AMD's Catalyst 15.10 Beta Optimizes 'Ashes Of The Singularity' DX12 Performance

[N.Broekhuijsen]

AMD's latest Beta Catalyst driver optimizes performance for "Ashes of the Singularity," and it comes with a couple of other fixes.

AMD's Catalyst 15.10 Beta Optimizes 'Ashes Of The Singularity' DX12 Performance : Read more

 

[turkey3_scratch]

Ashes of Singularity already has proved to perform excellently on AMD GCN cards, so with these optimizations they could even see larger FPS boosts.

 

[hammereditor]

AMD doesn't need drivers to perform well at DX12 async shaders.

 

[Johan Kryger Haglert]

I remember crashed drivers from Radeon 9600 days, is that something one still have to live with with AMD cards?

 

[salgado18]

I remember crashed drivers from Radeon 9600 days, is that something one still have to live with with AMD cards?

Short answer: no.

Long answer: it may happen with either Nvidia or AMD drivers, but it's like a 1% chance for either. This talk of "AMD drivers are bad" or "Nvidia drivers are bad" is in the deep past. Both have the same average quality, and most of the times, errors happen due to individual hardware and/or software configurations.

It's like saying: in the 90's, Fiat brought to my country the Tipo. Some cases of Tipos catching fire expontaneously came out, so the car was badly seen, and sold very little. No other cases of the type happened in any other model or brand (including Fiat) since then, so we can't just say "Fiat cars catch fire".

 

[Alec Mowat]

I remember crashed drivers from Radeon 9600 days, is that something one still have to live with with AMD cards?

Short answer: no.

Long answer: it may happen with either Nvidia or AMD drivers, but it's like a 1% chance for either. This talk of "AMD drivers are bad" or "Nvidia drivers are bad" is in the deep past. Both have the same average quality, and most of the times, errors happen due to individual hardware and/or software configurations.

It's like saying: in the 90's, Fiat brought to my country the Tipo. Some cases of Tipos catching fire expontaneously came out, so the car was badly seen, and sold very little. No other cases of the type happened in any other model or brand (including Fiat) since then, so we can't just say "Fiat cars catch fire".

Not sure I agree with this. AMD drivers are still bad. Crossfire is buggy compared to Nvidia SLI. If there is an issue, expect months before the next driver release.

And for DX12, one game doesn't cut it for me
http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

It's like saying: in the 90's, Fiat brought to my country the Tipo. Some cases of Tipos catching fire expontaneously came out, and now they own Chrysler. So I'll never buy a Chrysler, or a Fiat, because they don't perform as well as the competition.

 

[Alec Mowat]

I remember crashed drivers from Radeon 9600 days, is that something one still have to live with with AMD cards?

In that sense, it's been 10 years and Windows is a lot more stable than it used to be. I wouldn't necessarily blame the drivers for everything that happened back then.

 

[alextheblue]

And for DX12, one game doesn't cut it for me
http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

From the first page of that article you linked to:

The software provided to us is a prerelease version of Fable Legends, with early drivers, so ultimately the performance at this point is most likely not representative of the game at launch and should improve before release. What we will see here is more of a broad picture painting how different GPUs will scale when DX12 features are thrown into the mix. In fact, AMD sent us a note that there is a new driver available specifically for this benchmark which should improve the scores on the Fury X, although it arrived too late for this pre-release look at Fable Legends

Basically those numbers aren't very useful at this point except to compare Nvidia to Nvidia and AMD to AMD. After launch it will be interesting to take another look.

 

[Johan Kryger Haglert]

Short answer: no.

Long answer: it may happen with either Nvidia or AMD drivers, but it's like a 1% chance for either. This talk of "AMD drivers are bad" or "Nvidia drivers are bad" is in the deep past. Both have the same average quality, and most of the times, errors happen due to individual hardware and/or software configurations.

It's like saying: in the 90's, Fiat brought to my country the Tipo. Some cases of Tipos catching fire expontaneously came out, so the car was badly seen, and sold very little. No other cases of the type happened in any other model or brand (including Fiat) since then, so we can't just say "Fiat cars catch fire".

No, the driver talk is NOT deep.

I've read some likely fairly popular description of AMD vs Nvidia drivers and tolerance and I've used Linux since 1996 or something and I know the Nvidia drivers has been way superior there and they also supported Solaris and FreeBSD.

Also I've got the impression Nvidia is quicker to modify their drivers for specific games.

And I for sure didn't had those crashing issues that my friend had on my Nvidia card. I've never had problems with them (I've had with Noveau drivers but that's not Nvidias drivers.)

