News - AMD exec says NVIDIA neglecting gamers

[HPL]

The story from semiaccurate was a little difficult to take in. For me it seems quite biased, the writer seems to have a great dislike of nVidia. I'd prefer to be presented with facts and make up my own mind. It is better to stick to news from hexus.net

[shaithis]

ATi can't supply me with stereo, or PhysX, so I use NVIDIA. Not to mention I need to transcode video because I travel for work.

nVidia can't even supply Physx if it sees an AMD card in your machine.....funny that.

Why do you need a nVidia card for video transcoding?

[ajones]

These days they can't meet my needs, I'm a high end gamer. I have a Dell 3007WFP- HC on one box, and 3d Vision on the other. ATi can't supply me with stereo, or PhysX, so I use NVIDIA. Not to mention I need to transcode video because I travel for work.

You're making it hard not to argue against you....

1) You can be a high end gamer using ATi hardware. Your point is that the hardware you currently own is good enough for the needs of your gaming, so I'd argue that this point is moot
2) It's your choice to use 3D Vision, so you are locking yourself to nVidia. However, if you want to use nVidia proprietary features then that's your choice
3) Ahhh... PhysX.... Well you can still use PhysX with ATi graphics as long as you have an ickle nVidia card in your graphics subsystem to handle it.... And herein lies one of the fundamental arguaments in this thread. No one is complaining that you can't run PhysX on ATi hardware. They are complaining that if even if you have nVidia hardware capable of handling the PhysX, that nVidia drivers have blocked it's use because it has detected an ATi presence
4) Fairy snuff. Again, you are freely selecting a graphics subsystem based upon your needs

[Syllopsium]

It may or may not be a huge leap from DX10.1 to DX11, nevertheless it's a welcome improvement.
The power consumption has had a vast improvement.
Eyefinity is not just 'a driver trick'. Sure, presenting three monitors as a unified display space is basically a driver trick but a) No mainstream NVidia card can drive more than two monitors (they do have special cards that do) and b) It's not supported in their drivers even if it is 'a trick'.

ATI and Nvidia have taken different architectural decisions and time will tell which scales better. Currently the winner appears to be ATI, both in extending architecture (NVidia has DX11 where exactly - if it's such a minor tweak why doesn't the DX10.1 GT240 have it?) and implementation (as the new cards are low power and fast). I don't care, and neither should you, which is theoretically better.

Technically, AMD's Phenom architecture and HyperTransport should have destroyed Intel. In practice their implementation was poor and AMD failed to capitalise on a 2+ year period whilst Intel started to add on chip memory controllers and QPI..

It's hardly surprising there is some overlap between ATI's old and new products. Exactly the same tends to happen with other products. You weigh the better price/performance ratio of older hardware against the improved functionality and possible future proofing of new hardware.

A mild google will also tell you that 4870/5870 is not comparable to 8800GT/9800GTX+. Looks like worst case scenario the 5870 has the same lead as a best case scenario of the 9800GTX+ over the 8800GT. Best case scenario, the 5870 completely destroys the 4870.

[scaryjim]

I have but two things to say:

Does it seem odd to anyone else it takes 1440 ATi stream processors in a 5850 to finally best the GTX285 with 240 stream processors?

No odder than the fact that ATI can fit 1440 stream processors onto a smaller die, and then produce the die at lower cost, than NVidia with its 240 stream processors. The companies have each taken their own approach to unified shader architecure and stream processing. The simple fact is that the 5850 provides better framerates than the GTX285s, at a similar if not lower cost, and with significantly lower power draw. I think your sarcastic "efficiency" comments are aimed in the wrong direction...

Besides, this entire thread was won about 20 posts ago and really should have stopped with:

[shaithis]

As to EyeFinity being a gimmik....

Isn't that what the nVidia 3D is? A gimmik? Not only that but it's a gimmik that cripples framerate and makes you feel sick.

It's very easy to slate features.....

[Biscuit]

Rollos post conveniently brings us back to the original point though... Nvidia dont care about gamers.

The business IP might be the reason behind NVidias decisions/practices and maybes it is completely valid however this shift toward hardware specific software cripples the gaming industry and damages the gamers who use the hardware/software most of all giving credence to Mr Huddys statements.

[caveman-jim]

http://www.brightsideofnews.com/news....aspx?pageid=1

both ATI and nVidia recommend the following standard technique for enabling MSAA: "storing linear depth into the alpha channel of the RenderTarget and using that alpha value in the resolve pass. You only need to linearize it if you are storing something like 8-bit color so that you can get depth approximation, stuff it into alpha channel and then when you've finished rendering at high resolution you simply filter down the color values and use depth value maintain some kind of quality to your Anti-Aliasing so you don't just average."

What got AMD seriously aggravated was the fact that the first step of this code is done on all AMD hardware: "'Amusingly', it turns out that the first step is done for all hardware (even ours) whether AA is enabled or not! So it turns out that NVidia's code for adding support for AA is running on our hardware all the time - even though we're not being allowed to run the resolve code!
So… They've not just tied a very ordinary implementation of AA to their h/w, but they've done it in a way which ends up slowing our hardware down (because we're forced to write useless depth values to alpha most of the time...)!"

So Batman performs less than optimally on AMD cards because the 'nVidia IP' MSAA implementation (which is perfectly standard, and is how both companies recommend it be implemented) is half-performed on AMD card- they are just locked out of the render path to actually apply it.

The nVidia implementation of MSAA in Batman Arkham Asylum forces AMD cards to perform the first part of the calculations, which are never used and the result never shown (MSAA).

Wouldn't it have been better to stop the code execution before the first path? Why position the vendor lock out after beginning code execution of implementing MSAA?

I'm willing to concede there might be a valid technical reason that all cards, no matter of final render path, need to execute the nVidia MSAA alpha operations.... but that's really suspicious. I would love some clarification on why the Vendor ID lock out was placed where it was, so that it doesn't look like nVidia inserted standard MSAA code that has the benefit of limited the performance of their competitor.

[GheeTsar]

Looks like BrightSideOfNews have now covered this issue:

http://www.brightsideofnews.com/news...-analyzed.aspx

It even puts a face to Mr Huddy! It's a long article (5 pages), but here is the author's conclusions:

Conclusion: My own two cents
There are only two conclusions you can make. The First one is the subject of this analysis, Batmangate. After carefully reviewing statements released by all sides, talking to developers, there isn't much left. Forum members on Batman: Arkham Asylum Forums and AMD themselves all changed the vendor ID on their cards and got equal functionality as the one experienced on nVidia cards. This happened at the time when nVidia members both claimed that Batman AA code is proprietary to Eidos and that no vendor ID locks is implemented in the code. "Hands in a cookie jar" is the only parallel we can draw here.

With their engineering resources spread on DirectX 11 titles, it is natural that AMD could not dedicate themselves on making Batman: AA code. But what is unforgiveable for both sides is the fact that Unreal Engine 3 is out for four years now and nor AMD nor nVidia didn't develop a way for in-game AA selection for engine itself. Ultimately, the responsibility for FSAA doesn't lie neither with nVidia nor AMD. For a feature that became a standard in 2000, i.e. nearly 10 years ago, it is irresponsible that Epic Games didn't offer built-in AA support as the new versions of Unreal Engine rolled out. Being forced to develop multi-platform titles to survive, developers such as Rocksteady have thin resources to spend additional time on developing PC-only features, especially in a recession year when so many great studios closed their doors.

Secondly, if AMD is constantly criticizing nVidia as a company and deployed strategies, it would be nice to finally hear some positive AMD-related experience from the developers. Naturally, AMD needs to earn that respect first - while on USS Hornet, AMD disclosed the list of DirectX 11 titles that are coming out and if those developers don't start publicly saying that AMD's support is great - something will have to change.

[RichardHuddy]

Lars,

I'll quote a little of my own text and then some of yours. It seems I may have not been clear about one question I asked - and you inadvertantly responded to something I never intended to state.

I posted this:

Hi Lars,
...

• NVIDIA's behaviour in locking something as trivial as antialiasing to its own hardware (in Batman

...
If you don't agree with my fourth bullet point above then I'd guess you'd probably argue that AMD should lock DX11 functionality to its own hardware, yes? Something we haven't done!

And you replied with this:

With your comment regarding locking DX11, do you try to indicate that AMD invented DX11 and could have been an AMD-only feature?? DirectX 11 is a new version of DirectX, that will be fully supported by Fermi, as we announced at GTC. It seems that AMD tries to create the perception that DX11 is a AMD only feature. It is not.

Lars Weinand, NVIDIA

It was not my intention to claim that DX11 is an AMD-only feature, and I think if you re-read my queston you'll see I did not suggest that.

My point here is that if NVIDIA believes that it's right for NVIDIA to lock DX's antialiasing support in to NVIDIA-only hardware in a game - then I assume you'd agree that AMD locking DX11 functionality in upcoming games would also be right. Yes?

Doesn't doing this that undercut the whole of point of DirectX?

If you also take note of Tim Sweeney's statements (available on BrightSideOfNews) that the DX10 version of UE3 fully supports antialiasing then this whole situation is clearly just NVIDIA locking generic functionality to it's own hardware.

