Depends on how fancy you need to be, but I would check out the R bioconductor package snpstats. See page 16 of the LD Vignette.pdf. You'll need to convert your VCF data to ped format, you can do this easily with vcftools, I would make a tped if youf VCF is big (it will progably crash if you don't do a tped) eg:

vcftools --vcf your.vcf --plink-tped --out yourTped
plink --tfile yourTped --make-bed --out yourBed

You can also try your own hand at clustering, I gave a simple example here on biostar for ADMIXTURE data, but this is easily modified for SNPs in a tped file:Simple R clustering dendrogram

I would try Margarita:
http://seqanswers.com/wiki/20980557

It doesn't produce trees but instead it produces ancestral recombination graphs (ARGs) which theoretically will better reflect recombination between your SNPs.

If you can code -

create a binary matrix where rows are loci and columns are individuals. 1 means present 0 means not present (non reference allele).

In R use the APE package to create NJ bootstrapped trees. If your interested in more details send me a message.

UPDATE:

There are now "standard" methods for building a SNP tree. For example: http://www.biomedcentral.com/1471-2164/15/162