MSI Statement Regarding Reported Instability with GeForce RTX 30 Series Graphics Cards

[btarunr]

MSI became aware of reports from customers, reviewers, and system integrators that there may be instability when GeForce RTX 30 Series graphics cards core clocks exceeded a certain amount. The latest GeForce driver (456.55) includes fixes for the issue. As such, MSI recommends owners of GeForce RTX 30 Series graphics cards update to the latest driver release which can be downloaded from the NVIDIA GeForce website.

MSI stands behind its design decisions for its GeForce RTX 30 Series graphics cards catalog which consists of GAMING models and VENTUS models. MSI utilizes a mixed capacitor grouping in its designs to benefit from the strengths of both SP-Caps and MLCCs. All MSI GeForce RTX 30 Series cards that have shipped out since the beginning of production, which include media review samples, feature the PCB configurations as shown in the updated images below.



MSI would like to express its appreciation to customers, reviewers, and system integrators that raised awareness to the issue, and to NVIDIA for the cooperation in quickly resolving the problem.

Product image of GeForce RTX 3090 GAMING TRIO series that have shipped out since the beginning of production:


Product image of GeForce RTX 3090 VENTUS 3X series that have shipped out since the beginning of production:


Product image of GeForce RTX 3080 GAMING TRIO series that have shipped out since the beginning of production:


Product image of GeForce RTX 3080 VENTUS 3X series that have shipped out since the beginning of production:


View at TechPowerUp Main Site

 

[Vayra86]

Any news on what the driver fix is made of? I don't see a statement on it yet

 

[btarunr]

Any news on what the driver fix is made of? I don't see a statement on it yet

After the RTX 3090 review day, @W1zzard relocated to a new big castle. He's setting up his new lair, and will start testing soon. We have many things planned.

 

[Chrispy_]

Any news on what the driver fix is made of? I don't see a statement on it yet

The bypass capacitors only handle the very brief draw from the GPU core during load transitions, almost immediately after a load transition has occurred, the voltage regulators will catch up and adjust so that the capacitors aren't needed again.

I'm going to guess that the driver update slows down the frequency of changes between boost clock levels. There are two main causes of load transitions that would push the bypass capacitors behind a GPU core; Workload and clock/voltage changes. Nvidia don't have much control over what workload software devs are putting on the GPU, but they can control how much and how frequently the GPU boost clock changes, each jump of which causes a small deviation in the output of the voltage regulator and the input drawn by the GPU core. The bypass capacitors have to deal with that, and so the slower the boost clock changes, the less the bypass capacitors are stressed, and the smaller the jump between different clock states, the lower the draw on the bypass capacitor array.

TL;DR, I suspect the boost algorithm has been tuned to be less agressive, so it'll probably reach slightly lower peaks, and also the clock vs time chart will be a smoother line. I doubt it'll have much impact on average boost clocks over the course of any benchmarks or gaming sessions.

 

[clopezi]

It's strange, because in multiple websites, they have this picture on 3080 models, but MSI assures that this it's not correct, that the GPU's are the same on media reviews.

However, for me, if with the new drivers the are not CTD, it's fine.

This is the picture in multiple shops that are MSI partners:

 

[bug]

After the RTX 3090 review day, @W1zzard relocated to a new big castle. He's setting up his new lair, and will start testing soon. We have many things planned.

Wth, isn't he supposed to be in an ivory tower? This world is going down the drain...

 

[stimpy88]

Any news on what the driver fix is made of? I don't see a statement on it yet

I'm guessing that they have gimped 50MHz or so off the clock speed, or simply altered the mechanism that sets the boost clock, so that voltage has more of an effect on it's decisions, i.e. slightly more voltage at a slightly lower clockspeeds...

If true, it will be interesting to see the effect on overclockers.

 

[VSG]

Wth, isn't he supposed to be in an ivory tower? This world is going down the drain...

Hard to get GPUs up an ivory tower, much easier inside a castle with serfs around to help.

 

[NJM1564]

What are you talking about. The castle has lots of towers. Why would you think he would be always limited to one.

 

[Anymal]

Any news on what the driver fix is made of? I don't see a statement on it yet

You should look at guru3d, you are even in the same country, FFS!

 

[clopezi]

Okay, some new info.

I have ordered the MSI Trio X in Spanish store. This Spanish store it's a MSI partner and it has been working for many years, it is serious business.

I have asked about the shipping time of the card, because 15 days ago, they told me that 02/10 will be the launch date. They told me today that MSI is making the rev. V2 of the GPU, and it's the delay reason, so when it comes to Spain, they will deliver it.

This info open two options:

- MSI it's really making a V2 of the GPU (Maybe it's other thing, maybe it's version with two MLCCs groups)
- Store told me bullshit, but I doubt it.

 

[bug]

Hard to get GPUs up an ivory tower, much easier inside a castle with serfs around to help.

What, now he doesn't have telekinesis in his book either?

 

[Vayra86]

After the RTX 3090 review day, @W1zzard relocated to a new big castle. He's setting up his new lair, and will start testing soon. We have many things planned.

World domination obviously.

Can't wait!

The bypass capacitors only handle the very brief draw from the GPU core during load transitions, almost immediately after a load transition has occurred, the voltage regulators will catch up and adjust so that the capacitors aren't needed again.

I'm going to guess that the driver update slows down the frequency of changes between boost clock levels. There are two main causes of load transitions that would push the bypass capacitors behind a GPU core; Workload and clock/voltage changes. Nvidia don't have much control over what workload software devs are putting on the GPU, but they can control how much and how frequently the GPU boost clock changes, each jump of which causes a small deviation in the output of the voltage regulator and the input drawn by the GPU core. The bypass capacitors have to deal with that, and so the slower the boost clock changes, the less the bypass capacitors are stressed, and the smaller the jump between different clock states, the lower the draw on the bypass capacitor array.

TL;DR, I suspect the boost algorithm has been tuned to be less agressive, so it'll probably reach slightly lower peaks, and also the clock vs time chart will be a smoother line. I doubt it'll have much impact on average boost clocks over the course of any benchmarks or gaming sessions.

Sounds plausible.

I'm guessing that they have gimped 50MHz or so off the clock speed, or simply altered the mechanism that sets the boost clock, so that voltage has more of an effect on it's decisions, i.e. slightly more voltage at a slightly lower clockspeeds...

If true, it will be interesting to see the effect on overclockers.

Sounds too simple and too much like PR disaster. Never say never, but...

You should look at guru3d, you are even in the same country, FFS!

Ehm... yeah? And where does that detail what fix is applied then? Patch notes don't really say.

ffs?

 

[Julhes]

false news. are we surprised? crap site.

 

[laszlo]

false news. are we surprised? crap site.

this is the official MSI statement; if you think this site is crap why bother visiting and posting...??

 

[Deleted member 185088]

Maybe the 30xx series are pushed to their limits, from what I've seen they don't overclock well at all. Maybe that's the reason why they are crashing.

 

[bug]

Maybe the 30xx series are pushed to their limits, from what I've seen they don't overclock well at all. Maybe that's the reason why they are crashing.

All that was written in this thread flew right over your head, didn't it?

 

[Deleted member 185088]

All that was written in this thread flew right over your head, didn't it?

Blind me!