Can we have a poll on Hexus please asking the simple question "Do you feel that AMD would be right to lock any generic DirectX 11 functionality in forthcoming DirectX 11 games (like DiRT2 or AvP etc) so that these features are available only on AMD's hardware?"

I'd be astonished if the answer on a poll came out as yes. And right now I can tell you that AMD does _NOT_ plan to do it!

Thanks,


Richard Huddy [Graphics Developer Relations, AMD]

[caveman-jim]

The 5870 is the same lousy VLIW arch ATi has been selling since the 2900, with the addition of much needed TMUs and ROPs. Why do I say "lousy VLIW arch"? Does it seem odd to anyone else it takes 1440 ATi stream processors in a 5850 to finally best the GTX285 with 240 stream processors?

Lot's of efficiency going on there, only takes 5X as many ATi SPs to offer comparable performance.

The changes and improvements in the Fermi make the changes from 4870 to 5870 look like the difference in the 800GT and 9800GTX+ comparatively. It wasn't a huge leap from DX10.1 to DX11, and EyeFinity is just a driver trick that gives users the ability to stop worrying about things like whether they can actually see the difference in 4X and 8XAA, and gives them the ability to wonder if they can actually forget about the 2" thick opaque black bars running through their field of vision.

It seems odd to me that you would compare nVidia CUDA cores to ATi Stream Processors, yes. Internally they are of different architecture and both companies are promoting different approaches to GPGPU using the same name 'Stream Processors'. Note how I call it a CUDA core, in jest.

Your 3870 upgrade scenario is interesting, as typically I've found that people who buy a $200 card spend about the same on the next one. So if Joe 3870 buys a 5770 or a 4890 or a GTX260 he's getting a nice performance boost, and depending on choice a nice value-add feature set.

Having experienced Eyefinity in person I can tell you the experience is very immerse; the 'prison bars' really don't interfere with the gaming experience. I didn't have as nice an experience with 3DStereo, but that's because all stereoscopic solutions give me problems due to my vision; I struggle to maintain the focus and effect, especially without my glasses. But that's ok, Eyefinity and 3D glasses are out of my price range anyway.

All this is sort of off topic though- what's being discussed here is whether or not ATi has a right to expect Eidos to write AA just for them given NVIDIA had already given them code to provide it for over half of their market.

Or if the code NVIDIA wrote is fair game for ATi to use because it's based in DX10 calls.

Or if ATi has the right to lock out users from DX11 games because their development was begun on ATi cards.

The answers are no, no, and no. It's NVIDIA's IP, same as PhysX. The fact it uses DX is irrelevant.

If DX11 being started on ATi cards should be locked out, get ready to lose access to 90% of games on the market today ATi guys. ATi took a year and a half to bring SM3 to market, and beyond that TWIMTB is the predominant dev assist in the market today.

How do you feel about the assertion that ATi cards are suffering a performance penalty because only the render path of nVidia's code is being vendor restricted, and not the alpha path? Shouldn't there be no nVidia code run on ATi's cards at all? Or is this my lack of knowledge on this implementation (straight question, no malice)?


Off topic:
Oh, and someone at ATi Dev needs to throw Ray Adams a bone, he's doing great work with ATT and could use some support. It's not like his product isn't helping out ATi...

[mercyground]

Hey Mr Rollo... Wanna answer me this one?

Why did NV get MS to pull most of the specs for DX10 (And thus making it a joke).

The answer for those that dont know... Is they didnt have DX10.1 capable parts.

If you compare the ORIGINAL DX10 standard vs the crippled DX10 that was eventually released its laughable.

And adding insult to injury? Remember how GFX engine was suppost to be fully outside the core for Vista? MS pulled that requirement as NV couldnt get their drivers to work properly. (The bulk of issues with vista betas was purely NV drivers.)

Standards are there for a reason. If you cant be bothered to support them or get hardware that will run it? Then dont be surprised when ppl dont buy your products.

There is ZERO chance i will buy a fermi. Nor will i buy one of those HOT, OVEREXPENSIVE spaceheaters that NV is marketing as GFX cards. I have central heating... I dont need a sauna.

Get your 40nm process working and slim down that fat chick you call Fermi. I dont want to fathom the cure for cancer... I want to game. (Unless you cant slim her down becuase you need her to run Physx as well? )

Being bluntly honest. Given how poorly coded games are with multicore machines, isnt it better to use a CPU instead to do physics and just get great visuals from the cards instead? A Quad core machine is underutilised in gaming. For a good while its why gamers went for a fast dual core over a quadcore. Even Intel realise this with the iCore range. They drop the extracores and boost the primaries to get you the speed. Isnt it about time NV woke up to this too?

