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 :
At first, do not modify the power table and only apply the vram timings straps, when you apply the vram timings please follow the kindly attached video below,

Sorry, but the video you linked to me is incompatible with Apple strips (in the video strips all the same),
and with this post
 
Please check out these numbers.
The only card that is out of its optimal performance is the MSI/Micron.
I find the MSI to be a very difficult and unstable card. Therefore, I would NOT recommend getting the MSI 5700XT Gaming for mining.
For the MSI I believe the best is to keep the original vbios without modifying the power play and only modifying the vram setting by following the Test 1 from before after that applying the tREF as below,

- K4Z80325BC
1000 MHz (Copy) - - - > (Paste) 1250 MHz
1250 MHz - - - > (1500-1550) MHz
(1500-1550) MHz - - - > 1800 MHz
1800 MHz - - - > 2000 MHz
2000 MHz - - - > 2250 MHz

- MT61K256M32
1000 MHz - - - > 1250 MHz
1250 MHz - - - > (1500-1550) MHz
(1500-1550) MHz - - - > (1750 - 1800 - 1875) MHz
(1750 - 1800) MHz - - - > 2000 MHz

Memory Clock - tREF
1000 MHz - 3900 (Samsung & Micron)
1250 MHz - 4875 (Samsung & Micron)
1550 MHz - 6045 (Samsung & Micron)
1750 MHz - 6825 (Micron only)
1800 MHz - 7020 (Samsung & Micron)
1875 MHz - 7315 (Micron only)
2000 MHz - 7800 (Samsung & Micron)
2250 MHz - 8775 (Samsung only)
 
Sorry, but the video you linked to me is incompatible with Apple strips (in the video strips all the same),
and with this post
The video to show how to copy and paste the straps without editing the tCL or anything else,

The Apple straps is the same concept however by copying each clock strap to its respective one in the bios as below,

Apple straps - - - - - > Bios vram timings straps in RedBiosEdititor

1000 MHz - - - - - > 1000MHz
1250 MHz - - - - - > 1250 MHz
1550 MHz - - - - - > 1550 MHz

and the same for the rest for both Micron memory MT61K256M32 and Samsung memory K4Z80325BC.
 
I moved the AMD card to production. It was too good to keep testing. I had to get it to start making money :).

I am requesting if I can defer these tests until I am getting a card into my test rig.
Please, never worry about testing at all, you are the most who contributed in improving the mining process for RX 5700 and to me you will still that one,

I hope that the money keeps flowing to you until you are satisfied,

You may have a nice sleep with happy dreams,

Good night buddy.
 
[Updated]
This is the final power table locked at 140 watt for any XT card as below,

- Under Overdrive Limits tab, set:
GFX Maximum Clock 1600 MHz
Memory Maximum Clock 1000 MHz
Power Limit Maximum 0%
Power Limit Minimum 0%

- Under Power and Voltages, set:
Maximum Voltage GFX 900 mv
Maximum Voltage SoC 1200 mv
Minimum Voltage GFX 750 mv
Minimum Voltage SoC 750 mv
Power Limit GPU 140 W
TDC Limit GFX 140 A
TDC Limit SoC 12 A

- Under Frequency tap, set:
GFX Maximum 1200 MHz
Memory DPM 3 - 950 MHz

- Under Fan tab, set:
Zero RPM Enable - YES
Stop Temperature 60° C
Start Temperature 70° C

Apple straps for vram timings.

HiveOS stable settings are below, you can choose the one that you see fit, any decreasing yield to instability for the long run, the chip quality still play a role here, you can increase the voltages values by 5 mv in case the chip quality is average and a crash happened at the long run,

Core Clock @ VDD
1450 MHz @ 800 mv,
1440 MHz @ 795 mv,
1435 MHz @ 790 mv,
1425 MHz @ 785 mv,
1420 MHz @ 780 mv,
1410 MHz @ 775 mv,
1400 MHz @ 770 mv,
1390 MHz @ 765 mv,
1380 MHz @ 760 mv,
1360 MHz @ 750 mv,

Micron memory Clock @ VDDCI & MVDD
960 MHz @ 810 mv & 1360 mv,
950 MHz @ 800 mv & 1345 mv,
940 MHz @ 790 mv & 1335 mv
935 MHz @ 785 mv & 1325 mv,
930 MHz @ 780 mv & 1320 mv,
920 MHz @ 775 mv & 1305 mv,
915 MHz @ 770 mv & 1300 mv
910 MHz @ 765 mv & 1290 mv,
900 MHz @ 755 mv & 1275 mv,
890 MHz @ 750 mv & 1265 mv,

Apple straps for the vram timings.
First of all, this thread has been a fantastic learning experience, thank you to all who have contributed. With regards to this power table, are we meant to use the same respective core clock settings with the counterpart settings for the memory clock? For example if I use the settings from the 1st line on the core clock settings (1450 MHz @ 800 mv), would I then use the 1st line for the mem clock settings (960 MHz @ 810 mv & 1360 mv)? I'm running 8 x Gigabyte 5700 XT Gaming OC, with the apple straps that were uploaded in this thread and MPT values indicated above. I'm just in the process of dialing everything in.
 
