wei0339
Newbie
Offline
Activity: 12
Merit: 1
|
|
May 29, 2019, 10:54:36 PM |
|
I may have missed it, but: What does the GPU0 thread 0 interleave adjust xxx ms means? For dual thread CN mining to keep its highest hashrate over time you need to make sure the two threads are running with the appropriate separation in time. If they gravitate and become too close to running synchronously, the whole point of running dual threads disappears. Therefore, it's better to delay one of the threads a little before starting the next hash run. This is the "interleave adjustment" logged above. As long as your hashrate is stable (and high) over time, it's nothing to worry about, everything is working as intended. Asking, because i saw it on only one card. And this one tends to crash every week or two. I have the same problem. VEGA 56 is normal in the case of "mode 16+14". However, the same problem occurs after using AMD Mem Tweak & "--cn_config 1x*1x". And after a while, it will drop from 2000 h/s to 16xx h/s. Is it an overclocking problem? Is there any solution?
|
|
|
|
MinersRus
Member
Offline
Activity: 214
Merit: 24
|
|
May 30, 2019, 04:10:48 AM |
|
I may have missed it, but: What does the GPU0 thread 0 interleave adjust xxx ms means? For dual thread CN mining to keep its highest hashrate over time you need to make sure the two threads are running with the appropriate separation in time. If they gravitate and become too close to running synchronously, the whole point of running dual threads disappears. Therefore, it's better to delay one of the threads a little before starting the next hash run. This is the "interleave adjustment" logged above. As long as your hashrate is stable (and high) over time, it's nothing to worry about, everything is working as intended. Asking, because i saw it on only one card. And this one tends to crash every week or two. I have the same problem. VEGA 56 is normal in the case of "mode 16+14". However, the same problem occurs after using AMD Mem Tweak & "--cn_config 1x*1x". And after a while, it will drop from 2000 h/s to 16xx h/s. Is it an overclocking problem? Is there any solution? I have had issues with various Vega 56's that had the same problems that you mention (drop from 2000 h/s to 16xx h/s or just crash). The problems usually a few days to happen. The solution I have found is to increase the core voltage in small increments until the problems disappear. I also had to down the core clock speed. Example: Vega 56: Core 920mv and 1300 MHz core. I also run the mode as 14*14. I still get around 2000 H/s (no memory tweak done yet)
|
|
|
|
pbfarmer
Member
Offline
Activity: 340
Merit: 29
|
|
May 30, 2019, 05:35:34 AM |
|
I may have missed it, but: What does the GPU0 thread 0 interleave adjust xxx ms means? For dual thread CN mining to keep its highest hashrate over time you need to make sure the two threads are running with the appropriate separation in time. If they gravitate and become too close to running synchronously, the whole point of running dual threads disappears. Therefore, it's better to delay one of the threads a little before starting the next hash run. This is the "interleave adjustment" logged above. As long as your hashrate is stable (and high) over time, it's nothing to worry about, everything is working as intended. Asking, because i saw it on only one card. And this one tends to crash every week or two. I have the same problem. VEGA 56 is normal in the case of "mode 16+14". However, the same problem occurs after using AMD Mem Tweak & "--cn_config 1x*1x". And after a while, it will drop from 2000 h/s to 16xx h/s. Is it an overclocking problem? Is there any solution? watch your hbm temps. you may be over-temp throttling
|
|
|
|
kerney666
Member
Offline
Activity: 658
Merit: 86
|
|
May 30, 2019, 02:09:33 PM |
|
I may have missed it, but: What does the GPU0 thread 0 interleave adjust xxx ms means? For dual thread CN mining to keep its highest hashrate over time you need to make sure the two threads are running with the appropriate separation in time. If they gravitate and become too close to running synchronously, the whole point of running dual threads disappears. Therefore, it's better to delay one of the threads a little before starting the next hash run. This is the "interleave adjustment" logged above. As long as your hashrate is stable (and high) over time, it's nothing to worry about, everything is working as intended. Asking, because i saw it on only one card. And this one tends to crash every week or two. I have the same problem. VEGA 56 is normal in the case of "mode 16+14". However, the same problem occurs after using AMD Mem Tweak & "--cn_config 1x*1x". And after a while, it will drop from 2000 h/s to 16xx h/s. Is it an overclocking problem? Is there any solution? watch your hbm temps. you may be over-temp throttling Hi! When you exit the miner after this happens, do you get a clean shutdown or is one thread timing out? The 2000 to 1600 h/s drop is very close to the expected single threaded performance, I’m guessing one thread has died.
