AMD Radeon HD 6970 already benchmarked? Enough to beat GTX480 in Tesselation?

[Scali]

You need to take off the green shades...

What green shades? What is this, some kind of insult? Explain yourself!

I don't see anything in my post that makes me biased towards AMD....Except my sig. All I did was ask for your opinion, clarification of what you meant by your statement.

It was in the way which you asked.

 

[edplayer]

What green shades? What is this, some kind of insult? Explain yourself!

that is very convincing. Are you a movie actor?

 

[Kenmitch]

Originally Posted by Kenmitch
You need to take off the green shades...

What green shades? What is this, some kind of insult? Explain yourself!

Green shades....Biased towards Nvidia. Red shades....Biased towards AMD/ATI

To me you come accross as a hard core AMD/ATI hater....Just seems like most of your posting is anti AMD/ATI

Originally Posted by Kenmitch
I don't see anything in my post that makes me biased towards AMD....Except my sig. All I did was ask for your opinion, clarification of what you meant by your statement.

It was in the way which you asked.

Maybe it was in the way you read it. I didn't see anything offensive in it.

 

[ShadowOfMyself]

Yea, people like you are so horribly biased towards AMD that you see enemies everywhere.
I'm the most neutral guy on the planet. Anyone who's known me for more than a few years will know.

LOL Why do you get away with this? I have yet to see a single post from you that doesnt bash AMD/praise Nvidia, and then anyone who talks back is either clueless (because you are the only person on the forum that understands how tessellation works - LOL!!) or an AMD fanboy

I guess that makes me one too, but really, Ive seen that all too many times, from people like Wreckage and others before him... It seems nothing AMD does is ever good enough for you

And its pretty funny you say you are neutral, when you have already admitted you have a grudge against AMD for something they did in the past...

Oh, and I will be the one getting an infraction for this post I bet, but so be it


Member callouts are not acceptable...but I guess you knew that already

Moderator Idontcare

 

[SlowSpyder]

I was talking about AMD's talk of tessellation.
Don't you remember those nice slides they made, of how you shouldn't use smaller than 16-pixel triangles, because the rasterizer wouldn't like it, and all that?

But nice that you managed to turn it into yet another personal attack.

Where is the personal attack?


Your join date shows that you have been around for years, but I don't recall (maybe I'm wrong here) you being upset with Nvidia for not having any tessellation capability for years while AMD did. Now it's all about how Nvidia's tessellator is great and AMD's is no good. But AMD has had a tessellator for almost a decade now, you obviously feel very passionate about what it can do for gaming, but I don't recall posts by you being unhappy with Nvidia for the 8-9 years that they did not have a tessellator in any form while AMD did.

Also, as I understand it tessellation was supposed to be in DX10, but because Nvidia did not have that built into their hardware they were able to get Microsoft to push it back to DX11.

You stand out as particularly vocal about Nvidia's superior tessellation abilities, but I don't recall any posts by you lambasting Nvidia for not supporting it at all or delaying adoption.

 

[nitromullet]

When I first heard of the Hawx II fiasco before I saw the benchies, my first thoughts were the game is only playable on Nvidia HW. I thought AMD was getting like late teens, low 20s type framerates to Nvidia ~30+, then I saw that the 5770 could get 40 fps at at 19X12 res with 8AA.
The point being, we're not there yet, it will be a wee while before it becomes the difference maker. We on the PC are stuck with console ports with slapped on tessellation for good while yet

That wasn't really the point. For the past year we've been hearing that AMD brought us DX11 first and how much more efficient their gpus are than Fermi. They were first, and their cards do consume less power... but why?

When you really break down "first to DX11" what does that really mean? Well, it comes down to tessellation really. So, AMD was first to market with DX11 tessellation, but it wasn't nearly as fast as NV's subsequent DX11 chips. So, I'm starting to think that maybe AMD's engineers actually aren't magically able to squeeze blood from a stone... Maybe the reason why Fermi is bigger and hotter, etc is because it takes a big, powerful gpu to pull off the type of tessellation performance we see from Fermi.

...I'm not sure, I just find it odd that AMD would be playing catch up to NV with regards to tessellation considering they are on their 8th generation of tessellation hardware. It seems like they should have the know-how to be stronger with tessellation. Perhaps it was a choice to keep Evergreen chips small. I guess we'll see if the size and power requirements grow in accordance with Cayman's tessellation performance, or of Cayman is a smaller chip that performs as well as Fermi.

 

[GaiaHunter]

I'd say it's almost too much of a coincidence that after this recent talk of tessellation, the leaked benchmark of 6970 happens to be a tessellation-oriented benchmark...
That's just too convenient.

