Topic: AMD Mem Tweak XL - Read/modify timings/pp/straps on the fly - page 47. (Read 58906 times)
so what settings would be great on rx570s with samsungs?
Hey all.
I assume messing with polaris is not relevant here, eh but here's a question:
I cant apply changes to CAS and MISC2 timings (causes "random colored screen" driver crash) as I assume changing those also needs changes to MC_SEQ_MISC straps (IMC straps)
Hovewer Wattman can successfully apply it's "level 2" timings and it's changing CAS and MISC2 aswell as reported by --current command.
Maybe I'm missing something?
I assume messing with polaris is not relevant here, eh but here's a question:
I cant apply changes to CAS and MISC2 timings (causes "random colored screen" driver crash) as I assume changing those also needs changes to MC_SEQ_MISC straps (IMC straps)
Hovewer Wattman can successfully apply it's "level 2" timings and it's changing CAS and MISC2 aswell as reported by --current command.
Maybe I'm missing something?
Hello,
I have some bad card ( vega 56 reference with blower ) running at 1407 / 900 at 930mV and still not really stable ( no mod timing yet)
Can it be better with 64 bios and mod timing ?
Running on 18.6.1 driver.
I have some bad card ( vega 56 reference with blower ) running at 1407 / 900 at 930mV and still not really stable ( no mod timing yet)
Can it be better with 64 bios and mod timing ?
Running on 18.6.1 driver.
A QUICK TWEAK--
After running TRM for more than 5 hours with the AMDmemtweak values described in the quotes below, I made an on-the-fly clock change to my Vega/ethOS box. I issued the following command at the command line:
"$ sudo ethos-smi -c 1407 -m 1100"
This set my core clock to 1407 and my memory clock to 1100. The system had been running at 1507/945 for the initial run. A "stats" command showed a drop of about 20 watts per card to ~250W each. The suprising result was that the hash rate increased about 300H/s per card. See below:
Formerly, the cards had been running at about 2.08kh/s. The "avg" and "pool" values are low because they were calculated based on the initial ~5 hours of the run. Overall, not bad for an initial run!
A big Thank You! to Eliovp, pbfarmer, and the TRM coders Todd and Kerney, and to hammuh. whose clock settings I used.
To maybe make things simpler - you definitely don't need to set timings under load w/ vegas on linux. Locking your mem to p3 or whatever your top state will be, w/ something like
will be enough (replace [N] w/ your GPU number). Honestly, i'm not even sure if this is necessary, but i was already doing it for other reasons.
After running TRM for more than 5 hours with the AMDmemtweak values described in the quotes below, I made an on-the-fly clock change to my Vega/ethOS box. I issued the following command at the command line:
"$ sudo ethos-smi -c 1407 -m 1100"
This set my core clock to 1407 and my memory clock to 1100. The system had been running at 1507/945 for the initial run. A "stats" command showed a drop of about 20 watts per card to ~250W each. The suprising result was that the hash rate increased about 300H/s per card. See below:
Formerly, the cards had been running at about 2.08kh/s. The "avg" and "pool" values are low because they were calculated based on the initial ~5 hours of the run. Overall, not bad for an initial run!
A big Thank You! to Eliovp, pbfarmer, and the TRM coders Todd and Kerney, and to hammuh. whose clock settings I used.
AMDMEMTWEAK AND TEAMREDMINER V4.4 SUCCESS--
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
To maybe make things simpler - you definitely don't need to set timings under load w/ vegas on linux. Locking your mem to p3 or whatever your top state will be, w/ something like
Code:
$ echo "3" |sudo tee /sys/class/drm/card[N]/device/pp_dpm_mclk
will be enough (replace [N] w/ your GPU number). Honestly, i'm not even sure if this is necessary, but i was already doing it for other reasons.
AMDMEMTWEAK AND TEAMREDMINER V4.4 SUCCESS--
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
To maybe make things simpler - you definitely don't need to set timings under load w/ vegas on linux. Locking your mem to p3 or whatever your top state will be, w/ something like
Code:
$ echo "3" |sudo tee /sys/class/drm/card[N]/device/pp_dpm_mclk
will be enough (replace [N] w/ your GPU number). Honestly, i'm not even sure if this is necessary, but i was already doing it for other reasons.
