[ANN] TeamRedMiner v0.10.10 - Ironfish/Kaspa/ZIL/Kawpow/Etchash and More - page 86.

kerney666

member

Activity: 658

Merit: 86

Quote from: carlosmonaco on April 19, 2019, 09:44:22 AM

Is it better to flash vega 56 with 64 bios or stay with original one ( VEGA 56 reference BlowER ).
Some of them are already hard to get stable ( 1407clock /900 mem : 930 mV for the bad ones)

I’ve had all my ref V56 flashed to V64 for such a long time now, I honestly can’t say much about stock 56 bios with Samsung HBM.

I think you should try it though. The mem voltage is actually hardcoded in the bios, the names in eg OverdriveNTool are misnomers. The V64 bios drives mem with a higher voltage, so you might have a better (mem) experience. Just save the old bios and you can revert in a matter of minutes.

carlosmonaco

newbie

Activity: 105

Merit: 0

Is it better to flash vega 56 with 64 bios or stay with original one ( VEGA 56 reference BlowER ).
Some of them are already hard to get stable ( 1407clock /900 mem : 930 mV for the bad ones)

Kerppa77

newbie

Activity: 17

Merit: 1

Quote from: kerney666 on April 19, 2019, 09:13:54 AM

Quote from: Kerppa77 on April 19, 2019, 09:10:49 AM

Quote from: kerney666 on April 19, 2019, 08:58:18 AM

Quote from: Kerppa77 on April 19, 2019, 08:30:01 AM

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

Indeed,

I finally succeeded to get the --current values out from the rigs and it seems that the TRM 0.4.4. haven't managed to set updated timings on the GPUs that run with ~2100H/s... (all GPUs in the 3 rigs and one GPU from the 4th rig has still CL 20, REF 3900 RAS 33 RFC 260 etc. while other better performing 5 GPUs has CL 19, REF 15600, RAS 28, RFC 250).

What it the reason TRM 0.4.4. haven't managed to update all the timings? Or should I do manually a .bat to get things set properly? I have so far just run the TRM 0.4.4. just like TRM 0.4.2. w/o any added parameters or anything.

We don’t set those timings for you, everyone needs to download ElioVP’s tool themselves and include the timings mod in some .bat file Grin

. We don’t want to bundle it, it’s a slippery slope from a license perspective imho.

I am thinking a little about building some flight control gui for TRM though that would be fully open source. No idea when I’ll get the time though...

hmm.... Now I wonder how one of my rigs still have updated timings in 5 out of 6 GPUs Cheesy

I did indeed wonder the same thing. Quite certain we didn’t set them for you

.
[/quote]

HAHA I found the reason

My shortcut points to old .bat which have the config I used with TRM 0.4.2. Cheesy

Now I got all fixed.... thanks for the help... awesome miner and results! Thanks!!!

kerney666

member

Activity: 658

Merit: 86

Quote from: Kerppa77 on April 19, 2019, 09:10:49 AM

Quote from: kerney666 on April 19, 2019, 08:58:18 AM

Quote from: Kerppa77 on April 19, 2019, 08:30:01 AM

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

Indeed,

I finally succeeded to get the --current values out from the rigs and it seems that the TRM 0.4.4. haven't managed to set updated timings on the GPUs that run with ~2100H/s... (all GPUs in the 3 rigs and one GPU from the 4th rig has still CL 20, REF 3900 RAS 33 RFC 260 etc. while other better performing 5 GPUs has CL 19, REF 15600, RAS 28, RFC 250).

What it the reason TRM 0.4.4. haven't managed to update all the timings? Or should I do manually a .bat to get things set properly? I have so far just run the TRM 0.4.4. just like TRM 0.4.2. w/o any added parameters or anything.

We don’t set those timings for you, everyone needs to download ElioVP’s tool themselves and include the timings mod in some .bat file Grin

. We don’t want to bundle it, it’s a slippery slope from a license perspective imho.

I am thinking a little about building some flight control gui for TRM though that would be fully open source. No idea when I’ll get the time though...

hmm.... Now I wonder how one of my rigs still have updated timings in 5 out of 6 GPUs Cheesy

[/quote]

I did indeed wonder the same thing. Quite certain we didn’t set them for you

.

