Support NeoGAF

[DPB]

I haven't seen the kepler update for Witcher, are you sure it's even been released?

It was the 353.06 drivers.

"Implements optimizations and bug fixes that provide increased performance for Kepler-based GPUs."

I haven't tested them myself, but several people have reported some substantial improvements in The Witcher 3 performance thread.

 

[ss_lemonade]

Didn't Nvidia acknowledge gimped Kepler performance and promise a patch?

Don't think it did anything.

I didn't really see much of a difference with the new driver, but I also didn't really do any scientific testing. I think this was just for damage control, especially with all those kepler witcher 3 discussions in the geforce forums (which seems to have spawned a billion other topics with people complaining about everything they can think of)

 

[ZOONAMI]

As stated above - AMD takes more time to complete the drivers. Nvidia cards are not aging at all, this has nothing to do with Nvidia, what a misleading title!

AMD just aren't ready on launch

Um no. The amd drivers are good to go on release of a new card, they just make better optimizations over time for older cards vs nvidia. Nvidia doesn't focus on driver optimimizations for older cards because they want you to buy their new cards. It increases the perception of performance increase over their last gen cards. Makes a lot of sense from a business perspective, but is a bummer for people who have last gen cards.

 

[The Llama]

No, I'm pretty sure we all have that understanding.

They're saying that ^ happens because AMD is still rocking GCN for 3 generations and take longer to optimize while Nvidia is on a different architecture and has been focusing on the newer.

But that doesn't make any sense, because in older games they were just as well optimized.

 

[Yurikerr]

Both opinions that Nvidia is sabotaging older gpu owners and that amd drivers are the ultimate shit at launch look really silly.

Of course amd performance will improve as they keep working on the same architecture for so much time.

I don't know why everyone insists that only Nvidia have good software engineers.

 

[mugurumakensei]

Probably a combination of things.

First, Nvidia has better drivers in general compared to AMD so less room for improvement.
Second, Nvidia probably (or rather definitely does. see witcher 3) prioritizes improving performance for new cards when it comes to newer releases. In some cases, they may go back and improve performance for older cards later as was the case with witcher 3. However, you should expect that to be the exception rather than the norm as budgets and time are limited so focusing on current cards along with driver development for new cards is the most sane and profitable thing to do.

 

[Älg]

I'm a bit confused here are we saying that Radeon GPU's get more powerful over time? Or even weirder are we saying that Nvidia GPU's get weaker over time?

It's an issue with rotational velocidensity.

GTX 780
Lead over 290X = -7% (Oct 23, 2013)
Lead now = -16.3%
290X lead increased over 100%.

According to the graph it's only 10 percentage points now; a 3 percentage point difference it pretty much within margin of error. Making the number in percentages rather than percentage points makes them appear a lot bigger than they are.

 

[QaaQer]

No, not really. It is just that their drivers improve everything back to the 7xxx series, while NVIDIA generally only improves things for the current architecture.

and that makes amd drivers better: they improve more things.

 

[Serick]

But that doesn't make any sense, because in older games they were just as well optimized.

Which part doesn't make sense? 15.0 - 15.5 driver updates from AMD impact every single GCN card because they don't have any new architecture.

340-353 driver updates from Nvidia focused on Maxwell performance.

Kepler didn't suddenly start performing worse in your games 1-5.... they just weren't the area of focus for games 6-10.

It was the 353.06 drivers.

"Implements optimizations and bug fixes that provide increased performance for Kepler-based GPUs."

I haven't tested them myself, but several people have reported some substantial improvements in The Witcher 3 performance thread.

That's good to hear, I didn't pay attention to the release notes :3

But still, I don't think anyone here is disputing that Nvidia is putting less effort into improving Kepler than they are Maxwell.... We're just saying AMD users lucked out because the architecture hasn't changed in 3(or 4?) years.

 

[OmegaDL50]

I'm a bit confused here are we saying that Radeon GPU's get more powerful over time? Or even weirder are we saying that Nvidia GPU's get weaker over time?

I think it's more about cutoff features in driver support.

Look at the The Witcher 3 for example, Nvidia pushed the hell out of Maxwell with this game, but when it came to Kepler, it wasn't given the same regard.