AMDMEMTWEAK AND TEAMREDMINER V4.4 SUCCESS--
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
nice test, can you test cnr algo, is it necessary flash 56 to 64 for this algo?
Ok - now that the latest and greatest TRM is public... here are my results for vega 64s (power is ATW)
Base GPU settings: 1375 (effective) cclock / 1107 mclock / 837mv (range across 8x64s is 825-843mv)
TRM 0.4.3, 15+15, no timing mods: 2120 h/s @ 165w or 12.8 h/w
TRM 0.4.3, 15+15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2140 h/s (did not measure power)
TRM 0.4.4, 15*15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2440 h/s @ 178w or 13.7 hw
TRM 0.4.4, 15*15, --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600: 2420 h/s (did not measure power - assume same as previous)
As you can see:
1. Timings mods aren't doing much for CNR, w/o the latest TRM '*' mode
2. Many of the timing mods are just pushing at the edges... You can get ~95% of the gain w/ just 6 (maybe less) parameters mods.
After reading this thread at least once a day since it opened, I finally set up AMDmemtweak on my Vega/ethOS box. It was not a difficlult set up. I simply followed these steps:
1) Clone AMDmemtweak from git (https://github.com/eliovp). Just follow the instructions on the README.md page.
2) Compile the amdmemtweak.cpp file with g++. Watch out for typos. The instructions are in the README.md.
3) Copy the executable binary to "/opt/ethos/bin". This directory is in the system path and contains tools such as "OhGodADecode" and "ATIflash".
You should now be able to run AMDmemtweak as a system utility. After installing AMDmemtweak, I upgraded TeamRedMiner to version 4.4, setup a configuration using the new "15*15" mode, and launched TRM.
TRM boots and informs the miner when it is at full load. When the miner is at full load, it is ready to receive a command from AMDmemtweak. I ran the following command at the command prompt, utilizing pbfarmer's short string given in his attatched post below:
"$ sudo amdmemtweak --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600"
The result was an immediate ~0.2kh/s increase in each of my Vega64 cards' CNr hash rate. At about 2.08kh/s each, they are finally over the 2.0kh/s goal. I will worry about additional tweaks later, and post a detailed tutorial in my Notebook. --scryptr
First - quick results for cn-trtl, w/ efficiency-focused settings, using timings mostly borrowing from others here, w/ some minor tweaks:
Vega 64 air, ubuntu 18.04 + amdgpu-pro 18.50, TRM 0.4.3 (L18+18), 852 cclock (p0)/1107 mclock/818mv, power readings at the wall
stock timings:
--CL 20 --RAS 33 --RCDRD 16 --RCDWR 10 --RC 47 --RP 14 --RRDS 4 --RRDL 6 --RFC 260 (--REF 3900)
18.5 kh/s @ 135w (137 h/w)
modded timings 1
--CL 19 --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --RFC 248
19.75 kh/s @ 137w (144 h/w)
modded timings 2
same as above, plus --REF 15600
20.71 kh/s @ 137w (151 h/w)
Second - notes on power... I don't see any appreciable power differences - nor would I expect to. Clocks and voltages are untouched, we simply have a bit more data being transferred. Even the 2w difference I'm reporting here is conservative - taking natural fluctuations in my readings into account, my actual increase could be closer to <= 1w. People seeing large power increases (at least on vega 64) seem to have something else going on.
Last - some conjecture / educated guessing re: THAT --REF THO!!! I'm assuming --REF is the refresh frequency, in nanoseconds, and unlike most timings, a higher value (meaning less refreshing) is better. Refreshes steal bandwidth, and AMD seems to have gone majorly conservative (aggressive?) on this, probably due to the super high temps of the HBM during normal/gaming use. As leakage increases w/ temps, more refreshes would be required when running your GPU/HBM at high clocks/voltages. Since (efficient) miners tend to run cooler, the crazy high default refresh rate is really unnecessary. I found 4x the default to be around where returns quickly diminish, at least at my clocks - i can get maybe another 50 h/s (turtle) going 4.5x. HOWEVER - if you run super aggressive for max h/r, or just aren't effectively cooled in general, you may want to dial this back, or you may start seeing mem errors / bad shares from corrupted data due to insufficient refreshing / leakage.