Kerppa77

newbie

Activity: 17

Merit: 1

Quote from: kerney666 on April 19, 2019, 08:58:18 AM

Quote from: Kerppa77 on April 19, 2019, 08:30:01 AM

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

Indeed,

I finally succeeded to get the --current values out from the rigs and it seems that the TRM 0.4.4. haven't managed to set updated timings on the GPUs that run with ~2100H/s... (all GPUs in the 3 rigs and one GPU from the 4th rig has still CL 20, REF 3900 RAS 33 RFC 260 etc. while other better performing 5 GPUs has CL 19, REF 15600, RAS 28, RFC 250).

What it the reason TRM 0.4.4. haven't managed to update all the timings? Or should I do manually a .bat to get things set properly? I have so far just run the TRM 0.4.4. just like TRM 0.4.2. w/o any added parameters or anything.

We don’t set those timings for you, everyone needs to download ElioVP’s tool themselves and include the timings mod in some .bat file Grin

. We don’t want to bundle it, it’s a slippery slope from a license perspective imho.

I am thinking a little about building some flight control gui for TRM though that would be fully open source. No idea when I’ll get the time though...
[/quote]

hmm.... Now I wonder how one of my rigs still have updated timings in 5 out of 6 GPUs Cheesy

kerney666

member

Activity: 658

Merit: 86

Quote from: Kerppa77 on April 19, 2019, 08:30:01 AM

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

Indeed,

I finally succeeded to get the --current values out from the rigs and it seems that the TRM 0.4.4. haven't managed to set updated timings on the GPUs that run with ~2100H/s... (all GPUs in the 3 rigs and one GPU from the 4th rig has still CL 20, REF 3900 RAS 33 RFC 260 etc. while other better performing 5 GPUs has CL 19, REF 15600, RAS 28, RFC 250).

What it the reason TRM 0.4.4. haven't managed to update all the timings? Or should I do manually a .bat to get things set properly? I have so far just run the TRM 0.4.4. just like TRM 0.4.2. w/o any added parameters or anything.

[/quote]

We don’t set those timings for you, everyone needs to download ElioVP’s tool themselves and include the timings mod in some .bat file Grin

. We don’t want to bundle it, it’s a slippery slope from a license perspective imho.

I am thinking a little about building some flight control gui for TRM though that would be fully open source. No idea when I’ll get the time though...

Kerppa77

newbie

Activity: 17

Merit: 1

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?
[/quote]

Indeed,

I finally succeeded to get the --current values out from the rigs and it seems that the TRM 0.4.4. haven't managed to set updated timings on the GPUs that run with ~2100H/s... (all GPUs in the 3 rigs and one GPU from the 4th rig has still CL 20, REF 3900 RAS 33 RFC 260 etc. while other better performing 5 GPUs has CL 19, REF 15600, RAS 28, RFC 250).

What it the reason TRM 0.4.4. haven't managed to update all the timings? Or should I do manually a .bat to get things set properly? I have so far just run the TRM 0.4.4. just like TRM 0.4.2. w/o any added parameters or anything.

wacholek

newbie

Activity: 20

Merit: 0

Quote from: kerney666 on April 19, 2019, 08:07:32 AM

Quote from: wacholek on April 19, 2019, 07:46:26 AM

Quote from: todxx on April 18, 2019, 06:10:21 PM

Team Red Miner v0.4.4 released

https://github.com/todxx/teamredminer/releases

Changes in v0.4.4

Added * mode specifically for modded timings on Vega GPUs. Use with e.g. --cn_config=15*15. This mode is now the default for Vegas.
Introduced slow start/ramp-up. Threads increase their workload slowly at start or restart after e.g. a network outage.
Added interleave adjustment logic. Readjusts the two threads per gpu over time to make sure they don't gravitate and get stuck.
Added support for forcing colors (--force_color) for windows redirected consoles (git bash, cygwin, node.js).
Added hotkey menu system (show stats, enable/disable gpu).

