Raspberry Compute Module IO Board

A point to take note of

The physical layer is the part that is 10, 100 or 1000 Mbits, nothing above that even really cares and it is only relevent between the two directly connected devices (Pi and a HUB for instance). if any other devices are not compatable at the same speed when the connection is established then the two connected devices agree on a common speed, often with modern chips also compensating for non MX cables too if needed. if a PI connected to a hub is 100Mbit and on the hub is the PC at 1GBit, the HUB will buffer and speed shift between the two

The TCP/IP layer is completly seperate and it often has a bigger impact as it decides the packet size, typically 1500bytes per packet but this can again be over-ridden if desired and you know what your doing

USB is as correctly stated a 480MBit bus so the PI irrespective of its performance will never get to exceed that, meaning it will never be able to make full use of a 1GBit LAN but can participate on one, the higher speed will mean it gets its data out of the chip and on its way quicker and therefor will free up the LAN quicker. If the PI was uprated to USB3 then of course things improve quite a bit, or a LAN chip is more directly connected to the chip / CPU bus.

So does a faster LAN chip have benefits... yes. will the app get its data out quicker if everything else is the same, probably not or only marginally so is it really worth it, you can judge that yourself

as far as designing a board to be faster for networking and USB, have you looked at this :- Meet MinnowBoard MAX | minnowboard.org or this ODROID | Hardkernel, basically the job your trying to do has been done and for a good price too so unless your doing it purly for the intelectual excersize, your trying to replicate whats been done already

And thats OK, I just want you to understand what your facing

Well actually it would be  a bit like having  a switch between two computers where one uses 100Mbit Ethernet and the other uses 1Gbit. In this case the LANxxx chip will be buffering the data but the OP won't gain any great advanatge (if any) in throughput and will have the disadvantage of cutting himself off from any community support for the existing and normal solution. It sounds  a bad plan to me.

@Cody  - it looks to me as if you have neither experience or test gear - in which case your only chance of making the the high speed stuff work is to copy exactly the RPIi breakout board for the tracking of high speed signals. (ie - if they use 0..5 oz copper tracks 0.2mm wide, 0.5mm apart separated from a ground plane by xmm of FR4 board then you must too.) The problem you will find is that they don't give you a step by step guide. You may get lucky but this is more a case of trying to fly before you can walk !

Why are you using the compute board at all - isn't there any easier way of getting where you want be ?

MK

screamingtiger

     Actually it would be 1000 million bit per second theoretically which comes to a bit more than 100 million bytes per second which overwhelms USB's 480Mbit/s. I would say the problem comes when using external USB disk like SSD which could saturate the interface.

C

Edit corrected statement.

Since I couldn't remember off the top of my head I had to go back through the datasheets and it appears that it can only handle up to 480 Mbps at the High end. So it would appear that using the LAN7500 chip would be overkill in my project.

Oh, okay. When you reviewed the USB bus, what kind of throughput can it handle? (Genuinely interested to know.)

Yes the Ethernet chip still connects via the USB bus. I have already reviewed that. And will be reviewing the specs on the Broadcom processor.

Does the current Ethernet chip not connect over the USB bus?

I question if the Raspberry Pi could handle 1000MBS, it may not have enough CPU power to keep up.

Honestly I am still in the learning phases of circuitry and such. The board i'm looking to make is or should be pretty basic. It will have 1 HDMI, 1 LAN and 2 USB's to start. Eventually I will add an IR receiver for remote control and possibly on board WiFi. Any help and direction in learning more about designing and developing circuit boards will be greatly helpful.

A semi clone of the RPi IO board is quite a challenge to learn on !

If you want the video to work make sure you understand about impedance matching and tracking high speed controlled impedance differential pairs !

How will you test your board ?

MK