plenty of Z97 motherboards with Thunderbolt: https://thunderbolttechnology.net/products?tid=15&field_comp... I don't understand why it matters whether it's default or not.

Neither of my Z97 motherboards come with it, it is roughly a $100 premium to get it.

If it doesn't come with systems by default, no one is going to use it. I'm interested in it purely for IP-over-Thunderbolt.

Yeah, in retrospect Intel probably should have forced millions of people to pay extra so that DiabloD3 could save some money on his home LAN.

Not really. It costs about as much to put Thunderbolt on a board (especially when the board already has a DisplayPort port wired into the on-die GPU) as it does to put USB 3 on a board.

It is merely a few cents in parts. There is literally no reason why Intel just doesn't put Thunderbolt everywhere.

No, it doesn't. Intel has stringent certification requirements for Thunderbolt. It's made for higher-end devices with high-end prices.

>>It is merely a few cents in parts.

Parts make up a fraction of the cost.

> It is merely a few cents in parts

someone never worked with intel integrated.

That isn't the issue. Its an Intel platform and Intel _wants_ to lock everyone into their platform... which requires the platform to be as accessible and ubiquitous as possible.

Limiting Thunderbolt to a few computers no one owns (except Apple products, Thunderbolt is everywhere there) basically signs Thunderbolt's death.

Thunderbolt 3 is now fast enough that I can do 40gbit Ethernet over it. With the upcoming PCI-E 4.0 standard, that is 2-3 lanes worth of bandwidth.

Yet, Intel isn't storming the industry and trying to legitimately compete with USB 3.x.

Charging extra for something you obviously want is hardly the same thing as refusing to sell something to you.

of course not. they would have to double the price of the board to get it in!

thunderbolt is not innovative. it is just a canon. so far everyone is killing flies. so, no need for that. putting that extra $100 or so on each and every board would certainly just drive consumers away.

See my reply here: https://news.ycombinator.com/item?id=10029607

You can get it on motherboards where the manufacturer chooses to pay the licensing fee and submit to the Intel certification requirements/tests/fees. It's an option. Not every motherboard manufacturer chooses to do this and make consumers pay more for the motherboards for a feature that most consumers will never utilize.

Intel will include TB3 by default for the mobile Xeon lineup they just announced a few days ago.

The problem I foresee with Thunderbolt 3 is that it's serving a use that few people need at a price that no one wants to pay. The price-point issues might disappear with the USB-C compatibility, but even then, the only time I can ever foresee me actually reaching for Thunderbolt is for a display when DP 1.3 isn't an option or when huge data xfers were commonplace.

Unless it competes on price point with USB, I can't see this being any more popular than previous iterations of the protocol.

Maybe that'd change some if Intel actually put it on the enthusiast chipsets, but even then, I think the price point is a pretty big turnoff.

I don't think it will ever be as ubiquitous as USB, but it doesn't have to be useful. I'm a video editor and thunderbolt has saved me enough time copying files that i'll gladly pay the price. The cost typically gets passed along to clients and at rates of $100/hr or more for a single editor they'll happily pay the extra cost of the drives if it saves a few hours over the life of a project.

4K displays are going to become the new norm over the coming years. In which case those that use dual monitors (many of us) are going to want Thunderbolt 3 to drive them.

Also having multiple ports of the same type on laptops is going to be far better than the current status quo which is a nightmare for many on a day to day basis (you need which adapter for the work projector again ?).

Thinner, easier to use, more flexible laptops that can support multiple Retina displays seems pretty compelling to me. And you can sure as hell bet Apple will be all over this.

OMG external graphics card !

Reply on news.ycombinator.com