:edit:
Dammit i knew i forgot something.

3DVision. You HAVE to be kidding. Cheap glasses and stereo gfx? Do yourselfs and the consumer a favour. SCRAP IT and go the proper route. Even ppl who want it are screaming for support on NV's own forums. This cheap and nasty patch (cos that is what it is) is just poor form for a "Gaming" firm.

Do it right or dont do it at all. Its a basic engineering mantra.

[Grandpaq]

Lars,


Can we have a poll on Hexus please asking the simple question "Do you feel that AMD would be right to lock any generic DirectX 11 functionality in forthcoming DirectX 11 games (like DiRT2 or AvP etc) so that these features are available only on AMD's hardware?"

I'd be astonished if the answer on a poll came out as yes. And right now I can tell you that AMD does _NOT_ plan to do it!

Thanks,


Richard Huddy [Graphics Developer Relations, AMD]

I registered on this forum just to post this..

I already voted with my wallet the answer was NO.

I didn't buy Batman and won't buy any future games that limit available DX** functionality in this manner, period. I realize that my choice is a drop in the bucket but it's a start and I'm pretty sure I'm not alone. I also recently bought an ATI video card because it offered the best performance for my dollars (ATI not passing on a surcharge for controlling game developers I guess).

[Perfectionist]

http://www.forbes.com/2008/09/12/spo...0912spore.html
...Nuff said.

This is actually WORSE than the Spore debacle, at least the game still worked at least a few times if needed to reinstall, here you have a case where features (Anti-Aliasing, PhysX if you aren't using an nvidia graphics card, even if you have a damn nvidia PhysX card (!) I bet the people who bought those must be feeling really ripped off) are being summarily disabled just because you didn't buy the developers' favourite brand of card.

Batmangate: AMD vs. nVidia vs. Eidos fight analyzed - page 5 of 5

Forum members on Batman: Arkham Asylum Forums and AMD themselves all changed the vendor ID on their cards and got equal functionality as the one experienced on nVidia cards. This happened at the time when nVidia members both claimed that Batman AA code is proprietary to Eidos and that no vendor ID locks is implemented in the code. "Hands in a cookie jar" is the only parallel we can draw here.

[shaithis]

I cannot find the thread on the Official Batman: AA forums....

Is it me being blind or has nVidia forced it's removal?

[mercyground]

Originally Posted by Lars Weinand

With your comment regarding locking DX11, do you try to indicate that AMD invented DX11 and could have been an AMD-only feature?? DirectX 11 is a new version of DirectX, that will be fully supported by Fermi, as we announced at GTC. It seems that AMD tries to create the perception that DX11 is a AMD only feature. It is not.

Lars Weinand, NVIDIA

It was not my intention to claim that DX11 is an AMD-only feature, and I think if you re-read my queston you'll see I did not suggest that.

Well technically you are right anyway. AVAILABLE DX11 card ranges... well thats AMD then

Fermi still isnt in the hands of consumers. (And according to industry insiders wont be available till march 2010 at earliest.) - Care to comment Mr Rollo?

My point here is that if NVIDIA believes that it's right for NVIDIA to lock DX's antialiasing support in to NVIDIA-only hardware in a game - then I assume you'd agree that AMD locking DX11 functionality in upcoming games would also be right. Yes?

Doesn't doing this that undercut the whole of point of DirectX?

If you also take note of Tim Sweeney's statements (available on BrightSideOfNews) that the DX10 version of UE3 fully supports antialiasing then this whole situation is clearly just NVIDIA locking generic functionality to it's own hardware.

Can we have a poll on Hexus please asking the simple question "Do you feel that AMD would be right to lock any generic DirectX 11 functionality in forthcoming DirectX 11 games (like DiRT2 or AvP etc) so that these features are available only on AMD's hardware?"

I'd be astonished if the answer on a poll came out as yes. And right now I can tell you that AMD does _NOT_ plan to do it!

I'd be dissapointed if AMD did that. It does undermine the standards. Its a stepback to the old days of fighting drivers and trying to figure out what games will work with what cards.

Its the difference between innovation and anticompetitive behaviour.

Example. AMD is working with monitor manufacturers to bring the bezel issue down. NV could have gone the same route with their 3DVision. Instead they decided to take the cheap route and let the consumer eat costs and also suffer with poor support and a bad implementation.

A PROPER way to go would be to work with monitor manufacterers to get a better 3d solution that would work with the card and monitor.