Using 1000BASE-LX long wavelength optics over 62.5µm MMF, the IEEE 802.3z standard specifies a maximum distance of 550m (1804ft) using fiber with modal bandwidth of 500MHz-km @ 1300nm wavelength.
At any distance greater than 300m (984ft), Cisco says mode conditioning patch cords are required at each end of the LX link to reduce the possibility of differential mode delay.
Mode conditioning patch cords at distances less than 300m (984ft) with LX optics are optional. But at short link distances, i.e. less than 100m (328ft), Cisco also says using an ordinary patch cord with MMF can cause transceiver saturation resulting in an elevated bit error rate (BER); whereas using a mode-conditioning patch cord reduces the amount of optical power coupled into the MMF and avoids this situation.
I have read elsewhere that even worst-case MMF will support LX out to 700m (2296ft), and that the IEEE 802.3z distance was reduced to 550m (1804ft) to agree with the ISO/IEC 11801 structured cabling model.
Here's a Cisco link that quotes the 700m (2296ft) max. distance of LX over multimode; search for the number "700" to find it:
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/ethernet.htm
I generally deal with SMF on the networks I work on. But I do have a customer who tried to run Gig Ethernet LX/LH GBICs over a 1-2km (3280-6561ft) span of 62.5µm MMF with mode conditioning patch cords; it did NOT work. (Instead, they are running 4x100-megabit Fast EtherChannel now using media converters.)
Remember that when using mode-conditioning patch cords, the SMF bit of the patch cord plugs into the LX/LH GBIC's Tx port, not the Rx port.
Hope this helps.