Vega 64 air, ubuntu 18.04 + amdgpu-pro 18.50, TRM 0.4.3 (L18+18), 852 cclock (p0)/1107 mclock/818mv, power readings at the wall
stock timings:
--CL 20 --RAS 33 --RCDRD 16 --RCDWR 10 --RC 47 --RP 14 --RRDS 4 --RRDL 6 --RFC 260 (--REF 3900)
18.5 kh/s @ 135w (137 h/w)
modded timings 1
--CL 19 --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --RFC 248
19.75 kh/s @ 137w (144 h/w)
modded timings 2
same as above, plus --REF 15600
20.71 kh/s @ 137w (151 h/w)
Last - some conjecture / educated guessing re: THAT --REF THO!!! I'm assuming --REF is the refresh frequency, in nanoseconds, and unlike most timings, a higher value (meaning less refreshing) is better. Refreshes steal bandwidth, and AMD seems to have gone majorly conservative (aggressive?) on this, probably due to the super high temps of the HBM during normal/gaming use. As leakage increases w/ temps, more refreshes would be required when running your GPU/HBM at high clocks/voltages. Since (efficient) miners tend to run cooler, the crazy high default refresh rate is really unnecessary. I found 4x the default to be around where returns quickly diminish, at least at my clocks - i can get maybe another 50 h/s (turtle) going 4.5x. HOWEVER - if you run super aggressive for max h/r, or just aren't effectively cooled in general, you may want to dial this back, or you may start seeing mem errors / bad shares from corrupted data due to insufficient refreshing / leakage.
nice test, can you test cnr algo, is it necessary flash 56 to 64 for this algo?
Ok - now that the latest and greatest TRM is public... here are my results for vega 64s (power is ATW)
Base GPU settings: 1375 (effective) cclock / 1107 mclock / 837mv (range across 8x64s is 825-843mv)
TRM 0.4.3, 15+15, no timing mods: 2120 h/s @ 165w or 12.8 h/w
TRM 0.4.3, 15+15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2140 h/s (did not measure power)
TRM 0.4.4, 15*15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2440 h/s @ 178w or 13.7 hw
TRM 0.4.4, 15*15, --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600: 2420 h/s (did not measure power - assume same as previous)
As you can see:
1. Timings mods aren't doing much for CNR, w/o the latest TRM '*' mode
2. Many of the timing mods are just pushing at the edges... You can get ~95% of the gain w/ just 6 (maybe less) parameters mods.
Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
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
my current samsung timings for stock vega56 are not so aggressive
--rc 38 --rcdrd 12 --rcdwr 5 --rp 11 --faw 13 --RRDL 3 --rtp 4 --ref 15600
I like'em
--rc 38 --rcdrd 12 --rcdwr 5 --rp 11 --faw 13 --RRDL 3 --rtp 4 --ref 15600
I like'em
For Samsung 56 users (non-flashed). Should get pretty close to flashed, pretty aggressive though, and will cause errors on weaker gpus.
--CL 17 --RAS 26 --rp 10 --rc 36 --rfc 240 --rrds 3 --rrdl 3 --rcdrd 12 --rcdwr 5 --ref 14000 --faw 13 --rtp 4
cn_config 14*14 with 19.3.3 drivers.
--CL 17 --RAS 26 --rp 10 --rc 36 --rfc 240 --rrds 3 --rrdl 3 --rcdrd 12 --rcdwr 5 --ref 14000 --faw 13 --rtp 4
cn_config 14*14 with 19.3.3 drivers.
Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
Nice, from 1.75 to 2.03 kh/s increase for cn/r from 0.4.3 to 0.4.4. vega56 hynix 1200/940@775mv (gpu-z shown clocks and voltage) Same timings For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
How about power? How many watts?
config 16*16 doesn't work for my vegas64 and 19.4.1 drivers. 14*14 config is ok and I'm getting 2.33kh/s (1300/1090@825mv 120 watt gpu-z )
Should have been more clear reg the *-mode: turtle is fine with the old + mode. The new mode is for all 2MB pad variants.
