It will be helpful if you can further explain the following to better address your question:
(1). What specific GARCH model do you want to estimate? Is it a univariate GARCH model or a multivariate GARCH model?
(2). How is the GARCH model used in your VaR portfolio calculation? How is it related to 'the standard deviation and correlation coefficients required for VaR' that you want to compute?
PROC AUTOREG supports various types of univariate GARCH models as discussed here:
https://go.documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=etsug&docsetTarget=ets...
and PROC VARMAX supports multivariate GARCH modeling as discussed here:
https://go.documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=etsug&docsetTarget=ets...
Further clarifications on (1) and (2) above will be helpful to better understand what you want from the GARCH model estimation to get your desired statistics.