Comments Locked

18 Comments

Back to Article

  • AdditionalPylons - Thursday, June 20, 2019 - link

    Three questions:
    1. Many vendors now include multiple connectors such as storage and networking in eGPU cases, but they still only use one TB3 connector. How does the GPU handle the varying performance load from these other devices? Does it get dedicated static bandwidth assigned over PCIe or is it dynamic depending on the non-GPU devices' load?
    2. A back-of-the-envelope calculation of theoretical bandwidth used by SATA, GbE and USB3.0 (assuming one bus with shared bandwidth over all 5 ports) I get 6+1+5+3=15 Gbps. This leaves roughly 40-15=25 Gbps for the GPU. This seems like a pretty tough limitation, potentially neglecting the point of putting a powerful GPU in this box. Why wouldn't they rather go for a GPU only solution, or add a second TB controller for the non-GPU devices? Or am I missing something here?
    3. Are there any news on whether AMD will implement TB3? I'm mainly concerned about laptops, as in desktop machines we can always buy the Gigabyte GC-Alpine Ridge add-in card.
  • TheUnhandledException - Thursday, June 20, 2019 - link

    If it is anything like all existing eGPU enclosures the GBe and the SATA port as are just off the usb hub. That means it is is just 5 Gbps (or 10 Gbps in the case of usb3.1) reserved for the ports and the rest for the GPU.

    A few enclosures use a second TB3 controller to run all the ports that way the GPU has the full bandwidth but based on the specs I don't think that is the case here.
  • Skeptical123 - Monday, June 24, 2019 - link

    So I went down a little bit of a rabbit hole looking this up.

    "In the Thunderbolt mode, Thunderbolt 3 port has the ability to support at least one or two (4 lane) DisplayPort interface(s), and up to 4 lanes of PCI Express Gen 3" and lists USB Only Mode

    DisplayPort Only Mode - a single four lane (4 x 5.4 Gbps, or HBR2) link of DisplayPort

    DisplayPort and USB Mulit-Function Mode - one of the high speed connector pin pairs of signals will be dedicated to DisplayPort (now 2 lanes at 5.4 Gbps) and one to USB 3.1 This allows for a basic connectivity for data and display devices such as docking stations or data and display dongles.

    Thunderbolt 3 Mode - If a cable and device supporting it... the Thunderbolt silicon activates its highest capability mode and config- ures four high-speed links at either 10 Gbps or 20 Gbps (depending on cable and device support) to support the Thunderbolt transport. This provides bidirectional data rates of 20 or 40 Gbp."
    "Additionally, to fill this Thunderbolt link, the silicon extracts and routes up to 4 lanes of PCI Express Gen 3(4 x 8 Gbps) and up to two full (4 lane) links of DisplayPort out over the Thunderbolt cable and connector to the device(s) attached downstream from the host system."
    How these lanes are divided up depends on the the first device plugged into the host. It's safe to assume a solution like this will use the full thunderbolt spec and probably have all 4 lanes of the PCIe bus the to PCIe x16 slot and use the 4 remain pins for the rest of the io.
  • jordanclock - Thursday, June 20, 2019 - link

    For both 1 and 2, GPUs actually don't use as much bandwidth as you might be under the impression they use and even then it really depends on the game.

    https://www.techpowerup.com/review/nvidia-geforce-...

    Overall the GPU will just have to deal with having variable data rates, but I don't think it will have that big of an impact. The scenario you pointed out of having max utilization of all ports is extremely unlikely and those that would run into those scenarios are likely to be using a separate dock and TB/USB interface for those purposes.
  • Meaker10 - Friday, June 21, 2019 - link

    TB3 controllers hang off the chipset on mainstream platforms. Having a 2nd controller would not improve overall bandwidth as the chipset is limited to the same amount ,(basically 4x pcie 3.0)
  • jenesuispasbavard - Thursday, June 20, 2019 - link

    Any word on price? Hopefully this is an interesting alternative to the Razer Core X Chroma which is $400. That enclosure though has two Thunderbolt 3 controllers, one exclusively to handle the GPU and the other for USB+ethernet. I wonder if this is the same.
  • Flunk - Thursday, June 20, 2019 - link

    That's the important question here. Similar products exist already, so is this cheaper than those?
  • sorten - Thursday, June 20, 2019 - link

    I like the 100W PD, but 500W for the GPU is more than double the power needed for the 2080ti. Even last gen AMD GPUs don't use that much power.
  • mischlep - Thursday, June 20, 2019 - link

    Remember that power supplies hit their peak efficiency levels between 40 and 80% of supported capacity. Having the capacity for double the power of the video card sounds about right.
  • Flunk - Thursday, June 20, 2019 - link

    Also, they pretty much don't make decent quality power supplies under 450W anymore.
  • sorten - Thursday, June 20, 2019 - link

    Interesting. Silverstone has multiple 300W options at Gold or better.
  • sorten - Thursday, June 20, 2019 - link

    Well, the key is to be above 20%. The difference between 20% and 100% rarely varies by more than 2% efficiency.
  • The_Assimilator - Thursday, June 20, 2019 - link

    On the one hand, I appreciate the fact that the only speciality components inside seem to be the motherboard and the front USB board, which means this will *hopefully* be priced decently.

    On the other hand, they could've at least used a modular PSU so there aren't useless Molex connector cables floating around! Even better, an SFX PSU, because the one they've got in there is blocking the single SATA port.

    Sell it in two configurations - one with a PSU bundled, one with no PSU so users can (re-)use their own - and we'll see.
  • zodiacfml - Thursday, June 20, 2019 - link

    I can feel this will be reasonably priced unlike existing ones.
  • KalTorak - Thursday, June 20, 2019 - link

    Did anyone else read the headline and immediately think "That's silly. Why would you build a 750W box and only deliver 100W to it?"
  • Lord of the Bored - Friday, June 21, 2019 - link

    You're right, that IS what it looks like.
    *insert rant about USB-C here*
  • PeachNCream - Thursday, June 20, 2019 - link

    That's a bulky case. I get that it almost has to be in order to handle the PSU and GPU, but considering the end goal (typically just playing a few video games on a computer that otherwise doesn't have a powerful enough GPU) it seems like a lot of additional footprint and mobility compromise compared to just picking things your computer can run well on its existing graphics hardware.
  • Lord of the Bored - Friday, June 21, 2019 - link

    It is ridiculously large, and I feel like they're solving the wrong problem here.

    "If we use a standard ATX power supply, there's a whole bunch of dead space in the box because the card is way longer than the supply, but super-skinny. I guess we'll throw a hard drive mount and a USB hub in that extra space!"

    I mean, I get why they aren't making their own custom power supply. It isn't easy to do it well, and then you have a bunch of boring safety reviews. You can avoid both issues by using someone else's closed box power supply. But a full ATX supply, sourcing the better part of a kilowatt?

    Come on, guys! Get a 400-watt SFX supply.
    Or use a 12V supply in a long, narrow case. Some regulators on the board can generate 3.3v, you don't need a multi-voltage supply.
    You could make this thing so much smaller than an ENTIRE ITX SYSTEM.

Log in

Don't have an account? Sign up now