For power: we’ve seen increased efficiency across the board. A rule of thumb has been +2x% hashrate increase for +x% power, one example was a Vega 64 with +15% hashrate and +7.8% power draw.
vega56 hynix is feeling better with 14*14 (2.03 kh/s)
Yeah, 16*16 will never work, it would use the full 8GB of vram for pads. It’s a mess that the drivers allow the allocation. Use 15*15 instead. Have we written 16*16 somewhere, or even default to it?
The behavior of the 19.4.1 drivers is odd. Unless you run two massive monitors on that gpu, there’s no reason for 12*12 to be better, rather the opposite.
Now tested some ODD numbers.
And it helps! Best config for my vega56 is 13+13, samsung is doing 2.12kh/s, hynix 2.09kh/s
For vega64 15+15 is not the best. 14+14 is better (2.39kh/s with 1100mem clock).
Seems that 19.4.1 drivers don't like max mem load
anyone have vega 56 with samsung memory timings yet? If I use Hynix for it kills the rig. I did not flash it to a 64....
--rp 12 --rc 37 --rrds 3 --rrdl 3 --rcdrd 12 --rcdwr 5 --ras 25
This is what I use. Works on 3 out of 4 of mine.
First - quick results for cn-trtl, w/ efficiency-focused settings, using timings mostly borrowing from others here, w/ some minor tweaks:
Vega 64 air, ubuntu 18.04 + amdgpu-pro 18.50, TRM 0.4.3 (L18+18), 852 cclock (p0)/1107 mclock/818mv, power readings at the wall
stock timings:
--CL 20 --RAS 33 --RCDRD 16 --RCDWR 10 --RC 47 --RP 14 --RRDS 4 --RRDL 6 --RFC 260 (--REF 3900)
18.5 kh/s @ 135w (137 h/w)
modded timings 1
--CL 19 --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --RFC 248
19.75 kh/s @ 137w (144 h/w)
modded timings 2
same as above, plus --REF 15600
20.71 kh/s @ 137w (151 h/w)
Second - notes on power... I don't see any appreciable power differences - nor would I expect to. Clocks and voltages are untouched, we simply have a bit more data being transferred. Even the 2w difference I'm reporting here is conservative - taking natural fluctuations in my readings into account, my actual increase could be closer to <= 1w. People seeing large power increases (at least on vega 64) seem to have something else going on.
Last - some conjecture / educated guessing re: THAT --REF THO!!! I'm assuming --REF is the refresh frequency, in nanoseconds, and unlike most timings, a higher value (meaning less refreshing) is better. Refreshes steal bandwidth, and AMD seems to have gone majorly conservative (aggressive?) on this, probably due to the super high temps of the HBM during normal/gaming use. As leakage increases w/ temps, more refreshes would be required when running your GPU/HBM at high clocks/voltages. Since (efficient) miners tend to run cooler, the crazy high default refresh rate is really unnecessary. I found 4x the default to be around where returns quickly diminish, at least at my clocks - i can get maybe another 50 h/s (turtle) going 4.5x. HOWEVER - if you run super aggressive for max h/r, or just aren't effectively cooled in general, you may want to dial this back, or you may start seeing mem errors / bad shares from corrupted data due to insufficient refreshing / leakage.