|
|
|
|
kerney666
Member
Offline
Activity: 658
Merit: 86
|
|
May 30, 2019, 02:11:28 PM |
|
Hey Kerney. I had PMed you in the MMP OS Discord for a weird issue. 2 days ago, i experencied a crashed GPU, its a 570 4GB. System is Win 10 1703, AMD 19.2.1, 4GB RAM, i5-3470. The GPU would die for no reason (maybe timings). The GPU stats reporting will stop, and only the share submitting will remain. API is getting killed as well. But the miner itself does not seem to react properly with the dead GPU, as no SICK/DEAD is declared (and the API should still be responsive). Miner does not initiate properly the shutdown process, or at the very least, output is not shown. The GPUs did stop mining though. Trying to invoke the Task Manager or the AMD Radeon Settings freezes them almost right away. Even the driver does not know if its crashed or not EDIT2: Here are the screenshots. https://imgur.com/a/fu3ZkxjI presume, it could be overheating, since yesterday crashed again at 81*C. May EDIT: Also, i see that the Interleaving adjustment happens primarily on the GPU0 (shortly after the miner initialization). EDIT3: When the card was run with lower voltage (core + IMC), it crashed, and it was detected and declared as DEAD. Will try again, by raising the voltage, to see if the issue will repeat. Hi ku4eto! Thanks for the edit updates. Any more info after bumping voltage a little?
|
|
|
|
wei0339
Newbie
Offline
Activity: 12
Merit: 1
|
|
May 30, 2019, 04:44:22 PM |
|
I may have missed it, but: What does the GPU0 thread 0 interleave adjust xxx ms means? For dual thread CN mining to keep its highest hashrate over time you need to make sure the two threads are running with the appropriate separation in time. If they gravitate and become too close to running synchronously, the whole point of running dual threads disappears. Therefore, it's better to delay one of the threads a little before starting the next hash run. This is the "interleave adjustment" logged above. As long as your hashrate is stable (and high) over time, it's nothing to worry about, everything is working as intended. Asking, because i saw it on only one card. And this one tends to crash every week or two. I have the same problem. VEGA 56 is normal in the case of "mode 16+14". However, the same problem occurs after using AMD Mem Tweak & "--cn_config 1x*1x". And after a while, it will drop from 2000 h/s to 16xx h/s. Is it an overclocking problem? Is there any solution? I have had issues with various Vega 56's that had the same problems that you mention (drop from 2000 h/s to 16xx h/s or just crash). The problems usually a few days to happen. The solution I have found is to increase the core voltage in small increments until the problems disappear. I also had to down the core clock speed. Example: Vega 56: Core 920mv and 1300 MHz core. I also run the mode as 14*14. I still get around 2000 H/s (no memory tweak done yet) Core 920 mv, much better than before. Thank you.
