TPCDS-Q95 run slower with 32 CPU TiDB comparing with 16 CPU TiDB

## Bug Report

Please answer these questions before submitting your issue. Thanks!

### 1. Minimal reproduce step (Required)

1. deploy a tidb cluster
- 1 TiDB 32CPU
- 3 TiKV16 CPU
- 1 PD 4 CPU
- 1 Tiflash 32 CPU
2. run TPCDS query 95
3. deploy a second tidb cluster
- 1 TiDB 16 CPU
- 3 TiKV16 CPU
- 1 PD 4 CPU
- 1 Tiflash 32 CPU
4. run TPCDS query 95

### 2. What did you expect to see? (Required)
The elapsed time of Q95 with 16 CPU is around 97s. Q95 would run faster with 32 CPU TiDB.

### 3. What did you see instead (Required)
The execution of Q95 was finished in 122s with 32 CPU TiDB.

```
Olap_Detail_Log_ID: 2409311 Plan_Digest: 605145cd8b6149053d1dd4f7a6124883 Elapsed_Time (s): 122.90

+-----------------------------------------------+--------------+-----------+--------------+------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------+---------+
| ID | ESTROWS | ACTROWS | TASK | ACCESS OBJECT | EXECUTION INFO | OPERATOR INFO | MEMORY | DISK |
+-----------------------------------------------+--------------+-----------+--------------+------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------+---------+
| TopN_77 | 1.00 | 1 | root | | time:2m2.9s, loops:2 | Column#204, offset:0, count:100 | 3.18 KB | N/A |
| └─StreamAgg_83 | 1.00 | 1 | root | | time:2m2.9s, loops:3 | funcs:count(distinct tpcds50.web_sales.ws_order_number)->Column#204, funcs:sum(tpcds50.web_sales.ws_ext_ship_cost)->Column#205, funcs:sum(tpcds50.web_sales.ws_net_profit)->Column#206 | 129.6 KB | N/A |
| └─HashJoin_193 | 30633.76 | 3223 | root | | time:2m2.9s, loops:6, build_hash_table:{total:1m22.1s, fetch:1m22.1s, build:924µs}, probe:{concurrency:5, total:10m14.5s, max:2m2.9s, probe:662.3ms, fetch:10m13.8s} | inner join, equal:[eq(tpcds50.web_sales.ws_order_number, tpcds50.web_returns.wr_order_number)] | 657.4 KB | 0 Bytes |
| ├─HashJoin_88(Build) | 30633.76 | 4539 | root | | time:1m22.1s, loops:6, build_hash_table:{total:584ms, fetch:582.9ms, build:1.07ms}, probe:{concurrency:5, total:6m50.4s, max:1m22.1s, probe:1.51s, fetch:6m48.9s} | inner join, equal:[eq(tpcds50.web_sales.ws_order_number, tpcds50.web_sales.ws_order_number)] | 1.29 MB | 0 Bytes |
| │ ├─TableReader_120(Build) | 30633.76 | 4539 | root | | time:583.4ms, loops:6, cop_task: {num: 24, max: 0s, min: 0s, avg: 0s, p95: 0s, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_119 | N/A | N/A |
| │ │ └─ExchangeSender_119 | 30633.76 | 4539 | mpp[tiflash] | | tiflash_task:{time:554.7ms, loops:571, threads:24} | ExchangeType: PassThrough | N/A | N/A |
| │ │ └─HashJoin_89 | 30633.76 | 4539 | mpp[tiflash] | | tiflash_task:{time:540.7ms, loops:571, threads:24} | inner join, equal:[eq(tpcds50.web_sales.ws_ship_addr_sk, tpcds50.customer_address.ca_address_sk)] | N/A | N/A |
| │ │ ├─ExchangeReceiver_118(Build) | 30633.76 | 30497 | mpp[tiflash] | | tiflash_task:{time:89.7ms, loops:6, threads:24} | | N/A | N/A |
| │ │ │ └─ExchangeSender_117 | 30633.76 | 30497 | mpp[tiflash] | | tiflash_task:{time:15.6ms, loops:6, threads:1} | ExchangeType: Broadcast | N/A | N/A |
| │ │ │ └─Selection_116 | 30633.76 | 30497 | mpp[tiflash] | | tiflash_task:{time:15.6ms, loops:6, threads:1} | eq(tpcds50.customer_address.ca_state, "TX") | N/A | N/A |
| │ │ │ └─TableFullScan_115 | 383000.00 | 383000 | mpp[tiflash] | table:customer_address | tiflash_task:{time:9.64ms, loops:6, threads:1} | keep order:false | N/A | N/A |
| │ │ └─HashJoin_103(Probe) | 102376.77 | 57632 | mpp[tiflash] | | tiflash_task:{time:534.7ms, loops:571, threads:24} | inner join, equal:[eq(tpcds50.web_sales.ws_ship_date_sk, tpcds50.date_dim.d_date_sk)] | N/A | N/A |
