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 am going to write something that contradicts everything that I have wrote so far.

This report is based on a test with only one GPU 5700XT Nitro+. I will use these new assumptions with a rig of 9 cards.

I check the power consumption at the wall with HiveOS settings reported from 94W to 118W. The power difference at the wall is less than 5W (166-170W). At 94W in HiveOS, the power at the wall is 166W, and Hash at 55.9MH/s. At 118W in HiveOS, the power at the wall is 170W, and the hash is 57.9. With a gain of 2MH/s for a negligent increase of 6W, which is 3.5% of total power, I am getting an increase of 3.5% in Hash.

My conclusion, the run the rig at the MAXIMUM Hash, as long as the rig is stable. All the tweaking of VDDCO, MVDD etc. have little or NO impact on the power at the WALL.

The most significant variable on the power is VDD. I would try a new strategy that targets the best VDD/Core combination, with the highest memory clock. The purpose is to get the highest hash, which pays, with little increments of power AT THE WALL. There is a ratio of increments has to increased power, and I am looking for every 1MH, and increase on no more than 2-3 at the wall.

What takes lots of power are the fans 3 - 5 watts each. I will start focusing on reducing the temperatures so the fans can run slow. I will also look at replacing the risers, perhaps better risers reduce the overall power.
 
Hm, worth a shot, but I had the rig next to me for two weeks and doing anything with clocks or power made significant differences at the wall measurers. The crazy thing is, you raise the clock 10 Mhz and the power goes up, not touching the voltage! Then you touch the voltage, nothing happens. These cards work differently than previous generations.
 
This is crazy. I increased the core to 1450, vdd=79, Mem 945
Hm, worth a shot, but I had the rig next to me for two weeks and doing anything with clocks or power made significant differences at the wall measurers. The crazy thing is, you raise the clock 10 Mhz and the power goes up, not touching the voltage! Then you touch the voltage, nothing happens. These cards work differently than previous generations.
I am susp[ecting the SoC clock speed. Above memory speed 910MHz, the SoC jumps from 950MHz to 1085. This jump takes about 4-5W at the wall.
 
I am going to write something that contradicts everything that I have wrote so far.

This report is based on a test with only one GPU 5700XT Nitro+. I will use these new assumptions with a rig of 9 cards.

I check the power consumption at the wall with HiveOS settings reported from 94W to 118W. The power difference at the wall is less than 5W (166-170W). At 94W in HiveOS, the power at the wall is 166W, and Hash at 55.9MH/s. At 118W in HiveOS, the power at the wall is 170W, and the hash is 57.9. With a gain of 2MH/s for a negligent increase of 6W, which is 3.5% of total power, I am getting an increase of 3.5% in Hash.

My conclusion, the run the rig at the MAXIMUM Hash, as long as the rig is stable. All the tweaking of VDDCO, MVDD etc. have little or NO impact on the power at the WALL.

The most significant variable on the power is VDD. I would try a new strategy that targets the best VDD/Core combination, with the highest memory clock. The purpose is to get the highest hash, which pays, with little increments of power AT THE WALL. There is a ratio of increments has to increased power, and I am looking for every 1MH, and increase on no more than 2-3 at the wall.

What takes lots of power are the fans 3 - 5 watts each. I will start focusing on reducing the temperatures so the fans can run slow. I will also look at replacing the risers, perhaps better risers reduce the overall power.
Great findings, By the way you could try setting the TDC limit at 112 A and 122 A and 132 A, to determine the impact of the relationship between the Voltage and the Amperage at the wall,

Regarding the HiveOS settings, I was trying my best to explain that running the system at the recommended limits for a safe stable working system at 24/7 is the best choice as it is never wise to squeeze the components and put them at the edge of collapse however I believe only few who listen as I saw many videos on YouTube and read many minig community posts that made me surprised how they did not inquire sufficient knowledge before proceeding doing such trials and errors that most of them will degrade the cards and may render them useless on the long run.

