Voltages and Resolutions

[xFluing]

Is there a possibility of a certain voltage being unstable at any particular resolution? I run my HD 7850 at stock (920/1250), and undervolted it and it seemed stable, nothing wrong even in 1920x1200

 

However, today I was trying 1440p in Shadow Warrior and the driver violently crashed on me (temps were in check, yes), so is it possible that the voltage I set it at to be stable for 1080p but not stable enough for 1440p?

[Skiiwee29]

Yes. Its similar to a CPU, if you dont supply it enough Voltage for the task it requires, it likely will crash and burn on you. 

[Froody129]

They do also eventually lose stability, forcing you go use lower clocks or higher voltages 

[xFluing]

2 minutes ago, Froody129 said:

They do also eventually lose stability, forcing you go use lower clocks or higher voltages 

So you're saying a certain voltage will slowly lose stability all by itself?

[Froody129]

Just now, xFluing said:

So you're saying a certain voltage will slowly lose stability all by itself?

Eventually, the chip degrades. It varies from chip to chip. 

[xFluing]

Just now, Froody129 said:

Eventually, the chip degrades. It varies from chip to chip. 

Doesn't the chip degrade if it runs too hot or is overvolted too much?

[Mira Yurizaki]

7 minutes ago, xFluing said:

Doesn't the chip degrade if it runs too hot or is overvolted too much?

Correct.

 

Undervolting can cause issues with signal integrity and may not provide enough of a voltage on transistor gates to let enough current through. It doesn't degrade the part.

[BarackOBatman]

I think the question is being misinterpreted?

 

My understanding is that OP has a stable undervolt, which can run at 100 load% with no issues in 1920x1200.

 

The question is: is it possible that by changing only the resolution to 2560x1440 with the same exact undervolt settings, can it somehow become unstable.

 

I would argue that no this should not happen, and the GPU should still be stable. He isn't trying to change the undervolt... it was already stable using 100% of its resources before.

 

However... it may be possible that the higher resolution uses more VRAM and if the vram undervolt isn't stable this may be the issue??? (if it was never stable to begin with, but wasn't causing an issue because it wasn't used as much?) This is only hypothetical.

 

The only way that I could see degradation being an issue is if the chip was running too hot for too long. Considering this GPU was undervolted.. I doubt this was the issue. (although technically possible) the odds of being 100% stable one moment, and physical chip degradation occurring at the exact moment the resolution changed is unlikely.

[KarathKasun]

25 minutes ago, xFluing said:

Doesn't the chip degrade if it runs too hot or is overvolted too much?

Incorrect.  All chips at all settings degrade over time.

[KarathKasun]

2 minutes ago, BarackOBatman said:

I think the question is being misinterpreted?

 

My understanding is that OP has a stable undervolt, which can run at 100 load% with no issues in 1920x1200.

 

The question is: is it possible that by changing only the resolution to 2560x1440 with the same exact undervolt settings, can it somehow become unstable.

 

I would argue that no this should not happen, and the GPU should still be stable. He isn't trying to change the undervolt... it was already stable using 100% of its resources before.

 

However... it may be possible that the higher resolution uses more VRAM and if the vram undervolt isn't stable this may be the issue??? (if it was never stable to begin with, but wasn't causing an issue because it wasn't used as much?) This is only hypothetical.

 

The only way that I could see degradation being an issue is if the chip was running too hot for too long. Considering this GPU was undervolted.. I doubt this was the issue. (although technically possible) the odds of being 100% stable one moment, and physical chip degradation occurring at the exact moment the resolution changed is unlikely.

Yes, different resolution means different load distribution on the GPU subsystems (memory controller, cache, dispatch, etc.)

[BarackOBatman]

2 minutes ago, KarathKasun said:

Yes, different resolution means different load distribution on the GPU subsystems (memory controller, cache, dispatch, etc.)

So a different resolution will need a different undervolt, as each resolution requires different resources from the GPU?

 

In other words, the undervolt is stable at 1920x1200, but will need to be tweaked to be stable in 2560x1440

[BarackOBatman]

7 minutes ago, KarathKasun said:

Incorrect.  All chips at all settings degrade over time.

id say, no they don't? (at least not measurably, not unless its been seriously abused)

 

[xFluing]

1 minute ago, BarackOBatman said:

So a different resolution will need a different undervolt, as each resolution requires different resources from the GPU?