That tessellation slide was leaked some time ago and there was some Heaven benchmark leaked before too.

Although I've to admit it sounds like those Fermi talks before Cypress release or these GTX580 talk just before Cayman.

 

[AnandThenMan]

Yea, people like you are so horribly biased towards AMD that you see enemies everywhere.

LOL Why do you get away with this? I have yet to see a single post from you that doesnt bash AMD/praise Nvidia, and then anyone who talks back is either clueless (because you are the only person on the forum that understands how tessellation works - LOL!!) or an AMD fanboy

There is an ignore list for a reason, I suggest people use it. You will not get Scali to be even remotely objective. He admits that, "It's a pet peeve of mine - the Internet is filled with AMD fanboys"


As for the 6970, what is Scali going to say if it trumps the GTX480 in tessellation? That it's not important again, and that there are other ways to increase visual fidelity? Remember, this is what he was saying before Fermi was released, but suddenly changes his tune. Let's all hope that the 6970 has a very very strong tessellation unit, and that game devs use tessellation along with other DX11 features to push out visually exciting games, instead of console retreads. We all win.


Member callouts and personal attacks are not acceptable.

Re: "You will not get Scali to be even remotely objective"

Moderator Idontcare

 

[Spyhawk]

Yeah, that is pretty impressive. It only took ATI eight generations of building tessellators to supposedly catch up to NVIDIA's first gen tessellator. Seriously, didn't ATI provide most of the tessellation specs for DX11? It just seems to me that tessellation should be AMD's thing, instead they're playing catch up. I'm a little confused by that.

Up until recently AMD had no need to beef up thier tesselator. There simply wasnt any games using it. So they concentrated transistor budget on other aspects/features while keeping basic tessalation functionality.

Nvidia on the other hand makes a GPU with a super duper tesselating monster but...ooops...no games on the market is able to use it so in the end its wasted silicone.

 

[Idontcare]

Welcome to the forums Spyhawk! :thumbsup:

 

[Idontcare]

That wasn't really the point. For the past year we've been hearing that AMD brought us DX11 first and how much more efficient their gpus are than Fermi. They were first, and their cards do consume less power... but why?

When you really break down "first to DX11" what does that really mean? Well, it comes down to tessellation really. So, AMD was first to market with DX11 tessellation, but it wasn't nearly as fast as NV's subsequent DX11 chips. So, I'm starting to think that maybe AMD's engineers actually aren't magically able to squeeze blood from a stone... Maybe the reason why Fermi is bigger and hotter, etc is because it takes a big, powerful gpu to pull off the type of tessellation performance we see from Fermi.

...I'm not sure, I just find it odd that AMD would be playing catch up to NV with regards to tessellation considering they are on their 8th generation of tessellation hardware. It seems like they should have the know-how to be stronger with tessellation. Perhaps it was a choice to keep Evergreen chips small. I guess we'll see if the size and power requirements grow in accordance with Cayman's tessellation performance, or of Cayman is a smaller chip that performs as well as Fermi.

Kinda reminds you of the AMD of 2005-2006 leading up to Conroe release when they coasted for a while and it bit them in the ass when the competition caught up.

I hope not, but I get a dejavu'ee feeling about it. Tessellation should have been in the bag for AMD.

Here's my mental analogy. We all feel a little over-exposed to the Physx hype, right?

Well imagine how surprised we'd all be if Physx made it into the DX12 spec, Nvidia releases their first DX12 part before AMD (and does well in Physx as we'd all expect) and then AMD comes round late to the DX12 release scene but with their first DX12 (and first Physx chip) just clobbering Nvidia's best Physx parts despite the lengthy history Nvidia has had to perfect their Physx hardware?

That's how I personally feel about AMD and tessellation. I'm just gobsmacked that despite their experience with the technology and the history of making it work and so on that they got one-upped by Nvidia right out of the gate with Nvidia's first foray into the technology.

Being a CPU guy more so than a GPU guy, this would be like if Intel's first effort at integrating the memory controller (something AMD had been perfecting for years) came out of the gate above and beyond the performance of AMD's most advanced IMC at the time.

 

[AnandThenMan]

PhysX becoming part of the DX12 spec? That's highly, highly unlikely. Nvidia has quite the opposite corporate policy vs. AMD, they believe in proprietary "only us" features to help differentiate their products. AMD is much more apt to create, use and adopt open source initiates, for better or for worse. It would be nice to see something like PhysX become part of the DX spec, it would benefit all of us. But then Nvidia would lose a checkbox feature that they are definitely keen on.

