diff --git a/libs/keymanip.cu b/libs/keymanip.cu
index 6d2984d..107b19a 100644
--- a/libs/keymanip.cu
+++ b/libs/keymanip.cu
@@ -41,6 +41,7 @@ __device__ void getRawAddress(int lErase, Key32& InvertedPublicKey, Addr16& rawA
     memcpy(&rawAddr[2], &InvertedPublicKey[start], 14);
 }
 __device__ void invertKey(const unsigned char* key, unsigned char* inverted) {
+#pragma unroll
     for (int i = 0; i < 32; i += 4) {
         uchar4 k = *(reinterpret_cast<const uchar4*>(&key[i]));
         *(reinterpret_cast<uchar4*>(&inverted[i])) = make_uchar4(k.x ^ 0xFF, k.y ^ 0xFF, k.z ^ 0xFF, k.w ^ 0xFF);
diff --git a/sources/main.cu b/sources/main.cu
index 811747e..63db9d6 100644
--- a/sources/main.cu
+++ b/sources/main.cu
@@ -8,7 +8,8 @@
 #include <keymanip.cuh>
 __device__ unsigned d_high = 0x10;
 __device__ int parameters(const char* arg) noexcept {
-    if ((cstring_find(arg, "--altitude") == 0 && cstring_length(arg) == 10) || (cstring_find(arg, "-a") == 0 && cstring_length(arg) == 2)) {
+    if ((cstring_find(arg, "--altitude") == 0 && cstring_length(arg) == 10) ||
+        (cstring_find(arg, "-a") == 0 && cstring_length(arg) == 2)) {
         return 777;
     }
     int space_index = cstring_find(arg, " ");
@@ -49,7 +50,8 @@ __device__ __forceinline__ unsigned char getZeros(const unsigned char* v) noexce
     unsigned char leadZeros = 0;
 #pragma unroll
     for (int i = 0; i < 32; i += 4) {
-        unsigned word = (static_cast<unsigned>(v[i]) << 24) | (static_cast<unsigned>(v[i + 1]) << 16) | (static_cast<unsigned>(v[i + 2]) << 8) | (static_cast<unsigned>(v[i + 3]));
+        unsigned word = (static_cast<unsigned>(v[i]) << 24) | (static_cast<unsigned>(v[i + 1]) << 16) |
+            (static_cast<unsigned>(v[i + 2]) << 8) | (static_cast<unsigned>(v[i + 3]));
         if (word == 0)
             leadZeros += 32;
         else {
@@ -59,19 +61,32 @@ __device__ __forceinline__ unsigned char getZeros(const unsigned char* v) noexce
     }
     return leadZeros;
 }
+
 /*
 __global__ void initRandSeed(curandState* states, const unsigned long seed) {
     int idx = blockIdx.x * blockDim.x + threadIdx.x;
     curand_init(seed, idx, 0, &states[idx]);
 }
 */
-__global__ void initRand(curandState* rs) {
+__global__ void initRand(curandState* rs, unsigned int* d_seeds) {
     int id = blockIdx.x * blockDim.x + threadIdx.x;
     curand_init(clock64() + id * 7919ULL, id, 0, &rs[id]);
 #pragma unroll 10
     for (int i = 0; i < 10; i++) {
         curand(&rs[id]);
     }
+    unsigned int seed = curand(&rs[id]);
+    d_seeds[id] = seed;
+}
+int checkSeeds(unsigned int* seeds, int count) {
+    for (int i = 0; i < count; i++) {
+        for (int j = i + 1; j < count; j++) {
+            if (seeds[i] == seeds[j]) {
+                return 1;
+            }
+        }
+    }
+    return 0;
 }
 __device__ __forceinline__ void rmbytes(unsigned char* buf, curandState* state) {
 #pragma unroll 8
@@ -84,11 +99,6 @@ __device__ __forceinline__ void rmbytes(unsigned char* buf, curandState* state)
     }
 }
 __global__ void KeyGenKernel(curandState* randStates) {
-    __shared__ unsigned local_high;
-    if (threadIdx.x == 0) {
