Frage How to exceed 1.35v on 5700xt? suicide run OC's

[sebaskitten]

Hello good day!

i've been in the mood for some suicide run overclocking lately and got a few 5700xt's here to play around with.

I have noticed that if I set 1.4v core voltage in MPT that it bounces back down to 1.35v.

I have heard from a few people that there is a certain moddified version of "MPT" that allows users to set unsafe ranges of settings.

How do I obtain this special version of the software? do I need to toggle something myself in MPT to allow unsafe ranges ?

i'm well aware 1.4v and up is getting to the region of popping silicon but that's all part of the fun. and the card's that score well/survive might get their VRAM upgraded to some special chips


currently a few cards are holding steady at a maximum of around 2200mhz for timespy extreme run's on a custom watercooling block. i might be able to squeeze it a bit higher with revision 2 of the block. but i am just being limited by 1.35v core voltage .

thank you in advance for the info!

 

[hellm]

I think AMD did integrate this limit in the driver since a certain version. So without EVC you would have to use a previous version.

But you can try for yourself. All you need to do is to save 2 registry files with different voltage and compare them. The bytes are saved in little endian, that means reverse order.
Example: Your voltage is 1200mV, this value is stored x4, so 4800 in hex is 0x12C0; which would be found as "C0, 12" in the registry file.
..but as i said, the driver might not accept it.

 

[sebaskitten]

I think AMD did integrate this limit in the driver since a certain version. So without EVC you would have to use a previous version.

But you can try for yourself. All you need to do is to save 2 registry files with different voltage and compare them. The bytes are saved in little endian, that means reverse order.
Example: Your voltage is 1200mV, this value is stored x4, so 4800 in hex is 0x12C0; which would be found as "C0, 12" in the registry file.
..but as i said, the driver might not accept it.

i see. i'll give this a shot when i can. thanks for the suggestion.

i have been thinking of getting a elmors EVC along with some other toy's from them.


currently testing out a new hybrid "cryo cooler" i made to see if it nets to a more controller hotspot+vrm+mem temps

 

[hellm]

Sounds interesting. If you have trouble with a cold bug or something, you could try this:
offset 0x232 (0x332 in .mpt):
// SECTION: Throttler settings
FE 70 00 00 ThrottlerControlMask // See Throtter masks defines

// Throttler Control/Status Bits
#define THROTTLER_PADDING_BIT      0
#define THROTTLER_TEMP_EDGE_BIT    1
#define THROTTLER_TEMP_HOTSPOT_BIT 2
#define THROTTLER_TEMP_MEM_BIT     3
#define THROTTLER_TEMP_VR_GFX_BIT  4
#define THROTTLER_TEMP_VR_MEM0_BIT 5
#define THROTTLER_TEMP_VR_MEM1_BIT 6
#define THROTTLER_TEMP_VR_SOC_BIT  7
#define THROTTLER_TEMP_LIQUID0_BIT 8
#define THROTTLER_TEMP_LIQUID1_BIT 9
#define THROTTLER_TEMP_PLX_BIT     10
#define THROTTLER_TEMP_SKIN_BIT    11
#define THROTTLER_TDC_GFX_BIT      12
#define THROTTLER_TDC_SOC_BIT      13
#define THROTTLER_PPT0_BIT         14
#define THROTTLER_PPT1_BIT         15
#define THROTTLER_PPT2_BIT         16
#define THROTTLER_PPT3_BIT         17
#define THROTTLER_FIT_BIT          18
#define THROTTLER_PPM_BIT          19
#define THROTTLER_APCC_BIT         20

0x70FE would mean: 0111 0000 1111 1110 in binary
To avoid temperature issues at low temps you could try die zero the first half byte, this would mean: F0 70
Or this, in the feature menu:

#define FEATURE_THERMAL_BIT             33

Or this, Offset ox15 (0x115 in .mpt):
1B thermal_controller_type //one of SMU_11_0_PP_THERMALCONTROLLER
Here 0x00 means deactivated. This was a workaround for Fiji coldbug.

 

[Satori_Nori]

Hello good day!

i've been in the mood for some suicide run overclocking lately and got a few 5700xt's here to play around with.

I have noticed that if I set 1.4v core voltage in MPT that it bounces back down to 1.35v.

I have heard from a few people that there is a certain moddified version of "MPT" that allows users to set unsafe ranges of settings.

How do I obtain this special version of the software? do I need to toggle something myself in MPT to allow unsafe ranges ?

i'm well aware 1.4v and up is getting to the region of popping silicon but that's all part of the fun. and the card's that score well/survive might get their VRAM upgraded to some special chips


currently a few cards are holding steady at a maximum of around 2200mhz for timespy extreme run's on a custom watercooling block. i might be able to squeeze it a bit higher with revision 2 of the block. but i am just being limited by 1.35v core voltage .

thank you in advance for the info!

Good day Mr. Sebaskitten,

1.25V is the max safe voltage for RDNA1 RX 5000 series both core and SoC, your card is crippled by other parameters in the VBIOS,

If you like to unleash safely the potential of your card, please feel free to message me anytime, you will be surprised how much potential was hidden,

AMD - Radeon RDNA1 Series VBIOS Configuration

Greetings everyone, I am Satori, I have sufficient knowledge in properly configuring VBIOS of RDNA1, Polaris, Vega 2 and DDR3/DDR4/DDR5 RAMs in addition to unlocking hidden settings of any motherboard UEFI BIOS, here it is going to be for RDNA1 graphics cards, I decided to help the users of...

www.igorslab.de