Now with the asynchronous-computing issues for the Maxwell-cards I guess it's too easy and likely unfair to blame shitty drivers for the DX11-performance with AMD cards relative the performance using DX12, Mantle or Vulkan. I don't know how good AMD OpenGL-drivers has been on Windows.

It was an honest question from me though because I think the R9 380 especially hold a good spot against the GTX 960 currently and with the process shrink next year and twice the performance / watt for AMD cards as-well things should go better for AMD and in my case it still have the Linux drivers issue (well, unless Vulkan fixes that too and X/whatever starts to use that) but IF there was a lot of issues with crashing drivers or poor performance when it came to new titles that would be reconsider the cards anyway because I don't want poor drivers.

 

[none12345]

/shrug my current system amd cpu+gpu almost never crashes(every 3-4 years or so).

When it does crash, its not the drivers, or any other software. It means the heat sinks have wadded up with lint/dust and they need to be cleaned. As soon as take the system and blow out the dust, it wont crash again for another 3 or so years.

Assuming you didnt buy low level budget hardware, most systems are pretty darn stable these days. Crashes are very rare on any of the systems i own or use normally. If a system does crash its almost always from a power spike(lights flicker, on a system without a UPS), or its a thermal problem(fans stopped, heat sinks dusty)

If your gpu is crashing in games constantly, at least in mainstream games, its likely because of that overclock you think is stable really isnt stable. Its almost certainly a thermal problem not the drivers. Or in the case of a cheap power supply it could be unstable voltages at high load. One of those 2 things are almost always the case.

That said, i also have noscript on by default in my browser. And im blocking a few million ip address of ad sites, china sites, etc, etc. All those ones who like to push addware/spyware/virsues/etc on you. So i dont have a problem with that causing crashes.

 

[Alec Mowat]

And for DX12, one game doesn't cut it for me
http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

From the first page of that article you linked to:

The software provided to us is a prerelease version of Fable Legends, with early drivers, so ultimately the performance at this point is most likely not representative of the game at launch and should improve before release. What we will see here is more of a broad picture painting how different GPUs will scale when DX12 features are thrown into the mix. In fact, AMD sent us a note that there is a new driver available specifically for this benchmark which should improve the scores on the Fury X, although it arrived too late for this pre-release look at Fable Legends

Basically those numbers aren't very useful at this point except to compare Nvidia to Nvidia and AMD to AMD. After launch it will be interesting to take another look.

That's been AMD's excuse for the last decade, "AMD sent us a note that there is a new driver available specifically for this benchmark"

Just doesn't work in the real world. I'll never recommend AMD, unless at some point in the future, your games run on unfulfilled promises and false hopes. But you can probably just get an Nvidia card and see the performance now.

 

[JackOfAllBlades]

Short answer: no.

Long answer: it may happen with either Nvidia or AMD drivers, but it's like a 1% chance for either. This talk of "AMD drivers are bad" or "Nvidia drivers are bad" is in the deep past. Both have the same average quality, and most of the times, errors happen due to individual hardware and/or software configurations.

It's like saying: in the 90's, Fiat brought to my country the Tipo. Some cases of Tipos catching fire expontaneously came out, so the car was badly seen, and sold very little. No other cases of the type happened in any other model or brand (including Fiat) since then, so we can't just say "Fiat cars catch fire".

No, the driver talk is NOT deep.

I've read some likely fairly popular description of AMD vs Nvidia drivers and tolerance and I've used Linux since 1996 or something and I know the Nvidia drivers has been way superior there and they also supported Solaris and FreeBSD.

Also I've got the impression Nvidia is quicker to modify their drivers for specific games.

And I for sure didn't had those crashing issues that my friend had on my Nvidia card. I've never had problems with them (I've had with Noveau drivers but that's not Nvidias drivers.)

Now with the asynchronous-computing issues for the Maxwell-cards I guess it's too easy and likely unfair to blame shitty drivers for the DX11-performance with AMD cards relative the performance using DX12, Mantle or Vulkan. I don't know how good AMD OpenGL-drivers has been on Windows.

It was an honest question from me though because I think the R9 380 especially hold a good spot against the GTX 960 currently and with the process shrink next year and twice the performance / watt for AMD cards as-well things should go better for AMD and in my case it still have the Linux drivers issue (well, unless Vulkan fixes that too and X/whatever starts to use that) but IF there was a lot of issues with crashing drivers or poor performance when it came to new titles that would be reconsider the cards anyway because I don't want poor drivers.