Well everyone is acting at their own discretion, I always prefer to minimize the risks as much as possible and put my mind and heart at ease on the long term.
 
Great findings, By the way you could try setting the TDC limit at 112 A and 122 A and 132 A, to determine the impact of the relationship between the Voltage and the Amperage at the wall,

Regarding the HiveOS settings, I was trying my best to explain that running the system at the recommended limits for a safe stable working system at 24/7 is the best choice as it is never wise to squeeze the components and put them at the edge of collapse however I believe only few who listen as I saw many videos on YouTube and read many minig community posts that made me surprised how they did not inquire sufficient knowledge before proceeding doing such trials and errors that most of them will degrade the cards and may render them useless on the long run.

Well everyone is acting at their own discretion, I always prefer to minimize the risks as much as possible and put my mind and heart at ease on the long term.
I must confess; it is too tempting to try to tweak for best performance. No wonder it is called "mining" because we all have a crypto gold rush.

I needed the evidence to be convinced, and I was blown away to see the difference between HiveOS numbers and reality.

Thank you for being along with everyone, even if sometimes people ignore your recommendations. I stuck with the ratio 1.8 (sometimes I did 1.82 in secret), but now I am going to re-test the rig with the power meter.
 
This is crazy. I increased the core to 1450, vdd=79, Mem 945

I am susp[ecting the SoC clock speed. Above memory speed 910MHz, the SoC jumps from 950MHz to 1085. This jump takes about 4-5W at the wall.
By the way, when decreasing the TDC limit that will lead to set higher voltage to meet the required amount of power at specific clock,

for an example
TDC limit at 112 A, VDD for 1400 MHz may need higher voltage than TDC limit at 132 A,

Unfortunately, I heard that the TDC limit function might not work all the time and only the power limit one is working.
 

Anhänge

  • IMG_3675.JPG
    IMG_3675.JPG
    2 MB · Aufrufe : 54
  • IMG_3676.JPG
    IMG_3676.JPG
    3,9 MB · Aufrufe : 36
  • IMG_3677.JPG
    IMG_3677.JPG
    3,8 MB · Aufrufe : 30
  • IMG_3679.JPG
    IMG_3679.JPG
    3,8 MB · Aufrufe : 26
  • IMG_3680.JPG
    IMG_3680.JPG
    3,8 MB · Aufrufe : 52
I must confess; it is too tempting to try to tweak for best performance. No wonder it is called "mining" because we all have a crypto gold rush.

I needed the evidence to be convinced, and I was blown away to see the difference between HiveOS numbers and reality.

Thank you for being along with everyone, even if sometimes people ignore your recommendations. I stuck with the ratio 1.8 (sometimes I did 1.82 in secret), but now I am going to re-test the rig with the power meter.
Dear, it is not about the recommendations, everyone has the choice to do what they want, the problem is whenever someone discovers something that he believes it is good without knowledge about it and he spread it to the public without concerns and many take it.

Regarding increasing the ratio, it is possible however on the expense of stability and decreasing performance and increasing temperature.
 
Zuletzt bearbeitet :
I am going to write something that contradicts everything that I have wrote so far.

This report is based on a test with only one GPU 5700XT Nitro+. I will use these new assumptions with a rig of 9 cards.

I check the power consumption at the wall with HiveOS settings reported from 94W to 118W. The power difference at the wall is less than 5W (166-170W). At 94W in HiveOS, the power at the wall is 166W, and Hash at 55.9MH/s. At 118W in HiveOS, the power at the wall is 170W, and the hash is 57.9. With a gain of 2MH/s for a negligent increase of 6W, which is 3.5% of total power, I am getting an increase of 3.5% in Hash.

My conclusion, the run the rig at the MAXIMUM Hash, as long as the rig is stable. All the tweaking of VDDCO, MVDD etc. have little or NO impact on the power at the WALL.

