HIS Radeon HD 6970 2 GB

[Steevo]

He is looking for a benchmark that is repeatable, standard, and clearly defined in measurement.


A DX11 only benchmark will not allow someone to compare how much of a increase a game will get compared to their old 4xxx, or 2xx cards. Currently a upgrade from either company is almost a sideways move in performance. Metro is no more playable on a 580 than a 480, likewise dirt2 is no less playable on a 6970 than a 5870.

A few worthless benchmarks are the only place most users would notice any performance increase.

 

[pantherx12]

He is looking for a benchmark that is repeatable, standard, and clearly defined in measurement.


A DX11 only benchmark will not allow someone to compare how much of a increase a game will get compared to their old 4xxx, or 2xx cards. Currently a upgrade from either company is almost a sideways move in performance. Metro is no more playable on a 580 than a 480, likewise dirt2 is no less playable on a 6970 than a 5870.

A few worthless benchmarks are the only place most users would notice any performance increase.

I didn't mean a paid for benchmark, just meant he seems to want to see more opencl support in applications before commiting to adding it to his benchmarks. ( due to him already doing so many I imagine )

 

[Swamp Monster]

I think wiz is interested in programs you have to pay for in order for it to be worth his time kinda thing.

Seems that way anyway.

I didn't mean it for benchmarks, only for people to know that some apps use OpenCL

For benchmarks I've been using:
http://www.ngohq.com/graphic-cards/16920-directcompute-and-opencl-benchmark.html
Interesting program, but not suitable for TPU reviews I guess.

 

[Super XP]

I know ATI was first, I bought a 2900XT on launch day instead of an 8800 because of it, but that doesn't matter, all that matters is what's available now. Both have audio over HDMI.

I disagree on playback and render quality. They are both equal. I have also seen a lot more complaints about ATI drivers over the past few month than nVidia. Just look at how many unfixed Eyefinity problems there are. Sorry, but I just don't believe your crash report claims. Got proof of them?

And who cares if they use open standards, when there is nothing out there making use of them? OpenCL is all but useless. At least there are actually some apps written for CUDA, and nVidia actually lends a hand to devs to get stuff working on their cards. Can't say the same for ATI. I'd rather have a well supported proprietary, but free to use API, than a barely supported open API. If OpenCL actually gains ground, and ATI actually gets of off their asses and helps devs get it going on the hardware, I'll change my tune, until then, I'll take CUDA, thanks.

More complaints about ATI Drivers? Never have I heard of such a think as of late. They've been rock solid with a steady stream of wonderful updates.

 

[pantherx12]

I didn't mean it for benchmarks, only for people to know that some apps use OpenCL

For benchmarks I've been using:
http://www.ngohq.com/graphic-cards/16920-directcompute-and-opencl-benchmark.html
Interesting program, but not suitable for TPU reviews I guess.

Cool you should check out gluxmark also it's a openGL/CL benchmark.

 

[Magikherbs]

So its not as much about the drivers as I thought. ..
This awesome review by W1zzard shows that the GTX 580 is not the hands down winner some may say it is.

http://www.techpowerup.com/reviews/AMD/Catalyst_10.12_Performance/

Some games and benchmarks clearly favor the Nvidia gpus. In my view, they should not count. Otherwise you may aswell include Folding @Home stats !

 

[Wile E]

Playback quality has been done in a real test by W1zz, ATI won.
Render quality has been done on 5XXX, 6XXX series cards VS Nvidia, ATI won. To achieve the same effects as Nvidia we would have to re implement angle dependency, reduce LOD by 30% and not include the EQAA or MLA effects ATI has out.
Studies by Microsoft and Steam both show that per capita Nvidia has more driver faults than ATI.


You are right, there are apps for CUDA. 13 games use their proprietary physx, and even in those games Nvidia whitepapers do not reccomend the real time rendering of certain effects, in stead suggesting "precooking" effects.

ATI is supporting standardized techonology, they did support the beginnings of hardware acceleration, and had it first(F@H) (Avivo video transcoder). ATI had tessellation first (Nvidia didn't want to play along), ATI had DX10 specs met, Nvidia didn't want to play along.


