3 ms·

>Almost every motherboard, certainly every ATX motherboard of the past 10 years or so I've seen, has standard at least 6 SATA ports. Except that the SATA ports

>Almost every motherboard, certainly every ATX motherboard of the past 10 years or so I've seen, has standard at least 6 SATA ports. Except that the SATA ports on any motherboard are not equal. And you didn't even address the SAS point or highspeed U.2 drives for broadcast or data capture or... > There are plenty of excellent USB devices to accomplish this, you don't need a PCI Express-based solution. Evidently you don't work in broadcast or computer vision or machine vision inspection or... > Realtek cornered the market here with onboard audio, and Evidently you don't work in broadcast, or use a DAW, or desire higher grade audio, or special audio processors or low-latency MIDI... >> Legacy adapters for older devices that use serial, parallel, or PCI (non-Express) to connect to the computer >I'm not even sure what you'd attach that's that old... and more importantly, why... Evidently you don't work in robotics or machine vision inspection or industrial control or industrial logging or medical devices or... Before we go any further, USB is very consumer grade tech. We know this because it can be unplugged or work itself loose under vibration. There are locking connector options, but they are for the most part completely inadequate. USB is also prone to incredible cross-talk in noisy environments. I've touched all of these areas in my career, and they all require those connections you are so quick to dismiss out-of-hand. This is an incredibly parochial and naive mindset on exhibit here.

cbozeman 4y ago
Pity I can't downvote you. You deserve it. You didn't read the whole post, especially the most critical part. Lack of PCI Express LANES is the problem, not slots. All this great shit you're talking about is wonderful... as long as you have the PCI Express lanes to support all those add-in cards.
- justinlloyd 4y ago
  I did read your entire post. And the OP post. I have Intel Xeon Skylake processors, 48 lanes of PCIe 3.0, all of equal priority so they aren't tiered like in some of the earlier blue team processors or the red team processors. Across two CPUs that gives me 96 lanes. The slots are switched so either of the CPUs can use an individual PCIe card, or I can assign a PCIe card to a specific CPU. Alternatively if there are not enough PCIe slots I can run the PCIe cards through a separate switch. There are a lot of non-consumer grade motherboards out there that offer these features. It is rare that any deployment will require all 48 lanes on a CPU to be maxed out simultaneously, though I've been involved with use cases where that has taken place (8x 4K non-compressed video stream captures directly to storage), though what happens then is you run in to DRAM bandwidth issues and other problems. In my workstation I have dual RTX A5000 GPUs, dual Blackmagic quad 4K capture cards, dual Highpoint USB 3.0 expanders with four separate USB controllers per board, a Highpoint M.2 RAID controller with on-board PLX, along with the onboard M.2, six channel U.2 through a switch, onboard USB & SATA. What we should be asking for is better switching, not more lanes. Again, it is a rare use case where we can max out the bandwidth of all PCIe lanes of a CPU. We should also be asking for better switching for the direct DMA between cards which is a sorely neglected area across all architectures/motherboards. P.S. Forgot to mention the NIC PCIe card.
Underphil 4y ago
"I've touched all of these areas in my career, and they all require those connections you are so quick to dismiss out-of-hand." I hear you, and I'm usually first to remind people that their own use case is anecdotal and irrelevant. However, are you really suggesting that the use cases you mentioned make up a large enough percentage of the whole to warrant manufacturers catering to them?
- justinlloyd 4y ago
  Yeah, we're talking multiple multi-billion dollar industries. That if the main players won't service them will be serviced by niche players who do. You won't find a USB anything version of an SDI capture card worth (they exist, they're just universally not good) a damn or used in a professional broadcast environment because USB, by de facto, doesn't lock and is temperamental. You will struggle to find a USB multi HDMI capture card. Forget about putting multiple USB capture devices on the single shared USB connection integrated into your motherboard. A motherboard with 12+ ports usually has three distinct USB controllers, only one of which is worth a damn. There are PCIe machine vision capture cards that have onboard GPUs and dedicated co-processors so that the machine vision algorithms can run directly on the PCIe card and never involve the CPU nor have to move the captured video across the bus to main (usually far slower) DRAM. USB has incredibly high latency, and more importantly, non-deterministic latency, which is why USB MIDI on the desktop is fine for casual use, and lousy in an event setting or a professional recording studio.
  - stinos 4y ago
    USB has incredibly high latency, and more importantly, non-deterministic latency Seeing that RME's USB 2 interfaces manage to stream like 50 or more 24bit audio channels at 48kHz with buffer size small enough to get latencies in the mSec ranges, I always wonder: are other manufacturers just doing it wrong? I know that doesn't completely cover the non-determinism argument, but 'incredibly high' seems to be covered pretty well.
    - justinlloyd 4y ago
      The RME-Audio devices do indeed have low latency, at the limits of the USB 2.0 spec, 125 microseconds I believe. They crank up that USB poll rate. And are also using the Arasan chipset IIRC, the same to be found in some of the other prosumer and pro line-up of equipment, e.g. the Solid State Logic h/w. I am hazy on the details, it has been a few years since I was inside any of those devices, people from RME and SSL please feel free to correct me as to your chipsets. Some are using dedicated FPGAs to handle the data capture and processing before handing off to USB. RME's devices are definitely doing a bunch of on-board processing before giving it to the USB bus, and making sure the packets going out are as small as can be. Most non-integrated USB controllers are using VIA or Renesas. USB 2.0 has lower latency but less consistency (shared bus) vs USB 3.0 which has higher latency but is more consistent (point-to-point protocol). Obviously you don't want to go sharing your USB 2.0 port on your PC with an RME and a bunch of USB 3.0 devices, e.g. an external drive, because then you just end up with the worst of both specs, terrible latency and terrible consistency.