AMD is also guilty of this as well for example VSR the native downsampling functionality which AMD claimed was to be a feature of the R9 series due to a supposed hardware scaler was proven false when the R9 280 was stated to support this feature too, which was a nothing more then a rebranded HD 7950.

Someone did some snooping around and found out this VSR was nothing more than a driver level based downsampling and had nothing to do with a hardware scaler. AMD basically cut off the 7000 series of cards of this feature just to push people towards getting the R9 200 series.

Don't be surprised when R9 300 series comes out with some new feature, some of them will be missing from the driver support for the R9 200 series, just the same when Nvidia releases Pascal it will also push features that won't show up in the driver code for the Maxwell line.

Yes I agree it's misleading and I feel this is not the right way to go about pushing people to getting a newer card, if anything it pushes me away.

Cutting support and features at a driver level is the worst thing a GPU maker can do.

 

[wachie]

But according to the graph it's only 10 percentage points now; a 3 percentage point difference it pretty much within margin of error.

Math, how does it work?

Spoiler

10/61 = 16.3%

 

[Nikodemos]

Um no. The amd drivers are good to go on release of a new card, they just make better optimizations over time for older cards vs nvidia. Nvidia doesn't focus on driver optimimizations for older cards because they want you to buy their new cards. It increases the perception of performance increase over their last gen cards. Makes a lot of sense from a business perspective, but is a bummer for people who have last gen cards.

Pretty much. By deliberately not 'localizing' some of the improvements to their older cards, they (artificially) increase the perceived performance difference between generations.

The problem occurs when people smashed a cool grand on what used to be the absolute best in the world and now get roughly the same performance as somebody with a ~$400 card from the competition. It's unsurprising they get somewhat annoyed.

 

[QaaQer]

I don't know why everyone insists that only Nvidia have good software engineers.

I don't know that they do. but there does seem to be a "amd products are total shit and everyone who works there is incompetent so only buy nvidia and intel" prejudice. But whatever, people gonna people.

 

[Skyzard]

I haven't seen the kepler update for Witcher, are you sure it's even been released?

They mentioned something kerpler with a recent release, jack all difference. As another user said, their attempt at damage control and they didn't give a shit.

Second, Nvidia probably (or rather definitely does. see witcher 3) prioritizes improving performance for new cards when it comes to newer releases. In some cases, they may go back and improve performance for older cards later as was the case with witcher 3. However, you should expect that to be the exception rather than the norm as budgets and time are limited so focusing on current cards along with driver development for new cards is the most sane and profitable thing to do.

What that suggests is with AMD you get better long term value with performance cards for a £500 purchase, without all the benefits just in the first year....?

But go Nvidia if you plan to update every year.

 

[kitch9]

Fiji's architecture looks largely based on the GCN, so what you are hoping for (radical change and poor aging over time) doesnt seem likely. Time will tell.


Are you the AMD equivalent of Unknown Soldier? Or is this sarcasm?

I've got a 780 6gb it's very evident that Nvidia are only competing with themselves at the minute.

The best way to compete with yourself to push your brand forward is to make your old stuff look comparatively shit.

Which is what they are doing, subtly, but still, they are.

 

[thelastword]

I think for the most part (I could be very wrong) on paper AMD cards have always been superior?

Comparatively yes, price to performance, AMD always had the better performance, but software (drivers) have let AMD down quite a bit.

 

[Sijil]

I don't know that they do. but there does seem to be a "amd products are total shit and everyone who works there is incompetent so only buy nvidia and intel" prejudice. But whatever, people gonna people.

Considering the latest debacle with Project Cars and AMD drivers to blame, and in nearly every game on launch performing poorly in comparison to their Nvidia counterparts one cannot help but feel that AMD is lacking in their driver department.

 

[Älg]

Math, how does it work?

Spoiler

10/61 = 16.3%

What I'm trying to say is that the actual change in performace is very small. It could just as well be withing margin of error in their testing methods.

 

[Serick]

I've got a 780 6gb it's very evident that Nvidia are only competing with themselves at the minute.

The best way to compete with yourself to push your brand forward is to make your old stuff look comparatively shit.

Which is what they are doing, subtly, but still, they are.