The new * mode opens up tweaking Vega timings for 2MB CN algos like CN-R!
We've seen a few extremely tuned V64 LCs able to hit over 2400h/s on CN-R with 1400/1100 clocks!
For those of you playing with timings, please share your results! We love seeing how far you guys can push these cards

Setting L22*22 on my Vega 64 just decrease hashrate in CN_Turtle. On L22*22 I got 20.26kH/s and on L22+22 21.80kH/s on this timings:
--CL 19 --RAS 30 --RCDRD 12 --RCDWR 6 --RC 44 --RP 13 --RRDS 5 --RRDL 5 --RTP 4 --FAW 18 --CWL 6 --WTRS 4 --WTRL 9 --WR 15 --WRRD 1 --RDWR 18 --REF 12800 --RFC 248

CNR 15*15 2375h/s
CNR 14*16 2385h/s

Vega set at: 1400@900 1007@900.

Thanks for testing and reporting! Should have been better described: the *-mode is for the 2MB variants, not for turtle. Turtle doesn’t need any tweaks, it puts max load on the mem controller anyway.

Using this timings:
--RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 10 --RRDS 3 --RRDL 3 --RTP 5 --FAW 20 --CWL 7 --WTRS 4 --WTRL 9 --WR 16 --WRRD 1 --RDWR 19 --REF 12800 --MRD 8 --MOD 15 --PD 8 --CKSRE 10 --CKSRX 10 --RFC 250

CNR: 14*16 2430h/s
One error previously:
Vega set at 1400@900 and 1107@900

kerney666

member

Activity: 658

Merit: 86

Quote from: wacholek on April 19, 2019, 07:46:26 AM

Quote from: todxx on April 18, 2019, 06:10:21 PM

Team Red Miner v0.4.4 released

https://github.com/todxx/teamredminer/releases

Changes in v0.4.4

Added * mode specifically for modded timings on Vega GPUs. Use with e.g. --cn_config=15*15. This mode is now the default for Vegas.
Introduced slow start/ramp-up. Threads increase their workload slowly at start or restart after e.g. a network outage.
Added interleave adjustment logic. Readjusts the two threads per gpu over time to make sure they don't gravitate and get stuck.
Added support for forcing colors (--force_color) for windows redirected consoles (git bash, cygwin, node.js).
Added hotkey menu system (show stats, enable/disable gpu).

The new * mode opens up tweaking Vega timings for 2MB CN algos like CN-R!
We've seen a few extremely tuned V64 LCs able to hit over 2400h/s on CN-R with 1400/1100 clocks!
For those of you playing with timings, please share your results! We love seeing how far you guys can push these cards

Setting L22*22 on my Vega 64 just decrease hashrate in CN_Turtle. On L22*22 I got 20.26kH/s and on L22+22 21.80kH/s on this timings:
--CL 19 --RAS 30 --RCDRD 12 --RCDWR 6 --RC 44 --RP 13 --RRDS 5 --RRDL 5 --RTP 4 --FAW 18 --CWL 6 --WTRS 4 --WTRL 9 --WR 15 --WRRD 1 --RDWR 18 --REF 12800 --RFC 248

CNR 15*15 2375h/s
CNR 14*16 2385h/s

Vega set at: 1400@900 1007@900.

Thanks for testing and reporting! Should have been better described: the *-mode is for the 2MB variants, not for turtle. Turtle doesn’t need any tweaks, it puts max load on the mem controller anyway.

wacholek

newbie

Activity: 20

Merit: 0

Quote from: todxx on April 18, 2019, 06:10:21 PM

Team Red Miner v0.4.4 released

https://github.com/todxx/teamredminer/releases

Changes in v0.4.4

Added * mode specifically for modded timings on Vega GPUs. Use with e.g. --cn_config=15*15. This mode is now the default for Vegas.
Introduced slow start/ramp-up. Threads increase their workload slowly at start or restart after e.g. a network outage.
Added interleave adjustment logic. Readjusts the two threads per gpu over time to make sure they don't gravitate and get stuck.
Added support for forcing colors (--force_color) for windows redirected consoles (git bash, cygwin, node.js).
Added hotkey menu system (show stats, enable/disable gpu).

The new * mode opens up tweaking Vega timings for 2MB CN algos like CN-R!
We've seen a few extremely tuned V64 LCs able to hit over 2400h/s on CN-R with 1400/1100 clocks!
For those of you playing with timings, please share your results! We love seeing how far you guys can push these cards

Setting L22*22 on my Vega 64 just decrease hashrate in CN_Turtle. On L22*22 I got 20.26kH/s and on L22+22 21.80kH/s on this timings:
--CL 19 --RAS 30 --RCDRD 12 --RCDWR 6 --RC 44 --RP 13 --RRDS 5 --RRDL 5 --RTP 4 --FAW 18 --CWL 6 --WTRS 4 --WTRL 9 --WR 15 --WRRD 1 --RDWR 18 --REF 12800 --RFC 248

CNR 15*15 2375h/s
CNR 14*16 2385h/s

Vega set at: 1400@900 1007@900.