cbozeman 4y ago

Pity I can't downvote you. You deserve it. You didn't read the whole post, especially the most critical part. Lack of PCI Express LANES is the problem, not slots. All this great shit you're talking about is wonderful... as long as you have the PCI Express lanes to support all those add-in cards.

justinlloyd 4y ago
I did read your entire post. And the OP post. I have Intel Xeon Skylake processors, 48 lanes of PCIe 3.0, all of equal priority so they aren't tiered like in some of the earlier blue team processors or the red team processors. Across two CPUs that gives me 96 lanes. The slots are switched so either of the CPUs can use an individual PCIe card, or I can assign a PCIe card to a specific CPU. Alternatively if there are not enough PCIe slots I can run the PCIe cards through a separate switch. There are a lot of non-consumer grade motherboards out there that offer these features. It is rare that any deployment will require all 48 lanes on a CPU to be maxed out simultaneously, though I've been involved with use cases where that has taken place (8x 4K non-compressed video stream captures directly to storage), though what happens then is you run in to DRAM bandwidth issues and other problems. In my workstation I have dual RTX A5000 GPUs, dual Blackmagic quad 4K capture cards, dual Highpoint USB 3.0 expanders with four separate USB controllers per board, a Highpoint M.2 RAID controller with on-board PLX, along with the onboard M.2, six channel U.2 through a switch, onboard USB & SATA. What we should be asking for is better switching, not more lanes. Again, it is a rare use case where we can max out the bandwidth of all PCIe lanes of a CPU. We should also be asking for better switching for the direct DMA between cards which is a sorely neglected area across all architectures/motherboards. P.S. Forgot to mention the NIC PCIe card.

Underphil 4y ago

"I've touched all of these areas in my career, and they all require those connections you are so quick to dismiss out-of-hand." I hear you, and I'm usually first to remind people that their own use case is anecdotal and irrelevant. However, are you really suggesting that the use cases you mentioned make up a large enough percentage of the whole to warrant manufacturers catering to them?

justinlloyd 4y ago
Yeah, we're talking multiple multi-billion dollar industries. That if the main players won't service them will be serviced by niche players who do. You won't find a USB anything version of an SDI capture card worth (they exist, they're just universally not good) a damn or used in a professional broadcast environment because USB, by de facto, doesn't lock and is temperamental. You will struggle to find a USB multi HDMI capture card. Forget about putting multiple USB capture devices on the single shared USB connection integrated into your motherboard. A motherboard with 12+ ports usually has three distinct USB controllers, only one of which is worth a damn. There are PCIe machine vision capture cards that have onboard GPUs and dedicated co-processors so that the machine vision algorithms can run directly on the PCIe card and never involve the CPU nor have to move the captured video across the bus to main (usually far slower) DRAM. USB has incredibly high latency, and more importantly, non-deterministic latency, which is why USB MIDI on the desktop is fine for casual use, and lousy in an event setting or a professional recording studio.
- stinos 4y ago
  USB has incredibly high latency, and more importantly, non-deterministic latency Seeing that RME's USB 2 interfaces manage to stream like 50 or more 24bit audio channels at 48kHz with buffer size small enough to get latencies in the mSec ranges, I always wonder: are other manufacturers just doing it wrong? I know that doesn't completely cover the non-determinism argument, but 'incredibly high' seems to be covered pretty well.
  - justinlloyd 4y ago
    The RME-Audio devices do indeed have low latency, at the limits of the USB 2.0 spec, 125 microseconds I believe. They crank up that USB poll rate. And are also using the Arasan chipset IIRC, the same to be found in some of the other prosumer and pro line-up of equipment, e.g. the Solid State Logic h/w. I am hazy on the details, it has been a few years since I was inside any of those devices, people from RME and SSL please feel free to correct me as to your chipsets. Some are using dedicated FPGAs to handle the data capture and processing before handing off to USB. RME's devices are definitely doing a bunch of on-board processing before giving it to the USB bus, and making sure the packets going out are as small as can be. Most non-integrated USB controllers are using VIA or Renesas. USB 2.0 has lower latency but less consistency (shared bus) vs USB 3.0 which has higher latency but is more consistent (point-to-point protocol). Obviously you don't want to go sharing your USB 2.0 port on your PC with an RME and a bunch of USB 3.0 devices, e.g. an external drive, because then you just end up with the worst of both specs, terrible latency and terrible consistency.