And I think we should wait for Cayman before we declare Nvidia the "king" of tessellation land, not to mention tessellation on Fermi only shows a great advantage on synthetic benchmarks (especially ones that are sanctioned by Nvidia). BTW, for all we know, PhysX could be made to run better on AMD GPUs, we simply don't know because it obviously not coded to talk to Radeon hardware.

But as usual, being first with technology is in no way, shape or form a guarantee of success, nor does it necessarily help you keep the technology leadership. There are hundreds of such examples.

edit - I would like to add that 1)Nvidia is apparently stronger in tessellation but it comes at a cost, obviously. Nvidia is getting far worse performance/die size vs. AMD, there are always trade offs. 2)It has yet to be determined if Fermi is able to show higher tessellation performance when the GPU is called on to do other parallel tasks. In other words, it really depends on the game and how it is designed and coded to take advantage of the hardware. Why do you think Nvidia (and AMD) are so keen to have developers use their coding techniques? Obviously they both want the games to play to their strengths.

 

[SlowSpyder]

PhysX becoming part of the DX12 spec? That's highly, highly unlikely. Nvidia has quite the opposite corporate policy vs. AMD, they believe in proprietary "only us" features to help differentiate their products. AMD is much more apt to create, use and adopt open source initiates, for better or for worse. It would be nice to see something like PhysX become part of the DX spec, it would benefit all of us. But then Nvidia would lose a checkbox feature that they are definitely keen on.

Just a mental exercise, a comparrison. I don't think anyone thinks DX12 will include Physx.

While Nvidia rules the synthetic benches that include tessellation, I do wonder how this all stacks up in the real world. How many games are out there or will be out there in the next year or so that AMD's tessellator cannot handle? Is it really a bottle neck that makes games slow or unplayable, or is it just slower in benches but still enough for games? I really do not know.

I generally play older games, right now I'm leveling up in Titan Quest and playing Dragon Age: Origins. I don't even need DX10 much less fancy tessellation.

 

[Genx87]

Green shades....Biased towards Nvidia. Red shades....Biased towards AMD/ATI

To me you come accross as a hard core AMD/ATI hater....Just seems like most of your posting is anti AMD/ATI



Maybe it was in the way you read it. I didn't see anything offensive in it.

Scali blasted Nvidia big time over the NV30. He will blast both companies when they deserve it.

 

[Arkadrel]

Kinda reminds you of the AMD of 2005-2006 leading up to Conroe release when they coasted for a while and it bit them in the ass when the competition caught up.

I hope not, but I get a dejavu'ee feeling about it. Tessellation should have been in the bag for AMD.

Here's my mental analogy. We all feel a little over-exposed to the Physx hype, right?

Well imagine how surprised we'd all be if Physx made it into the DX12 spec, Nvidia releases their first DX12 part before AMD (and does well in Physx as we'd all expect) and then AMD comes round late to the DX12 release scene but with their first DX12 (and first Physx chip) just clobbering Nvidia's best Physx parts despite the lengthy history Nvidia has had to perfect their Physx hardware?

That's how I personally feel about AMD and tessellation. I'm just gobsmacked that despite their experience with the technology and the history of making it work and so on that they got one-upped by Nvidia right out of the gate with Nvidia's first foray into the technology.

Being a CPU guy more so than a GPU guy, this would be like if Intel's first effort at integrating the memory controller (something AMD had been perfecting for years) came out of the gate above and beyond the performance of AMD's most advanced IMC at the time.

I believe amd could if they wanted to push out some insane tessellation unit at any given time they wanted. I believe they know more about tessellation than nvida and how to get it working.

I believe amd choose not to go overboard with tessellation on purpose, because they rather have their cards strong in other areas and considered maybe the power/die size increases it would cost.

nvidia probably didnt want to end up looking bad in tessellation compaired to AMD, and because amd had like 9years experiance with it, they decided to AIM for the sky when they implimented it, so they where sure they wouldnt be behinde in this area.

The question is.... does anyone benefit from nvidias stronger tessellation units outsides of benchmarks currently? Hawx2 doesnt count because they optimised the code to run badly on amd hardware (faul play so doesnt accuratly represent how well the cards are one against another).

but if you have a code that works equally well on both... and you compair nvidia vs amd cards in a game, how big is the differnce? Very small in all likely hood.

 

[tannat]

It's all about area priorities from where I see it. AMD gave tesselation the die area they deemed necessary. Sofar this has worked out well, except for beatings in synthetically benchmarks by nvidias non-restricted area approach.

