diff --git a/.SRCINFO b/.SRCINFO
index d7889e410153e..345e2b5fd9002 100644
--- a/.SRCINFO
+++ b/.SRCINFO
@@ -1,39 +1,31 @@
 pkgbase = llama.cpp-hip
 	pkgdesc = Port of Facebook's LLaMA model in C/C++ (with AMD ROCm optimizations)
 	pkgver = b5669
-	pkgrel = 1
+	pkgrel = 2
 	url = https://github.com/ggerganov/llama.cpp
 	arch = x86_64
 	arch = armv7h
 	arch = aarch64
 	license = MIT
-	makedepends = clblast
 	makedepends = cmake
 	makedepends = git
-	makedepends = rocm-hip-runtime
 	makedepends = rocm-hip-sdk
-	depends = blas-openblas
-	depends = blas64-openblas
 	depends = curl
 	depends = gcc-libs
 	depends = glibc
-	depends = hip-runtime-amd
-	depends = hipblas
-	depends = openmp
+	depends = libggml-hip
 	depends = python
 	depends = python-numpy
 	depends = python-sentencepiece
-	depends = rocblas
 	optdepends = python-pytorch
 	provides = llama.cpp
 	conflicts = llama.cpp
 	options = lto
+	options = !debug
 	source = git+https://github.com/ggerganov/llama.cpp#tag=b5669
-	source = git+https://github.com/nomic-ai/kompute.git
 	source = llama.cpp.conf
 	source = llama.cpp.service
 	sha256sums = eb46ae341d2f89395c4c86186c25be504a9e4db3369e1a826ccb3bc50743d21c
-	sha256sums = SKIP
 	sha256sums = 53fa70cfe40cb8a3ca432590e4f76561df0f129a31b121c9b4b34af0da7c4d87
 	sha256sums = 0377d08a07bda056785981d3352ccd2dbc0387c4836f91fb73e6b790d836620d
 
diff --git a/PKGBUILD b/PKGBUILD
index a9652b42207b8..44d0a6988578a 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -3,55 +3,38 @@
 pkgname=llama.cpp-hip
 _pkgname="${pkgname%-hip}"
 pkgver=b5669
-pkgrel=1
+pkgrel=2
 pkgdesc="Port of Facebook's LLaMA model in C/C++ (with AMD ROCm optimizations)"
 arch=(x86_64 armv7h aarch64)
 url='https://github.com/ggerganov/llama.cpp'
 license=('MIT')
 depends=(
-  blas-openblas
-  blas64-openblas
   curl
   gcc-libs
   glibc
-  hip-runtime-amd
-  hipblas
-  openmp
+  libggml-hip
   python
   python-numpy
   python-sentencepiece
-  rocblas
 )
 makedepends=(
-  clblast
   cmake
   git
-  rocm-hip-runtime
   rocm-hip-sdk
 )
 optdepends=(python-pytorch)
 provides=(${_pkgname})
 conflicts=(${_pkgname})
-options+=(lto)
+options=(lto !debug)
 source=(
   "git+${url}#tag=${pkgver}"
-  "git+https://github.com/nomic-ai/kompute.git"
   llama.cpp.conf
   llama.cpp.service
 )
 sha256sums=('eb46ae341d2f89395c4c86186c25be504a9e4db3369e1a826ccb3bc50743d21c'
-            'SKIP'
             '53fa70cfe40cb8a3ca432590e4f76561df0f129a31b121c9b4b34af0da7c4d87'
             '0377d08a07bda056785981d3352ccd2dbc0387c4836f91fb73e6b790d836620d')
 
-prepare() {
-  cd "${_pkgname}"
-
-  git submodule init
-  git config submodule.kompute.url "${srcdir}/kompute"
-  git -c protocol.file.allow=always submodule update
-}
-
 build() {
   export CC=/opt/rocm/llvm/bin/clang
   export CXX=/opt/rocm/llvm/bin/clang++
@@ -60,15 +43,10 @@ build() {
     -S "${_pkgname}"
     -DCMAKE_BUILD_TYPE=Release
     -DCMAKE_INSTALL_PREFIX='/usr'
-    -DGGML_ALL_WARNINGS=OFF
-    -DGGML_ALL_WARNINGS_3RD_PARTY=OFF
     -DBUILD_SHARED_LIBS=ON
-    -DGGML_STATIC=OFF
-    -DGGML_LTO=ON
-    -DGGML_RPC=ON
     -DLLAMA_CURL=ON
-    -DGGML_BLAS=ON
-    -DGGML_HIP=ON
+    -DLLAMA_BUILD_TESTS=OFF
+    -DLLAMA_USE_SYSTEM_GGML=ON
     -Wno-dev
   )
   cmake "${_cmake_options[@]}"
@@ -77,8 +55,6 @@ build() {
 
 package() {
   DESTDIR="${pkgdir}" cmake --install build
-  rm "${pkgdir}/usr/include/"ggml*
-  # rm "${pkgdir}/usr/lib/"lib*.a
 
   install -Dm644 "${_pkgname}/LICENSE" "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE"