don't feel bad, it took me a looooong time to get 6 cards working in windows 10. There is no method to it, I just kept swapping pci-e positions and risers and it finally worked, for no reason at all.
"set" modifies the current shell's (window) environment values, and the change is available immediately, but it is temporary. The change will not affect other shells that are running, and as soon as you close the shell, the new value is lost until such time as you run set again.
"setx" modifies the value permenantly, which affects all future shells, but does not modify the environment of the shells already running. You have to exit the shell and reopen it before the change will be available, but the value will remain modified until you change it again.
this is what your bat file should look like, SETX or SET is fine afaik, i use both just to be safe.
setx GPU_FORCE_64BIT_PTR 0 setx GPU_MAX_HEAP_SIZE 100 setx GPU_USE_SYNC_OBJECTS 1 setx GPU_MAX_ALLOC_PERCENT 100 setx GPU_SINGLE_ALLOC_PERCENT 100 set GPU_FORCE_64BIT_PTR 0 set GPU_MAX_HEAP_SIZE 100 set GPU_USE_SYNC_OBJECTS 1 set GPU_MAX_ALLOC_PERCENT 100 set GPU_SINGLE_ALLOC_PERCENT 100 EthDcrMiner64.exe -esm 2 -ethi 16 -allcoins 1 -allpools 1 -epool us-east1.ethereum.miningpoolhub.com:17020 -ewal "@[email protected]" -eworker "@[email protected]" -epsw x -dpool stratum+tcp://dcr.coinmine.pl:2222 -dwal "@[email protected]" -dpsw x
except replace with your relevant pools and login/wallet
what size SSD are you running windows off of? clear off 16gb or see if you can set the pagefile on another HDD/SSD/maybe even a USB flashdrive or hard drive
I'm getting a constant 100 Mh/s for 4 cards. 85W per card at 56-60C temps. At this point, not sure it's worth an extra few hashes to flash new bios timings.
The slowest card is Elpida memory..running at 23.4 Mh/s.
Using only Afterburner tweaks...all four cards: Core: 1110 Mem: 2070
Try 1100/1900, strap 1500 copy to 1650 and up. Most important thing is you must increase power consumption limits to 110w also even most of the time GPU works at 85w consumption. By doing this you card may achieve 26mhs
Im sorry dude, but you're just plain wrong, I have been doing bios mod since the launch of the RX series, as i'm doing it for a company I keep counts and I had personally mod more than 367 GPUS (over 52 7 GPU rigs) including rx 480 8 gb (xfx and msi both reference and gamings), rx 480 4gb (msi gaming , xfx and sapphire nitro), rx 470 4gb (msi gaming and armor, asus strix). Modding the bios gives like 20-25% extra performance without major power consumption difference so they aren't optional if you are trying to make a living out of mining. However there is something that you should know, the latest batch of gpu SUUUCCKS like since 2-3 months, I used to easily reach 27-28mhs on the 470 4GB gpus, now I struggle like hell to get 26.5 because mem clocks use to reach 1925 without any need of increasing the power limit or mv. But now most gpus struggle (it doesn't matter the brand) to reach 1850, and many need to be set to 1750. I have seen than the asic quality of those new cards are substantially lower than the first batches that used to be better than 70% of other gpu... now I see myself better than 0%, 3%, 15%, and i dont know why, but one cards has asic quality high enough to be better than 70% other cards and still it just doesn't reach 1925 without going into limp mode
Im sorry dude, but you're just plain wrong, I have been doing bios mod since the launch of the RX series, as i'm doing it for a company I keep counts and I had personally mod more than 367 GPUS (over 52 7 GPU rigs) including rx 480 8 gb (xfx and msi both reference and gamings), rx 480 4gb (msi gaming , xfx and sapphire nitro), rx 470 4gb (msi gaming and armor, asus strix). Modding the bios gives like 20-25% extra performance without major power consumption difference so they aren't optional if you are trying to make a living out of mining. However there is something that you should know, the latest batch of gpu SUUUCCKS like since 2-3 months, I used to easily reach 27-28mhs on the 470 4GB gpus, now I struggle like hell to get 26.5 because mem clocks use to reach 1925 without any need of increasing the power limit or mv. But now most gpus struggle (it doesn't matter the brand) to reach 1850, and many need to be set to 1750. I have seen than the asic quality of those new cards are substantially lower than the first batches that used to be better than 70% of other gpu... now I see myself better than 0%, 3%, 15%, and i dont know why, but one cards has asic quality high enough to be better than 70% other cards and still it just doesn't reach 1925 without going into limp mode
Impressive, and I believe you...just trying to learn how best to mod is all...I'm happy to give it a go again at some point. Licking my wounds at the moment.
Currently, I'm getting 152Mh/s (6 card rig)...with just tweaking settings. No bios flash and using Crimson 16.9.2.
Settings: Core: 1100 Mem: 2060 Watts: 85 average Temps: 60C average
try lowering your settings on that card, GPU 2, seems like it is overclocked beyond stability? I used to get that on my 290s and 290x when i was pushing them too hard or undervolting them too much
also @thepixelgeek you are using windows 10, correct? how long did claymore run stable before GPU 2 started hanging.
@thepixelgeek you have to go into the settings and unlock voltage control in afterburner, i usually just use the "Reference design" option. I disable ULPS (ultra low power states) as well, though I am not sure if this is necessary.
Alternatively, you can use sapphire's TRIXX program. it is their equivalent of MSI afterburner, though not as full featured. With TRIXX i am able to undervolt right away.
I undervolt for heat purposes mostly. My power is cheap so it is not a big deal for me, but for others it is more necessary.
It is just for saving money over time by consuming less power, sometimes significantly less. Also reducing heat. Slightly. Which can prolong the life of the cards and the other surrounding hardware.
It is just for saving money over time by consuming less power, sometimes significantly less. Also reducing heat. Slightly. Which can prolong the life of the cards and the other surrounding hardware.
I see...and what settings are advisable. -10, -20...+20?
it is different for most cards. On my 280x, 290s, 290x i tried to undervolt as much as possible because those are HOT, High-consumption electricity addicts (especially the reference 290x). I think I got them all anywhyere from -72 to -96mv without losing performance. My 470 I have at -50mv in Trixx with core at 1200, mem at 1900 with the biosmod. But the power difference on the 470 isnt much.
Try 1100/1900, strap 1500 copy to 1650 and up. Most important thing is you must increase power consumption limits to 110w also even most of the time GPU works at 85w consumption. By doing this you card may achieve 26mhs
@tinroof By "increase power consumption limits" do you mean "power limit %" in Afterburner? If so, what is that setting? +5, +10...+100?
lol memory hardware errors basically dont count for shit I have 3 RX 480s producing millions of hardware errors each and they are mining at 30+ mh/s and the rig was on for 645 hours before I manually took it down to add cards to it It was a rig of a 390x, 2 RX 480s, and 3 nanos
Are you sure that's true though?
I had a moment when i got the rig running stable at 166 mhs but after running it for a few hours I was actually finding less shares than running at stock settings at 130 mhs. Every card had tons of memory errors and I was finding only about 140-160 shares an hour vs the 210-240 at stock.
210-240 seems way too high for 1h shares for even 166mhs.. will you check this again? I have been ignoring mem errors as it seems the consensus is that they don't matter for shares
I think you will be stable at MUCH less than -30mV. Flashing is easy and highly recommended
What's a good -mV target? Why go lower?
As for flashing, tried it once and bricked the card. Looking for solutions to bring back from dead.
I am planning on flashing again...want to do more research first. Please share you steps to flash...and by steps, I mean did you flash one card via 16x by itself and then repeat the process for everycard? Did you flash one at a time, and then update Crimson driver after all cards were flashed and connected to the mobo? I'm still learning the process.