All said though, I agree, if ATI and open CL doesn't get moving I am going green on my next card.

Playback quality is not better on ATI. I disable ALL effects in video playback. I want accuracy, not flashy. I don't do "enhancements". They are both equal.

I don't like what I've seen of EQAA or MLA. I like very light AA.

I said nothing about Physx.

I don't care about open standards, I care what relevant to me as an end user.

And more driver faults over what span of time, with what hardware? I still don't buy it. A couple years ago nVidia had really bad drivers, but now it's ATI, and it has been since a couple months after the 4k series release. (8.11 is where it all went to crap, iirc.)

More complaints about ATI Drivers? Never have I heard of such a think as of late. They've been rock solid with a steady stream of wonderful updates.

The updates have been all but useless. Still stupid crappy bugs like scaling. Then there's still the shitty multi-gpu scaling in some games. Borderlands comes to mind for me. It's not exactly like my X2 is a weak card.

So its not as much about the drivers as I thought. ..
This awesome review by W1zzard shows that the GTX 580 is not the hands down winner some may say it is.

http://www.techpowerup.com/reviews/AMD/Catalyst_10.12_Performance/

Some games and benchmarks clearly favor the Nvidia gpus. In my view, they should not count. Otherwise you may aswell include Folding @Home stats !

You are misunderstanding that review. First, the 580 is faster in most of the benchmarks, second, the performance summary is not comparing card vs card, it is comparing launch driver vs newest driver, grouped. 100% for both 6970 and 580 doesn't mean they are equal, it means that both have the same performance with launch drivers as they do the newest drivers.

 

[pantherx12]

I don't care about open standards, I care what relevant to me as an end user.

.

That's sort of contradicting yourself dude, open standards = better for the consumer as it breeds competition as all companies need to be competitive with their hardware if the software is open to all.

Where as when it's propietry you don't have to be competitive simply because another company can't use your software. I.E Phsyx could actually be really shitty and we have no idea if it is as we have no real comparison except CPU at the moment and a few AMD demos.

Open standards are precisely what consumers should feel most strongly about, if everyone pushed forward open standards then hardware would get better as companies are really forced to make better hardware than the other company

 

[Wile E]

That's sort of contradicting yourself dude, open standards = better for the consumer, propietry = charge what ever they like.

Excuse my spelling.

No spell check XD

It's not contradictory at all. Open standards aren't better for the consumer when nobody adopts them into their software. OpenCL is completely irrelevant for me, because nothing uses it.

So, buy into an Open standard with no software, or buy into a free to use, but proprietary standard that does have software?

As an end user, I know which one I am picking.

 

[pr0n Inspector]

Playback quality is not better on ATI. I disable ALL effects in video playback. I want accuracy, not flashy. I don't do "enhancements". They are both equal.

This.

I don't understand the fascination over shitty filters. or wide-gamut.

 

[Bjorn_Of_Iceland]

He is looking for a benchmark that is repeatable, standard, and clearly defined in measurement.


A DX11 only benchmark will not allow someone to compare how much of a increase a game will get compared to their old 4xxx, or 2xx cards. Currently a upgrade from either company is almost a sideways move in performance. Metro is no more playable on a 580 than a 480, likewise dirt2 is no less playable on a 6970 than a 5870.

A few worthless benchmarks are the only place most users would notice any performance increase.

Ok so unigine heaven doesnt count

 

[Magikherbs]

You are misunderstanding that review. First, the 580 is faster in most of the benchmarks, second, the performance summary is not comparing card vs card, it is comparing launch driver vs newest driver, grouped. 100% for both 6970 and 580 doesn't mean they are equal, it means that both have the same performance with launch drivers as they do the newest drivers.

I understand it completely lol.. You seem not to understand my end statement. Games/app/benchmarks that benefit from Cuda, Physx and/or high shader speeds will score higher on Nvidia gpus. I was trying to use Folding @Home as an example.

How many of those results fall into that catagory ?

That's sort of contradicting yourself dude, open standards = better for the consumer as it breeds competition as all companies need to be competitive with their hardware if the software is open to all.

Where as when it's propietry you don't have to be competitive simply because another company can't use your software. I.E Phsyx could actually be really shitty and we have no idea if it is as we have no real comparison except CPU at the moment and a few AMD demos.

Open standards are precisely what consumers should feel most strongly about, if everyone pushed forward open standards then hardware would get better as companies are really forced to make better hardware than the other company

Exactly ! hah.. or.. you may aswell own an Apple ! LmfreakinAO

 

[Wile E]

I understand it completely lol.. You seem not to understand my end statement. Games/app/benchmarks that benefit from Cuda, Physx and/or high shader speeds will score higher on Nvidia gpus. I was trying to use Folding @Home as an example.

How many of those results fall into that catagory ?



Exactly ! hah.. or.. you may aswell own an Apple ! LmfreakinAO

The fact that nVidia scores higher due to Cuda, Physx and especially higher shader speeds is irrelevant. All that matters is end results. If an nVidia card does better on your favorite games, it does better on your favorite games. Reasons why do not matter. And F@H performance is poor on ATI because ATI is not devoting any time to Stanford to help them get it running better on their cards. nVidia does. ATI hardware has the potentialto do much better, but their software team lets it down.

I'd rather have Windows or OS X than Linux. Windows and OS X are not open, and Linux is, but Linux isn't better from the standpoint of most end users. I actually do own an older iMac, btw.

Open =/= better. Sometimes proprietary standards are just better for an end user, because it is better supported in the real world.

 

[bear jesus]

If cuda and physx ran on ATI/AMD cards i think that could be compared to something like windows as it may be closed but it runs on all hardware so there is no limit to what company you have to buy a product from, that will never happen so i would assume as openCL is newer there will be a while until there is anywhere near as many programs that use cuda but with time the number will grow so yes openCL is of little use to us right now but in the near future it should become more useful.

I think the cuda sdk came out near 4 years ago yet ATI/AMD released their openCL sdk a little over a year ago, is there any surprise there is more things that use cuda? how many programs used cuda within a year of the sdk release?

And F@H performance is poor on ATI because ATI is not devoting any time to Stanford to help them get it running better on their cards. nVidia does. ATI hardware has the potential to do much better, but their software team lets it down.

I wish AMD/ATI could see folding performance is a selling point for high end cards, sure not a massive market but if AMD cards could fold as well as nvidia but with less power usage then I'm sure many people running their own folding farms would happily look to them for a future upgrade due to reduced power usage/running cost and possibly heat, that is of course assuming AMD's architecture could be efficient at folding given good enough software.

i used to fold with my 4870 but the performance was hardly better than people running 3850's yet used loads of power and put out so much heat that i just had to give up on it, i only intend to fold once i get a Nvidia card in my htpc as i have no faith in a useful AMD folding client as things stand right now.

 

[Wile E]

If cuda and physx ran on ATI/AMD cards i think that could be compared to something like windows as it may be closed but it runs on all hardware so there is no limit to what company you have to buy a product from, that will never happen so i would assume as openCL is newer there will be a while until there is anywhere near as many programs that use cuda but with time the number will grow so yes openCL is of little use to us right now but in the near future it should become more useful.

I think the cuda sdk came out near 4 years ago yet ATI/AMD released their openCL sdk a little over a year ago, is there any surprise there is more things that use cuda? how many programs used cuda within a year of the sdk release?

Then compare it to OSX instead. OSX only runs on approved hardware, but is still better for most end users compared to linux. Although I have a feeling Steve Jobs is going to ruin that at some point, and move to the walled garden system used on iPad/iPhone/iPod touch.

And when OpenCL becomes more useful, it will become relevant. By that time, we'll probably have new hardware out, so the point is moot as far as reviews are concerned for the current generation.

I wish AMD/ATI could see folding performance is a selling point for high end cards, sure not a massive market but if AMD cards could fold as well as nvidia but with less power usage then I'm sure many people running their own folding farms would happily look to them for a future upgrade due to reduced power usage/running cost and possibly heat, that is of course assuming AMD's architecture could be efficient at folding given good enough software.

i used to fold with my 4870 but the performance was hardly better than people running 3850's yet used loads of power and put out so much heat that i just had to give up on it, i tried other software that ATI cards work better with but always struggled to get anything to process so fully gave up and only intend to fold once i get a Nvidia card in my htpc.

I wish ATI would devote more time to the software side of things, period. Not just folding.

 

[Magikherbs]

The fact that nVidia scores higher due to Cuda, Physx and especially higher shader speeds is irrelevant. All that matters is end results. If an nVidia card does better on your favorite games, it does better on your favorite games. Reasons why do not matter. And F@H performance is poor on ATI because ATI is not devoting any time to Stanford to help them get it running better on their cards. nVidia does. ATI hardware has the potentialto do much better, but their software team lets it down.

I'd rather have Windows or OS X than Linux. Windows and OS X are not open, and Linux is, but Linux isn't better from the standpoint of most end users. I actually do own an older iMac, btw.

Open =/= better. Sometimes proprietary standards are just better for an end user, because it is better supported in the real world.

I highly doubt a Nvidia card will score the same with Physx disabled, with 3DMark06/11, Nfs Shift, FFXIV benchmark and others. Nfs shift goes so far as to install Physx, even though my gpu is not Nvidia. ahha.......

If Physx had never sold out to Nvidia, and stayed neutral, we would not be having this converstation right now lolz

pantherx12 is right ! "Open standards are precisely what consumers should feel most strongly about, if everyone pushed forward open standards then hardware would get better as companies are really forced to make better hardware than the other company."

We as consumers should demand this.

 

[Wile E]

I highly doubt a Nvidia card will score the same with Physx disabled, with 3DMark06/11, Nfs Shift, FFXIV benchmark and others. Nfs shift goes so far as to install Physx, even though my gpu is not Nvidia. ahha.......

If Physx had never sold out to Nvidia, and stayed neutral, we would not be having this converstation right now lolz

pantherx12 is right ! "Open standards are precisely what consumers should feel most strongly about, if everyone pushed forward open standards then hardware would get better as companies are really forced to make better hardware than the other company."

We as consumers should demand this.

GPU Physx doesn't even effect those games. On or off makes no difference. if nVidia is faster, it's purely because the card is faster in those games/benches. Most games that use Physx use the CPU, not the gpu, just like havok does. Go here: http://www.geforce.com/#/GamesandApps and click on the Physx link. Only 13 games use gpu acceleration.

And you seemed to miss the fact that I was primarily talking about CUDA, not Physx anyway. There are no physics engines written in OpenCL to my knowledge, so there is nothing here to compare Physx against.

Why should we feel strongly about open standards? CUDA is free to use. Open standards hasn't made Linux the better choice for most people. You aren't getting all worked up over OpenGL instead of Microsoft's proprietary DirectX. I just want well supported APIs that costs me nothing as an end user. I don't care who developed them. APIs with no apps are useless.

 

[pantherx12]

YO wille, bullet phsyx supports opencl, ( or will do very soon) download their pre compiled demos, got a few GPU based demos on there.

Also mathmatica uses opencl, mathmatica is a program that's 20 years old ( although opencl support since 2010) it's one of THE maths research programs.

CUDA had a two year headstart is all man, I think most people are going to use the open standard now it's starting to become a bit more polished as well, it's stupid not to lol it restricts how many systems their software can run on other wise : ].


You make it sound like Cuda is inheriantly better ( may of spelt that wrong he he) where as it's not it's just more stable, a bit more grown up.

But opencl should have a massive growth spurt soon as it CAN do everything cuda can, but it can do it running on ANY hardware.

That is a HUUUUUUUUUUUUGEEEEEEEEEE thumbs up when your a software designer

Us end users are the least important when it comes to stuff like this, remember that.