Yeah, the "980ti is 3x faster than the 680ti" advertising they did kind of struck me as odd. They didn't mention performance vs. a single AMD card.

Considering the latest debacle with Project Cars and AMD drivers to blame, and in nearly every game on launch performing poorly in comparison to their Nvidia counterparts one cannot help but feel that AMD is lacking in their driver department.

It took AMD 12+ months to see these kind of performance gains on GCN? If that performance was there all along held back by shitty drivers doesn't that point to AMD's software engineers being kind of meh?

 

[kitch9]

Considering the latest debacle with Project Cars and AMD drivers to blame, and in nearly every game on launch performing poorly in comparison to their Nvidia counterparts one cannot help but feel that AMD is lacking in their driver department.

There's two horses in the race, one is lame and has lost its rider....

The other isn't lame, hasn't lost its rider and doesn't feel bad about cheating to increase the margin of its win in the belief that the competition will be sent to the glue factory

PC graphics market in a nutshell.

Cpu market is similar only one of the one in the lead is deliberately slowing down so to not get too far in front too quickly.

 

[wachie]

What I'm trying to say is that the actual change in performace is very small. It could just as well be withing margin of error in their testing methods.

How is it small? The deficit is basically doubled, it's well out of the "margin of error" lol

 

[Sephiroth_VII]

I know for a fact that new drivers gimp older cards. Back when I was still using a GTX9800, I played a lot of warband. I didn't update my driver for years, and the game ran great except for occasional slowdowns on the world map. Thinking that 2-3 years of driver development must have led to SOME progress, I finally updated my driver, only to find that performance dropped like a rock. The game was completely unplayable.

I downgraded the driver by using the xD I got with the card, and performance was back to normal. I tried slightly older versions of the driver too: Performance was never as good as with the original driver. Since then, I've stopped upgrading my GPU driver once my card is no longer the latest and greatest.

 

[kitch9]

Man, I got my 780Ti for 1440p gaming, but it's starting to look like I might have to get the 980Ti in the next few months lol

This is the reason Nvidia are doing what they are doing.

 

[Skyzard]

This is the reason Nvidia are doing what they are doing.

Yep.

 

[hesido]

I think you're both misunderstanding the problem. I'll try to explain it. The claim is this: Let's say, on equivalent settings, GPU X (Kepler) from nVidia and GPU Y (GCN) from AMD average 60 FPS in games 1-5. Then nVidia releases GPU Z (Maxwell). In games 1-5, GPU Z average 70 FPS. Then games 6-10 launch. On average (assuming all are running the same settings), GPU X averages 50 FPS, GPU Y averages 60 FPS, and GPU Z averages 70 GPS. Owners of GPU X are now complaining "Hey, I used to average the same FPS as owners of GPU Y, but now that GPU Z launches, we get less."

Ok, that was probably not very clear XD, but I tried...

Hmm, that's a bit "shady".

 

[nubbe]

Would not be surprised if Nvidia throttle performance on older cards too when new are released

Nvidia is a scumbag company

 

[napata]

Considering the latest debacle with Project Cars and AMD drivers to blame, and in nearly every game on launch performing poorly in comparison to their Nvidia counterparts one cannot help but feel that AMD is lacking in their driver department.

Project Cars also runs like shit on Kepler and Nvidia never fixed it. A 960 outperforms a 780 for example.

 

[spuckthew]

I'm still rockin a 680 like a peasant

I don't have the cash to upgrade mine

I could probably get around £100 for my MSI Lightning easy and then get a second-hand GTX980 or something (more bang for buck that way).

 

[Skyzard]

Project Cars also runs like shit on Kepler and Nvidia never fixed it. A 960 outperforms a 780 for example.

Dat tessellation excuse, good for everything. I wonder what Pascal's will be. Even faster tessellation.

 

[OmegaDL50]

Project Cars also runs like shit on Kepler and Nvidia never fixed it. A 960 outperforms a 780 for example.

This is the kind of shit Class action lawsuits are made from.

Nvidia intentionally decreasing performance of an older card through drivers just to incentivise people to upgrade.

It's bullshit.

 

[Octavianus]

Hmm... so you're saying I should upgrade my 660?