All chips from GTX460 and up do have scalability but sofar we have not seen them scale very much - all chips are larger than AMDs largest chip.They might even need to break the scalability in order to go down on similar diesize as AMD and stay competitive in performance.

Advantage in technology is directly correlated to die size. Overuse of die area may directly eradicate illusions of technological superiority and possibly put you in a dead end when you find yourself making too big chips which anyway don't perform as well as the competition.

 

[blastingcap]

I dunno, to me, it looks like in order to meet their self-imposed deadline to launch Cypress by the time Windows 7 and DX11 came out, AMD bolted a fixed tessellator to the GPU. NV decided to make a scalable new design with GPGPU in mind and almost pulled it off until 40nm sunk their ship. They were half a year late by the time they semi-fixed their problems. Their architecture is more elegant than AMD's, so they can reuse it at 28nm. AMD's architecture is getting old, but maybe they also did a radical redesign with Cayman. Doubtful, given David Hoff's comments, but possible.

It's all about area priorities from where I see it. AMD gave tesselation the die area they deemed necessary. Sofar this has worked out well, except for beatings in synthetically benchmarks by nvidias non-restricted area approach.

All chips from GTX460 and up do have scalability but sofar we have not seen them scale very much - all chips are larger than AMDs largest chip.They might even need to break the scalability in order to go down on similar diesize as AMD and stay competitive in performance.

Advantage in technology is directly correlated to die size. Overuse of die area may directly eradicate illusions of technological superiority and possibly put you in a dead end when you find yourself making too big chips which anyway don't perform as well as the competition.

 

[OCGuy]

Up until recently AMD had no need to beef up thier tesselator. There simply wasnt any games using it. So they concentrated transistor budget on other aspects/features while keeping basic tessalation functionality.

Nvidia on the other hand makes a GPU with a super duper tesselating monster but...ooops...no games on the market is able to use it so in the end its wasted silicone.

You completely contradicted yourself.


Why does AMD need to beef up its tesselator if no games on the market (are) able to use it? Wouldn't they be wasting silicon as well?

 

[dust]

Well imagine how surprised we'd all be if Physx made it into the DX12 spec, Nvidia releases their first DX12 part before AMD (and does well in Physx as we'd all expect) and then AMD comes round late to the DX12 release scene but with their first DX12 (and first Physx chip) just clobbering Nvidia's best Physx parts despite the lengthy history Nvidia has had to perfect their Physx hardware?

This is your Final Fantasy MCLXVII

 

[Lonbjerg]

That chart seems fishy:

When you look at this:
http://www.geeks3d.com/20100326/geforce-gtx-480-unigine-heaven-2-0-score/

 

[badb0y]

You completely contradicted yourself.


Why does AMD need to beef up its tesselator if no games on the market (are) able to use it? Wouldn't they be wasting silicon as well?

I think what he is saying is up until now there was no need for tessellator to be strong now that games use it, they beefed it up. When he referred to nVidia he probably meant nVidia "over-beefed up" the tesselator.

Judging by most games available today it seems that way but you never know what next year may bring.

 

[darkswordsman17]

That chart seems fishy:

When you look at this:
http://www.geeks3d.com/20100326/geforce-gtx-480-unigine-heaven-2-0-score/

Could the CPU used make for the difference?

 

[blastingcap]

Could the CPU used make for the difference?

Maybe different settings were used in each instance. The 5870's performance is better at the same site Lon linked to: http://www.geeks3d.com/20100525/qui...-4-0-and-direct3d-11-in-extreme-tessellation/

 

[Lonbjerg]

Maybe different settings were used in each instance. The 5870's performance is better at the same site Lon linked to: http://www.geeks3d.com/20100525/qui...-4-0-and-direct3d-11-in-extreme-tessellation/

Resolution and tesselation levels are the same, according to this AA settings also match:
http://www.hexus.net/content/item.php?item=27053&page=10

And the CPU is even more beefy in that review:
http://www.hexus.net/content/item.php?item=27053&page=8

I think it's safe to say that just plugging those numbers into a chart dosn't give any clue to actual performance.

But I already see posts on forums claming that the charts shows 6970 beating the GTX480...I'd would be carefull making such a claim right now....if the chart is the "source".

 

[blastingcap]

I'd would be carefull making such a claim right now....if the chart is the "source".

I agree.

I haven't even kept track of Cayman rumors lately, I only cared about Barts and when I saw how much price gouging was going on (aside from newegg, which I don't want to buy from), I gave up and got a GTX460-768. I'll see what happens at 28nm with GPUs. So I have little interest in Cayman except in what it implies for future architectures.