Kerppa77

newbie

Activity: 17

Merit: 1

Quote from: kerney666 on April 19, 2019, 05:20:11 AM

Quote from: Kerppa77 on April 19, 2019, 04:43:14 AM

I wonder what is going on.... Shocked

I have 4 x 6 Vega64 rigs. All running with the same setup (OS, drivers, versions etc.)

When I applied the TRM 0.4.4 only one of my rigs got significant performance boost.

In this very rig 5 out of 6 Vega64 has ~2400H/s but one still has ~2100H/s and the same one has also regular interleave adjustments.

All the other GPUs in 3 rigs all have ~2100H/s and regular interleave adjustments.

I would like all the rigs to perform like ~2400H/s per GPU, but wonder what could be wrong with the most of the GPUs and rigs?

I tested 14*14 and seems that interleave adjustments are gone but hashrate remains ~2100H/s level per GPU.

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

I have optimized all the GPUs one by one. Rigs are stable now and Core is 1407Mhz with 860-970mV and Mem 1090-1100Mhz with 860-970mV. Power draw from the wall in these win10 rigs is 1320w in one rig and 1360-1370w rest of the rigs.

All the rigs have mixed set of GPUs in terms of Core and Mem clocks. There is also one weird thing too. The rig that got the hashrate boost with TRM 0.4.4. have overall like 70w lower power draw from the wall than the other 3 rigs.

Based on the mixed setup all the rigs should have ~ even power draw from the wall. I checked the HWinfo -> no signs of other differencies.

My GPUs did like 2020-2060H/s with TRM 0.4.2. and w/o timing adjustments. So 0.4.4. gives slight improvement in all the rest 3 rigs but major boost only one of the rigs (5 out of 6 GPUs in that rig).

Power draw from the wall increased 1320 -> 1420w and hashrate ~12%. In other 3 rigs hashes per GPU went only slightly up to ~2100H/s and rig power draw up like ~20-30w.

kerney666

member

Activity: 658

Merit: 86

Quote from: Kerppa77 on April 19, 2019, 04:43:14 AM

I wonder what is going on.... Shocked

I have 4 x 6 Vega64 rigs. All running with the same setup (OS, drivers, versions etc.)

When I applied the TRM 0.4.4 only one of my rigs got significant performance boost.

In this very rig 5 out of 6 Vega64 has ~2400H/s but one still has ~2100H/s and the same one has also regular interleave adjustments.

All the other GPUs in 3 rigs all have ~2100H/s and regular interleave adjustments.

I would like all the rigs to perform like ~2400H/s per GPU, but wonder what could be wrong with the most of the GPUs and rigs?

I tested 14*14 and seems that interleave adjustments are gone but hashrate remains ~2100H/s level per GPU.

The interleave adjustments aren’t always an indication of error, I have a few setups that need to adjust every 2 min or so per gpu.

The 2100 h/s is the normal Vega 64 hashrate at 14*14 or 15*15 before memory timings are applied. Have you verified that the memory tweak tool applied your timings correctly?

Last, what clocks are you using for your rigs?

Kerppa77

newbie

Activity: 17

Merit: 1

I wonder what is going on.... Shocked

I have 4 x 6 Vega64 rigs. All running with the same setup (OS, drivers, versions etc.)

When I applied the TRM 0.4.4 only one of my rigs got significant performance boost.

In this very rig 5 out of 6 Vega64 has ~2400H/s but one still has ~2100H/s and the same one has also regular interleave adjustments.

All the other GPUs in 3 rigs all have ~2100H/s and regular interleave adjustments.

I would like all the rigs to perform like ~2400H/s per GPU, but wonder what could be wrong with the most of the GPUs and rigs?

I tested 14*14 and seems that interleave adjustments are gone but hashrate remains ~2100H/s level per GPU.