I am applying all the tricks I found in this thread and all the advice received but I can't get beyond this

it is as if there is something (which I cannot understand) that it limits to 59.6.
I would like to see + 60Mh / s (it would be a psychological satisfaction)
 

Anhänge

  • Screenshot_20201103-131126~2.png
    Screenshot_20201103-131126~2.png
    61,5 KB · Aufrufe : 123
Thank you for the great work you are doing in this thread, in particular for the Sapphire Pulse with Micron that I have. I used 1550 straps copied up, and after that the Apple Inc straps set here, gaining around 1mh/s. However, there are two problems, one of them MAJOR. 740mv and 1350 mh/z for the core, 910(1820) mh/z for the memory. Factory memory voltage.

The small issue is that I can't get past 910 mh/z on memory, which is 57 mh/s. That's ok.

The major problem is the power consumption. Software shows around 105w, great. Whole gpu stack at 1360w, great. But at the wall, 13 card rig is at around 1950 W.

I used hiveos to update the bios, that was the only way. Asus b250 mining expert, new risers, new system. 2x Seasonic 1300w gold. I tested every card individually in windows and in hiveos and got a stable 57.0 mh/s with 165w for the whole system(one card) in its 16x PCI-E slot. Perfect stability and locked at ACTUAL 105w. System idles at around 55w so the card draw in software at 105w puts it there or there about. But once put in the rig, which is open air and with decent spacing, it all goes to hell.
 
Here is a screen from the rig, and a photo of the wall measurers.
 

Anhänge

  • Screenshot_2.jpg
    Screenshot_2.jpg
    140,5 KB · Aufrufe : 106
  • 20201104_030019.jpg
    20201104_030019.jpg
    2,3 MB · Aufrufe : 97
I am applying all the tricks I found in this thread and all the advice received but I can't get beyond this

it is as if there is something (which I cannot understand) that it limits to 59.6.
I would like to see + 60Mh / s (it would be a psychological satisfaction)
You can go beyond that with the settings below after applying Apple straps,

Core clock at 1500 MHz, VDD at 825 mv,
Memory clock at 960, VDDCI at 820 mv, MVDD at 1360 mv.
 
Zuletzt bearbeitet :
First of all, this thread has been a fantastic learning experience, thank you to all who have contributed. With regards to this power table, are we meant to use the same respective core clock settings with the counterpart settings for the memory clock? For example if I use the settings from the 1st line on the core clock settings (1450 MHz @ 800 mv), would I then use the 1st line for the mem clock settings (960 MHz @ 810 mv & 1360 mv)? I'm running 8 x Gigabyte 5700 XT Gaming OC, with the apple straps that were uploaded in this thread and MPT values indicated above. I'm just in the process of dialing everything in.
You can choose the core clock and memory clock that suits your needs, there are no restrictions when choosing, for an example,

Core clock 1450 MHz @ VDD 800 mv,
Memory clock 950 MHz @ VDDCI 800 mv & 1350 mv.
 
Thank you for the great work you are doing in this thread, in particular for the Sapphire Pulse with Micron that I have. I used 1550 straps copied up, and after that the Apple Inc straps set here, gaining around 1mh/s. However, there are two problems, one of them MAJOR. 740mv and 1350 mh/z for the core, 910(1820) mh/z for the memory. Factory memory voltage.

The small issue is that I can't get past 910 mh/z on memory, which is 57 mh/s. That's ok.

The major problem is the power consumption. Software shows around 105w, great. Whole gpu stack at 1360w, great. But at the wall, 13 card rig is at around 1950 W.

I used hiveos to update the bios, that was the only way. Asus b250 mining expert, new risers, new system. 2x Seasonic 1300w gold. I tested every card individually in windows and in hiveos and got a stable 57.0 mh/s with 165w for the whole system(one card) in its 16x PCI-E slot. Perfect stability and locked at ACTUAL 105w. System idles at around 55w so the card draw in software at 105w puts it there or there about. But once put in the rig, which is open air and with decent spacing, it all goes to hell.
At the wall is different than from software as the HiveOS show the power consumption for the gpu only, at the wall it shows the power consumption for the whole system, gpu, motherboard, cpu, RAM and SSD,

Please could you provide your rig specification regarding cpu, RAM and SSD,

By the way, your memory temperature is high, please could you provide the values of VDDCI and MVDD that being used for the 910 MHz and the fan speed as well, If you like I could help in decreasing the temperature without affecting the stability and performance of the rig.
 
currently testing on xt flashed with non xt and modified straps, and -openclLocalWork 64
doesn't seems to fire rejected shares yet but only 21 minutes with that configuration...
edit: after 30 min started getting some rejected
1604468375177.png1604468354518.png
 
Zuletzt bearbeitet :
Thank you for the great work you are doing in this thread, in particular for the Sapphire Pulse with Micron that I have. I used 1550 straps copied up, and after that the Apple Inc straps set here, gaining around 1mh/s. However, there are two problems, one of them MAJOR. 740mv and 1350 mh/z for the core, 910(1820) mh/z for the memory. Factory memory voltage.