-        local_high = d_high;
-    }
-    __syncthreads();
     int idx = blockIdx.x * blockDim.x + threadIdx.x;
     curandState localState = randStates[idx];
     while (true) {
@@ -96,15 +106,13 @@ __global__ void KeyGenKernel(curandState* randStates) {
         KeysBox32 keys;
         rmbytes(seed, &localState);
         ed25519_keygen(keys.PrivateKey, keys.PublicKey, seed);
-        if (unsigned zeros = getZeros(keys.PublicKey); zeros > local_high) {
-            if (zeros > atomicMax(&d_high, zeros)) {
-                Addr16 raw;
-                Key32 inv;
-                invertKey(keys.PublicKey, inv);
-                getRawAddress(zeros, inv, raw);
-                printf("\nIPv6:\t%s\nPK:\t%s\nSK:\t%s\n", getAddr(raw).data, ktos(keys.PublicKey).data, ktos(keys.PrivateKey).data);
-                local_high = zeros;
-            }
+        if (unsigned zeros = getZeros(keys.PublicKey); zeros > atomicMax(&d_high, zeros)) {
+            Addr16 raw;
+            Key32 inv;
+            invertKey(keys.PublicKey, inv);
+            getRawAddress(zeros, inv, raw);
+            printf("\nIPv6:\t%s\nPK:\t%s\nSK:\t%s\n", getAddr(raw).data, ktos(keys.PublicKey).data, ktos(keys.PrivateKey).data);
+            d_high = zeros;
         }
     }
 }
@@ -123,18 +131,30 @@ int main(int argc, char* argv[]) {
         cudaMemcpy(d_str, argv[i], len, cudaMemcpyHostToDevice);
         cudaMemcpy(&d_argv[i], &d_str, sizeof(char*), cudaMemcpyHostToDevice);
     }
-    args<<<1, 1>>>(d_argv, argc, d_result);
+    args<<<1, 1 >>>(d_argv, argc, d_result);
     cudaDeviceSynchronize();
     cudaMemcpyFromSymbol(&h_high, d_high, sizeof(unsigned));
-    cudaGetDeviceProperties_v2(&prop, 0);
+    cudaGetDeviceProperties(&prop, 0);
     cudaOccupancyMaxActiveBlocksPerMultiprocessor(&mBpSM, KeyGenKernel, THREADS_P_B, 0);
     const int tTh = mBpSM * prop.multiProcessorCount * THREADS_P_B;
-    printf("High addrs: 2%02x+\nSMs: %d\nTotalThreads: %d\nBlocksThreads: %d:%d\n", h_high, prop.multiProcessorCount, tTh, tTh / THREADS_P_B, THREADS_P_B);
+    printf("High addrs: 2%02x+\nSMs: %d\nTotalThreads: %d\nBlocks: %d (Threads: %d)\n", h_high, prop.multiProcessorCount, tTh, tTh / THREADS_P_B, THREADS_P_B);
     cudaMalloc(&rst, tTh * sizeof(curandState));
-    initRand<<<tTh / THREADS_P_B, THREADS_P_B>>>(rst);
-    cudaDeviceSynchronize();
-    KeyGenKernel<<<tTh / THREADS_P_B, THREADS_P_B >>>(rst);
+    unsigned int* d_seeds;
+    cudaMalloc(&d_seeds, tTh * sizeof(unsigned int));
+    initRand<<<tTh / THREADS_P_B, THREADS_P_B >>>(rst, d_seeds);
     cudaDeviceSynchronize();
+    unsigned int* h_seeds = (unsigned int*)malloc(tTh * sizeof(unsigned int));
+    cudaMemcpy(h_seeds, d_seeds, tTh * sizeof(unsigned int), cudaMemcpyDeviceToHost);
+    if (checkSeeds(h_seeds, tTh)) {
+        fprintf(stderr, "Duplicate seeds found!\n");
+        free(h_seeds);
+        cudaFree(d_seeds);
+        cudaFree(rst);
+        return EXIT_FAILURE;
+    }
+    free(h_seeds);
+    cudaFree(d_seeds);
+    KeyGenKernel<<<tTh / THREADS_P_B, THREADS_P_B>>>(rst);
     cudaFree(rst);
     return 0;
 }
\ No newline at end of file