SamAlackass

newbie

Activity: 27

Merit: 1

Quote from: ingyenfrag on April 19, 2019, 03:31:53 AM

Is there any option to make the miner restart after detecting a dead gpu? It's weird for me that the minert doesn't do it automatically.

just add --watchdog_script in your start.bat file and create a watchdog.bat with the command "shutdown /r /f" in the same directory.

ingyenfrag

newbie

Activity: 84

Merit: 0

Is there any option to make the miner restart (automatically) after detecting a dead gpu? It's weird for me that the minert doesn't do it automatically.

ku4eto

jr. member

Activity: 194

Merit: 4

Quote from: carlosmonaco on April 19, 2019, 01:49:17 AM

I'm on Windows and far from these result ! Is it possible to mid timing on Windows too ?

The MMPOS Agent is also under testing for WIndows, so you can try that as well.

kerney666

member

Activity: 658

Merit: 86

Quote from: carlosmonaco on April 19, 2019, 01:49:17 AM

I'm on Windows and far from these result ! Is it possible to mid timing on Windows too ?

Absolutely. I tried to write a sort of tuning guide that is included in the release. ElioVP’s tool is available for Windows as well and I’ve reproduced all of the presenter hashrates both on win and Linux.

carlosmonaco

newbie

Activity: 105

Merit: 0

I'm on Windows and far from these result ! Is it possible to mid timing on Windows too ?

hammuh

jr. member

Activity: 41

Merit: 1

Quote from: todxx on April 18, 2019, 06:10:21 PM

Team Red Miner v0.4.4 released

https://github.com/todxx/teamredminer/releases

Changes in v0.4.4

Added * mode specifically for modded timings on Vega GPUs. Use with e.g. --cn_config=15*15. This mode is now the default for Vegas.
Introduced slow start/ramp-up. Threads increase their workload slowly at start or restart after e.g. a network outage.
Added interleave adjustment logic. Readjusts the two threads per gpu over time to make sure they don't gravitate and get stuck.
Added support for forcing colors (--force_color) for windows redirected consoles (git bash, cygwin, node.js).
Added hotkey menu system (show stats, enable/disable gpu).

The new * mode opens up tweaking Vega timings for 2MB CN algos like CN-R!
We've seen a few extremely tuned V64 LCs able to hit over 2400h/s on CN-R with 1400/1100 clocks!
For those of you playing with timings, please share your results! We love seeing how far you guys can push these cards

Quick test...

Vega 64 timing via pbfarmer and others:
--RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --RFC 248 --REF 15600

Settings:
1401 mhz core
1107 mhz memory
870 mv voltage

TRM v0.4.4 results:
cnr (monero): ~2.400 kh/s
cnv8_rwz (graft): ~3.150 kh/s
cnv8_half (masari): ~4.600 kh/s
cnv8_dbl (x-cash): ~1.200 kh/s

rednoW

legendary

Activity: 1510

Merit: 1003

Quote from: todxx on April 18, 2019, 06:10:21 PM

Team Red Miner v0.4.4 released

https://github.com/todxx/teamredminer/releases

Changes in v0.4.4

Added * mode specifically for modded timings on Vega GPUs. Use with e.g. --cn_config=15*15. This mode is now the default for Vegas.
Introduced slow start/ramp-up. Threads increase their workload slowly at start or restart after e.g. a network outage.
Added interleave adjustment logic. Readjusts the two threads per gpu over time to make sure they don't gravitate and get stuck.
Added support for forcing colors (--force_color) for windows redirected consoles (git bash, cygwin, node.js).
Added hotkey menu system (show stats, enable/disable gpu).

The new * mode opens up tweaking Vega timings for 2MB CN algos like CN-R!
We've seen a few extremely tuned V64 LCs able to hit over 2400h/s on CN-R with 1400/1100 clocks!
For those of you playing with timings, please share your results! We love seeing how far you guys can push these cards

Max I got is 2470h/s with vega64lc 1428/1140@880mv and 147watt gpu power - all according to gpu-z.
Not worth it ))
2390h/s at 1320/1100@825mv and 127watt is better I think
Or even better vega56 (both samsung and hynix) 2100h/s at 1220/940@775mv and 97watt in gpu-z

Topic: [ANN] TeamRedMiner v0.10.10 - Ironfish/Kaspa/ZIL/Kawpow/Etchash and More - page 86. (Read 211762 times)