Question GIGABYTE 5700 XT Bios mod fails

AlleyCat

Veteran
Mitglied seit
Okt 9, 2020
Beiträge
290
Bewertungspunkte
17
Punkte
17
Hi,
I am failing to update Gigabyte 5700 XT. The same procedure updates the BIOS on MSI cards with no problem. I follow the instructions on Igors Lab.

The sign of having trouble with the card bios flash shows in GPU-Z. After flash using amdvbflash the values of GPU and memory frequencies is empty. With stock bios there are Mhz frequencies.

Is there any known problems with flashing Gigabyte cards?
Any suggestions on what other forums I may ask for assistance?
Is it possible that the OEM bios is signed, and any modified bios will be rejected?
If the bios is signed, any tools to resign, or would I need to buy a card from a different vendor?

Thanks,

Alley Cat
 
Zuletzt bearbeitet :
I was able to replace the tREF value to 12000. I searched the text string for 810B, and replaced it with E02E.

Windows will not start with tREF value set to 12000 (E02E).
I had to remove the card, remove the AMD drivers, re-flash with a working bios.

What is the next step?
Alright, tREF to 9998 if not then 5998.
 
Zuletzt bearbeitet :
I was able to replace the tREF value to 12000. I searched the text string for 810B, and replaced it with E02E.

Windows will not start with tREF value set to 12000 (E02E).
I had to remove the card, remove the AMD drivers, re-flash with a working bios.

What is the next step?
The Gigabyte RX 5700 XT are not easy to modify because of their bios, patience is needed to find the best stable vram timings for it.
 
Zuletzt bearbeitet :
The Gigabyte RX 5700 XT are not easy to modify, patience is needed to find the best stable vram timings for it.
The problem was with the tCL settings, and not the tREF.
I had to revert to tCL = 20.

I an successful setting the tREF = 12000 (E02E) and the works.

I am getting 60.08 MH / s with the following configuration:
Platform: Windows 10
Miner: Phoenix 5.1c
Power: 129W

I achieved my hash speed target.
I must reduce the power to below 110W.

What are the next steps?
 

Anhänge

  • IMG_3435.JPG
    IMG_3435.JPG
    2,1 MB · Aufrufe : 354
The problem was with the tCL settings, and not the tREF.
I had to revert to tCL = 20.

I a successful setting the tREF = 12000 (E02E) and the works.

I am getting 60.08 MH / s with the following configuration:
Platform: Windows 10
Miner: Phoenix 5.1c
Power: 129W

I achieved my hash speed target.
I must reduce the power to below 110W.

What are the next steps?
Great results, now try to reduce the core volt of 850 mv by 25 mv and test until you reach a sweet spot of performance to power.
 
The problem was with the tCL settings, and not the tREF.
I had to revert to tCL = 20.

I a successful setting the tREF = 12000 (E02E) and the works.

I am getting 60.08 MH / s with the following configuration:
Platform: Windows 10
Miner: Phoenix 5.1c
Power: 129W

I achieved my hash speed target.
I must reduce the power to below 110W.

What are the next steps?
By the way, what about the temperature values.
 
I am starting the optimization on HiveOS.

Memory Temp: 88

The power doesn't go bellow 117W.

What are the next changes?
 

Anhänge

  • IMG_3436.JPG
    IMG_3436.JPG
    2,6 MB · Aufrufe : 328
  • Screen Shot 2020-10-14 at 11.47.46 AM.png
    Screen Shot 2020-10-14 at 11.47.46 AM.png
    78,6 KB · Aufrufe : 302
  • Screen Shot 2020-10-14 at 11.48.03 AM.png
    Screen Shot 2020-10-14 at 11.48.03 AM.png
    136,4 KB · Aufrufe : 312
How about returning the core voltage to 750 mv and decreasing the core clock until you reach 110 W.
I have achieved 106-108 W at 55.7 without any miner special configurations.

Mem temp: 90c - 92c

I had to test many configurations to achieve power below 110W. I reduced the core v to 715mw. The memory volt is down to 750 (bellow it crashes. )

To increase the hash rate I used the following Phoenix command line:

-coin %COIN%
-fanmin 0 -amd -acm
-mi 1 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 64 -openclGlobalMultiplier 4096

With the -openclLocalWork 64 switch I am getting close to 60MH/s, BUT I am getting invalid shares.

1) What parameter I could change to reduce INVALID shares?
 

Anhänge

  • Screen Shot 2020-10-14 at 12.36.09 PM.png
    Screen Shot 2020-10-14 at 12.36.09 PM.png
    297,7 KB · Aufrufe : 284
  • Screen Shot 2020-10-14 at 12.53.17 PM.png
    Screen Shot 2020-10-14 at 12.53.17 PM.png
    81,3 KB · Aufrufe : 240
  • Screen Shot 2020-10-14 at 12.53.26 PM.png
    Screen Shot 2020-10-14 at 12.53.26 PM.png
    133 KB · Aufrufe : 366
I backed the memory clock to 860MHz to try reducing the number of invalid shares. Even at memory clock 860, and miner option -openclLocalWork 64 I am getting:

Power: 104W
Hash: 59.9-60.4MH/s.

I will let the system run for a while to test if I reduced the number of invalid shares. I will also monitor on the pool that I am recording ~60MH/s.

Without the miner option -openclLocalWork 64 the hash rate will be around 55MH/s.

I think the bios configuration is now complete. I will start a batch updates to all cards.
 
I backed the memory clock to 860MHz to try reducing the number of invalid shares. Even at memory clock 860, and miner option -openclLocalWork 64 I am getting:

Power: 104W
Hash: 59.9-60.4MH/s.

I will let the system run for a while to test if I reduced the number of invalid shares. I will also monitor on the pool that I am recording ~60MH/s.

Without the miner option -openclLocalWork 64 the hash rate will be around 55MH/s.

I think the bios configuration is now complete. I will start a batch updates to all cards.
You can try first as advised before to keep the voltage of the core at 750 mv and the memory voltage at 1350 mv and reduce the memory by 10 MHz from 950 MHz to 940 MHz and the core by 50 MHz from 1400 MHz to 1350 MHz to reduce the rejected share and keep the power less than 115 W with hash rate around 58.5 MH/s for best performance to power.
The rejected share are inevitable and too much users in the pool will increase that, our goal to get best hash rate with minimum possible power consumption, performance to power, with safe temperature and minimum rejected share.
 
Zuletzt bearbeitet :
I backed the memory clock to 860MHz to try reducing the number of invalid shares. Even at memory clock 860, and miner option -openclLocalWork 64 I am getting:

Power: 104W
Hash: 59.9-60.4MH/s.

I will let the system run for a while to test if I reduced the number of invalid shares. I will also monitor on the pool that I am recording ~60MH/s.

Without the miner option -openclLocalWork 64 the hash rate will be around 55MH/s.

I think the bios configuration is now complete. I will start a batch updates to all cards.
After you finish the Gigabyte card, how about we start optimizing the Sapphire one.
 
I am going to take a break. I put the Gigabytes in production and I need to reduce the invalid shares errors. I am going to work on testing ways to reduce the errors. I will get to start the mods on the Sapphire tomorrow. I could use information in advance if you happen to have.

Regards,
AlleyCat
 
I am going to take a break. I put the Gigabytes in production and I need to reduce the invalid shares errors. I am going to work on testing ways to reduce the errors. I will get to start the mods on the Sapphire tomorrow. I could use information in advance if you happen to have.

Regards,
AlleyCat
My last advice regarding the Gigabyte cards before moving to the Sapphire ones,
In order to easily reduce the invalid/rejected shares, we need to focus on one reason and eliminate the others,
Invalid/Rejected shares caused by either the hardware or the software or both,
Regarding the hardware side, too much undervolting or overclocking or bad optimization will lead to invalid/rejected shares even if we tried our best to optimize the software, the reduction of the invalid/rejected shares will be very small or even it will be the same and not effected by that,
In order to eliminate the culprit from the hardware side, we need to properly optimize it, when it comes to pc components the main element that cause the system temperature and instability rise up is the current (amperes) and not the voltage, according to the equation below:
Power (Watt) = Current (Ampere) x Voltage (Volt) x Power Factor

Because it is a DC system the Power Factor is 1,
From the equation we see that by raising either the current or the voltage or both it will increase the power, more power mean more performance,
In our case, when we increase the clock we increase the required power, to reduce the power while maintaining the stability and safety of the system and achieving the required performance is by fixing the voltage to the minimum stable amount and decreasing the current by decreasing the clock until we reach the desired amount of the power while maintaining the stability and safety of the system,

Therefore, I advise you to fix the voltage of the core at 750 mv and the voltage of the memory at 1350 mv at the memory controler voltage at 750 mv as advised by professionals and decrease the clock for both until you reach the desired power of 110 W or less while maintaining stable performance at lower temperature, by that we achieved our goal while preserving the stability and the safety of the system,
After we eliminate any possible cause of the invalid/rejected share from the hardware side, we tune the software to reduce the invalid/rejected share as much as possible.

