RX 550 hash problem after hardfork

[minimalist83]

Type of the GPU: Gigabyte RX 550 2GB Windows 10 Pro I did not compile the miner by my own. GPUs are one-click time patch overclocked. Using version 2.5.1.

I have two rigs each containing 6 RX 550 (2GB). My hashrate for each was about 2.5KH/s. I was using dual thread for 6 months with no problem. Now it's 1.6KH/s just after the hardwork. There is no other problem. Even without dual thread, it's so slow.

[huddsonpereira]

Mine too.

Went from avg 450 H/s to 305 H/s each RX 550 after the fork.

[keefstone]

Same for me with Vega rigs. I'm seeing much lower hashrates on xmr-stak 2.5.1 with Vega 64, Vega FE and Vega 56.

Running AMD 18.6.1 drivers with Windows 10.

[huddsonpereira]

Changed my intensity from 422 to 432 and worksize from 8 to 16 and the hashrate is now 410+ H/s

[keefstone]

Changed my intensity from 422 to 432 and worksize from 8 to 16 and the hashrate is now 410+ H/s

Ha! I was just going to respond that bumping the worksize from 8 to 16 seems to have fixed it for me as well.

[minimalist83]

It's better after changing worksize from 8 to 16 but it's still 10% slow.

[keefstone]

It's better after changing worksize from 8 to 16 but it's still 10% slow.

Looks like that is normal.

[Tafaz]

Hi, can i ask you to see your amd config? i'm not able to push my rx550 over 390H/s

[huddsonpereira]

Hi, can i ask you to see your amd config? i'm not able to push my rx550 over 390H/s

this is my amd config right now

{ "index" : 0, "intensity" : 448, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2, "unroll" : 8, "comp_mode" : true }, { "index" : 0, "intensity" : 448, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2, "unroll" : 8, "comp_mode" : true },

[Tafaz]

Thanks man, i'll try and let you know!

[Tafaz]

my cards with your config are slower

---

HASHRATE REPORT - AMD | ID | 10s | 60s | 15m | ID | 10s | 60s | 15m | | 0 | 149.7 | (na) | (na) | 1 | 149.7 | (na) | (na) | Totals (AMD): 299.4 0.0 0.0 H/s

Totals (ALL): 299.4 0.0 0.0 H/s Highest: 299.5 H/s

which model is yours? mine is a Sapphire RX550 2gb

[huddsonpereira]

Mine is RX 550 Gigabyte Gaming OC 2gb (https://www.gigabyte.com/Graphics-Card/GV-RX550GAMING-OC-2GD) with One time-clicking Patch and no Overclock

[Tafaz]

Thanks ... maybe there are some HW differences ... with some config testing now i'm around 400H/s on each card, so if anyone is interested here's my config

"gpu_threads_conf" : [ // gpu: Baffin memory:1395 // compute units: 10 { "index" : 0, "intensity" : 640, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2, "unroll" : 8, "comp_mode" : true }, // gpu: Baffin memory:1395 // compute units: 10 { "index" : 1, "intensity" : 640, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2, "unroll" : 8, "comp_mode" : true }, // gpu: Baffin memory:1395 // compute units: 10 { "index" : 2, "intensity" : 640, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2, "unroll" : 8, "comp_mode" : true }, ],

Bios modded with One Click Timing Patch and memory @1940 Mhz

Here's the results

HASHRATE REPORT - AMD | ID | 10s | 60s | 15m | ID | 10s | 60s | 15m | | 0 | 399.5 | 397.9 | (na) | 1 | 401.9 | 401.9 | (na) | | 2 | 401.6 | 401.5 | (na) | Totals (AMD): 1203.1 1201.3 0.0 H/s

Totals (ALL): 1203.1 1201.3 0.0 H/s Highest: 1203.7 H/s

[NmxMilk]

Hi, Have same problem here with my rx 550. It's even worst in fact. Since after several minutes the hash rate will drop to ZERO.

I downgraded to adrenalin 18.5.1 from later version but the problem is still here.

When the hashrate drops, I always have this event in windows: System : Event ID: 4101 : Display driver amdkmdap stopped responding and has successfully recovered.

Last night I reduced intensity from 432 to 256 and this morning it is still hashing. Combined with the fork, I loose about 15% of my hash rate though !

Cheers

[CryptoBroke550]

I used the following settings to stop my hash rate from dropping to zero. RX 550 4GB 2 threads 400-407H/s 30 hours so far after the change. "intensity" : 640, "worksize" : 16, "affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2, "unroll" : 4, "comp_mode" : false

Does anyone know why some use dual threads with two different intensity values? I have not tested this as i'm not sure where to start.

[Spudz76]

In cryptonight_v7 with older versions of xmr-stak there was a limit on intensity (1280?) which has since been fixed, so in order to get best rates (on old algo in old times on old drivers) you would put the most intensity it would do without crashing on one thread and then put the "overflow" work on a second thread, and that was how it worked best.

BUT none of that even remotely matters anymore. Algo works equipment in new ways and might as well be unknown, burn all old configuration 'tips' they will tend to be misleading. Cross-comparisons will also be useless, if deviceA was 30% of deviceB in v7 it will probably not be the same in v8 performance. deviceB could be faster than deviceA now, the work load is different. Unless devices are identical don't use relative-differential scaling for your targets of what 'should be' with new algo.

[CryptoBroke550]

Thanks for the info, saved me some time.