diff --git a/src/query/inc/qExecutor.h b/src/query/inc/qExecutor.h
index 4db86c25aac5554fdc2cd820a3ccc97fc7da87b5..f6d92fb8d12b9e48cddc08ff8b245080c5408da4 100644
--- a/src/query/inc/qExecutor.h
+++ b/src/query/inc/qExecutor.h
@@ -210,6 +210,7 @@ typedef struct SQuery {
 
   int32_t          srcRowSize;       // todo extract struct
   int32_t          resultRowSize;
+  int32_t          intermediateResultRowSize; // intermediate result row size, in case of top-k query.
   int32_t          maxSrcColumnSize;
   int32_t          tagLen;           // tag value length of current query
   SSqlGroupbyExpr* pGroupbyExpr;
diff --git a/src/query/inc/qResultbuf.h b/src/query/inc/qResultbuf.h
index 080e3f09bbfe2231657efa3e6ef29979a0c3b407..e46af8c29891cf52fbbac97c6afb42e0d0571215 100644
--- a/src/query/inc/qResultbuf.h
+++ b/src/query/inc/qResultbuf.h
@@ -76,7 +76,7 @@ typedef struct SDiskbasedResultBuf {
   SResultBufStatis statis;
 } SDiskbasedResultBuf;
 
-#define DEFAULT_INTERN_BUF_PAGE_SIZE  (256L)                          // in bytes
+#define DEFAULT_INTERN_BUF_PAGE_SIZE  (1024L)                          // in bytes
 #define PAGE_INFO_INITIALIZER         (SPageDiskInfo){-1, -1}
 
 /**
diff --git a/src/query/inc/qUtil.h b/src/query/inc/qUtil.h
index ac9740e7ba5bf112bbaeaed433765c2d7e7c7c0a..c5bdf28817e84877bf50fe59f5ec23bc7b8abba8 100644
--- a/src/query/inc/qUtil.h
+++ b/src/query/inc/qUtil.h
@@ -54,9 +54,9 @@ static FORCE_INLINE SResultRow *getResultRow(SResultRowInfo *pResultRowInfo, int
 static FORCE_INLINE char *getPosInResultPage(SQuery *pQuery, tFilePage* page, int32_t rowOffset, int16_t offset) {
   assert(rowOffset >= 0 && pQuery != NULL);
 
-//  int32_t realRowId = (int32_t)(rowId * GET_ROW_PARAM_FOR_MULTIOUTPUT(pQuery, pQuery->topBotQuery, pQuery->stableQuery));
+  int32_t numOfRows = GET_ROW_PARAM_FOR_MULTIOUTPUT(pQuery, pQuery->topBotQuery, pQuery->stableQuery);
 //  return ((char *)page->data) + offset * numOfRowsPerPage + bytes * realRowId;
-  return ((char *)page->data) + rowOffset + offset;
+  return ((char *)page->data) + rowOffset + offset * numOfRows;
 }
 
 bool isNullOperator(SColumnFilterElem *pFilter, const char* minval, const char* maxval, int16_t type);
@@ -82,7 +82,7 @@ void interResToBinary(SBufferWriter* bw, SArray* pRes, int32_t tagLen);
 SArray* interResFromBinary(const char* data, int32_t len);
 void freeInterResult(void* param);
 
-void    initGroupResInfo(SGroupResInfo* pGroupResInfo, SResultRowInfo* pResultInfo, int32_t offset);
+void    initGroupResInfo(SGroupResInfo* pGroupResInfo, SResultRowInfo* pResultInfo);
 void    cleanupGroupResInfo(SGroupResInfo* pGroupResInfo);
 bool    hasRemainDataInCurrentGroup(SGroupResInfo* pGroupResInfo);
 bool    hasRemainData(SGroupResInfo* pGroupResInfo);
diff --git a/src/query/src/qExecutor.c b/src/query/src/qExecutor.c
index 86bbfb32d7e9e5fb84237b2fc62214cd16030e81..4c9edb5008bb41a699365e1ca69df0355cccb1b6 100644
--- a/src/query/src/qExecutor.c
+++ b/src/query/src/qExecutor.c
@@ -653,7 +653,8 @@ static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SResultRow
 
   // not assign result buffer yet, add new result buffer
   if (pResultRow->pageId == -1) {
-    int32_t ret = addNewWindowResultBuf(pResultRow, pResultBuf, (int32_t) groupId, pRuntimeEnv->pQuery->resultRowSize);
+
+    int32_t ret = addNewWindowResultBuf(pResultRow, pResultBuf, (int32_t) groupId, pRuntimeEnv->pQuery->intermediateResultRowSize);
     if (ret != TSDB_CODE_SUCCESS) {
       return -1;
     }
@@ -2444,8 +2445,8 @@ static SColumnInfo* doGetTagColumnInfoById(SColumnInfo* pTagColList, int32_t num
 static void doSetTagValueInParam(void* pTable, int32_t tagColId, tVariant *tag, int16_t type, int16_t bytes);
 
