rob53

About

Username
rob53
Joined
Visits
273
Last Active
Roles
member
Points
9,125
Badges
2
Posts
3,383
  • Microsoft details macOS vulnerability that allowed protected data access

    And we’re congratulating the largest vendor of malware for what reason? 
    F_Kent_D12Strangerscat52williamlondonleighrMplsPwatto_cobra
  • Everything we know about the redesigned MacBook Air with M2 processor

    Is that two TB ports sharing the same limited power so you can’t run two unpowered external TB drives or will they fix this? Any work on fixing the limited write speeds of TB drives?

    Haven’t seen any testing on M1Pro/Max showing faster external write speeds. Please, someone tell me this has been fixed. 
    lkrupp
  • Intel's Alder Lake chips are very powerful, and that's good for the entire industry

    I like to compare Apple vs Intel with Tesla vs GM (possibly Ford). GM is definitely done, they have no way to even try and catch Tesla. They can talk the talk, like Intel, but talk is cheap. Sure, GM will continue to make vehicles until the money runs out but like Intel they really don’t have anything that compares favorably in the long term to Tesla/Apple. Game, set already has happened for both. We’ll see how long it takes for match. 
    williamlondonpatchythepirateAlex_Vbyronlwatto_cobra
  • Intel's Alder Lake chips are very powerful, and that's good for the entire industry

    It’s funny Intel calls them mobile processors when the only way to get their documented performance is to have them plugged in while in cold storage. Every comparison I’ve seen shows Intel processors dropping 70-80% when unplugged. I bet Apple could dial up the M1’s power and beat Intel’s scores but no reason to. The M1 runs the same whether plugged in or on batteries yet I doubt Intel will admit their results were all done with the computer plugged in. You might as well call it a desktop. 

    Once Apple releases its desktop versions Intel will have to go back to the drawing board (sorry, I grew up with these) and try and figure out just how high they can turn the power up especially since they can’t seem to reduce the production facilities like others have. 
    mwhitewilliamlondonscstrrfpatchythepirate9secondkox2pscooter63Alex_VJWSCbyronlwatto_cobra
  • New Sonnet Echo 5 hub has three downstream Thunderbolt 4 ports, one USB-A

    Detnator said:
    rob53 said:
    This hub is like OWC's 4-port Thunderbolt hub. It also says it can drive two 4K displays and costs $20 less. The Sonnett hub will also have limitations when connecting too many devices that want to use as much of the TB bandwidth as possible. I've seen questions about using a hub like this to create a RAID using three NVMe external drives and software RAID. It might be fun to try but I wouldn't expect the full 5000Mb/s TB3 bandwidth. This hub will be used to connect multiple devices that don't require simultaneous use. 

    I'd love to be proved wrong but I only have a 2020 M1 MBA to test with and it has Thunderbolt limitations.
    I have two of the OWC hub you mentioned, and a number of high end NVMe drives in various enclosures, and I've tried to do exactly this.  Sadly, there are some limitations.

    For one, I'm not exactly sure of the details, but I believe the 40Gb/s TB channel (in TB3 at least, and I believe this part is unchanged in TB4) is split into some specifically for data and the rest specifically for video, and neither gets the full 40. It's likely there's some part of that sentence that isn't quite right, but the upshot, as I understand it, is that the maximum SSD throughput you can get out of a single TB3 channel is about 2800MB/s or 2.8GB/s  (note: MBytes/GBytes, not Mbits/Gbits).

    This 2800 is usually stated on TB drives and enclosures from companies who are honest with their marketing (while everyone else says misleading statements like "full 40Mb/s bandwidth"), and I would say it mostly agrees with my experience in practice... or at least that's the case under ideal conditions, though most conditions are short of ideal (eg. background processes doing stuff that chews up some of the bandwidth etc).

    Based on those numbers I bought myself two OWC 4M2 TB3 enclosures and I get about 2200 MBps throughput on each of those, and when I stripe them all I get at most about 3500, usually not quite that much. That's the case on both my maxed out 2020 M1 13" MBP and my maxed out 2019 Intel 16" MBP.  And I've generally got similar performance out of other striped NVMe SSD configurations I've tried (eg. with some of the bus powered single drive enclosures).  When I had an iMac Pro I got better speeds out of the same configurations.  But despite trying various different configurations and setups, no setup has ever got me anywhere near the 5000MBps you'd expect/hope from a theoretically maxed out single 40Gbps TB channel, and in fact, I've never even got that much out of two channels striped.