Vega 64 air, ubuntu 18.04 + amdgpu-pro 18.50, TRM 0.4.3 (L18+18), 852 cclock (p0)/1107 mclock/818mv, power readings at the wall
stock timings:
--CL 20 --RAS 33 --RCDRD 16 --RCDWR 10 --RC 47 --RP 14 --RRDS 4 --RRDL 6 --RFC 260 (--REF 3900)
18.5 kh/s @ 135w (137 h/w)
modded timings 1
--CL 19 --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --RFC 248
19.75 kh/s @ 137w (144 h/w)
modded timings 2
same as above, plus --REF 15600
20.71 kh/s @ 137w (151 h/w)
Last - some conjecture / educated guessing re: THAT --REF THO!!! I'm assuming --REF is the refresh frequency, in nanoseconds, and unlike most timings, a higher value (meaning less refreshing) is better. Refreshes steal bandwidth, and AMD seems to have gone majorly conservative (aggressive?) on this, probably due to the super high temps of the HBM during normal/gaming use. As leakage increases w/ temps, more refreshes would be required when running your GPU/HBM at high clocks/voltages. Since (efficient) miners tend to run cooler, the crazy high default refresh rate is really unnecessary. I found 4x the default to be around where returns quickly diminish, at least at my clocks - i can get maybe another 50 h/s (turtle) going 4.5x. HOWEVER - if you run super aggressive for max h/r, or just aren't effectively cooled in general, you may want to dial this back, or you may start seeing mem errors / bad shares from corrupted data due to insufficient refreshing / leakage.
nice test, can you test cnr algo, is it necessary flash 56 to 64 for this algo?
Ok - now that the latest and greatest TRM is public... here are my results for vega 64s (power is ATW)
Base GPU settings: 1375 (effective) cclock / 1107 mclock / 837mv (range across 8x64s is 825-843mv)
TRM 0.4.3, 15+15, no timing mods: 2120 h/s @ 165w or 12.8 h/w
TRM 0.4.3, 15+15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2140 h/s (did not measure power)
TRM 0.4.4, 15*15, --RAS 28 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --RRDS 3 --RRDL 3 --REF 15600 --RFC 248 --FAW 14: 2440 h/s @ 178w or 13.7 hw
TRM 0.4.4, 15*15, --RAS 32 --RCDRD 12 --RCDWR 5 --RC 44 --RP 12 --REF 15600: 2420 h/s (did not measure power - assume same as previous)
As you can see:
1. Timings mods aren't doing much for CNR, w/o the latest TRM '*' mode
2. Many of the timing mods are just pushing at the edges... You can get ~95% of the gain w/ just 6 (maybe less) parameters mods.
anyone have vega 56 with samsung memory timings yet? If I use Hynix for it kills the rig. I did not flash it to a 64....
Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
Nice, from 1.75 to 2.03 kh/s increase for cn/r from 0.4.3 to 0.4.4. vega56 hynix 1200/940@775mv (gpu-z shown clocks and voltage) Same timings For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
How about power? How many watts?
config 16*16 doesn't work for my vegas64 and 19.4.1 drivers. 14*14 config is ok and I'm getting 2.33kh/s (1300/1090@825mv 120 watt gpu-z )
Should have been more clear reg the *-mode: turtle is fine with the old + mode. The new mode is for all 2MB pad variants.
For power: we’ve seen increased efficiency across the board. A rule of thumb has been +2x% hashrate increase for +x% power, one example was a Vega 64 with +15% hashrate and +7.8% power draw.
vega56 hynix is feeling better with 14*14 (2.03 kh/s)
Yeah, 16*16 will never work, it would use the full 8GB of vram for pads. It’s a mess that the drivers allow the allocation. Use 15*15 instead. Have we written 16*16 somewhere, or even default to it?
The behavior of the 19.4.1 drivers is odd. Unless you run two massive monitors on that gpu, there’s no reason for 12*12 to be better, rather the opposite.
My bad, you right, it is the same as others, looked on wrong place, that was score for both cards, and 570 is overheating 
one card is rx 570 msi
other is vega 56 from powercolor (samsung memory) only difference to other vega card is that this vega has artic cooling 3 cooler, because crapy blower vapor chamber exploded year ago haha
Well, that would be a world record with that 2700 h/s gpu. I’m guessing one of the two threads is broken and you need to dial down clocks and/or timings. But, poolside hash wins, what does the miner say for “pool” hash?