| │ │ ├─ExchangeReceiver_114(Build) | 61.00 | 61 | mpp[tiflash] | | tiflash_task:{time:90.7ms, loops:1, threads:24} | | N/A | N/A |
| │ │ │ └─ExchangeSender_113 | 61.00 | 61 | mpp[tiflash] | | tiflash_task:{time:2.94ms, loops:1, threads:1} | ExchangeType: Broadcast | N/A | N/A |
| │ │ │ └─Selection_112 | 61.00 | 61 | mpp[tiflash] | | tiflash_task:{time:3.94ms, loops:1, threads:1} | ge(tpcds50.date_dim.d_date, 1999-05-01 00:00:00.000000), le(tpcds50.date_dim.d_date, 1999-06-30) | N/A | N/A |
| │ │ │ └─TableFullScan_111 | 73049.00 | 8192 | mpp[tiflash] | table:date_dim | tiflash_task:{time:2.94ms, loops:1, threads:1} | keep order:false | N/A | N/A |
| │ │ └─HashJoin_104(Probe) | 3263567.27 | 2997937 | mpp[tiflash] | | tiflash_task:{time:510.7ms, loops:571, threads:24} | inner join, equal:[eq(tpcds50.web_site.web_site_sk, tpcds50.web_sales.ws_web_site_sk)] | N/A | N/A |
| │ │ ├─ExchangeReceiver_108(Build) | 2.00 | 2 | mpp[tiflash] | | tiflash_task:{time:89.7ms, loops:1, threads:24} | | N/A | N/A |
| │ │ │ └─ExchangeSender_107 | 2.00 | 2 | mpp[tiflash] | | tiflash_task:{time:1.67ms, loops:1, threads:1} | ExchangeType: Broadcast | N/A | N/A |
| │ │ │ └─Selection_106 | 2.00 | 2 | mpp[tiflash] | | tiflash_task:{time:1.67ms, loops:1, threads:1} | eq(tpcds50.web_site.web_company_name, "pri") | N/A | N/A |
| │ │ │ └─TableFullScan_105 | 22.00 | 22 | mpp[tiflash] | table:web_site | tiflash_task:{time:1.67ms, loops:1, threads:1} | keep order:false | N/A | N/A |
| │ │ └─Selection_110(Probe) | 35970181.33 | 35981281 | mpp[tiflash] | | tiflash_task:{time:320.7ms, loops:571, threads:24} | not(isnull(tpcds50.web_sales.ws_ship_addr_sk)), not(isnull(tpcds50.web_sales.ws_ship_date_sk)), not(isnull(tpcds50.web_sales.ws_web_site_sk)) | N/A | N/A |
| │ │ └─TableFullScan_109 | 35996850.00 | 35996850 | mpp[tiflash] | table:ws1 | tiflash_task:{time:234.7ms, loops:571, threads:24} | keep order:false | N/A | N/A |
| │ └─HashAgg_177(Probe) | 2402536.42 | 3000000 | root | | time:1m21.8s, loops:2933, partial_worker:{wall_time:1m19.87680233s, concurrency:5, task_num:347986, tot_wait:6m0.539587527s, tot_exec:37.654757277s, tot_time:6m39.219180565s, max:1m19.876726352s, p95:1m19.876726352s}, final_worker:{wall_time:1m22.082867266s, concurrency:5, task_num:25, tot_wait:6m39.13925395s, tot_exec:11.024193393s, tot_time:6m50.16346278s, max:1m22.08279734s, p95:1m22.08279734s} | group by:tpcds50.web_sales.ws_order_number, funcs:firstrow(tpcds50.web_sales.ws_order_number)->tpcds50.web_sales.ws_order_number | 1.04 GB | N/A |
| │ └─Selection_178 | 345006182.27 | 356337144 | root | | time:1m18.6s, loops:347987 | not(isnull(tpcds50.web_sales.ws_order_number)) | 24.7 KB | N/A |
| │ └─CTEFullScan_179 | 431257727.84 | 356337144 | root | CTE:ws_wh | time:50.3s, loops:348287 | data:CTE_0 | 16.4 GB | 0 Bytes |
| └─HashAgg_180(Probe) | 2083584.00 | 2119776 | root | | time:2m2.8s, loops:2074, partial_worker:{wall_time:2m1.648780808s, concurrency:5, task_num:455912, tot_wait:9m19.286196323s, tot_exec:47.522441542s, tot_time:10m8.163112057s, max:2m1.648674522s, p95:2m1.648674522s}, final_worker:{wall_time:2m2.898362017s, concurrency:5, task_num:25, tot_wait:10m8.091950295s, tot_exec:6.119080067s, tot_time:10m14.21105335s, max:2m2.898215129s, p95:2m2.898215129s} | group by:tpcds50.web_returns.wr_order_number, funcs:firstrow(tpcds50.web_returns.wr_order_number)->tpcds50.web_returns.wr_order_number | 756.0 MB | N/A |
| └─HashJoin_182 | 516808411.86 | 466851188 | root | | time:2m0.5s, loops:455913, build_hash_table:{total:1.13s, fetch:105.3ms, build:1.02s}, probe:{concurrency:5, total:10m6.9s, max:2m1.4s, probe:6m1.4s, fetch:4m5.5s} | inner join, equal:[eq(tpcds50.web_returns.wr_order_number, tpcds50.web_sales.ws_order_number)] | 172.1 MB | 0 Bytes |
| ├─TableReader_186(Build) | 3598924.00 | 3598924 | root | | time:83.6ms, loops:3518, cop_task: {num: 8, max: 128.3ms, min: 3.52ms, avg: 90.9ms, p95: 128.3ms, rpc_num: 8, rpc_time: 726.7ms, copr_cache_hit_ratio: 0.00, distsql_concurrency: 15} | data:TableFullScan_185 | 27.5 MB | N/A |
| │ └─TableFullScan_185 | 3598924.00 | 3598924 | cop[tiflash] | table:web_returns | tiflash_task:{proc max:69.8ms, min:990µs, avg: 47.7ms, p80:62.6ms, p95:69.8ms, iters:69, tasks:8, threads:8} | keep order:false | N/A | N/A |
| └─Selection_189(Probe) | 345006182.27 | 356337144 | root | | time:1m24.6s, loops:347987 | not(isnull(tpcds50.web_sales.ws_order_number)) | 24.7 KB | N/A |
| └─CTEFullScan_190 | 431257727.84 | 356337144 | root | CTE:ws_wh | time:48.6s, loops:348287 | data:CTE_0 | 0 Bytes | 0 Bytes |
| CTE_0 | 431257727.84 | 356337144 | root | | time:50.3s, loops:348287 | Non-Recursive CTE | 16.4 GB | 0 Bytes |
| └─Projection_35(Seed Part) | 431257727.84 | 356337144 | root | | time:25.4s, loops:348286, Concurrency:5 | tpcds50.web_sales.ws_order_number, tpcds50.web_sales.ws_warehouse_sk, tpcds50.web_sales.ws_warehouse_sk | 171.0 KB | N/A |
| └─TableReader_48 | 431257727.84 | 356337144 | root | | time:24.2s, loops:348286, cop_task: {num: 572, max: 0s, min: 0s, avg: 0s, p95: 0s, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_47 | N/A | N/A |
| └─ExchangeSender_47 | 431257727.84 | 356337144 | mpp[tiflash] | | tiflash_task:{time:21.8s, loops:571, threads:24} | ExchangeType: PassThrough | N/A | N/A |
| └─HashJoin_36 | 431257727.84 | 356337144 | mpp[tiflash] | | tiflash_task:{time:2.48s, loops:571, threads:24} | inner join, equal:[eq(tpcds50.web_sales.ws_order_number, tpcds50.web_sales.ws_order_number)], other cond:ne(tpcds50.web_sales.ws_warehouse_sk, tpcds50.web_sales.ws_warehouse_sk) | N/A | N/A |
| ├─ExchangeReceiver_42(Build) | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:117.6ms, loops:569, threads:24} | | N/A | N/A |
| │ └─ExchangeSender_41 | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:462.8ms, loops:571, threads:24} | ExchangeType: HashPartition, Hash Cols: [name: tpcds50.web_sales.ws_order_number, collate: N/A] | N/A | N/A |
| │ └─Selection_40 | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:129.8ms, loops:571, threads:24} | not(isnull(tpcds50.web_sales.ws_warehouse_sk)) | N/A | N/A |
| │ └─TableFullScan_39 | 35996850.00 | 35996850 | mpp[tiflash] | table:ws1 | tiflash_task:{time:120.8ms, loops:571, threads:24} | keep order:false | N/A | N/A |
| └─ExchangeReceiver_46(Probe) | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:36.6ms, loops:571, threads:24} | | N/A | N/A |
| └─ExchangeSender_45 | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:12.7s, loops:571, threads:24} | ExchangeType: HashPartition, Hash Cols: [name: tpcds50.web_sales.ws_order_number, collate: N/A] | N/A | N/A |
| └─Selection_44 | 35988060.45 | 35988084 | mpp[tiflash] | | tiflash_task:{time:126.4ms, loops:571, threads:24} | not(isnull(tpcds50.web_sales.ws_warehouse_sk)) | N/A | N/A |
| └─TableFullScan_43 | 35996850.00 | 35996850 | mpp[tiflash] | table:ws2 | tiflash_task:{time:117.4ms, loops:571, threads:24} | keep order:false | N/A | N/A |
+-----------------------------------------------+--------------+-----------+--------------+------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------+---------+

```
### 4. What is your TiDB version? (Required)

commit f0d4a9696fcdb878d2a4a50f128da8ddc9a1b0ab