I am the guy who discovered the "micron" bug and started the long thread at Nvidia.com under the ID brm65. In that thread, Nvidia acknowledged the bug and started working with the partners to resolve the bug with the production of a bios update.
In case anyone has not seen that thread, it is here https://forums.geforce.com/default/topic/963768/geforce-drivers/gtx-1070-memory-vrm-driver-or-bios-bug-in-micron-memory-1070-cards-/
I have seen so much misinformation and fear mongering about what it actually means to have Micron memory installed posted here in the 1070 owners club thread, on reddit and various other forums and tech blogs. I thought I would post a definitive note here in an attempt to clear up the various misconceptions and half truths that so many people seem to have.
1. There is nothing physically wrong with Micron memory chips as a product. If your card crashes at stock speeds, without any overclocking being applied, firstly remove all old drivers and do a clean install. if it still crashes at stock speeds then your card is possibly faulty and you might consider returning it.
2. The White Checkerboard artiface and Crash is caused by a timing issue between the micron chips ability to switch to running at higher clock speeds and the ability of the memory VRMs (Voltage regulator) ability to increase the voltage from a low idle state to being under a full 3D load. The end result, in certain circumstances, is that the memory chips clock up and end up being starved of power causing the checkerboard artifacts, lockup and subsequent BSOD crash.
3. Nvidia 1070 cards have 4 user accessible "P-States" (Performance States) that they can operate at. The three idle states are configured via the Nvidia control Panel under the setting "Power Management Mode" and are described as "Optimal Power", "Adaptive" and "Prefer Maximum Performance". The fourth P-State is accessed when you place the card under a 3d load such as when you are gaming or running a benchmark. A "P-State" determines the voltage and performance levels that the GPU operates at.
4. in Optimal, Adaptive modes and Prefer Maximum performance modes, the card will idle at .285v, .625v or .811v respectively at default core clock speeds.
5. The idle voltages will vary if you use the core clock slider to overclock your card, the higher you set the slider, the idle voltages are determined by the P-State and are fixed to a specific core frequency as determined by the bios. As you increase the slider, the point on the curve that represents the fixed base clock frequency reduces the associated voltage point on the curve.
6. the Memory VRM levels, while not directly reported by Afterburner or Precision XOC etc. appear to be proportional to the Voltage levels for the GPU that are reported by those overclock utilities.
7. the Timing issues that cause the crash only show themselves if the voltage has to increase from idle levels of about 0.775v or below to full power with a large memory overclock that typically exceeds +400 from the reference memory clock speed.
8. there are a number of methods, some of which I describe here that you can use you can use to mitigate the bug: a) lock the voltage on the curve using Afterburner or use K boost in Precision XOC. b) Set the performance setting to "prefer max performance mode in the control panel and make sure that the card is idling at .811v or above. c) if you do not use the lock voltage method, do not use the slider to increase core clock frequency, use the curve to increase clocks at the high end of the curve while leaving the low end of the curve at stock.
9. The published datasheets for both Samsung and Micron, where they have comparable specifications, both show identical 1.35-1.5v voltage requirements and other physical specifications. The exception to that statement is that Samsung has not published ram timings where Micron has.
If you discover any other workarounds or have news about the new vbioses for your brand of card, please feel free to include them in this thread for other peoples reference.
In case anyone has not seen that thread, it is here https://forums.geforce.com/default/topic/963768/geforce-drivers/gtx-1070-memory-vrm-driver-or-bios-bug-in-micron-memory-1070-cards-/
I have seen so much misinformation and fear mongering about what it actually means to have Micron memory installed posted here in the 1070 owners club thread, on reddit and various other forums and tech blogs. I thought I would post a definitive note here in an attempt to clear up the various misconceptions and half truths that so many people seem to have.
1. There is nothing physically wrong with Micron memory chips as a product. If your card crashes at stock speeds, without any overclocking being applied, firstly remove all old drivers and do a clean install. if it still crashes at stock speeds then your card is possibly faulty and you might consider returning it.
2. The White Checkerboard artiface and Crash is caused by a timing issue between the micron chips ability to switch to running at higher clock speeds and the ability of the memory VRMs (Voltage regulator) ability to increase the voltage from a low idle state to being under a full 3D load. The end result, in certain circumstances, is that the memory chips clock up and end up being starved of power causing the checkerboard artifacts, lockup and subsequent BSOD crash.
3. Nvidia 1070 cards have 4 user accessible "P-States" (Performance States) that they can operate at. The three idle states are configured via the Nvidia control Panel under the setting "Power Management Mode" and are described as "Optimal Power", "Adaptive" and "Prefer Maximum Performance". The fourth P-State is accessed when you place the card under a 3d load such as when you are gaming or running a benchmark. A "P-State" determines the voltage and performance levels that the GPU operates at.
4. in Optimal, Adaptive modes and Prefer Maximum performance modes, the card will idle at .285v, .625v or .811v respectively at default core clock speeds.
5. The idle voltages will vary if you use the core clock slider to overclock your card, the higher you set the slider, the idle voltages are determined by the P-State and are fixed to a specific core frequency as determined by the bios. As you increase the slider, the point on the curve that represents the fixed base clock frequency reduces the associated voltage point on the curve.
6. the Memory VRM levels, while not directly reported by Afterburner or Precision XOC etc. appear to be proportional to the Voltage levels for the GPU that are reported by those overclock utilities.
7. the Timing issues that cause the crash only show themselves if the voltage has to increase from idle levels of about 0.775v or below to full power with a large memory overclock that typically exceeds +400 from the reference memory clock speed.
8. there are a number of methods, some of which I describe here that you can use you can use to mitigate the bug: a) lock the voltage on the curve using Afterburner or use K boost in Precision XOC. b) Set the performance setting to "prefer max performance mode in the control panel and make sure that the card is idling at .811v or above. c) if you do not use the lock voltage method, do not use the slider to increase core clock frequency, use the curve to increase clocks at the high end of the curve while leaving the low end of the curve at stock.
9. The published datasheets for both Samsung and Micron, where they have comparable specifications, both show identical 1.35-1.5v voltage requirements and other physical specifications. The exception to that statement is that Samsung has not published ram timings where Micron has.
If you discover any other workarounds or have news about the new vbioses for your brand of card, please feel free to include them in this thread for other peoples reference.