 

[kitch9]

Would not be surprised if Nvidia throttle performance on older cards too when new are released

Nvidia is a scumbag company

I'm a Nvidia owner.

They are dirty business people alright, I agree.

Not a sniff of morals in the entire damn brand, unfortunately those practices mean they won.

It's how it is.

 

[Serick]

This is the kind of shit Class action lawsuits are made from.

Nvidia intentionally decreasing performance of an older card through drivers just to incentivise people to upgrade.

It's bullshit.

Where is the evidence showing that they intentionally decreased performance?

Not spending time on optimizing drivers for a new release on their older architecture isn't the same thing as purposely gimping performance.

 

[kitch9]

This is the kind of shit Class action lawsuits are made from.

Nvidia intentionally decreasing performance of an older card through drivers just to incentivise people to upgrade.

It's bullshit.

They don't have to decrease anything. They just have to drop support and not look for optimisation in new games.

Boom, give me 650 dollar for a video card that's essentially the same as the last one and not as much ram as you think.

Thanks.

 

[Skyzard]

Not spending time on optimizing drivers for a new release on their older architecture isn't the same thing as purposely gimping performance.

Performance wise, it turns out to be the same, so what's the difference for the consumer?

They shouldn't expect the company they brought from to actively work against them, just that the service they provide to get it up to scratch expires after 1 year, and it ends up as the same thing.

 

[kitch9]

Performance wise, it turns out to be the same, so what's the difference for the consumer?

A lot of dollars.

 

[thelastword]

My Titan is holding up pretty well, after all.

Lately I have been playing AC unity and I must say that I'm impressed by the performance I get: everything maxed at 1440p (2xmsaa)I get frames from 30 to 45.

Cutscenes sometimes tank a bit in the 15-20 region but gameplay is perfectly fine thanks to gsync.

Still, I already ordered a 980ti xD

I wouldn't call that holding up, what is in these cutscenes that require such an expensive GPU to be at 15-20fps.....You're not even running the game at 4k.

 

[Serick]

Performance wise, it turns out to be the same, so what's the difference for the consumer?

They shouldn't expect the company they brought from to actively work against them, just that the love expires after 1 year, and that's the same thing.

The difference to the consumer is that it doesn't warrant a class action lawsuit, which is what was mentioned in the post I quoted and responded to.

If you can't show proof that the 340.52 Drivers (last pre-maxwell drivers) were substantially superior in performance to the Maxwell release drivers Nvidia didn't "purposely decrease performance" of Kepler. You're also assuming that Kepler has room for improvement which has yet to be shown by anyone.

You guys keep painting AMD in this "pro-consumer light" like they would never do anything similar but that's kind of hard to say when they've been rocking GCN for so long. I highly doubt AMD would keep supporting GCN the way they do now if they released a brand new architecture. Hell they can barely keep their drivers up with current game releases.

 

[LQX]

More like being intentionally crippled. Sucks.

 

[Seanspeed]

Post-release performance increases via driver updates. Seems like AMD upgrades their performance more than Nvidia.

Or just had more room for improvement.

 

[Skyzard]

If it's with new games how are you going to show performance improvements on the same card. You have to rely on other branded cards and compare performance differences.

 

[Serick]

If it's with new games how are you going to show performance improvements on the same card. You have to rely on other branded cards and compare performances differences.

Again, you're assuming Kepler has room for improvement. This has yet to be demonstrated.

 

[OmegaDL50]

The difference to the consumer is that it doesn't warrant a class action lawsuit, which is what was mentioned in the post I quoted and responded to.

If you can't show proof that the 340.52 Drivers (last pre-maxwell drivers) were substantially superior in performance to the Maxwell release drivers Nvidia didn't "purposely decrease performance" of Kepler.

You guys keep painting AMD in this "pro-consumer light" like they would never do anything similar but that's kind of hard to say when they've been rocking GCN for so long. I highly doubt AMD would keep supporting GCN the way they do now if they released a brand new architecture. Hell they can barely keep their drivers up with current game releases.

Where is anyone saying about AMD being "pro-consumer"

I even criticized AMD also does cut off support for it's drivers in the very first post I made in this thread.

Let's not start with victim blaming with Nvidia here. Both companies have their issues, it's not isolated to just a single manufacturer here.