|
|
|
|
ku4eto
Jr. Member
Offline
Activity: 194
Merit: 4
|
|
May 30, 2019, 08:28:18 PM |
|
Hey Kerney. I had PMed you in the MMP OS Discord for a weird issue. 2 days ago, i experencied a crashed GPU, its a 570 4GB. System is Win 10 1703, AMD 19.2.1, 4GB RAM, i5-3470. The GPU would die for no reason (maybe timings). The GPU stats reporting will stop, and only the share submitting will remain. API is getting killed as well. But the miner itself does not seem to react properly with the dead GPU, as no SICK/DEAD is declared (and the API should still be responsive). Miner does not initiate properly the shutdown process, or at the very least, output is not shown. The GPUs did stop mining though. Trying to invoke the Task Manager or the AMD Radeon Settings freezes them almost right away. Even the driver does not know if its crashed or not EDIT2: Here are the screenshots. https://imgur.com/a/fu3ZkxjI presume, it could be overheating, since yesterday crashed again at 81*C. EDIT: Also, i see that the Interleaving adjustment happens primarily on the GPU0 (shortly after the miner initialization). EDIT3: When the card was run with lower voltage (core + IMC), it crashed, and it was detected and declared as DEAD. Will try again, by raising the voltage, to see if the issue will repeat. Hi ku4eto! Thanks for the edit updates. Any more info after bumping voltage a little? Okay, after some testing, i got in the past 2 days, 3-4 more crashes in total. After playing with the voltage for that card (retesting the stability, in case it has degraded), but the voltage initially used was indeed correct. Bumping up the voltage by 2 steps (2x6.25mV) had no difference. I set the fan today to 95% for above 75*C (was for 85*C before that). I have not observed the issue still. I will check for 2 more days, before calling it 100% an thermal issue. Its a RX 570 Gigabyte Aorus. Not sure, but i believe some cards had a different thermal limit for some elements, compared to the rest. Although, i did compare the BIOS, and it had the same temp limits defined, as the rest of my cards.
|
|
|
|
katar14
Newbie
Offline
Activity: 20
Merit: 0
|
|
May 31, 2019, 07:38:42 PM |
|
algo Cryptonight HeavyX and MTP when? HeavyX for X-cash is already available, it's called cnv8_dbl in TRM. We're still debating how to prioritize MTP. The issue is the big investment in time for building the custom stratum protocol, submitting those huge proofs etc, all for a single coin and algo. there will be new coins on mtp np; very good project tecracoin, mtp has a future
|
|
|
|
alexcrys1
Newbie
Offline
Activity: 60
Merit: 0
|
|
May 31, 2019, 09:56:30 PM |
|
algo Cryptonight HeavyX and MTP when? HeavyX for X-cash is already available, it's called cnv8_dbl in TRM. We're still debating how to prioritize MTP. The issue is the big investment in time for building the custom stratum protocol, submitting those huge proofs etc, all for a single coin and algo. there will be new coins on mtp np; very good project tecracoin, mtp has a future how do you know MTP is the future? what do I know progpow -maybe it's the future.. why MTP it is
|
|
|
|
fluxy12
Jr. Member
Offline
Activity: 145
Merit: 1
|
|
June 01, 2019, 04:28:02 AM |
|
algo Cryptonight HeavyX and MTP when? HeavyX for X-cash is already available, it's called cnv8_dbl in TRM. We're still debating how to prioritize MTP. The issue is the big investment in time for building the custom stratum protocol, submitting those huge proofs etc, all for a single coin and algo. there will be new coins on mtp np; very good project tecracoin, mtp has a future how do you know MTP is the future? what do I know progpow -maybe it's the future.. why MTP it is He said "mtp HAS a future" not "is the future" that's not exactly the same meaning
|
|
|
|
zsoca72
Newbie
Offline
Activity: 14
Merit: 0
|
|
June 05, 2019, 02:17:14 PM |
|
TRM supported WOWnero mining?
|
|
|
|
PIOUPIOU99
Copper Member
Member
Offline
Activity: 293
Merit: 11
|
|
June 05, 2019, 02:22:59 PM |
|
WoWnero will soon fork in RandomX it seems to me
|
|
|
|
kerney666
Member
Offline
Activity: 658
Merit: 86
|
|
June 05, 2019, 07:36:30 PM |
|
WoWnero will soon fork in RandomX it seems to me
RandomX is a very different beast, and yes wownero will fork shortly. We're not making any promises around RandomX support at this point, we'll observe what happens.
|
|
|
|
todxx (OP)
Member
Offline
Activity: 176
Merit: 76
|
|
June 05, 2019, 08:08:08 PM Merited by PIOUPIOU99 (1) |
|
Team Red Miner v0.5.0 releasedhttps://github.com/todxx/teamredminer/releasesChanges in v0.5.0 - Added cryptonight 4MB variants: heavy, haven and saber.