    Important to note that all this is per CHANNEL, not per port. The Intel Macs share two ports per channel, so the four port Intel MBPs and iMac Pro have two channels. The two port Intel 13" MBP and MBA have one TB channel. On all Apple Silicon Macs so far I believe it's one port per channel, which means the low end 13" M1 MBP and MBA have two channels, despite only two ports -- which is an upgrade over the 2-port Intel machines they replaced.  And on the new 2021 14" and 16" M1 Pro/Max MBP's, the three ports, although a downgrade in number of ports are a 50% upgrade in bandwidth over the previous 4-port (2-channel) 13" and 16" Intel MBPs.

    That's a bit of a ramble, to include a lot of detail.  The point is it's my understanding that the maximum theoretical data (ie. SSD drive) throughput of any single thunderbolt channel is 2800MBytes/s.  And in practice it's generally significantly slower.  I believe there are some other overheads interfering with drive performance, which no doubt include background processes trying to do stuff among other things.  The other factor I've found (and this might be where the iMac Pro benefited most) is that these ports get very hot, and if allowed to get too hot will burn out the electronics, and they have built in throttling to slow them down if they get there.  Things like ice packs or keeping your Mac in a refrigerator might help with that... that's something I want to try one day.

    So needless to say... unfortunately these hubs don't really improve speed because the bottleneck is the channel more than anything else.  Some drive enclosures are slower than the channel bottleneck, but the best these hubs will give us is bringing multiples of those enclosures, striped over one channel, up to similar speeds as the already fast enclosures can do. There might also be some benefit in spreading the heat around, although there's still the heat at the single host port, the management of which may or may not have improved in the newer MBPs.

    I'm keen to get a new M1 Max MBP and see if that's the case, and how fast I can get drives striped across all three channels to go. I'll probably do a brand new clean install to minimize other overheads, and for theoretical/experimental interest I might try it in the fridge and see if that helps. 😉   I'll report back my findings when I can get my hands on one of those machines.


    Thanks for correcting my Mb/s to MB/s. The 4M2 has a single channel interface while the Envoy Express has a dual and the Envoy Pro I believe has a four-channel TB interface. As you know the 4M2 has a single TB input and might have a hardware RAID, I didn't check. Nothing wrong with the 2000 MB/s but it can go higher. My 2020 M1 MBA (same hardware in MBP, I bought 512GB storage so got 8+8 cores) has hardware limitations imposed by Apple on write speeds to external drives. This has been discussed on the OWC forums extensively. I'm still trying to get OWC/MacSales or a youtube person to actually so the same tests on the 2021 MBPs.

    If you look at the fine print on the OWC/MacSales website you'll see that these fast speeds are produced on a Windows PC with a special Thunderbolt interface card. If they can get to this speed on a Windows PC they can get there on a Mac as long as Apple doesn't restrict it. I know the Envoy Pro running on a 21.5" Intel iMac as the system disk can go over 2000GB/s because I put one on to get rid of the internal 5400 rpm sloooooooooooow HDD. 

    There's some tricky nomenclature about PCIe channels (might be called lanes). There's the x1, x2, x4 and not sure how high it goes. These are different from PCI buses. The 2020 M1 Macs share one PCIe bus for the two ports. The 2021 MBPs each use their own PCIe bus for each USB4 port. BTW: USB4/TB4 have the same Thunderbolt speed, 40Gb/s, as Thunderbolt 3. TB4 has a few extra things it can do. I have the OWC Envoy Express with the Aura/Phison NVMe blade. It has two channels. I also have the Fledging which uses the same Phison blade but it's enclosure interface is PCIe x4 (four channels). The fledging does run faster but nowhere near 2800 MB/s on my MBA. 

    I agree with your statement in the second to last paragraph which is why I suggested the hub is good for multiple devices that aren't running at same time. I saw a SSD with hub that allows four computers to connect to it but also lets you connect two TB ports to it (using separate buses/channels/????) to get more speed when striping the NVMe blades. I hope the 2021 MBPs let you do the same thing but unless you're using a powerful RAID box with 2 or more TB inputs I don't think we'll ever get anything approaching the full bandwidth. That said the M1 Pro and M1 Max are really fast when using the unified memory so Apple is doing something right that hopefully will be made available to external storage using that "hidden" bus discovered on the latest M-series SoCs.
    watto_cobrakurai_kage