Note, I have Polaris editor, MSI, GPUz. I've read already about copy/paste 1500 straps to 1650 and above for mod bios.
I think you will be stable at MUCH less than -30mV. Flashing is easy and highly recommended
What's a good -mV target? Why go lower?
As for flashing, tried it once and bricked the card. Looking for solutions to bring back from dead.
I am planning on flashing again...want to do more research first. Please share you steps to flash...and by steps, I mean did you flash one card via 16x by itself and then repeat the process for everycard? Did you flash one at a time, and then update Crimson driver after all cards were flashed and connected to the mobo? I'm still learning the process.
Note, I have Polaris editor, MSI, GPUz. I've read already about copy/paste 1500 straps to 1650 and above for mod bios.
I use WattTool and set voltages manually, nothing higher than 920mv which I think is more than -100mv.. maybe even like -130mv.. and sometimes I can get cards stable at 900mv or less (even lower for 470s generally).
You save money on electric cost and there is less heat output going lower voltage.
The main thing I change in flashing is copying the 1500 memory strap timing to the higher ones if hynix mem (or if with samsung mem copy the 1750 to higher). I usually flash them all together (well 1 at a time but I don't reboot in between flashes or do anything with drivers). I reboot when im done flashing.
You don't have to mess with drivers at all, just get the driver you want (like 16.9.2) installed before you flash, or after, doesn't matter. I'd have to know exactly what you did to have any idea how you bricked your card
Note that beginning with Crimson ReLive 16.12.2 WHQL driver (Windows Hardware Quality Labs) BIOS signature testing is started up again so the installation of a patched BIOS will result in a driver error during startup because the validated signature is not present in the patched BIOS. This prevents OpenCL from using the GPU since there will not be a supported driver to support it. I discovered this the hard way when I upgraded to 16.12.2 and suddenly none of my GPU's functioned and a default driver was used instead. This can be "fixed" in one of two ways. 1. Install and use an older driver that does not perform the signature validation so that a patched BIOS can be used with no startup issues. 2. Run a workaround that bypasses the BIOS signature check that is implemented in 16.12.2 and newer drivers. This is the method I deployed downloading a running a patch that runs against the installed driver so that the next reboot will bypass the BIOS signature check. It works perfectly fine and you just have to wait until the system marks all of the GPU's in error status before running the patch. This can take several minutes to complete during startup with multiple GPU's installed. Also, if a new GPU is added all of the old GPU's will be marked in error and the patch will need to be rerun. This program is called the Pixel Clock Patcher written and maintained by ToastyX which can be found at https://monitortests.com/forum/Thread-AMD-ATI-Pixel-Clock-Patcher
@Drawde Thanks... My current driver is Crimson 16.6, according to GPUz. I had 16.9.2 installed but Windows did an update and must have changed the driver. Odd that it went backwards, judging by the number.
Assuming I'm ok to begin flashing the drives, having 16.6, without the need for the workaround?
Looking in Atiflash, it show three cards, however I'm not sure what mem type they are in order to flash the samsung mod bios with the samsung mem card. Na mean?
Comments
I wonder if it has to do with virtual memory? I have Win installed on SSD that only has 2.9GB space free. Virtual memory is currently at 1920MB...
Only have 4926MB space available for Virtual memory.
"set" modifies the current shell's (window) environment values, and the change is available immediately, but it is temporary. The change will not affect other shells that are running, and as soon as you close the shell, the new value is lost until such time as you run set again.