 //TODO refactor
-int32_t loadDataBlockOnDemand_rv(SQueryRuntimeEnv *pRuntimeEnv, STableScanInfo* pTableScanInfo, SSDataBlock *pBlock,
-    uint32_t *status) {
+int32_t loadDataBlockOnDemand_rv(SQueryRuntimeEnv* pRuntimeEnv, STableScanInfo* pTableScanInfo, SSDataBlock* pBlock,
+                                 uint32_t* status) {
   *status = BLK_DATA_NO_NEEDED;
   pBlock->pDataBlock = NULL;
   pBlock->pBlockStatis = NULL;
@@ -2502,19 +2503,23 @@ int32_t loadDataBlockOnDemand_rv(SQueryRuntimeEnv *pRuntimeEnv, STableScanInfo*
       int32_t         numOfOutput = pTableScanInfo->numOfOutput;
       SQLFunctionCtx* pCtx = pTableScanInfo->pCtx;
 
-      for (int32_t i = 0; i < numOfOutput; ++i) {
-        int32_t functionId = pCtx[i].functionId;
-        int32_t colId = pTableScanInfo->pExpr[i].base.colInfo.colId;
-
-        // group by + first/last should not apply the first/last block filter
-        if (!pQuery->groupbyColumn && (functionId == TSDB_FUNC_FIRST_DST || functionId == TSDB_FUNC_LAST_DST)) {
-          (*status) |= aAggs[functionId].dataReqFunc(&pTableScanInfo->pCtx[i], &pBlock->info.window, colId);
-          if (((*status) & BLK_DATA_ALL_NEEDED) == BLK_DATA_ALL_NEEDED) {
+      if (pQuery->groupbyColumn) {
+        (*status) = BLK_DATA_ALL_NEEDED;
+      } else {
+        for (int32_t i = 0; i < numOfOutput; ++i) {
+          int32_t functionId = pCtx[i].functionId;
+          int32_t colId = pTableScanInfo->pExpr[i].base.colInfo.colId;
+
+          // group by + first/last should not apply the first/last block filter
+          if (functionId != TSDB_FUNC_FIRST_DST && functionId != TSDB_FUNC_LAST_DST) {
+            (*status) |= aAggs[functionId].dataReqFunc(&pTableScanInfo->pCtx[i], &pBlock->info.window, colId);
+            if (((*status) & BLK_DATA_ALL_NEEDED) == BLK_DATA_ALL_NEEDED) {
+              break;
+            }
+          } else {
+            (*status) |= BLK_DATA_ALL_NEEDED;
             break;
           }
-        } else {
-          (*status) |= BLK_DATA_ALL_NEEDED;
-          break;
         }
       }
     }
@@ -3495,12 +3500,6 @@ static int32_t doCopyToSData_rv(SQueryRuntimeEnv* pRuntimeEnv, SGroupResInfo* pG
 
     int32_t numOfRowsToCopy = pRow->numOfRows;
 
-    //current output space is not enough to accommodate all data of this page, prepare more space
-//    if (numOfRowsToCopy > (pRuntimeEnv->resultInfo.capacity - numOfResult)) {
-//      int32_t newSize = pRuntimeEnv->resultInfo.capacity + (numOfRowsToCopy - numOfResult);
-//      expandBuffer(pRuntimeEnv, newSize, pRuntimeEnv->qinfo);
-//    }
-
     pGroupResInfo->index += 1;
 
     tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, pRow->pageId);
@@ -4169,7 +4168,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, SArray* prevResult, void *ts
 
   int32_t ps = DEFAULT_PAGE_SIZE;
   int32_t rowsize = 0;
-  getIntermediateBufInfo(pRuntimeEnv, &ps, &rowsize);
+  getIntermediateBufInfo(pRuntimeEnv, &ps, &pQuery->intermediateResultRowSize);
   int32_t TENMB = 1024*1024*10;
 
   if (isSTableQuery && !onlyQueryTags(pQuery)) {
@@ -4630,7 +4629,7 @@ static SSDataBlock* doSTableAggregate(void* param) {
   updateWindowResNumOfRes_rv(pRuntimeEnv, pInfo->pCtx, pOperator->numOfOutput, &pInfo->resultRowInfo,
                              pInfo->rowCellInfoOffset);
 
-  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pInfo->resultRowInfo, 0);
+  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pInfo->resultRowInfo);
 
   toSSDataBlock(&pRuntimeEnv->groupResInfo, pRuntimeEnv, pInfo->pRes);
   if (pInfo->pRes->info.rows == 0 || !hasRemainDataInCurrentGroup(&pRuntimeEnv->groupResInfo)) {
@@ -4785,7 +4784,7 @@ static SSDataBlock* doIntervalAgg(void* param) {
   setQueryStatus(pRuntimeEnv, QUERY_COMPLETED);
   finalizeQueryResult_rv(pOperator, pIntervalInfo->pCtx, &pIntervalInfo->resultRowInfo, pIntervalInfo->rowCellInfoOffset);
 
-  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pIntervalInfo->resultRowInfo, 0);
+  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pIntervalInfo->resultRowInfo);
   toSSDataBlock(&pRuntimeEnv->groupResInfo, pRuntimeEnv, pIntervalInfo->pRes);
 
   if (pIntervalInfo->pRes->info.rows == 0 || !hasRemainDataInCurrentGroup(&pRuntimeEnv->groupResInfo)) {
@@ -4894,7 +4893,7 @@ static SSDataBlock* doHashGroupbyAgg(void* param) {
 
   updateWindowResNumOfRes_rv(pRuntimeEnv, pInfo->binfo.pCtx, pOperator->numOfOutput, &pInfo->binfo.resultRowInfo, pInfo->binfo.rowCellInfoOffset);
 
-  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pInfo->binfo.resultRowInfo, 0);
+  initGroupResInfo(&pRuntimeEnv->groupResInfo, &pInfo->binfo.resultRowInfo);
   toSSDataBlock(&pRuntimeEnv->groupResInfo, pRuntimeEnv, pInfo->binfo.pRes);
 
   if (pInfo->binfo.pRes->info.rows == 0 || !hasRemainDataInCurrentGroup(&pRuntimeEnv->groupResInfo)) {
diff --git a/src/query/src/qUtil.c b/src/query/src/qUtil.c
index dc8f6d56c08e705309bf82f659d10d52679dba9c..67ecc518929869cdd4d63bbbe90c104abcbedc75 100644
--- a/src/query/src/qUtil.c
+++ b/src/query/src/qUtil.c
@@ -344,13 +344,13 @@ void cleanupGroupResInfo(SGroupResInfo* pGroupResInfo) {
   pGroupResInfo->index     = 0;
 }
 
-void initGroupResInfo(SGroupResInfo* pGroupResInfo, SResultRowInfo* pResultInfo, int32_t offset) {
+void initGroupResInfo(SGroupResInfo* pGroupResInfo, SResultRowInfo* pResultInfo) {
   if (pGroupResInfo->pRows != NULL) {
     taosArrayDestroy(pGroupResInfo->pRows);
   }
 
   pGroupResInfo->pRows = taosArrayFromList(pResultInfo->pResult, pResultInfo->size, POINTER_BYTES);
-  pGroupResInfo->index = offset;
+  pGroupResInfo->index = 0;
 
   assert(pGroupResInfo->index <= getNumOfTotalRes(pGroupResInfo));
 }
diff --git a/tests/script/general/parser/topbot.sim b/tests/script/general/parser/topbot.sim
index f5c78d07a1d362ffdd5ebe5c989d88cc35a33e72..b81a1f9b34209e34e198977cc30d4bfe0a22cbe2 100644
--- a/tests/script/general/parser/topbot.sim
+++ b/tests/script/general/parser/topbot.sim
@@ -73,6 +73,60 @@ if $row != 100 then
   return -1
 endi
 
+sql select bottom(c3, 5) from tb_tb1 interval(1y);
+if $rows != 5 then
+  return -1
+endi
+
+if $data01 != 0.00000 then
+  print expect 0.00000, actual:$data01
+  return -1
+endi
+
+if $data11 != 0.00000 then
+  return -1
+endi
+
+if $data21 != 0.00000 then
+  return -1
+endi
+
+if $data31 != 0.00000 then
+  return -1
+endi
+
+sql select top(c4, 5) from tb_tb1 interval(1y);
+if $rows != 5 then
+  return -1
+endi
+
+if $data01 != 9.000000000 then
+  print expect 9.000000000, acutal:$data01
+  return -1
+endi
+
+if $data11 != 9.000000000 then
+  return -1
+endi
+
+if $data21 != 9.000000000 then
+  return -1
+endi
+
+if $data31 != 9.000000000 then
+  return -1
+endi
+
+sql select top(c3, 5) from tb_tb1 interval(40h)
+if $rows != 25 then
+  return -1
+endi
+
+if $data01 != 9.00000 then
+  print expect 9.00000, actual:$data01
+  return -1
+endi
+
 sql select last(*) from tb_tb9
 if $row != 1 then
   return -1