04-15-2020, 02:32 PM
While I am disappointed in the PineTab delay, I also completely understand the current-logistics decision. Thank you so much for dedicating a brief section about it. The attention-currency it has received has been, IMNSHO, underwhelming.
On the RockPro router idea:
That's great! Why can't we use off-the-shelf PCIe cards? (I don't have a PCIe-capable Pine64 board, so I'm not following this closely) I think it makes more sense to include a 3rd party card choice in the Pine Store, and maintain a list of "officially supported" (ie: thoroughly tested) other cards, than it does to develop your own expansion card(s). Pick a kernel driver developer with existing and mature code, send them a SBC and ask for better ARM support. Don't limit this to network cards. Maybe I want to plug in a video card. Maybe I want to plug in a card with M.2 ports; or a shortwave radio (HAM) antenna; or AM/FM tuner, or TV capture; or live-stream capture/export; or sound cards; etc etc. Yes, some of these are more cheaply available [new] using the GPIO pins, but there's also a whole bunch of legacy cards installed in aging gear; and new hardware coming out all the time with greater capabilities than what's available via GPIO header. Seriously, you should market and support that PCIe slot much stronger than you already do.
If the case you're designing includes a standard PCI plate slot, there's no reason why it would need custom openings for a custom card.
If you're going to do a custom card, why not a host for SoPines?
On the subject of the SBCs:
I really think SBCs should be (still are?) the core of the Pine64 business and should routinely see updates in the blog. I'm not in a position to write SBC-dedicated blog posts; and I'm not sure that's necessary. However, I appreciate seeing even short mentions of development happenings.
It seems to me that the SBCs have suffered in the attention-currency department since the announcement of the PinePhone and PineBook Pro. The rk3399 is getting long in the tooth and is RAM-throttled. Meanwhile, application suites have become more resource hungry. Google TV devices in the same price ranges as the Pine64 product line are considerably more compute-capable using AMLogic or other SoCs, and come pre-installed in an enclosure, with power-supply, and don't need an SD card! Even the Raspberry Pi now outperforms everything on offer from Pine64, at a lower price.
SoPine hasn't changed significantly since its initial introduction. The Clusterboard could use better spacing to allow for (passive) heat sinks on the modules, and it's own processor for module management; and what if it were a standardized PCB size (and port cluster) to fit available enclosures?
Other PCBs:
Actual desktop-grade boards with expandable RAM? Nobody's doing it at a price competitive with (sub-$1000) solutions from Intel/AMD systems. Yet, processors have been available for four or five years, now.
Servers? How about a rack-mount chassis with backplane to support, say, 10 ClusterBoards. Or even the earlier suggested desktop-grade PCBs. There are already ARM servers in the wild, but they are all very expensive, very custom, very proprietary systems. Would anyone change out their existing ATX-based, ancient, Intel server motherboard for shiny, new, lower-electricity ARM board in the same form factor from Pine64?
Other thoughts and questions:
Will you make available that back-plane in the server-chassis photo?
Power supplies. How many people with SBC clusters have had to kluge together their own? (me! My A64 cluster is powered from a USB hub; awkward but effective) Also handy for any PCIe card(s) that need power not available from the SBC.
Okay, this has run long enough and wandered way off-topic.
On the RockPro router idea:
That's great! Why can't we use off-the-shelf PCIe cards? (I don't have a PCIe-capable Pine64 board, so I'm not following this closely) I think it makes more sense to include a 3rd party card choice in the Pine Store, and maintain a list of "officially supported" (ie: thoroughly tested) other cards, than it does to develop your own expansion card(s). Pick a kernel driver developer with existing and mature code, send them a SBC and ask for better ARM support. Don't limit this to network cards. Maybe I want to plug in a video card. Maybe I want to plug in a card with M.2 ports; or a shortwave radio (HAM) antenna; or AM/FM tuner, or TV capture; or live-stream capture/export; or sound cards; etc etc. Yes, some of these are more cheaply available [new] using the GPIO pins, but there's also a whole bunch of legacy cards installed in aging gear; and new hardware coming out all the time with greater capabilities than what's available via GPIO header. Seriously, you should market and support that PCIe slot much stronger than you already do.
If the case you're designing includes a standard PCI plate slot, there's no reason why it would need custom openings for a custom card.
If you're going to do a custom card, why not a host for SoPines?
On the subject of the SBCs:
I really think SBCs should be (still are?) the core of the Pine64 business and should routinely see updates in the blog. I'm not in a position to write SBC-dedicated blog posts; and I'm not sure that's necessary. However, I appreciate seeing even short mentions of development happenings.
It seems to me that the SBCs have suffered in the attention-currency department since the announcement of the PinePhone and PineBook Pro. The rk3399 is getting long in the tooth and is RAM-throttled. Meanwhile, application suites have become more resource hungry. Google TV devices in the same price ranges as the Pine64 product line are considerably more compute-capable using AMLogic or other SoCs, and come pre-installed in an enclosure, with power-supply, and don't need an SD card! Even the Raspberry Pi now outperforms everything on offer from Pine64, at a lower price.
SoPine hasn't changed significantly since its initial introduction. The Clusterboard could use better spacing to allow for (passive) heat sinks on the modules, and it's own processor for module management; and what if it were a standardized PCB size (and port cluster) to fit available enclosures?
Other PCBs:
Actual desktop-grade boards with expandable RAM? Nobody's doing it at a price competitive with (sub-$1000) solutions from Intel/AMD systems. Yet, processors have been available for four or five years, now.
Servers? How about a rack-mount chassis with backplane to support, say, 10 ClusterBoards. Or even the earlier suggested desktop-grade PCBs. There are already ARM servers in the wild, but they are all very expensive, very custom, very proprietary systems. Would anyone change out their existing ATX-based, ancient, Intel server motherboard for shiny, new, lower-electricity ARM board in the same form factor from Pine64?
Other thoughts and questions:
Will you make available that back-plane in the server-chassis photo?
Power supplies. How many people with SBC clusters have had to kluge together their own? (me! My A64 cluster is powered from a USB hub; awkward but effective) Also handy for any PCIe card(s) that need power not available from the SBC.
Okay, this has run long enough and wandered way off-topic.
