News Crash Tested: Nvidia Improves RTX 30-Series Stability With New Driver

[Admin]

We tested several RTX 3080 and 3090 cards with the latest drivers, showing before and after performance, clocks, and power.

Crash Tested: Nvidia Improves RTX 30-Series Stability With New Driver : Read more

 

[spongiemaster]

the 456.55 drivers are 0.2-0.5% slower overall

I hope Nvidia set aside money for the class action lawsuit.

 

[helper800]

I hope Nvidia set aside money for the class action lawsuit.

Thats within the margin of error. I completely doubt any lawyer worth their salt would sue over that specifically.

 

[Kamen Rider Blade]

So OCing is largely meaningless when you're this close to the ragged edge of stability.

 

[Tanquen]

Still a pretty good slap in the face considering gaming cards cost 2-3-4 times what they should.

 

[ForwardUntoDawn]

Tom's, could you please list which cards from which manufacturers do and don't have the MLCC blocks?

 

[drivinfast247]

I hope Nvidia set aside money for the class action lawsuit.

Lol. Where did it guarantee a 2ghz clock?

 

[nofanneeded]

It seems that EVGA KINGPIN has no chance at all this Generation of RTX cards ...

 

[spongiemaster]

Thats within the margin of error. I completely doubt any lawyer worth their salt would sue over that specifically.

Seriously?

 

[helper800]

Seriously?

Seriously. 0.5-1% is within the margin of error in any testing with many many variables like hardware. Especially since boosting behaviors are never going to be the same between benchmarking runs.

 

[deesider]

I hope Nvidia set aside money for the class action lawsuit.

They fixed the problem for free - how are you going to sue for that

 

[JarredWaltonGPU]

Tom's, could you please list which cards from which manufacturers do and don't have the MLCC blocks?

My understanding is:

Asus: 6xMLCC
EVGA: Varies, 6xMLCC on top models? 4xSP-Cap/2xMLCC on others? (I don't have a card yet)
Gigabyte: 6xSP-Cap on some models, not sure on others
MSI: 5xSP-Cap/1xMLCC on some, 4xSP-Cap/2xMLCC on others
Nvidia: 6xMLCC
PNY: I don't know
Zotac: I don't know

 

[spongiemaster]

Seriously. 0.5-1% is within the margin of error in any testing with many many variables like hardware. Especially since boosting behaviors are never going to be the same between benchmarking runs.

No, seriously as in you didn't get that was a joke? Of course no one would sue over 0.2% performance loss.

 

[JarredWaltonGPU]

No, seriously as in you didn't get that was a joke? Of course no one would sue over 0.2% performance loss.

I'm suing over your comment for defamation, and implying no one considers my 0.2% testing results difference meaningful! sniff!

 

[Phaaze88]

I completely doubt any lawyer worth their salt would sue over that specifically.

😒
Well, people fight/argue over petty crap all the time - the justice system isn't exempt from this.

I'm not saying no one's actually going to do it, but that it's still a possibility.

 

[spongiemaster]

I'm suing over your comment for defamation, and implying no one considers my 0.2% testing results difference meaningful! sniff!

You don't need to sue me. Don't you have three 3080's? That's like $45,000 on Ebay.

Jokes aside, the article was very useful. Basically shows that if you want a 3080, and you can find one at a reasonable price, to go for it. The problems are getting fixed, and the performance isn't affected.

 

[Shadowclash10]

I'm suing over your comment for defamation, and implying no one considers my 0.2% testing results difference meaningful! sniff!

So, aren't the results here well within the margin of error? Something as samll as, oh I don't know, the weather being 10 degrees hotter during one test, or say the internals being warmer after a previous test? Not criticizing, simply asking. Regardless, I think that if the difference is so small it is basically inconsequential?

 

[sizzling]

I was reading elsewhere that SP-Caps have a better life expectancy with less chance of cracking. Therefore the decision between the 2 types was more about life expectancy vs a few % higher performance. Does anyone know if their is any truth to this? Definitely would put a different spin on this whole capacitor discussion.

 

[Blacksad999]

Considering the cards technically only need hit their advertised boost clock, anything even near 2k is pretty good. So, the Tuf boost advertises as: 1740MHz, the EVGA FTW 1800MHz, etc.

 

[excalibur1814]

Does this mean that the next cards, from the assembly line, will include a 'hardware' fix?

 

[sizzling]

Does this mean that the next cards, from the assembly line, will include a 'hardware' fix?

The more I read on this the more I come to the conclusion this isn’t a hardware problem. Expert reviews say the caps can only make <=50Mhz difference. I saw a video where 2 caps were replaced and it made a 30mhz difference. The other factor is the boost in question only operates for a few seconds, I would guess that over 99% of the time the gpu isn’t using this max boost anyway.

 

[vinay2070]

I'm suing over your comment for defamation, and implying no one considers my 0.2% testing results difference meaningful! sniff!

When I went through the article, I was like, he has a lot of patience!!!

 

[GetSmart]

From https://www.pcworld.com/article/3583894/nvidia-fix-rtx-3080-crashes-new-drivers-clock-speed.html the author found a repeatable crash scenario (in Horizon Zero Dawn) and tested it out with new drivers.. 🤓

 

[JarredWaltonGPU]

When I went through the article, I was like, he has a lot of patience!!!

It's absolutely required for testing hardware. I have 100+ hours logged in multiple games on Steam, for games that I basically never played but only benchmarked. Strange Brigade is a good example of this, or Shadow of the Tomb Raider -- I've played that for maybe 10-15 hours, but certainly not the 225 hours I have on record! And the great thing is that about once a quarter, or at least semi-annually, I need to go and retest every GPU to keep the GPU Benchmarks Hierarchy up to date. But as far as jobs go, writing about new hardware is pretty fun for me.

 

[bigdragon]

My takeaway is that gaming performance increased slightly, synthetic benchmark performance dropped significantly, and temps improved slightly. I use GPUs for gaming and game dev so the improvements are more appealing than losing synthetic numbers.

Nvidia should have delayed Ampere, however. There is no excuse for rushing untested or barely-tested designs to market like this. I am enjoying seeing scalpers panic due to early-adopter customers trying to return scalper cards! LOL!