- Added x16 algo suite: x16r, x16s, x16rt (both gin and veil).
- Auto-tuning mode for all CN variants, see bundled guide.
- Manual key-driven CN tuning mode available inside the miner.
- Additional data in miner stats console output.
- Watchdog now detecting single stuck thread when mining CN.
- Fix: in rare cases, poolside hash for compute algos (lyra2z, phi2, lyra2rev3) only reached ~95% of expected value.
We've added new algos for 4MB/heavy CN variants! Now you can mine haven/bittube with TRM We've also added more finetuning options, as well as built-in auto-tuning to find the best config!
|
|
|
|
Iamtutut
|
|
June 05, 2019, 08:25:57 PM |
|
Let's try !
|
|
|
|
UnclWish
|
|
June 05, 2019, 10:11:26 PM |
|
Team Red Miner v0.5.0 releasedhttps://github.com/todxx/teamredminer/releasesChanges in v0.5.0 - Added cryptonight 4MB variants: heavy, haven and saber.
- Added x16 algo suite: x16r, x16s, x16rt (both gin and veil).
- Auto-tuning mode for all CN variants, see bundled guide.
- Manual key-driven CN tuning mode available inside the miner.
- Additional data in miner stats console output.
- Watchdog now detecting single stuck thread when mining CN.
- Fix: in rare cases, poolside hash for compute algos (lyra2z, phi2, lyra2rev3) only reached ~95% of expected value.
We've added new algos for 4MB/heavy CN variants! Now you can mine haven/bittube with TRM We've also added more finetuning options, as well as built-in auto-tuning to find the best config! Wow! Exccelent update! Thanx! Let's try it!
|
|
|
|
GKumaran
Member
Offline
Activity: 204
Merit: 10
|
|
June 06, 2019, 01:59:43 AM Last edit: June 06, 2019, 02:19:57 AM by GKumaran |
|
Awesome Update :Results for Vega 64, Aircooled, Reference On Windows 10, 18.6.1 Driver Algorithm | Hashrate | Power at wall (Watt) | Efficiency (Hash/Watt) | ODT Config | TRM Config | x16r, x16s, x16rt | (average- mh/s) 21.5 | (average) 185 | 116.22 | 1082/ 800/812 | | CN R | 2345 | 205 | 11.44 | 1408/1100/875 | 16*14:AAA | CN Saber
| 2185 | 177 | 12.35 | 1408/1100/875 | 15+15:CBB | CN Haven | 2170 | 173 | 12.54 | 1408/1100/875 | 16+14:CBB | CN Turtle | 22340 | 198 | 112.83 | 1216/1100/875 | L26+26:AAA | CN UPX2 | 84940 | 208 | 408.37 | 1216/1100/875 | L26+26:AAA or L26+26:BAA | CN HeavyX | 1207 | 186 | 6.49 | 1408/1100/875 | 15*14:AAA |
Note: The TRM config is just for info, use the auto tuning and get the best config for your card/clock/timings.
|
|
|
|
UnclWish
|
|
June 06, 2019, 10:03:16 AM |
|
It's just wanderfull! Auto and manual config works fine!
|
|
|
|
Mashy81
Jr. Member
Offline
Activity: 225
Merit: 1
|
|
June 06, 2019, 10:25:18 AM |
|
Great update. Auto tune works great. Pickedup an extra 400hs on 4xvega rig and only an extra 20watts. Have only tried cnr so far
Vega 56 1407/930/875mv 2240hs 13*13:AAA
|
|
|
|
Iamtutut
|
|
June 06, 2019, 10:42:48 AM |
|
Very impressive upgrade, a few extra hundred hashes with Turtlecoin and very good implementation of CN Heavy algos.
Excellent work, kudos to the devs.
|
|
|
|
|