search

MaynardMiner / SWARM

Profit Switching Mining Administrator For HiveOS/Linux & Windows: HiveOS Integrated
MIT License
92 stars 40 forks source link

Hash rates wrongly associated on HiveOS dashboard #419

Closed UserDC-LeGrand closed 1 year ago

UserDC-LeGrand commented 1 year ago

This issue is strange.

On Gminer-N ETHASH, somehow the hash rates for the 3070 GPU#2, wind up in GPU#0. The first two 2060 GPU's (0 an1) should hover around the 30-32 Mh/s range, and the last 3070 GPU around the 60-63 MH/s range.

The issue is in the way you build your scripts ... Look at the order of the devices you are passing on the command line - -d 6 7 4 2 0 3 5 1 . This should be removed or properly sequence such as this: 0 1 2 3 4 5 6 7

root@rig4:/hive/miners/custom/SWARM.3.6.7.linux/bin/gminer-a-1# /hive/miners/custom/SWARM.3.6.7.linux/bin/gminer-a-1/miner -d 6 7 4 2 0 3 5 1 --api 33001 --server etchash.auto.nicehash.com --port 9200 --algo etchash --proto stratum --user xxxxxxxxxxxxxxxxxxxxxxx.rig4 --logfile '/hive/miners/custom/SWARM.3.6.7.linux/logs/gminer-a-1.log' --pass x --templimit 61

image

image

MaynardMiner commented 1 year ago

Go to .\debug\busdata.txt. It is a .txt file, but should be just normal json.

Please confirm that it is sorting the gpus correctly by its bus_id. It should appear in order based on its bus_id.

If that is correct, check agent-screen and see if bus_numbers are correct in minerstats [ ]

If that is correct, check order of hashrate hs:[ ] presented in agent-screen. If it's showing the 64.61 mh/s as first, then this is issue with Gminer, not SWARM.

MaynardMiner commented 1 year ago

Follow up- Did you ever check? I was searching code, and I don't see how Swarm atm could display hashrates in wrong slot currently (not to say bug isn't there).

Most miners don't show what hashrate goes to which gpu, just a list of hashrates in api. It is assumed it is displayed in order of gpu bus.

UserDC-LeGrand commented 1 year ago

I did but the sequence was OK for the miner/algo combo at the time I checked. I need ot do this again – and benchmark that specific miner/algo combo, while stairng at the background screens and pointers you mentioned. You are probably bang on with the specific miner having the issue ...

From: @.> Sent: October 27, 2022 10:11 AM To: @.> Cc: @.>; @.> Subject: Re: [MaynardMiner/SWARM] Hash rates wrongly associated on HiveOS dashboard (Issue #419)

Follow up- Did you ever check? I was searching code, and I don't see how Swarm atm could display hashrates in wrong slot currently (not to say bug isn't there).

Most miners don't show what hashrate goes to which gpu, just a list of hashrates in api. It is assumed it is displayed in order of gpu bus.

— Reply to this email directly, view it on GitHubhttps://github.com/MaynardMiner/SWARM/issues/419#issuecomment-1293585783, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AXXBVDPH2L43C7IUDVDVVZDWFKEQTANCNFSM6AAAAAARNP6BVI. You are receiving this because you authored the thread.Message ID: @.***>

UserDC-LeGrand commented 1 year ago

Gminer or is it HiveOS: displaying the hashrates out of sequence on another rig this morning ...

The 63.5 MH/S ... should be assigned to the 3070LHR GPU 0 and 2 ...

image

image

image

UserDC-LeGrand commented 1 year ago

Found the issue, and perhaps it is in the way you build your scripts ... Look at the order of the devices you are passing on the command line - -d 6 7 4 2 0 3 5 1 . This should be removed or properly sequence such as this: 0 1 2 3 4 5 6 7

root@rig4:/hive/miners/custom/SWARM.3.6.7.linux/bin/gminer-a-1# /hive/miners/custom/SWARM.3.6.7.linux/bin/gminer-a-1/miner -d 6 7 4 2 0 3 5 1 --api 33001 --server etchash.auto.nicehash.com --port 9200 --algo etchash --proto stratum --user xxxxxxxxxxxxxxxxxxxxxxx.rig4 --logfile '/hive/miners/custom/SWARM.3.6.7.linux/logs/gminer-a-1.log' --pass x --templimit 61

image

MaynardMiner commented 1 year ago

Can you post your debug\devicelist.txt file here? I just want to know how SWARM is recording them.

MaynardMiner commented 1 year ago

.\debug\busdata.txt too. If you can post both files, I can figure out issue.