one card is rx 570 msi
other is vega 56 from powercolor (samsung memory) only difference to other vega card is that this vega has artic cooling 3 cooler, because crapy blower vapor chamber exploded year ago haha
this is comedy, 56 flashed to 64,
11/12 runs 2250
and one vega runs guesh what 2660 (insane stabilized at 2700h/s wtf, but this one card only) everithing is the same
Samsung memory gpu 1408/ mem 1100 on 900mv
11/12 runs 2250
and one vega runs guesh what 2660 (insane stabilized at 2700h/s wtf, but this one card only) everithing is the same
Samsung memory gpu 1408/ mem 1100 on 900mv
Well, that would be a world record with that 2700 h/s gpu
. I’m guessing one of the two threads is broken and you need to dial down clocks and/or timings. But, poolside hash wins, what does the miner say for “pool” hash?But i have a general question for trying to run the program, whenever i try to start, i just get a quick pop up window and then it exits out. I tried already setting it to windows 7 compatibility mode and still get nothing. Can anyone help me get this thing going? My pc is windows 10 and using the latest firmware for amd gpu.
put this in a .bat file:
Quote
c:\\WinAMDTweak.exe --current
pause
pause
and run it as admin
Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
Nice, from 1.75 to 2.03 kh/s increase for cn/r from 0.4.3 to 0.4.4. vega56 hynix 1200/940@775mv (gpu-z shown clocks and voltage) Same timings For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
How about power? How many watts?
config 16*16 doesn't work for my vegas64 and 19.4.1 drivers. 14*14 config is ok and I'm getting 2.33kh/s (1300/1090@825mv 120 watt gpu-z )
Should have been more clear reg the *-mode: turtle is fine with the old + mode. The new mode is for all 2MB pad variants.
For power: we’ve seen increased efficiency across the board. A rule of thumb has been +2x% hashrate increase for +x% power, one example was a Vega 64 with +15% hashrate and +7.8% power draw.
vega56 hynix is feeling better with 14*14 (2.03 kh/s)
Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
Glad to see TRM support on this!
But i have a general question for trying to run the program, whenever i try to start, i just get a quick pop up window and then it exits out. I tried already setting it to windows 7 compatibility mode and still get nothing. Can anyone help me get this thing going? My pc is windows 10 and using the latest firmware for amd gpu.
this is comedy, 56 flashed to 64,
11/12 runs 2250
and one vega runs guesh what 2660 (insane stabilized at 2700h/s wtf, but this one card only) everithing is the same
Samsung memory gpu 1408/ mem 1100 on 900mv
11/12 runs 2250
and one vega runs guesh what 2660 (insane stabilized at 2700h/s wtf, but this one card only) everithing is the same
Samsung memory gpu 1408/ mem 1100 on 900mv
Well, that would be a world record with that 2700 h/s gpu
. I’m guessing one of the two threads is broken and you need to dial down clocks and/or timings. But, poolside hash wins, what does the miner say for “pool” hash? Hi!
For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
Nice, from 1.75 to 2.03 kh/s increase for cn/r from 0.4.3 to 0.4.4. vega56 hynix 1200/940@775mv (gpu-z shown clocks and voltage) Same timings For everyone testing modded mem timings for CN using TeamRedMiner:
TRM v0.4.4 is released on GitHub (https://github.com/todxx/teamredminer/releases).
The release contains a new *-mode specifically targeted towards Vega GPUs with modded mem timings and running normal CN variants (not Turtle). It’s the new default mode for Vegas.
The release also contains a new tuning guide with recipes for Vegas. Using it we can produce seriously sexy hashrates on all Vegas without increasing core clock above 1407 MHz (CN/r: 2440 h/s V64, 2250 h/s V56@64, 2130 h/s 56 Hynix).
The release will be in alpha state for a few hours before we post in our Bitcointalk topic, if anyone has the chance to take it for a spin, please do so and let us know your CN/r results.
Cheers, K
How about power? How many watts?
config 16*16 doesn't work for my vegas64 and 19.4.1 drivers. 14*14 config is ok and I'm getting 2.33kh/s (1300/1090@825mv 120 watt gpu-z )
Should have been more clear reg the *-mode: turtle is fine with the old + mode. The new mode is for all 2MB pad variants.
For power: we’ve seen increased efficiency across the board. A rule of thumb has been +2x% hashrate increase for +x% power, one example was a Vega 64 with +15% hashrate and +7.8% power draw.
Jump to: