<eddyb>
but its comparison table says 0 for SERDES, that seems a bit disappointing? is it just not exposing SERDES available in the FPGA or does it only come with FPGAs without SERDES?
<eddyb>
oh huh "ECP5 85F" is confusingly eliding the LFE5U part which has an UM instead of U to indicate it comes with SERDES
<daveshah>
Yup no SERDES and the pins aren't broken out
<daveshah>
(the non SERDES parts are obviously the same die with a possibly-out-of-spec SERDES, but if the pins aren't accessible that doesn't really help)
<eddyb>
also, 1 SERDES for Versa? not sure that seems right given everything it offers
<daveshah>
Yeah it has four SERDES channels in total
<daveshah>
One for PCIe, one broken out to SMAs, two used for a SGMII connection to the Ethernet PHYs (these also have RGMII)
<daveshah>
Also, it has 1024Mb of RAM not 1024MB
<eddyb>
daveshah: and each channel is duplex?
<daveshah>
Yes
<daveshah>
The EVN also has four duplex channels, not "9 SERDES" (the one extra pair is a reference clock input)
<eddyb>
(the off-by-half errors can get frustrating when it's not obvious what is being counted :P)
<eddyb>
thanks!
<flea86>
eddyb: To give you a better understanding of their comparison table, their board and my own FleaFPGA Ohm board don't break out serdes in any way :)
<flea86>
(but I think all of the others do)
<eddyb>
flea86: I'm kinda curious what these boards are designed for. mostly to take advantage of the higher LUT count (compared to iCE40)? parallel LVDS?
<flea86>
Well, the ECP5 series is cheap, yet useful, in the manner you describe. Yes.
<eddyb>
I kinda just want a tiny-ass board that's mostly just a PCIe x1 board edge and an ECP5 :P
<flea86>
It's also fun :D
<flea86>
heh
<eddyb>
bonus points for having a version that can plug into my X230 thinkpad
<eddyb>
I forget what standard that is even using (at this point it might be easier to design myself such a board, doubt anyone is going to)
<eddyb>
flea86: wait hang on I just noticed the cardedge - what's that?
<flea86>
Well, there are two micro-USB host ports wired directly to the FPGA (for use with USB host softcores) as well as one USB slave/Power in
<flea86>
PCIe x1 :)
<flea86>
(the form factor - no serdes sorry)
<eddyb>
erm I need to count the pads for some reason I thought PCIe was smaller
<flea86>
"I've only danced around with more GPIO breakout"
<flea86>
Pretty sure that's mini PCIe?
<flea86>
I also tried the same thing with m2 :P
<flea86>
but the PCB thickness is too narrow for good mechanical support IMHO
<flea86>
for m2
<flea86>
*m2 version
<eddyb>
flea86: ooooh I haven't seen miniPCIe in a while, and I was just looking at the Versa. that makes sense now
<eddyb>
flea86: but really, what are you doing with all of that USB? :D
<flea86>
eddyb: Well, one for USB mouse, one for USB keyboard... for retro applications (for now)
<flea86>
I'm just a hobbyist myself
<flea86>
certainly not in the same league as dave :)
<flea86>
just fooling around with the tech because it is fun
<eddyb>
huh I would've thought all 3 connectors are microUSB
<eddyb>
if you can't tell the scale is really messing with me here :P
<flea86>
They are. Right-most connector is USB slave / Power-in
<flea86>
It is modelled on the raspi-zero
<eddyb>
okay I'm just not used to an USB host with micro USB, but I can see the practical reasons for doing that
<flea86>
haha yeah try assembling them :D
<eddyb>
so you have to use one of those "OTG adapters" or w/e?
<eddyb>
to actually connect a mouse & keyboard?
<flea86>
yeah, to Type-A host adapter cables
<flea86>
one sec.
rohitksingh has quit [Ping timeout: 258 seconds]
<eddyb>
daveshah: oh, one more thing: I was looking at XAUI earlier and it sounded like even the non-5G ECP5s would be able to do it, with 4 duplex channels, does that sound right?
rohitksingh has joined ##openfpga
<daveshah>
Yes, XAUI would be fine for the slower SERDES too
<daveshah>
The 5G parts have faster fabric though
<eddyb>
(I incorrectly assumed "4 SERDES" meant two in each direction or something)
<eddyb>
assuming you can't trust the I2C lines, you get up to 5 twisted pairs for HDMI, which is one more than RJ45
<flea86>
eddyb: Well, you can also treat it as straight-up four/five LVDS pairs/lanes (GPDI)
<daveshah>
On the ULX3S revisions I have (may have been fixed), it uses top bank IO pins that can't be differential inputs or true differential outputs (but pseudo differential output is fine for HDMI)
<flea86>
daveshah: I see.
<eddyb>
flea86: right, that's what I meant. I guess the advantage to HDMI is they're optimizing for video output (and it can do other stuff), and HDMI is larger anyway
<eddyb>
so it would be annoying if you had RJ45 connected straight to the FPGA and you needed to mess around with the other HDMI lines, you'd have a really ugly adaptor
<whitequark>
you need a PHY for ethernet over copper though, generally
<whitequark>
100BASE-T is analog
<eddyb>
whereas the other way around it seems like you should be able to use the main 4 differential pairs in a HDMI cable to pass... oh dammit
<daveshah>
Unless you are azonenberg (still needs a fair few resistors IIRC) :p
<eddyb>
hang on... I just looked at the DP data rates, it seems like you could output 1440p @ 60Hz or 4K @ 30Hz from an ECP5. maybe I've got something wrong?
ym has quit [Quit: Leaving]
<eddyb>
AFAICT you can't do the same with HDMI because it goes above 5Gbps
<eddyb>
(but, again, I might be misreading the tables on wikipedia)
<eddyb>
someone motivated enough might even build something cool on top of DP DSC and Adaptive Sync
<eddyb>
hardware cursor rendering could be really smooth and not need that high of a data rate
Asu has quit [Quit: Konversation terminated!]
wpwrak has quit [Remote host closed the connection]