UserDC-LeGrand commented 1 year ago

Im not using -GPUdevices1 or other... I just use -Type NVIDIA1,CPU or AMD1,CPU argument depending on rig.

From: @.> Sent: October 30, 2022 10:20 PM To: @.> Cc: @.>; @.> Subject: Re: [MaynardMiner/SWARM] Hash rates wrongly associated on HiveOS dashboard (Issue #419)

Are you using the -GPUDevices1 argument?

— Reply to this email directly, view it on GitHubhttps://github.com/MaynardMiner/SWARM/issues/419#issuecomment-1296452447, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AXXBVDL7LPXZO7EUZE74JTLWF4UFHANCNFSM6AAAAAARNP6BVI. You are receiving this because you authored the thread.Message ID: @.***>

UserDC-LeGrand commented 1 year ago

As requested. The busid is correctly sequenced. The device list is not ... Should it not be 0, 1 ,2 ?

root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug# cat busdata.txt [ { "busid": "03:00.0", "name": "NVIDIA TU106 [GeForce RTX 2060 Rev. A]", "brand": "nvidia", "subvendor": "ASUS", "mem": "6144 MiB", "vbios": "90.06.46.00.89", "plim_min": "105.00 W", "plim_def": "190.00 W", "plim_max": "238.00 W" }, { "busid": "06:00.0", "name": "NVIDIA TU106 [GeForce RTX 2060 Rev. A]", "brand": "nvidia", "subvendor": "ASUS", "mem": "6144 MiB", "vbios": "90.06.46.00.89", "plim_min": "105.00 W", "plim_def": "190.00 W", "plim_max": "238.00 W" }, { "busid": "08:00.0", "name": "NVIDIA GA104 [GeForce RTX 3070 LHR]", "brand": "nvidia", "subvendor": "EVGA", "mem": "8192 MiB", "vbios": "94.04.46.40.60", "plim_min": "100.00 W", "plim_def": "240.00 W", "plim_max": "270.00 W" } ] root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug# cat devicelist.txt { "AMD": {}, "CPU": { "11": 11, "9": 9, "7": 7, "6": 6, "2": 2, "13": 13, "12": 12, "3": 3, "4": 4, "8": 8, "10": 10, "0": 0, "5": 5, "1": 1 }, "NVIDIA": { "2": "2", "0": "0", "1": "1" } } root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug#

MaynardMiner commented 1 year ago

As requested. The busid is correctly sequenced. The device list is not ... Should it not be 0, 1 ,2 ?

root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug# cat busdata.txt [ { "busid": "03:00.0", "name": "NVIDIA TU106 [GeForce RTX 2060 Rev. A]", "brand": "nvidia", "subvendor": "ASUS", "mem": "6144 MiB", "vbios": "90.06.46.00.89", "plim_min": "105.00 W", "plim_def": "190.00 W", "plim_max": "238.00 W" }, { "busid": "06:00.0", "name": "NVIDIA TU106 [GeForce RTX 2060 Rev. A]", "brand": "nvidia", "subvendor": "ASUS", "mem": "6144 MiB", "vbios": "90.06.46.00.89", "plim_min": "105.00 W", "plim_def": "190.00 W", "plim_max": "238.00 W" }, { "busid": "08:00.0", "name": "NVIDIA GA104 [GeForce RTX 3070 LHR]", "brand": "nvidia", "subvendor": "EVGA", "mem": "8192 MiB", "vbios": "94.04.46.40.60", "plim_min": "100.00 W", "plim_def": "240.00 W", "plim_max": "270.00 W" } ] root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug# cat devicelist.txt { "AMD": {}, "CPU": { "11": 11, "9": 9, "7": 7, "6": 6, "2": 2, "13": 13, "12": 12, "3": 3, "4": 4, "8": 8, "10": 10, "0": 0, "5": 5, "1": 1 }, "NVIDIA": { "2": "2", "0": "0", "1": "1" } } root@rig5:/hive/miners/custom/SWARM.3.6.7.linux/debug#

This is not same rig.

Rig above has 9 gpus. This devicelist.txt shows only 3.

MaynardMiner commented 1 year ago

I think I found issue. It's being Powershell, and adding to the table out of order.

MaynardMiner commented 1 year ago

https://github.com/MaynardMiner/SWARM/commit/801bdb5b2bd8411a115045a73957646675f43b12

This should fix it, just I don't know the full extent/implications it will have particularly with cross-platform (AMD & NVIDIA) on rig. But it will order them by their device slot, and by extension the -d argument should be put correctly.

If you want to forward test and confirm while I'm working on other stuff, would appreciate it. It's a simple edit to the file notated in the commit.