The small issue is that I can't get past 910 mh/z on memory, which is 57 mh/s. That's ok.

The major problem is the power consumption. Software shows around 105w, great. Whole gpu stack at 1360w, great. But at the wall, 13 card rig is at around 1950 W.

I used hiveos to update the bios, that was the only way. Asus b250 mining expert, new risers, new system. 2x Seasonic 1300w gold. I tested every card individually in windows and in hiveos and got a stable 57.0 mh/s with 165w for the whole system(one card) in its 16x PCI-E slot. Perfect stability and locked at ACTUAL 105w. System idles at around 55w so the card draw in software at 105w puts it there or there about. But once put in the rig, which is open air and with decent spacing, it all goes to hell.
Unfortunately this is normal... for some reason AMD isnt including power usage trough the risers on its power consumption number. And Im not talking about power usage of risers, which is just few watts. But what power riser pins are feeding to GPU.

If you buy one of these:
1604473017864.png
And measure 12V wires going to GPU, you'll notice that reading what windows/hiveos is displaying is the sum of PCIE power drain. But if you measure power drain of that riser feed (usually molex or PCIE), its ~39W. So what real power usage of 5700 (XT) GPU's typically are is ~140-160W

Here is quick summary what measrements means what:

- measured current from ALL +12V lines that goes to GPU = true power consumption of GPU
- (somehow, dunno how) measured power output of PSU = whole system: GPU's, ~15W to CPU idling, ~5W to motherboard
- measured power from wall: whole systems power usage * efficiency factor of your PSU.
 
Zuletzt bearbeitet :
I must politely disagree on that one :unsure:
This is absolutely rock solid

Anhang anzeigen 6908
Great, your update is highly appreciated and now your rig is running nicely and stably with good performance to power ratio which is really a good thing to hear,
What AlleyCat meant is the XT version as most companies put restrictions toward modification at the same time the XT version does not have much tolerance towards undervolting like the non XT version, for an example, your core clock at 1400 MHz is running stable at VDD 760 mv while on the XT version to run the 1400 MHz stable on the long run then VDD at least at 770 mv.

Please stay safe and healthy and keep up the good work,

Best Regards
 
Zuletzt bearbeitet :
Unfortunately this is normal... for some reason AMD isnt including power usage trough the risers on its power consumption number. And Im not talking about power usage of risers, which is just few watts. But what power riser pins are feeding to GPU.

If you buy one of these:
Anhang anzeigen 6907
And measure 12V wires going to GPU, you'll notice that reading what windows/hiveos is displaying is the sum of PCIE power drain. But if you measure power drain of that riser feed (usually molex or PCIE), its ~39W. So what real power usage of 5700 (XT) GPU's typically are is ~140-160W

Here is quick summary what measrements means what:

- measured current from ALL +12V lines that goes to GPU = true power consumption of GPU
- (somehow, dunno how) measured power output of PSU = whole system: GPU's, ~15W to CPU idling, ~5W to motherboard
- measured power from wall: whole systems power usage * efficiency factor of your PSU.
You are correct, many thanks for the detailed explanation.
 
At the wall is different than from software as the HiveOS show the power consumption for the gpu only, at the wall it shows the power consumption for the whole system, gpu, motherboard, cpu, RAM and SSD,

Please could you provide your rig specification regarding cpu, RAM and SSD,

By the way, your memory temperature is high, please could you provide the values of VDDCI and MVDD that being used for the 910 MHz and the fan speed as well, If you like I could help in decreasing the temperature without affecting the stability and performance of the rig.
That would be awesome, as the memory temps really are pretty high and that's with only 20 minutes of running. Default memory temperature in one card rig (16 x pcie) was also high, 90 at the best card, 98 at most cards at the same settings.

CPU is i3-7100 currently. RAM - Kingston 8 gb. SSD- in hive, not installed (flash drive), in Windows-Kingston 128 gb ssd.

Memory controller voltage and memory voltage are untouched, they are default. I don't know where to see the reading of memory voltages in hiveos. Card is Micron Sapphire Pulse XT.

Fan speeds are set with target memory temp of 90 in HiveOs, and they range from 50% to 80%. Cards at the edge of the rig are 50%, cards in the middle are 80%. Attached are screenshots and the photo of the rig.

Core voltage is 740, with core speed at 1350.
 

Anhänge

  • 20201104_124733.jpg
    20201104_124733.jpg
    3,1 MB · Aufrufe : 62
  • fans.jpg
    fans.jpg
    36,4 KB · Aufrufe : 89
  • Screenshot_3.jpg
    Screenshot_3.jpg
    141,1 KB · Aufrufe : 104
Oben Unten