Also Wille, I don't think you work for AMD you can't really comment on what they do behind the scenes.
Loads of people say " ahh they don't give dev support !" where as they just didn't make a song and dance about it, they helped with loads of the launch dx11 titles but no one knew because there was no AMD logos splattered all over the place. (AVP for example, can't remember the rest, I'll need to watch the video again where all the devs met up)

I suspect that could be the case with their software development.

You should check out their developer forums, their reps are ALWAYS on them ( their avatars have headsets) so they're even helping normal people using their sdks, not just companys.

( sure nvidia does the same, but everyone assumes AMD doesn't)


I could go there now and get help developing an app that I've designed myself...... If I knew the first thing about coding that is!


( by the way, don't let my specs fool you, not a fanboy, I like what ever company gives me the best stuff! he he)

 

[Wile E]

There have been more than a few devs that specifically mentioned that ATI would not help them, but nVidia would.

And again, just because it's open, does not mean it will gain more support. OpenGL has less support than DirectX, despite being open and cross platform. OS X has more support than Linux, despite being a closed platform. Open is not always better for the consumer.

It will all come down to ease of developing for it. For some reason, open platforms seem to be harder to code for, and many devs find it not worth doing.

Like I said, I don't care about the behind the scenes, I only care what I can do with it. Mathmatica serves me no purpose. How many Bullet Physics titles have OpenCL acceleration? Again, completely not relevant to an end user. OpenCL will not be relevant until more things use it.

And look at my specs, I'm no fanboy either.

 

[pantherx12]

I was only saying about the fanboy bit to avoid being excused of being one for the long post

lol

Opengl is less sucessful because microsoft have directx and most people use their operating system anyway,(I.E it has a massive market share so open source has no advantage) no point using opengl just to please a a tiny market segment ( linux and osx)

So it's a bit of a different kettle but I can see what you mean.

And they were just examples to show that people will be using opencl and are using opencl.

And when something big like mathmatica uses it that's a good sign of things to come : ]

 

[Magikherbs]

GPU Physx doesn't even effect those games. On or off makes no difference. if nVidia is faster, it's purely because the card is faster in those games/benches. Most games that use Physx use the CPU, not the gpu, just like havok does. Go here: http://www.geforce.com/#/GamesandApps and click on the Physx link. Only 13 games use gpu acceleration.

And you seemed to miss the fact that I was primarily talking about CUDA, not Physx anyway. There are no physics engines written in OpenCL to my knowledge, so there is nothing here to compare Physx against.

Why should we feel strongly about open standards? CUDA is free to use. Open standards hasn't made Linux the better choice for most people. You aren't getting all worked up over OpenGL instead of Microsoft's proprietary DirectX. I just want well supported APIs that costs me nothing as an end user. I don't care who developed them. APIs with no apps are useless.

Why don't I don't see Need for Speed Shift on that list of Physx games lol..

Pardon the slight change of topic from Cuda to Physx. Maybe, I should not have included Cuda in my initial statements b/c Cuda drivers/code are written by Nvidia and not the software developers.

 

[wahdangun]

It's not contradictory at all. Open standards aren't better for the consumer when nobody adopts them into their software. OpenCL is completely irrelevant for me, because nothing uses it.

So, buy into an Open standard with no software, or buy into a free to use, but proprietary standard that does have software?

As an end user, I know which one I am picking.

but if no one pick up the open standard then how do you expect it to became mainstream??

 

[bear jesus]

As i said "cuda sdk came out near 4 years ago yet ATI/AMD released their openCL sdk a little over a year ago" how about we see how opencl is doing in 3 years?

 

[W1zzard]

As i said "cuda sdk came out near 4 years ago yet ATI/AMD released their openCL sdk a little over a year ago" how about we see how opencl is doing in 3 years?

http://en.wikipedia.org/wiki/Close_to_Metal around late 2005

 

[bear jesus]

http://en.wikipedia.org/wiki/Close_to_Metal around late 2005

wow i don't think i have ever been so wrong in my life my bad