 

[Skyzard]

Again, you're assuming Kepler has room for improvement. This has yet to be demonstrated.

Kepler's had drivers improve game performance for years now. They reached the 100% cap when maxwell was released huh? As shown by the performance differences...give me a break.

 

[Älg]

How is it small? The deficit is basically doubled, it's well out of the "margin of error" lol

In the real world the difference is a couple of frames at most. If I have 2 dollars and you have 1 dollar, my wealth is 200 % that of yours, this doesn't mean the difference is actually big.

 

[Serick]

Where is anyone saying about AMD being "pro-consumer"

I even criticized AMD also does cut off support for it's drivers in the very first post I made in this thread.

Let's not start with victim blaming with Nvidia here. Both companies have their issues, it's not isolated to just a single manufacturer here.

YOU criticized AMD. People here are using performance gains of GCN as evidence to assume Nvidia is purposely gimping Kepler. Just read the posts. Based on your evidence AMD has been the one purposely gimping their cards with future driver releases, not Nvidia . Outside of moar FPS! what is Kepler lacking feature-wise that Maxwell has?

No I'm assuming the drivers improve performance.

Right, and if Kepler has reached its full potential drivers can't improve performance. Drivers aren't secret magic sauce that gives GPUs infinite scaling possibility. At the end of the day if the hardware/software has hit its peak performance its hit its peak performance.

People in this thread are assuming there is untapped GPU potential in Kepler not being taken advantage of by new driver releases. Ass-u-ming.

Kepler's had drivers improve game performance for years now. They reached the 100% cap when maxwell was released huh? As shown by the performance differences...give me a break.

And if they've had improvements for years is it not logical to think that the timeerformance gain ratio has reached the point where it's no longer worthwhile to invest in?

Again, there literally is not a single person in here saying Nvidia hasn't spent less time developing kepler's performance since Maxwell came out. I (we) are saying Nvidia did not intentionally gimp Kepler in previous and future releases.

I'm also saying AMD would do the same damn thing if they weren't on GCN still.

 

[fizzelopeguss]

I've got a 780 6gb it's very evident that Nvidia are only competing with themselves at the minute.

The best way to compete with yourself to push your brand forward is to make your old stuff look comparatively shit.

Which is what they are doing, subtly, but still, they are.

Market is fucked, AMD lost. what's the marketshare at currently? nearly 80-20? What is the realistic chance of coming back from that to tackle a company that has greater marketing spend AND a larger warchest?

As someone who gamed primarily on Nvidia since 3dfx died, "ATi bad drivers" is the most disgusting and damaging of lies, and i can't believe i ever fell for it. Especially considering the greens have released some absolute dogshit software and hardware over the years. (bumpgate anyone?)

What i really think won it was Physx and the later Pubicworks stuff. PC gamers are batshit OCD when it comes to their precious Ultra settings. Genius marketing.

 

[Skyzard]

Right, and if Kepler has reached its full potential drivers can't improve performance. Drivers aren't secret magic sauce that gives GPUs infinite scaling possibility. At the end of the day if the hardware/software has hit its peak performance its hit its peak performance.

People in this thread are assuming there is untapped GPU potential in Kepler not being taken advantage of by new driver releases. Ass-u-ming.

AMD manages to make their cards more efficient, Nvidia already had max efficiency! Except it's lower now, just in time for their new cards.

Guess I should just stop downloading drivers for my 780ti less than 1 year after purchase too.

 

[Serick]

AMD manages to make their cards more efficient, Nvidia already had max efficiency! Except it's lower now, just in time for their new cards.

Efficiency is not lower. There is absolutely zero proof/evidence that Kepler has been intentionally gimped by driver releases. You're also equating AMD release drivers to Nvidia release drivers. So now AMD's inability to write proper drivers for their own hardware equates to nvidia gimping kepler? That's laughable.

AMD is still on the same architecture. Nvidia is not.

How are you not understanding this?

 

[Skyzard]

Nvidia releases a new line and gets to ditch all previous support and just give us fake support bullshit? Not without me whining about it.

780ti costed over £550 less than a year ago.


And I'm not even salty about that specifically since Amazon is like a bro that secretly loves you.