Brocade ICX Series (cheap & powerful 10gbE/40gbE switching)

kyled

New Member
Jan 25, 2019
5
0
1
I've actually tried that one, a Noctua NF-A4x20 FLX, and a Sunon KDE1204PKVX, all have the same problem once it switches to a lower speed
 

fohdeesha

Kaini Industries
Nov 20, 2016
1,816
1,582
113
29
fohdeesha.com
Really? I know multiple users use that exact fan with the ICX6450 and I'm 99% sure the "speed 1" voltage on both the icx6xxx and icx7xxx is 4.5v

edit: yeah looking at my old 7250 notes:

7250-2#dm fan-speed
Fan 1 Speed at 164 RPM (unplugged)
Fan 2 Speed at 6490 RPM

Fan header voltage at full speed - 12V
Fan header voltage at idle speed = 4.5V
 

kyled

New Member
Jan 25, 2019
5
0
1
Yeah that's why I got those, was expecting it to be the same

My fan is running a bit slower though

SSH@icx7250#dm fan-speed
Fan 1 Speed at 5895 RPM.
Fan 2 Speed at 164 RPM.
 

Zervun

Member
Feb 2, 2019
44
6
8
Oregon
Yes, I have used that exact breakout with a 6610 with no issues.
For that breakout cable - this may be a stupid question. I know that to get fiber stats you have to use the brocade SFPs (non brocade brands don't get fiber statistics on the SFPs from the 6610 readme). Does the breakout cable affect being able to get fiber stats or is it just specific to the SPFs?

I was thinking of coupling BRAND NEW OEM Avago AFBR-7IER02Z 40G QSFP+to 4x10G SFP+ Breakout AOC | eBay with this Brocade 57-0000075-01 10GbE SR SFP+ for FCOE and VDX switches 8904725111763 | eBay
 

Zervun

Member
Feb 2, 2019
44
6
8
Oregon
the 6610 never gets silent/very quiet, but with rev B power supplies it should spin WAY down after boot, and you shouldn't be able to hear it behind a closet door. I suspect the people with very loud 6610's in this thread have rev A PSUs
Just for some info - I just got my 6610 up and running (thanks to you). I was only expecting my 6610 to come with one power supply as it showed only one in the seller listing but it did come with two as well as two fans.

I pulled both of them to check the rev number and one is an A and one is a B.

Just out of curiosity I booted them with only one of each rev in it a couple of times. I really can't tell much of a difference in noise. Both spin up loud on boot but then settle down. B may be slightly quieter but it's hard to tell.

So those with screaming fans might not have updated the firmware or something (I'm on the latest via your config/update readme).

My switches are going in a rack in the crawlspace so it's not an issue for me. Much quieter than my r720xd w/ dual E5-2697v2 which are 130w a piece. Would still do a fan swap anyway but it seems near impossible.
 

PGlover

Active Member
Nov 8, 2014
461
51
28
53
Finally pulled the trigger on 2 Ruckus R720 APs. Will install the unleashed controller firmware on both APs. I'm still confused on the 803.3af vs 802.3at POE option on the AP and which port on the R720 to connect to my ICX6610 switch.

In addition, on the ICX6610, it looks like LLDP needs to be turned on and the switch needs to advertise power via LLDP.
 
Last edited:

fohdeesha

Kaini Industries
Nov 20, 2016
1,816
1,582
113
29
fohdeesha.com
that's not the correct cable - if you have to remove MTP guide pins to make it work then it's definitely not the right configuration, the fiber breakout is probably not the same as what's required for a 40gb optic to 4x 10gb lc - the pinout in that cable is probably mirrored if I had to guess, expecting a female connection on one end

edit: yeah just looked, the packaging for that fanout cable shows it breaks out strands 1 through 8 on the MTP connector, while 40gb uses strands 1-4 and 9-12 https://i.imgur.com/SXAxQGp.png

edit2: just realized that auction was listed in one my posts, not sure how that made it past me. removed, sorry about that
 
Last edited:

Zervun

Member
Feb 2, 2019
44
6
8
Oregon
for longer distances active DACs are generally cheapest IBM 59Y1930 30M QLogic Infiniband QDR QSFP Optical Cable | eBay
Do they have any shorter cables of those under 10m? I found IBM 59Y1926 10M QLogic Infiniband QDR-QSFP Optical Cable 883436055796 | eBay but it's more expensive than the 30m and couldn't find any shorter ones. 3m would be perfect.

They do have the copper DACs in shorter distances (I have a couple) but I wanted to do fiber just because I could ;)

I'm going to other servers in the rack with the HP flashed FCBT cards.
 

maes

Member
Nov 11, 2018
64
30
18
In regard to fan speeds, I started working on a small circuit based on an ATTiny25/45/85 that could either outright fake a valid RPM based on detected voltage at the fan connector, or multiply a detected RPM signal by the appropriate factor to match the factory-required fan. It's the same circuit for both options, just different programming. Of course the circuit would give a '0 RPM' flatline if it sees the replacement fan has stopped turning.