 

In other words, the undervolt is stable at 1920x1200, but will need to be tweaked to be stable in 2560x1440

Well yeah, my thoughts exactly, I won't be doing that though because I want my card to run as cool as possible, trying to keep it below 75 (and relatively silent) at all costs, even opened up the case so that it has more room to exhaust the hot air (which did help by 2-3 degrees which is the difference between a feather falling and a jet engine taking off).

[KarathKasun]

10 minutes ago, BarackOBatman said:

id say, no they don't? (at least not measurably, not unless its been seriously abused)

 

Id you are looking at the wrong thing there guy.  Any silicon chip degrades over time, and by time I mean hours/days/years powered on.  Eventually voltage needs to go up to maintain speed.  GPU's tend to be hurt more by this because of the higher temps and higher transistor density.  They can see the results of this degradation in ~10 years if they are used very often, but most GPUs run no load or outright turned off most of their lives.

 

Newer parts on smaller processes also degrade faster.

 

Electron migration in the CMOS process is the root of this issue.  The current going through the chip slowly re-shapes the conductive pathways.

[BarackOBatman]

17 minutes ago, KarathKasun said:

Id you are looking at the wrong thing there guy.  Any silicon chip degrades over time, and by time I mean hours/days/years powered on.  Eventually voltage needs to go up to maintain speed.  GPU's tend to be hurt more by this because of the higher temps and higher transistor density.  They can see the results of this degradation in ~10 years if they are used very often, but most GPUs run no load or outright turned off most of their lives.

 

Newer parts on smaller processes also degrade faster.

 

Electron migration in the CMOS process is the root of this issue.  The current going through the chip slowly re-shapes the conductive pathways.

What is wrong about the using this video for this topic? Seems to address degration in the same way it is here.

It also talks about electron migration and how it does not cause gradual degredation, but rather total failure.

[KarathKasun]

1 minute ago, BarackOBatman said:

What is wrong about the using this video for this topic? Seems to address degration in the same way it is here.

It also talks about electron migration and how it does not cause gradual degredation, but rather total failure.

It ends in total failure.  Once you get close to that point it does start to become unstable and you can add voltage/reduce clocks to keep it going for a short while.

 

Once the voltage requirements begin to rise, you are likely in the last 1% (or less) of the chips lifespan, which statistically is "total immediate failure".

[BarackOBatman]

Just now, KarathKasun said:

It ends in total failure.  Once you get close to that point it does start to become unstable and you can add voltage/reduce clocks to keep it going for a short while.

 

Once the voltage requirements begin to rise, you are likely in the last 1% (or less) of the chips lifespan, which statistically is "total immediate failure".

ok.. well the LTT video still holds in this case.

 

And I can agree with the "within the last 1% of its lifespan thing", but in most troubleshooting cases , including the OP's, degradation isn't an issue then.

and if it is, its time to buy a new video card cause its about to die completely.

[xFluing]

14 hours ago, BarackOBatman said:

ok.. well the LTT video still holds in this case.

 

And I can agree with the "within the last 1% of its lifespan thing", but in most troubleshooting cases , including the OP's, degradation isn't an issue then.

and if it is, its time to buy a new video card cause its about to die completely.

Which is what I fear might have happened, it started being unstable even at stock voltage even in stuff that used to be stable before, like Heaven, it's ok in idle, but under load the screen fills with the same looking artifact repeated, and it crashes, good thing I still have my R7 250 to hold its place while I save up for a new card.

[BarackOBatman]

6 hours ago, xFluing said:

Which is what I fear might have happened, it started being unstable even at stock voltage even in stuff that used to be stable before, like Heaven, it's ok in idle, but under load the screen fills with the same looking artifact repeated, and it crashes, good thing I still have my R7 250 to hold its place while I save up for a new card.

It could very well be completely done 

[xFluing]

5 hours ago, BarackOBatman said:

It could very well be completely done 

Remains to be seen, I'll visit the guy I got it from, and maybe I can get my money back, or we sort out the problem and we find out it was probably the power supply all along.