diff --git a/Cargo.toml b/Cargo.toml
index 05e49d0..ef9efca 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -29,8 +29,6 @@ derivative = "*"
 bincode = "*"
 serde = { version = "1.0", features = ["derive"] }
 
-nvtx = "1.2"
-
 
 [dev-dependencies]
 serial_test = "^2"
diff --git a/src/poly.rs b/src/poly.rs
index 19581b0..235d4cc 100644
--- a/src/poly.rs
+++ b/src/poly.rs
@@ -22,8 +22,6 @@ pub(crate) struct PrecomputedBasisForBarycentric {
     pub(crate) bases: DVec<F>,
 }
 
-use nvtx::{range_pop, range_push};
-
 impl PrecomputedBasisForBarycentric {
     pub fn precompute(domain_size: usize, point: EF) -> CudaResult<Self> {
         let mut bases = dvec!(2 * domain_size);
@@ -290,9 +288,7 @@ impl<'a, P: PolyForm> ComplexPoly<'a, P> {
 
 impl<'a> Poly<'a, LDE> {
     pub fn intt(mut self) -> CudaResult<Poly<'a, MonomialBasis>> {
-        range_push!("Poly<LDE> intt");
         ntt::lde_intt(self.storage.as_mut())?;
-        range_pop!();
         Ok(Poly {
             storage: self.storage,
             marker: std::marker::PhantomData,
diff --git a/src/prover.rs b/src/prover.rs
index be02c86..d96283d 100644
--- a/src/prover.rs
+++ b/src/prover.rs
@@ -32,8 +32,6 @@ use crate::{
 
 use super::*;
 
-use nvtx::{range_pop, range_push};
-
 pub fn gpu_prove_from_external_witness_data<
     P: boojum::field::traits::field_like::PrimeFieldLikeVectorized<Base = F>,
     TR: Transcript<F, CompatibleCap = [F; 4]>,
@@ -745,10 +743,8 @@ fn gpu_prove_from_trace<
         )?;
     }
 
-    range_push!("quotient evals to monomial");
     let quotient_monomial = quotient.intt()?;
-    range_pop!();
-    // quotient memory is guaranteed to allow batch ntts for cosets of the quotinet parts
+    // quotient memory is guaranteed to allow batch ntts for cosets of the quotient parts
     let quotient_chunks = quotient_monomial.clone().into_degree_n_polys(domain_size)?;
 
     let quotient_monomial_storage = GenericComplexPolynomialStorage {