If I didn't screw up my math, with a 8MHz clock and appropriate prescaling, it should be able to emulate fan speeds between ~915RPM and ~234,375RPM. With a 1MHz clock I could get that down to a range of 114RPM-29,297RPM.

Either way should cover the range of just about every computer fan out there. (Please tell me there's nothing above a quarter million RPM)

So far, I managed to shrink it down to ~0.85"x0.6" (including an ICSP card-edge header). It would piggyback right on a compatible fan connector. Early hardware design is all done, I just have to get some parts to begin prototyping the coding.

Haven't coded for microcontrollers in nearly a decade so I'm all kinds of rusty. Should be a fun little project.

edit: I might switch to an ATTiny202, since it uses 16-bit timers it would give far more fine-grained control over the emulated RPM.
 
Last edited:
  • Like
Reactions: mathiastro

747builder

Active Member
Dec 17, 2017
103
50
28
In regard to fan speeds, I started working on a small circuit based on an ATTiny25/45/85 that could either outright fake a valid RPM based on detected voltage at the fan connector, or multiply a detected RPM signal by the appropriate factor to match the factory-required fan. It's the same circuit for both options, just different programming. Of course the circuit would give a '0 RPM' flatline if it sees the replacement fan has stopped turning.

If I didn't screw up my math, with a 8MHz clock and appropriate prescaling, it should be able to emulate fan speeds between ~915RPM and ~234,375RPM. With a 1MHz clock I could get that down to a range of 114RPM-29,297RPM.

Either way should cover the range of just about every computer fan out there. (Please tell me there's nothing above a quarter million RPM)

So far, I managed to shrink it down to ~0.85"x0.6" (including an ICSP card-edge header). It would piggyback right on a compatible fan connector. Early hardware design is all done, I just have to get some parts to begin prototyping the coding.

Haven't coded for microcontrollers in nearly a decade so I'm all kinds of rusty. Should be a fun little project.

edit: I might switch to an ATTiny202, since it uses 16-bit timers it would give far more fine-grained control over the emulated RPM.
you gonna make the board available beyond yourself? i know i would be interested in a few.
 

maes

Member
Nov 11, 2018
64
30
18
you gonna make the board available beyond yourself? i know i would be interested in a few.
Not sure if I'll sell them, but I will definitely make the code and schematics available to everyone. Problem with selling something that has relatively low value from here is that shipping will often end up being more than the board itself.
 

747builder

Active Member
Dec 17, 2017
103
50
28
Not sure if I'll sell them, but I will definitely make the code and schematics available to everyone. Problem with selling something that has relatively low value from here is that shipping will often end up being more than the board itself.
you might be suprised, if you can make the pcb board files available too would be cool after you have the PCB debugged and your code working.

i got a ICX6650 with 15k fans that are noisy for sure. louder than my Dell r710 and would love to quieten them down a bit.
 

747builder

Active Member
Dec 17, 2017
103
50
28
Finally pulled the trigger on 2 Ruckus R720 APs. Will install the unleashed controller firmware on both APs. I'm still confused on the 803.3af vs 802.3at POE option on the AP and which port on the R720 to connect to my ICX6610 switch.

In addition, on the ICX6610, it looks like LLDP needs to be turned on and the switch needs to advertise power via LLDP.
if the Ruckus r720's are anything like my r700's you will be very happy with them. you gonna run the ZD (Zone Director) to manage them? im in a 2600 sq ft house and ended up only needing 2 and have 3 AP's sitting here.

my r700's hang off a Juniper Ex4200-48PX since my ICX6650 has no copper ports..
 

fitzpattywagon

New Member
Feb 15, 2019
12
0
1
Curious is anyone using a 6450 just as a l2 switch? I'm debating being lazy and keeping my ISP's router and using a 6450 to replace a basic dell switch and give me 10g goodness.
 

neb50

New Member
Aug 28, 2018
20
2
3
Curious is anyone using a 6450 just as a l2 switch?
90% of the people in this thread
What he said.

I am using mine for the switching, POE, and L3 VLAN routing/ACL. I am leaving the firewall/routing/NAT and DHCP to an Ubiquiti edgerouter (including the DHCP servers for the VLAN's)

I have replaced the fans in a 7250-48P with some KDE1204PKVX fans. They will run mostly on the low speed and switch to the high speed for around 5 minutes around 20-30 times a day. Is it okay to let it run that close to the upper limit on heat and I should put some faster fans in there or just let it run warm?