im a lucky boy

"Instructions on how to file a claim for the Nvidia class action settlement were not immediately available. Keep checking TopClassActions.com or sign up for our free newsletter for the latest updates. You can also “Follow” this case using your free Top Class Actions account to receive notifications when this article is updated." For anyone looking to claim

Can my GPU be made by another company to get the $30? Mine is made by MSI.

#4
The GPU isn't made by MSI. MSI just slapped a different cooler on it. It's still the same GPU, "made" by nVidia (it's actually "built" by TSMC but let's not go there), it got the same problem so nVidia still has to pay you those 30$.

i have a 960 god damn it

nvidia users even get free money now? smh

(if it's a 970 and you bought it before this, it's eligible; the lawsuit came from the fact that the "4GB" of vram was basically 3.5GB in practice since the last .5GB was complete garbage) (also afaik it's american-only, as almost all class action settlements end up being)

Doesn't cover the cost of a $300 brick.

It is actually far worse than that. The game still thinks it has access to 4gb vram so it tries to fill up the full 4gb (unused memory is wasted memory). When it tries to access this .5gb memory it causes MASSIVE stuttering due to lack of bandwidth.This actually causes it to have WORSE performance than if they just disabled the .5 entirely.

Now you may be thinking "my 970 works just fine". That is because they use some driver workarounds to keep games from putting important assets on the .5gb. Nvidia isn't known for there great long term driver support (Kepler is in legacy support already and is getting beaten by 7970 in DX12/Vulkan). This will become even more of an issue in the near future as games are using more and more vram and maxwell gets moved to legacy support. Nvidia also lied about native dx12 support but that is a discussion for another time (async driver promised over a year ago/performance regression on DX12 because it is only software and the entire point of DX12 is the performance boost).

my friend has a GTX 970M. is that covered too or no

https://i.imgur.com/1nxiSju.jpg
Amazon offering 20% refund.

#8
The driver hides it fairly well, but yes, it's still terrible in games that just grab all the VRAM they can. The funny thing is I think it would work best with not all of those 0.5GB disabled. Windows/Desktop need a bit as well, so having an extra 256MB for stuff that won't be used anyway is probably beneficial so I can see why they kept it.

The real issue is that both parts can't be accessed simultaneously. Sure read on one, write on the other works fine, but not reads or writes on both at the same time. If that weren't the case no one would notice. Normal usage 7/8 of the 980's bandwidth, in line with the 970's performance. Absolute worst case the 0.5GB swapping at full speed it steals you the same bandwidth from the fast section so you're still at 6/8=3/4. The way it is as soon as something needs to be swapped half your bandwidth is gone.

To be honest though on 1080p any halfway decent game won't need even 3GB. An those that do wouldn't run at acceptable framerates no matter how much VRAM the 970 had. I just don't understand why those people that complain about "3.5GB isn't enough, you need at least 4GB" didn't just get a 390 in the first place.

There's also two sides to this. From an engineering standpoint it's pretty impressive and probably got them really good yields.
From a business perspective it is either a embarassing fuck up, if you actually believe that nVidia's PR and marketing team are that incompetent or horribly anti consumer if they tried to hide it.

Now the rest is just you complaining about DX12 support which you apparently still don't understand.
nVidia does support DX12 natively. They even support some features that AMD doesn't support yet. Fun fact: Skylake iGPUs have the most comprehensive DX12 support, doesn't change the fact that they are still shit. There is nothing evil going on there. Yet.
So nVidia cards do have native DX12 support, no problems here.
They only thing where bullshit happens is Async Compute. Async Compute is not required for DX12. However it is a very nice feature that can get you a performance boost. Now the thing is Maxwell does not support AC in hardware. They wrote a weird software work around but it's bullshit since it actually costs performance. So any developer worth their salt (which they if they use AC) won't use AC on Maxwell, even though nVidia supports it. This is the kind of hair-splitting marketing loves. They support it but it doesn't help at all. But they can rightfully say they support it.

Now onto Pascal, where it gets interesting. If I recall correctly Maxwell lost about 1% performance with AC, Pascal gains 2%. Here's the thing: Pascal supports AC in hardware. This is as good as it'll get without significant architectural changes. nVidia's architecture has a limitation that really screws up AC: Each GPC can only work on one problem at a time. That means if there's only one GPC (the coming GP107 and GP108) then AC is completely useless. If there's only 2 like on the GP106 (GTX 1060) then half of the GPU has to work on the same thing. That's terrible granularity. You can see why they didn't bother implementing AC in hardware before DX12 was released.
nVidia already gets really good utilization so they probably figure that adding ACEs wasn't worth it. Simply using that space for more units gives them better performance even without AC.

They still get a performance boost from DX12, just not much from AC. The lower CPU overhead is also still there, just like on AMD.

So apart from marketing bending the truth (which you should expect) or in the case of the 970 either fucking up or outright lying, both of which are unacceptable, nVidia did nothing wrong.

However there's still reasons why you should buy an AMD card instead if there is one that is competitive.
1. If the cards were virtually interchangeable you should get the AMD card. Why? To raise their market share. Look at the Titans. Look at the Founders Editions. This is the kind of price gouging that happens because is so close to a monopolistic position. That's just business. If you let them jack up the prices they will.
2. AMD isn't financially well off. That means they aren't developing many new architectures. GCN is 5 years old at this point and is going to stay for a while. So a lot of the driver improvements for new cards still trickle down to the old cards. Both AMD and nVidia will only optimize for the newer cards, nothing wrong with that. But on AMD you still get some improvements on older cards anyway.
3. AMD drivers are terrible on launch. How is that a good thing? Well it means that if a card is just as fast as an nVidia card now (and priced accordingly) it will be faster in the long run. Because the optimizations over time don't yield nearly as good results on nVidias side, since those drivers are already pretty good.
4. DX12 / Async Compute and most importantly Vulkan. AMD cards can generally do a lot more than you can see at first because compilers don't even know these things exist and therefore don't use it. So low level APIs that do know about them automatically mean a significant performance boost. AC also helps with utilization.

2-4 are basically all about the same thing. AMD cards are technically sold "below value". They are priced to compete now, but they get faster.

Keep in mind all of this only matters if the cards are close. Same price nVidia 5% faster? I'd say AMD hands down. 30% faster? Yeah, not happening, sorry AMD.

tl;dr
You're attacking nVidia for the wrong reasons.
AMD is the better deal most of the time though anyway.

EDIT:
#9
Nope, they cut off all that shit on the 970M. Only 3GB (or 6GB) so everything working as expected.

Is someone just gonna bump this thread when we have a way to get the money or