You have a cite for the LD50?

Pardon me. Cntrl-C failure. The link in the grandparent was supposed to be to the source but I evidently copied the wrong tab. In any case the measure was likely based on rats---there doesn't seem to be good data for humans. It also seems that the more 'official' (in social authority not academic authority) a source is the worse the drug is made to appear.

Further investigation brought up a few things worth mentioning:

• The cause of death from GHB overdose is respiratory failure from CNS depression. If respiratory support is supplied the lethal dose (from whatever critical thing fails next) is several times higher. This suggests increased expected valueof having a sober friend who knows Expired Air Resuscitation and the emergency services number relative to other drugs of abuse.
• Many of the deaths that are attributed to "GHB overdose" can (allegedly) actually be better attributed to a different cause of death (including the GHB and alcohol cocktail mentioned).
• More 'official' sources suggest a lower LD50 than less official sources. In this case 'official' refers to social authority and not academic authority and there is good cause to doubt the credibility of sources for which writing a high LD50 could get the author fired by some moralizer with power.