The most significant variable on the power is VDD. I would try a new strategy that targets the best VDD/Core combination, with the highest memory clock. The purpose is to get the highest hash, which pays, with little increments of power AT THE WALL. There is a ratio of increments has to increased power, and I am looking for every 1MH, and increase on no more than 2-3 at the wall.

What takes lots of power are the fans 3 - 5 watts each. I will start focusing on reducing the temperatures so the fans can run slow. I will also look at replacing the risers, perhaps better risers reduce the overall power.
You should build yourself something like this, and yes it is significantly more what the OS reports!
The build from spacegoats panda uses in the video has still one mistake...it doesnt account for the power the riser draws through the USB cable.
AFAIK they use a USB 3.0 cable as connection between the riser and board because of its power rating (if im not mistaken its 5A).
So to get the whole "real" power draw we would need a box like in the video PLUS another connection for the USB to meassure.

It would be interesting to know if some settings just shuffle around the current draw from one plug to another resulting in no difference in wattage.
 
You should build yourself something like this, and yes it is significantly more what the OS reports!
The build from spacegoats panda uses in the video has still one mistake...it doesnt account for the power the riser draws through the USB cable.
AFAIK they use a USB 3.0 cable as connection between the riser and board because of its power rating (if im not mistaken its 5A).
So to get the whole "real" power draw we would need a box like in the video PLUS another connection for the USB to meassure.

It would be interesting to know if some settings just shuffle around the current draw from one plug to another resulting in no difference in wattage.
Unfortunately they have another mistake when they stated that backing up the vbios using gpu-z may brick the card which is not true as all the vbios on the TechPowerUp were backed up using the gpu-z and me myself never have a problem with it, however many wiyh dual bios card misuse it and after that they put the blame on it without thinking that they might missed something regarding it which is to put the switch on the bios they want to back up it first before they use the tool and the same when they want to reflash.

Is anyone available to help trouble shoot? Maybe in a discord? I speak English only though.
Please state what you need?
 
Please state what you need?
I am implementing the overclocks you suggested. Or trying to. And after I set the settings in more power tool, I cannot set the correct over clocks in OverdriveNtool

Memory will not go lower than 1900

This is the memory file I have added to the card
 

Anhänge

  • 16058233387426012493907862667081.jpg
    16058233387426012493907862667081.jpg
    1,2 MB · Aufrufe : 58
  • 5Tmod_msi_5700xt.rom
    1 MB · Aufrufe : 11
I recommend not use the overdrive n tool at the meantime as it is not working flawlessly with the newer version of adrenaline driver instead use the wattman of amd raedon software to tune your card or MSI afterburner.

I was able to adjust it by messing with the settings in More power tool. I am at the suggested settings, but seeing far less than I should. Power is not much different from original either. I am going to download hive and try it there.
 
I was able to adjust it by messing with the settings in More power tool. I am at the suggested settings, but seeing far less than I should. Power is not much different from original either. I am going to download hive and try it there.
I hope that you know what you are doing,
take your time and test things out.
 
yea this is why I am going to hive, because Ido not know what I am doing. I was able to bios mod just fine, because I have done that before and it is just copy and paste. but the other settings Iam not sure of.
 
I'm not surprised, because in Windows I had tremendous problems changing settings and they were getting reset and conflicted often, even with driver being installed without wattman. For example, not even miner could change the memory clock, only wattman.
 
I test the bios, don´t work, the GPU is bricked ! :(
Most of the times, it is possible to restore the original bios. Can you describe what is happening? Does your computer post? Does the OS starts? What os do you run? Do you see the GPU with GPU-Z?

in general, when my windows 10 will not start, remove the amd drivers and program the original bios. AMD has a tool that removes all the drivers.

another option is to crate a dos boot USB drive. In most cases, removing the amd driver will allow you to flash the GPU.
 
Oben Unten