Tesselation done properly..according to AMD

[minddripper]

OK.

http://www.xbitlabs.com/articles/video/display/hardware-tesselation.html#

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had. It's clear that NVIDIA is doing tessellation right and AMD is in trouble.

I don't see how your statement is supported by those examples. In most of the tests, the performance hit for tessellation on the 5870 and 460 is very similar. Almost all of the benches where the 460 matches or outperforms the 5870 with tessellation on are the same benches where it matches or outperforms the 5870 with tessellation off. This mostly occurs at the lowest resolution, but at the higher resolutions the 5870 gains on the 460. The only significant differences in the tessellation performance hit between the two cards seems to be in the extreme settings on the Stone Giant and Unigine Heaven benchmarks.

If the 460 matches the 5870 with tessellation on at the same settings and in the same games that it matches the 5870 with tessellation off, what significance does tessellation have and why is AMD in trouble because of it?

Also, in that review the 5870 appears to beat the 460 in Metro 2033 at all three resolutions.

 

[lavaheadache]

OK.

http://www.xbitlabs.com/articles/video/display/hardware-tesselation.html#

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had. It's clear that NVIDIA is doing tessellation right and AMD is in trouble.

lol, if you look in that article the metro 2033 bench clearly states what i said and proves what you just said in the other post to be utterly wrong.

 

[Wreckage]

lol, if you look in that article the metro 2033 bench clearly states what i said and proves what you just said in the other post to be utterly wrong.

So you are just going to ignore all the other benchmarks eh? Ok, I agree that tessellation and AA in Metro2033 really hurt AMD cards. Not sure that something to cheer about though.

 

[cusideabelincoln]

OK.

http://www.xbitlabs.com/articles/vid...selation.html#

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had. It's clear that NVIDIA is doing tessellation right and AMD is in trouble.

So you are just going to ignore all the other benchmarks eh? Ok, I agree that tessellation and AA in Metro2033 really hurt AMD cards. Not sure that something to cheer about though.

Nope, let's not ignore all the other benchmarks. I took the liberty of compiling a mix of recent benches, so you won't cherry-pick a four-month-old review.

https://spreadsheets.google.com/ccc...E5iek5NdzBFa0VsTmRCbVE&hl=en&authkey=CILr7d4D

 

[lavaheadache]

um, I just don't understand your continuous misunderstanding of just about everthing. If you want I can cherry pick some benchmarks where a 5870 beats a GTX 480? You see, I have nothing to lose in this discussion with you. I have in my possesion right now, a 5870, a gtx 580 and i'm trying to get this gtx 460 I just picked up tonight to work as a physx card with the 580 but seems to be some driver issue that keeps shutting off the 580 when i put in the 460. My experience I'm sure grossly out weighs yours, yet you don't see me coming up with the same absurd conclusions you do, do you? everything I say is from experience yet everything you say is from reading tidbits here and tidbits there piecing together some warped perception of reality. keep it up.

 

[Wreckage]

Nope, let's not ignore all the other benchmarks. I took the liberty of compiling a mix of recent benches, so you won't cherry-pick a four-month-old review.

https://spreadsheets.google.com/ccc...E5iek5NdzBFa0VsTmRCbVE&hl=en&authkey=CILr7d4D

You are missing the whole "tessellation" argument aren't you? :hmm:

 

[cusideabelincoln]

You are missing the whole "tessellation" argument aren't you?

Deflection? Is that all you can do - ignore the data presented right in front of your face? Try to insult less, present substance more.

Your argument so far is weak. You use one review sample and you make a declaritive statement without any supportive statements. Tessellation isn't the only thing that matters in a GPU, and this is why the HD 5870 is providing superior performance than the GTX 460 in DX 11 games that use tessellation.

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had.

https://spreadsheets.google.com/ccc...E5iek5NdzBFa0VsTmRCbVE&hl=en&authkey=CILr7d4D

Look at AvP and Stalker again. Every review shows the 5870 ahead of the GTX 460, and more times than not very considerably ahead. So before you claim them as part of your argument, you should actually verify the data you're using representative of the situation.

Hence why I provided the handy benchmarks, or did you just miss that point entirely?

 

[Troll Trolling]

OK.

http://www.xbitlabs.com/articles/video/display/hardware-tesselation.html#

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had. It's clear that NVIDIA is doing tessellation right and AMD is in trouble.

(OMG! AMD's more than one year old GPU is matching Fermi! Run to the hills!)

(GTX 470 x HD 6870)

Well, I guess if AMD is doing tesselation wrong, nVidia is doing the whole GPU wrong.
(See, I can cherry-pick graphs too!)

 

[SlowSpyder]

OK.

http://www.xbitlabs.com/articles/video/display/hardware-tesselation.html#

Look at AvP, Stalker, Stone Giant, Uniheven. NVIDIA's midrange chip is matching the fastest chip AMD has ever had. It's clear that NVIDIA is doing tessellation right and AMD is in trouble.

I don't see how this link supports what you are saying. If anything, it shows the opposite, that AMD's tessellator is keeping up with Nvidia pretty well. The hit cards from both camps take when tessellation is turned in is pretty similar.

 

[Will Robinson]

Dude,the GTX570 just came out and looks to be a pretty decent part....so,acting with clarity and decisiveness he ran out and bought a GTX470.:sneaky:

 

[Wreckage]

I don't see how this link supports what you are saying. If anything, it shows the opposite, that AMD's tessellator is keeping up with Nvidia pretty well. The hit cards from both camps take when tessellation is turned in is pretty similar.

The funny thing is that even AMD says that NVIDIA's tessellator is too powerful. I swear some people will defend an idea long beyond its death.

Either way my point is clear NVIDIA's superior tessellation allows its midrange card to keep up with the best AMD has to offer. To me that is proof that NVIDIA is tessellation done right.