I know of a lot of people who have no issues using their AMD cards on Linux, I use mine mainly on windows ( Crossfire R9 290). I haven't had a problem with any drivers that AMD has released, I have been using AMD drivers over the last 5 years. The only issues I have ever had is games not properly supporting crossfire, War Thunder for example.

 

[Alec Mowat]

AMD drivers in Linux are not great, they don't spend a lot of time developing them. There isn't much GUI either if you need to make modifications to the desktop configuration. It's command line heavy still.

Nvidia is miles ahead in Linux, and that's been known for awhile.

 

[Bloob]

I remember crashed drivers from Radeon 9600 days, is that something one still have to live with with AMD cards?

Sometimes yes, although I've had it happen with nVidia drivers as well. The thing is, with DX12 and Vulkan, the control is mostly relinquished to application developers, instead of the driver, which is likely to result in more crashes, luckily operating systems nowadays can usually recover from that.

edit: should point out that "crashing" here refers to the driver crashing, not necessarily a game, although likely, or operating system crashing

 

[Johan Kryger Haglert]

It's been known ~forever.

Nvidias Linux drivers has likely been good for like 15 years or so.

The AMD ones are getting better and the open-source variant is likely still better than the Nvidia one but I don't care for that one.

Just because of that the AMD drivers have improved CS:GO performance by a lot at-least:
http://www.phoronix.com/scan.php?page=article&item=amd-catalyst-159&num=2

Whatever it's competitive with Nvidia I don't know.

Nah, 120 FPS with a 5960x and a R9 290 or Fury doesn't sound all that impressive.

 

[blazorthon]

That's been AMD's excuse for the last decade, "AMD sent us a note that there is a new driver available specifically for this benchmark"

Just doesn't work in the real world. I'll never recommend AMD, unless at some point in the future, your games run on unfulfilled promises and false hopes. But you can probably just get an Nvidia card and see the performance now.

Something else that doesn't work in the real world is making recommendations towards AMD or Nvidia based on false claims regarding unfulfilled promises and false hopes and using far from release software as a counter-argument, granted I will agree that one game is too few to base our thoughts on DX12 performance.

Saying that AMD is giving unfulfilled promises and false hopes regarding AMD giving a note about a new driver that works better means you are refuting AMD's claim about the new driver being better. If the new driver is already available, then it isn't a false hope or an unfulfilled promise unless the new driver doesn't work as AMD says or at least does not improve the performance/stability well enough.

Without testing the new driver, the only conclusion you can make is that the driver should have been out slightly sooner so that the review of what it fixes could have used it. This is still a weak argument against AMD when the new driver is meant to fix pre-release software. Yes, it is annoying for devs and beta/alpha testers, but it is not representative of how well things actually work for the customers of the released game and that is what matters for someone looking to buy a card to play the game. Most people aren't looking to buy cards based on performance in a game's beta with what is now an older driver.

 

[Alec Mowat]

Saying that AMD is giving unfulfilled promises and false hopes regarding AMD giving a note about a new driver that works better means you are refuting AMD's claim about the new driver being better.

Do you have the benchmarks?

The driver is still beta, and will be for another month. Another typical AMD move to get out of accountability for any driver issues they get.

 

[blazorthon]

Do you have the benchmarks?

The driver is still beta, and will be for another month. Another typical AMD move to get out of accountability for any driver issues they get.

No, I don't have any benchmarks for it. That's why I didn't confirm nor deny it. You denied it and that requires benchmarks.

Yes, the driver spoken of in this article is in beta. This driver that was mostly to improve performance in DX 12 software that is also in beta. Said software was already not poorly performing even in your worst case example link. Where are you going with this?

Even most of the other fixes made were for prerelease software (some of which was mentioned as having work done on in Nvidia's latest driver on nvidia.com) or niche situations. Furthermore, it isn't like Nvidia doesn't use betas either, so why it it bad when AMD uses them?

 

[Alec Mowat]

Do you have the benchmarks?

The driver is still beta, and will be for another month. Another typical AMD move to get out of accountability for any driver issues they get.

No, I don't have any benchmarks for it. That's why I didn't confirm nor deny it. You denied it and that requires benchmarks.

Yes, the driver spoken of in this article is in beta. This driver that was mostly to improve performance in DX 12 software that is also in beta. Said software was already not poorly performing even in your worst case example link. Where are you going with this?

Even most of the other fixes made were for prerelease software (some of which was mentioned as having work done on in Nvidia's latest driver on nvidia.com) or niche situations. Furthermore, it isn't like Nvidia doesn't use betas either, so why it it bad when AMD uses them?

Nvidia releases updated drivers almost weekly. There's a huge difference in the quality of support, and the ease of use of Nvidia drivers. Not to mention, they don't include bundle Raptr software.

