281–290 of 333 posts
I find it amusing that every single one of our high-speed standards (ethernet, HDMI, PCIE, displayport, USB) are basically just the same high-speed differential pairs under the hood with different framing. I don't know exactly what we can do about this situation, but at some point it probably makes sense for all of them to converge into a multi-lane, differential-pair standard (or just support nesting each other arbi…
I wonder whether it's a software or hardware problem? Let's say you have the thunderbolt connector, shouldn't all those supported standards (esp. the usb) then just be software problems? This does not make it easier for apples side, it might still be a huge undertaking, but then it would be upgradable in principle.
The block size used isn't specified, and that has a significant impact on performance. The exact model drives used aren't specified, but elsewhere in the comments someone references the Crucial X8. With 32KB block sizes, the transfer speed is ~520 MB/s. https://www.anandtech.com/show/16186/crucial-portable-ssd-x6...
Earlier quoted context omitted.
How many parallel differential pairs can you have before it's no longer serial? Hah.
You can have arbitrarily many, so long as the serial lanes are independent of each other. PCIe can have lots, for example. Because they are independent, they won't suffer from clockskew problems that made parallel signaling impractical at high speeds and it's often not required to connect all of the wires (e.g. an x16 card hanging out of an x8 or smaller slot) .
Earlier quoted context omitted.
They still don't support DP MST, so you have to use multiple TB ports or use a monitor that supports TB daisy-chaining.
Or a dock. Not disputing the DP MST, just noting there are docks that can drive 2 x 6K. CalDigit's new dock takes a single TB4 cable to drive, and can do: - 2x TB4 - 1x DP 1.4 - supporting - 2x 5K60 OR - 1x 8K60 OR - 1x 4K144 OR - 1x 1440p240 According to https://dancharblog.wordpress.com/2021/02/05/usb4-tb4-docks/ Or, as CalDigit puts it (see final paragraph): Single 8K or Dual 6K Monitors Users on Windows can conne…
The only true 5K monitors I know of are the LG 5k and the Apple Studio Display which are both TB or USB-C monitors, not DP. 5K60 takes ~25gb/sec, multiply times two and the single TB cable as input can’t handle 50gb/sec.
I find it amusing that every single one of our high-speed standards (ethernet, HDMI, PCIE, displayport, USB) are basically just the same high-speed differential pairs under the hood with different framing. I don't know exactly what we can do about this situation, but at some point it probably makes sense for all of them to converge into a multi-lane, differential-pair standard (or just support nesting each other arbi…
IIRC Thunderbolt essentially is "PCIE but in a cable"
They wouldn't be able to do daisy chaining if it was bare PCIe, but it does sound horrible.
Earlier quoted context omitted.
You can have arbitrarily many, so long as the serial lanes are independent of each other. PCIe can have lots, for example. Because they are independent, they won't suffer from clockskew problems that made parallel signaling impractical at high speeds and it's often not required to connect all of the wires (e.g. an x16 card hanging out of an x8 or smaller slot) .
That's just kicking the can down the road to even less efficient software that must now manage synchronizing all the lanes.
Earlier quoted context omitted.
> Feel free to point out exactly which argument you feel was ignored. The retroactive renaming of speed+versions. I'm not talking about the Mac. > If you pick up a USB3.1 device you know beforehand it won't support SuperSpeed+ 2x or 4x. My whole argument is that this confusion wouldn't be an issue if the USB Consortium had reserved USB 3.1 for 10 Gb/sec speeds exclusively . In other words, this: 3.0: 5 Gb/s "SuperSpe…
Versions are not speeds. 3.2 continues to describe everything in 3.0, which means it continues to describe how to make devices supporting 5 gbps over USB-A/B
The inability of the USB standards body to stick to a versioning scheme beggars belief: what in the world could "USB 3.1 Gen 2" possibly mean? Versions are free, why not simply name it "USB 3.2" and make the next one 3.3? "Gen 2" strikes me as the "Essay FINAL (final version 2).docx" of standards revisions.
The USB-IF is a group consisting of implementers (IF = Implementers Forum) and obviously it's more profitable for them if they can put "USB 3.2 support" on the box, but keep using the same hardware (newer and more capable => more expensive => lower profit). So in a perverse way - the USB-IF actually has a financial incentive to keep making more confusing standards , because consumer confusion can be exploited for hig…
Well, seems like because physics™ it's hard to do both 420W charging and 69Gbps data over the same cable. And also people want thinner and cheaper charging cables. It's still better to have a unified connector and different capability cables.
> most cables get it wrong to a dangerous degree
That was the case in the early days. These days it's getting hard to find bad cables.
> negotiation is more complex and expensive to implement as well, even if you're only doing a low-capability device
No need for negotiation in a low capability device. Two resistors. It's not rocket science. Raspberry Pi did screw up and use one resistor but they made that mistake so famous that everyone knows not to repeat it now.
Earlier quoted context omitted.
Isn't any filesystem that runs on top of a SSD or flash drive at risk from the SSD/flash drive controller corrupting the disk or making it inaccessible, when the volume is not cleanly unmounted? I can remember losing a USB drive that way, even though it was using a journaling filesystem.
Potentially yes, realistically no. This is the old "if the storage controller is lying about flushing to disk, there's not a lot you can do about it" problem. ZFS does have defensive mechanisms, like doing a read after write to try and be sure that what is written is actually committed to disk, but if the storage controller chooses to serve that out of cache then that could be a lie too. It's the old "trusting trust"…