I will prepare the profile for the Sapphire cards in advance whenever it is possible,

You may have a good and peaceful day,
Please stay safe and healthy,

Best Regards.
 
Zuletzt bearbeitet :
I am going to take a break. I put the Gigabytes in production and I need to reduce the invalid shares errors. I am going to work on testing ways to reduce the errors. I will get to start the mods on the Sapphire tomorrow. I could use information in advance if you happen to have.

Regards,
AlleyCat
Greetings and good day,

I would like to inform you that the best Adrenaline driver for mining in windows until now is 20.3.1 from the official link below:


or 19.7.1

Best Wishes.
 
Zuletzt bearbeitet :
Greetings and good day,

I would like to inform you that the best Adrenaline driver for mining until now is 20.4.2, it can be downloaded from the link below as AMD took it down:


Best Wishes.
Thanks. I downloaded these drivers a few days ago, because I had a few times to remove the drivers to recover a card. It is convenient to reinstall the drivers with the standalone file.
 
My last advice regarding the Gigabyte cards before moving to the Sapphire ones,
In order to easily reduce the invalid/rejected shares, we need to focus on one reason and eliminate the others,
Invalid/Rejected shares caused by either the hardware or the software or both,
Regarding the hardware side, too much undervolting or overclocking or bad optimization will lead to invalid/rejected shares even if we tried our best to optimize the software, the reduction of the invalid/rejected shares will be very small or even it will be the same and not effected by that,
In order to eliminate the culprit from the hardware side, we need to properly optimize it, when it comes to pc components the main element that cause the system temperature and instability rise up is the current (amperes) and not the voltage, according to the equation below:
Power (Watt) = Current (Ampere) x Voltage (Volt) x Power Factor

Because it is a DC system the Power Factor is 1,
From the equation we see that by raising either the current or the voltage or both it will increase the power, more power mean more performance,
In our case, when we increase the clock we increase the required power, to reduce the power while maintaining the stability and safety of the system and achieving the required performance is by fixing the voltage to the minimum stable amount and decreasing the current by decreasing the clock until we reach the desired amount of the power while maintaining the stability and safety of the system,

Therefore, I advise you to fix the voltage of the core at 750 mv and the voltage of the memory at 1350 mv at the memory controler voltage at 750 mv as advised by professionals and decrease the clock for both until you reach the desired power of 110 W or less while maintaining stable performance at lower temperature, by that we achieved our goal while preserving the stability and the safety of the system,
After we eliminate any possible cause of the invalid/rejected share from the hardware side, we tune the software to reduce the invalid/rejected share as much as possible.

I will prepare the profile for the Sapphire cards in advance whenever it is possible,

You may have a good and peaceful day,
Please stay safe and healthy,

Best Regards.
Good morning.
Great explanation, Your description provides the logical steps to explore and test the optimal power settings.

My test system run steady overnight with no errors with the following settings:
My power is around 115W, because I chose to get over 59MH/s. If I reduce the clock speed to 1400, the power drops to 112W, but so is the harsh rate to 57.4.
Running at 115W at 59MH/s i think is more efficient than running at 112W at 57MH/s.

As far at the invalid shares, the blame is on the Phoenix miner configuration switch.

-openclLocalWork 64 is what causing the errors. According the the HiveOS forum reply, it is not recommended using this switch. Once I removed the option, I was able to tune the card down to 115W.

https://forum.hiveos.farm/t/pheonix-miner-invalid-shares/17714/2
 

Anhänge

  • Screen Shot 2020-10-15 at 6.48.24 AM.png
    Screen Shot 2020-10-15 at 6.48.24 AM.png
    90,6 KB · Aufrufe : 368
I am going to take a break. I put the Gigabytes in production and I need to reduce the invalid shares errors. I am going to work on testing ways to reduce the errors. I will get to start the mods on the Sapphire tomorrow. I could use information in advance if you happen to have.

Regards,
AlleyCat
I am ready to start the bios mods for the Sapphire. Looking forward to the memory timing suggestions. I am curious how you find the optimized values for different cards.
 
Oben Unten