"setx" modifies the value permenantly, which affects all future shells, but does not modify the environment of the shells already running. You have to exit the shell and reopen it before the change will be available, but the value will remain modified until you change it again.
this is what your bat file should look like, SETX or SET is fine afaik, i use both just to be safe.
setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
set GPU_FORCE_64BIT_PTR 0
set GPU_MAX_HEAP_SIZE 100
set GPU_USE_SYNC_OBJECTS 1
set GPU_MAX_ALLOC_PERCENT 100
set GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -esm 2 -ethi 16 -allcoins 1 -allpools 1 -epool us-east1.ethereum.miningpoolhub.com:17020 -ewal "@[email protected]" -eworker "@[email protected]" -epsw x -dpool stratum+tcp://dcr.coinmine.pl:2222 -dwal "@[email protected]" -dpsw x
except replace with your relevant pools and login/wallet
Thank you !! @cryptorex, @shutfu
@cryptorex possible to post those steps again? Or message it to me?
Currently, I'm getting 152Mh/s (6 card rig)...with just tweaking settings. No bios flash and using Crimson 16.9.2.
Settings:
Core: 1100
Mem: 2060
Watts: 85 average
Temps: 60C average
Restarts claymore but nothing happens.
I used to get that on my 290s and 290x when i was pushing them too hard or undervolting them too much
also @thepixelgeek you are using windows 10, correct? how long did claymore run stable before GPU 2 started hanging.
Yes, windows 10 64bit. @shutfu
Settings are crashing, and not stable. Lowering it now.
Alternatively, you can use sapphire's TRIXX program. it is their equivalent of MSI afterburner, though not as full featured. With TRIXX i am able to undervolt right away.
I undervolt for heat purposes mostly. My power is cheap so it is not a big deal for me, but for others it is more necessary.
No flashing yet...BUT all 6 cards stable at 148 MH/s. Would like to see if I can get to 160 Mh/s. Assuming I'll need to flash to get 160 Mh/s.
Current settings:
-30 mV
1140 Core
2000 Mem
78-90 Watts per card...total watts at the wall 870.
60 temps on all cards
I have it set at +0 at the moment.
As for flashing, tried it once and bricked the card. Looking for solutions to bring back from dead.
I am planning on flashing again...want to do more research first. Please share you steps to flash...and by steps, I mean did you flash one card via 16x by itself and then repeat the process for everycard? Did you flash one at a time, and then update Crimson driver after all cards were flashed and connected to the mobo? I'm still learning the process.
Note, I have Polaris editor, MSI, GPUz. I've read already about copy/paste 1500 straps to 1650 and above for mod bios.
You save money on electric cost and there is less heat output going lower voltage.
The main thing I change in flashing is copying the 1500 memory strap timing to the higher ones if hynix mem (or if with samsung mem copy the 1750 to higher). I usually flash them all together (well 1 at a time but I don't reboot in between flashes or do anything with drivers). I reboot when im done flashing.
You don't have to mess with drivers at all, just get the driver you want (like 16.9.2) installed before you flash, or after, doesn't matter. I'd have to know exactly what you did to have any idea how you bricked your card
1. Install and use an older driver that does not perform the signature validation so that a patched BIOS can be used with no startup issues.
2. Run a workaround that bypasses the BIOS signature check that is implemented in 16.12.2 and newer drivers. This is the method I deployed downloading a running a patch that runs against the installed driver so that the next reboot will bypass the BIOS signature check. It works perfectly fine and you just have to wait until the system marks all of the GPU's in error status before running the patch. This can take several minutes to complete during startup with multiple GPU's installed. Also, if a new GPU is added all of the old GPU's will be marked in error and the patch will need to be rerun. This program is called the Pixel Clock Patcher written and maintained by ToastyX which can be found at https://monitortests.com/forum/Thread-AMD-ATI-Pixel-Clock-Patcher
My current driver is Crimson 16.6, according to GPUz. I had 16.9.2 installed but Windows did an update and must have changed the driver. Odd that it went backwards, judging by the number.
Assuming I'm ok to begin flashing the drives, having 16.6, without the need for the workaround?
Oddly, the ATI program had garbled text which made the buttons unreadable.
-48mV
1200 core
2020 mem
60 temps / 95W per...although I have two cards at 86W.
Not sure if it's stable yet. How long before you know?