Re: [pfSense] New intel atom board
On Apr 5, 2014, at 12:48 PM, Ugo Bellavance u...@lubik.ca wrote: http://techcrunch.com/2014/04/03/intel-releases-99-minnowboard-max-an-open-source-single-board-computer/?utm_campaign=fbncid=fb An interesting platform for pfSense? It looks like it only has 1 NIC though. I looked at this earlier in the week when it was released. It’s interesting, (AES-NI and VT-x support! http://ark.intel.com/products/78475/Intel-Atom-Processor-E3845-2M-Cache-1_91-GHz) and Circuitco is just up the highway in Richardson, TX. I’ve considered driving up and seeing what it would take to take the schematics (when they are available) and have a board built with 2 Ethernets (rather than one), and maybe a miniPCIe socket (for an 802.11 NIC, as pfSense 2.2 should make a lot more of these work, or possibly an m-sata drive), in addition to pulling the expansion header off, and connectorizing the serial ‘debug’ header for a proper console. We would need a simple enclosure as well.Painted (or powder-coated) steel is less expensive than anodized aluminum, but I think the anodized aluminum looks nicer, and it can be laser engraved. The other issue is single or dual core and 1GB or 2GB ram (4GB?)? How interesting is the m-sata / miniPCIe option? How you can help: Indicate your level of interest. This board would without a doubt cost more than the minnow board. I don’t know how much more, but we’re not going to hit the same volumes as the minnow board. (I could be wrong.) The minnow board could be subsidized by Intel. (I could be wrong.) It’s going to require a significant investment (up-front NRE), an investment in getting a run of these made, and some return on those investments (profit). How important is form-factor? Larger PCBs cost more, but can sometimes relax routing enough to not need additional layers (fewer layers tend to cost less). - miniPCIe is going to require a connector (these cost money to both buy and place) - m-sata also requires a switch, such that if the m-sata drive is in-place it is connected to the SATA controller - RAM costs. At these densities, 2GB of ram costs twice as much as 1GB of ram. 4GB of ram costs 4X as much as 1GB of ram. making lots of different variants of the boards costs extra to both manufacture (stop the line, load the new parts, run the new SKU) and inventory. - dual core or single core?Remember that pfSense 2.2 (which is based on FreeBSD 10) supports a pf capable of multi-threading. Jim ___ List mailing list List@lists.pfsense.org https://lists.pfsense.org/mailman/listinfo/list
Re: [pfSense] New intel atom board
On 14-04-05 02:02 PM, Jim Thompson wrote: http://techcrunch.com/2014/04/03/intel-releases-99-minnowboard-max-an-open-source-single-board-computer/?utm_campaign=fbncid=fb An interesting platform for pfSense? It looks like it only has 1 NIC though. I looked at this earlier in the week when it was released. It’s interesting, [...] and Circuitco is just up the highway in Richardson, TX. I’ve considered driving up and seeing what it would take to take the schematics (when they are available) and have a board built with 2 Ethernets (rather than one), and maybe a miniPCIe socket (for an 802.11 NIC, as pfSense 2.2 should make a lot more of these work, or possibly an m-sata drive), in addition to pulling the expansion header off, and connectorizing the serial ‘debug’ header for a proper console. Given the high up-front costs to produce a variant board, wouldn't it be easier, faster and cheaper to just use the expansion header, which IIRC includes two PCIe 1x lanes? If a breakout cable existed that provided 2 PCIe slots, it would be possible to simultaneously have much more flexibility in enclosure design (e.g. PCIe cards underneath the board?) as well as flexibility in choice of add-on. I don't see that a breakout cable exists yet for the high-speed expansion bus, so there's that minor (*cough*) problem... but that seems a much smaller problem than re-tooling the board. We would need a simple enclosure as well.Painted (or powder-coated) steel is less expensive than anodized aluminum, but I think the anodized aluminum looks In case you don't have a local firm you're happy with, talk to Protocase for sample qtys. I've seen them be cheaper than mass mfg for small runs of simple cases (e.g. interlocked-U style). The other issue is single or dual core and 1GB or 2GB ram (4GB?)? The stock 2GB version should be adequate (barely) IMHO for most applications that function with that class of CPU/ethernet/storage anyway. Much more interesting to me would be if a small, low-cost board like that were available with ECC. That CPU does support ECC RAM, after all... How interesting is the m-sata / miniPCIe option? Not to me, as I tend to deploy pfSense at the higher-end of the spectrum, but *some* way to add WiFi would probably be important for the putative target audience. USB probably won't cut it for an AP, so mPCIe is probably needed. Again, expansion-header-to-mPCIe should be possible instead of reworking the board... and unlike PCIe 1x sockets, that wouldn't take up much more room than putting the mPCIe headers on the board. How you can help: Indicate your level of interest. Neat, but not commercially interesting to me right now. Linksys/ASUS/D-Link make cheaper gateways that are good enough for home users, and commercial users will either get a FortiWiFi (or equivalent) or if pfSense, re-use an existing rackmount server. This board would without a doubt cost more than the minnow board. I don’t know how much more, but we’re not going to hit the same volumes as the minnow board. (I could be wrong.) The minnow board could be subsidized by Intel. (I could be wrong.) See above comments :-). I'm not sure if a breakout cable is 100% workable, but if so it's a faster/cheaper option than mPCIe. It’s going to require a significant investment (up-front NRE), an investment in getting a run of these made, and some return on those investments (profit). How important is form-factor? Larger PCBs cost more, but can sometimes relax routing enough to not need additional layers (fewer layers tend to cost less). Smaller is better. Otherwise I may as well just deploy a miniITX or 1U system. Which, yes, argues *against* using a breakout cable for PCIe. - dual core or single core?Remember that pfSense 2.2 (which is based on FreeBSD 10) supports a pf capable of multi-threading. Good question - optimize for today or for tomorrow? -- -Adam Thompson athom...@athompso.net ___ List mailing list List@lists.pfsense.org https://lists.pfsense.org/mailman/listinfo/list
Re: [pfSense] New intel atom board
On Apr 5, 2014, at 5:06 PM, Adam Thompson athom...@athompso.net wrote: On 14-04-05 02:02 PM, Jim Thompson wrote: http://techcrunch.com/2014/04/03/intel-releases-99-minnowboard-max-an-open-source-single-board-computer/?utm_campaign=fbncid=fb An interesting platform for pfSense? It looks like it only has 1 NIC though. I looked at this earlier in the week when it was released. It’s interesting, [...] and Circuitco is just up the highway in Richardson, TX. I’ve considered driving up and seeing what it would take to take the schematics (when they are available) and have a board built with 2 Ethernets (rather than one), and maybe a miniPCIe socket (for an 802.11 NIC, as pfSense 2.2 should make a lot more of these work, or possibly an m-sata drive), in addition to pulling the expansion header off, and connectorizing the serial ‘debug’ header for a proper console. Given the high up-front costs to produce a variant board, wouldn't it be easier, faster and cheaper to just use the expansion header, which IIRC includes two PCIe 1x lanes? If a breakout cable existed that provided 2 PCIe slots, it would be possible to simultaneously have much more flexibility in enclosure design (e.g. PCIe cards underneath the board?) as well as flexibility in choice of add-on. The expansion header only includes one PCIex1 2.0 lane, 1x SATA2, 1x USB 2.0 host, I2C, GPIO, JTAG, +5VDC, GND http://www.minnowboard.org/meet-minnowboard-max/ I don't see that a breakout cable exists yet for the high-speed expansion bus, so there's that minor (*cough*) problem... but that seems a much smaller problem than re-tooling the board. We would need a simple enclosure as well.Painted (or powder-coated) steel is less expensive than anodized aluminum, but I think the anodized aluminum looks In case you don't have a local firm you're happy with, talk to Protocase for sample qtys. I've seen them be cheaper than mass mfg for small runs of simple cases (e.g. interlocked-U style). We have a local firm we’re pretty happy with. We also have a lot of experience in injection molding now (smallworks.com) The other issue is single or dual core and 1GB or 2GB ram (4GB?)? The stock 2GB version should be adequate (barely) IMHO for most applications that function with that class of CPU/ethernet/storage anyway. Much more interesting to me would be if a small, low-cost board like that were available with ECC. That CPU does support ECC RAM, after all… yes it does. ECC ram is also a lot more expensive. How interesting is the m-sata / miniPCIe option? Not to me, as I tend to deploy pfSense at the higher-end of the spectrum, but *some* way to add WiFi would probably be important for the putative target audience. USB probably won't cut it for an AP, so mPCIe is probably needed. Again, expansion-header-to-mPCIe should be possible instead of reworking the board... and unlike PCIe 1x sockets, that wouldn't take up much more room than putting the mPCIe headers on the board. see above. How you can help: Indicate your level of interest. Neat, but not commercially interesting to me right now. Linksys/ASUS/D-Link make cheaper gateways that are good enough for home users, and commercial users will either get a FortiWiFi (or equivalent) or if pfSense, re-use an existing rackmount server. This board would without a doubt cost more than the minnow board. I don’t know how much more, but we’re not going to hit the same volumes as the minnow board. (I could be wrong.) The minnow board could be subsidized by Intel. (I could be wrong.) See above comments :-). I'm not sure if a breakout cable is 100% workable, but if so it's a faster/cheaper option than mPCIe. It’s going to require a significant investment (up-front NRE), an investment in getting a run of these made, and some return on those investments (profit). How important is form-factor? Larger PCBs cost more, but can sometimes relax routing enough to not need additional layers (fewer layers tend to cost less). Smaller is better. Otherwise I may as well just deploy a miniITX or 1U system. Which, yes, argues *against* using a breakout cable for PCIe. - dual core or single core?Remember that pfSense 2.2 (which is based on FreeBSD 10) supports a pf capable of multi-threading. Good question - optimize for today or for tomorrow? Back when I was a teenager, I liked to hang out in the local speed shop. There was a plaque on the wall, with a very bent connecting rod, and the following lettered below it: “Speed costs money, son. How fast do you want to go?” This was before Mad Max appropriated it: http://www.imdb.com/title/tt0079501/quotes?item=qt0427399 Jim ___ List mailing list List@lists.pfsense.org https://lists.pfsense.org/mailman/listinfo/list