Well, maybe in 3 months when AMD makes another driver.. we can see some benchmark improvements in another game...
This is a laughable defense of AMD. Fan bois need to give it up already.

 

[turkey3_scratch]

While Nvidia driver support is better, that should not be the primary determining factor in purchasing the GPU. Any games older than 3 months will have optimized drivers, and any game that is younger than 3 months on PC sucks anyway from bad optimization on Nvidia and AMD.

 

[Alec Mowat]

While Nvidia driver support is better, that should not be the primary determining factor in purchasing the GPU. Any games older than 3 months will have optimized drivers, and any game that is younger than 3 months on PC sucks anyway from bad optimization on Nvidia and AMD.

Actually, it would be bad optimization for 1 week for Nvidia, and 3 months to forever, depending on if AMD ever contacts the company to fix the issues.

 

[alextheblue]

That's been AMD's excuse for the last decade, "AMD sent us a note that there is a new driver available specifically for this benchmark"

Just doesn't work in the real world. I'll never recommend AMD, unless at some point in the future, your games run on unfulfilled promises and false hopes. But you can probably just get an Nvidia card and see the performance now.

Great, you're biased, we get it. Even so, basing your "analysis" on alpha-stage games is asinine. Once we get more DX12 titles released we'll be able to see better comparisons. Anyway, as I've said before, the relative performance of Nvidia and AMD cards in DX12 will depend on how heavily the developers use async shaders. If they use them lightly, both vendors should do well enough.

 

[blazorthon]

Nvidia releases updated drivers almost weekly. There's a huge difference in the quality of support, and the ease of use of Nvidia drivers. Not to mention, they don't include bundle Raptr software.

Well, maybe in 3 months when AMD makes another driver.. we can see some benchmark improvements in another game...
This is a laughable defense of AMD. Fan bois need to give it up already.

Well let's see. Going to nividia.com and doing a search for windows 10 x64 drivers for the 900 series, there are 6 WHQL and 1 beta. Ignoring the beta because you hate betas, there was an average 24.167 days between releases with the minimum days between releases being 15.

Raptr software is optional.

Is AMD slower with the drivers? Yes. Their last full release was about two and a half months ago (in comparison to Nvidia's longest recent breaks between releases being 38 days closely followed by 37 days). Does Nvidia have new drivers anywhere near weekly? Well, their website says that's more like every three and a half weeks, so I'd say no, Nvidia is nowhere near weekly.

Laughable defense? I'm not defending AMD against you. I'm merely pointing out fallacies in your argument. You can call me an AMD fanboy, but the majority of my recommendations this generation have been Maxwell cards and you can see that in the forums.

 

[Alec Mowat]

Well let's see. Going to nividia.com and doing a search for windows 10 x64 drivers for the 900 series, there are 6 WHQL and 1 beta. Ignoring the beta because you hate betas, there was an average 24.167 days between releases with the minimum days between releases being 15.

Actually, the nvidia experience just notifies you, you open it, click "download driver", and hit express, reboot. Done.

Also, all Nvidia cards use the same drivers. Unless you need legacy drivers, 64 bit of any card on your current windows version works fine.

You're explaining the AMD method. Except that the last driver I had from them was in July, and there still isn't a official release yet.

 

[blazorthon]

Actually, the nvidia experience just notifies you, you open it, click "download driver", and hit express, reboot. Done.

What does this have to do with anything we've talked about? Are you saying Nvidia Experience has drivers that are different from geforce.com?

Also, all Nvidia cards use the same drivers. Unless you need legacy drivers, 64 bit of any card on your current windows version works fine.

Nvidia lists in some of the drivers' information specific cards the driver was made to add support for. For example, 355.98 starts by saying that it is meant to support the notebook GTX 980 (not GTX 980M). I'm not saying an older driver won't work, but if it worked as well as the new one made for the card, why would Nvidia state that the new driver is made for the card?

In what way do Nvida's drivers have an advantage over AMD's drivers in what cards they support? Like Nvidia, AMD's latest drivers support all of AMD's cards from the same generations. In fact, since they support the Radeon 5000 family, they support cards that are up to about six months older than the oldest cards supported by Nvidia's current driver.

You're explaining the AMD method. Except that the last driver I had from them was in July, and there still isn't a official release yet.

Where am I "explaining the AMD method"? What do you mean there still isn't an official release? July did have an official release (which I already mentioned in my previous post) and this beta driver is only three days old. Nvidia's beta driver on their site didn't get an official release for over two weeks (16 days to more more precise).