Now if the 6970 or whatever they rebadge the next card, matches NVIDIA in tessellation then they are agreeing with me. If not then they will be behind in the DX11 game.

 

[Mistwalker]

Dude,the GTX570 just came out and looks to be a pretty decent part....so,acting with clarity and decisiveness he ran out and bought a GTX470.:sneaky:

To be fair, for $200 that's a fantastic deal. Context people, it's important.

Similarly, it's too early to declare Nvidia's implementation of tessellation as "right." AMD is gambling on developers implementing the tech more gradually, and choosing to keep its tessellation performance in line with the general power of the card. If that also translates to leaner cards that are cheaper and/or cooler running, then as a consumer it's a perfectly valid approach. Until the 6900 series cards and the games of 2011 come out, it's impossible to say whether Nvidia's extreme emphasis gives it longer legs or is in fact not widely utilized any more than AMD's offerings.

It's easy to say that all else being equal, having too much potential performance is always better than too little, but that thinking led to the hot and loud 480--there are other factors to consider. Again, context.

 

[SlowSpyder]

The funny thing is that even AMD says that NVIDIA's tessellator is too powerful. I swear some people will defend an idea long beyond its death.

Either way my point is clear NVIDIA's superior tessellation allows its midrange card to keep up with the best AMD has to offer. To me that is proof that NVIDIA is tessellation done right.

Now if the 6970 or whatever they rebadge the next card, matches NVIDIA in tessellation then they are agreeing with me. If not then they will be behind in the DX11 game.

I think the link you posted shows that the GTX460, GTX480, and 5870 all take pretty close to the same percentage hit when tessellation is used in the games they benched. So, AMD's 'weak' tessellator is doing fine. Nvidia's part may be able to handle a higher tessellation load, but as I have said a few times throughout this thread, it doesn't seem like any software will take advantage of that. Your link shows that perfectly. By the time games do require the level of tessellation that Fermi can deliver, it is my opinion that we won't want to play those games on current hardware due to other limiting factors.

It's like getting 1GB on a Radeon 3870 or 8800GT. By the time you really need a 1GB frame buffer, it's not likely we'd want to game on those cards anyway.

And what do you mean rebadge the next card? The 6970 is a new architecture, not just a new sticker with a new name.

 

[happy medium]

Similarly, it's too early to declare Nvidia's implementation of tessellation as "right." AMD is gambling on developers implementing the tech more gradually, and choosing to keep its tessellation performance in line with the general power of the card.

Is that why they made the 6900 series with 3/4x the tessalation power of the 5870?
How do you exspain that.?

 

[SlowSpyder]

Is that why they made the 6900 series with 3/4x the tessalation power of the 5870?
How do you exspain that.?

The same reason Nvidia rushed out a tripple monitor gaming solution when the Nvidia fans said that is such a small niche. Or the same reason Nvidia tried to quiet down and provide better power use/heat charactersitics on their current parts when Nvidia fans said no high end gamer cares about that stuff. If I'm AMD or Nvidia I wouldn't want to appear to be treading water while the other moves forward. Can you imagine the marketing advantage the other would have?

 

[happy medium]

The same reason Nvidia rushed out a tripple monitor gaming solution when the Nvidia fans said that is such a small niche. Or the same reason Nvidia tried to quiet down and provide better power use/heat charactersitics on their current parts when Nvidia fans said no high end gamer cares about that stuff. If I'm AMD or Nvidia I wouldn't want to appear to be treading water while the other moves forward. Can you imagine the marketing advantage the other would have?

So they upped the tessalation performance for marketing puposes?
Ok, sounds reasonable.

 

[lavaheadache]

Now if the 6970 or whatever they rebadge the next card, matches NVIDIA in tessellation then they are agreeing with me. If not then they will be behind in the DX11 game.

what the hell are you talking about? rebadge again? you lost me *yet* again! There has been no rebadge thus far. Dare i even nod to the fact that Nvidia has done this quite a bit in the not so distant past?

 

[lavaheadache]

I swear some people will defend an idea long beyond its death

You said it man, you said it.....

 

[SlowSpyder]

So they upped the tessalation performance for marketing puposes?
Ok, sounds reasonable.

Not just marketing, but for the natural progression of their GPU. As performance increases AMD will want want tessellation performance to increase. I guess I could have mentioned that too, but I thought it would have been a given.

 

[Skurge]

Not just marketing, but for the natural progression of their GPU. As performance increases AMD will want want tessellation performance to increase. I guess I could have mentioned that too, but I thought it would have been a given.

3Dmark11 looks absolutely beautiful and it uses tessellation that doesn't choke AMD's cards. So is that tessellation done right? I would think so.

 

[Janooo]

6970 should be between 480 and 580 in Heaven

 

[cusideabelincoln]

The funny thing is that even AMD says that NVIDIA's tessellator is too powerful

People keep saying this but I do not see an official quote or source anywhere. Spin for the win.

I swear some people will defend an idea long beyond its death.

I swear some people will attack an idea long beyond its death. They'll even make up an idea and attack it!

 

[Skurge]

6970 should be between 480 and 580 in Heaven

You know too much.

 

[Leadbox]

I wonder how much of that 500mm^2+ is wasted on excessive tessellation

 

[taserbro]

I wonder how much of that 500mm^2+ is wasted on excessive tessellation

That's like asking how much of that cpu's surface area is wasted on excessive fp throughput just because you don't run anything that requires it. Shader cores using a polyvalent design capable of doing more than one thing is hardly a waste of space since when tessellation is unneeded, it can be dedicated to other tasks.

If anything, using a significant chip area dedicated solely for tesselation would technically be a waste when running applications that require little or any at all.