AMD V Nvidia by Richard Huddy

[Skurge]

Yea right... It's standard DX11 code. AMD has had DX11 drivers out for about a year now, with plenty of updates. Lots of software with tessellation has been around, including various samples from AMD in the early days.
You think they STILL didn't get around to optmizing the tessellation? Nope, it's the hardware.



That's the same pattern we've been seeing in all heavy tessellation scenario's... The tessellator is bottlenecking the rest of the GPU. Same tessellator in the 5770 as in the 5870, hence same performance.

Drivers for the game, Im sure you've seen in the past drivers speed up certain cards in certain games.

oh, I was under the impression the everything in the 5770 is halved, including the tesselator.

The 5770 scores almost half what the 5870 scores in heaven.

 

[Scali]

Drivers for the game, Im sure you've seen in the past drivers speed up certain cards in certain games.

Lol yea, we've already discussed how they 'optimize' for games in another thread.
Really, how gullible can you be? Standard DX11 code just needs to run. No excuses. Game-specific improvements are ALWAYS fishy, especially when a product has already been on the market for a year.
I'm sure that AMD can come up with some drivers that 'optimize' the performance in HAWX 2.
I'm also sure that this means we're not comparing apples-to-apples.

oh, I was under the impression the everything in the 5770 is halved, including the tesselator.

The 5770 scores almost half what the 5870 scores in heaven.

If you test tessellation performance specifically, you'll see that they're pretty much the same speed: http://www.geeks3d.com/20100826/tessmark-opengl-4-gpu-tessellation-benchmark-comparative-table/
As you can see, only at the moderate setting is the 5870 faster.
At normal, extreme and insane, there is hardly any performance difference. Conclusion: It's the same tessellator.

 

[Skurge]

Lol yea, we've already discussed how they 'optimize' for games in another thread.
Really, how gullible can you be? Standard DX11 code just needs to run. No excuses. Game-specific improvements are ALWAYS fishy, especially when a product has already been on the market for a year.
I'm sure that AMD can come up with some drivers that 'optimize' the performance in HAWX 2.
I'm also sure that this means we're not comparing apples-to-apples.



If you test tessellation performance specifically, you'll see that they're pretty much the same speed: http://www.geeks3d.com/20100826/tessmark-opengl-4-gpu-tessellation-benchmark-comparative-table/
As you can see, only at the moderate setting is the 5870 faster.
At normal, extreme and insane, there is hardly any performance difference. Conclusion: It's the same tessellator.

remember the last remnant? AMD cards performed so poorly even though it used standard DX9 code, after a new driver they were all much faster and I don't remember anything in the game changing. The same story with the Resident Evil 5 benchmark, AMD cards performed poorly up until the final game was released.

So hawks 2 is just a tessellation benchmark?

 

[Scali]

remember the last remnant? AMD cards performed so poorly even though it used standard DX9 code, after a new driver they were all much faster and I don't remember anything in the game changing. The same story with the Resident Evil 5 benchmark, AMD cards performed poorly up until the final game was released.

I have 2 things to say about that:
1) Did nVidia suffer from the same poor performance, until a driver update? If not, why do you think this is?
2) The point of shader replacement, texture downgrading etc is that you don't need to change anything to the game, the driver silently does that FOR you.

So hawks 2 is just a tessellation benchmark?

If tessellation is such a bottleneck as on AMD's hardware, then anything using tessellation will quickly become a tessellation benchmark.
On nVidia's hardware it's just a game.

 

[Skurge]

I have 2 things to say about that:
1) Did nVidia suffer from the same poor performance, until a driver update? If not, why do you think this is?
2) The point of shader replacement, texture downgrading etc is that you don't need to change anything to the game, the driver silently does that FOR you.

Nvidia has had the same issues with different games in the past. As for number 2, You do graphics so that might piss you off, but as a regular user it doesn't bother me at all.

If tessellation is such a bottleneck as on AMD's hardware, then anything using tessellation will quickly become a tessellation benchmark.
On nVidia's hardware it's just a game.

There are tons of games that use tessellation and we don't see a disparity like that. The game runs incredibly fast on anything even with the weaker tessellation on the AMD cards. So what I'm saying is does this BENCHMARK (as we haven't seen the game yet) include anything, but tessellation?

 

[badb0y]

Lol that's one game out of like 20 and even then AMD have said they will release drivers to fix it but really you can't even look at H.A.W.X. 2 and say the tessellator in the Radeon is not good enough.

As for AMD bottleneck causing devs not to develop a game with a lot of tessellation is nonsense. Nothing stopped crysis from going crazy in DX 10 when cards couldn't even handle that game.

Simply put devs will use as much tessellation as they want, they don't care if Radeon can't keep up with it and so far we have just been speculating everything.

 

[Skurge]

Lol that's one game out of like 20 and even then AMD have said they will release drivers to fix it but really you can't even look at H.A.W.X. 2 and say the tessellator in the Radeon is not good enough.

As for AMD bottleneck causing devs not to develop a game with a lot of tessellation is nonsense. Nothing stopped crysis from going crazy in DX 10 when cards couldn't even handle that game.

Simply put devs will use as much tessellation as they want, they don't care if Radeon can't keep up with it and so far we have just been speculating everything.

Unless someone pays them to put in a specific amount of tessellation.

 

[Scali]

Nvidia has had the same issues with different games in the past.

In a very distant past perhaps (GeForce FX?).
If nVidia doesn't have the same issues in the same game, then apparently AMD has driver issues... or has to cover up hardware issues in the drivers.

As for number 2, You do graphics so that might piss you off, but as a regular user it doesn't bother me at all.

You're not bothered by the fact that GPU vendors are fooling us all, and are manipulating games and benchmarks at the cost of image quality, just so they can pretend to have the better product?

There are tons of games that use tessellation and we don't see a disparity like that.

Think man, think!
Why do most current games not have problems with tessellation? Because most of them were developed for Radeons (lack of nVidia DX11 cards, remember?).

ALL synthetic benchmarks have shown huge performance differences in tessellation. Why are you in denial?
You think a game can not possibly run into the same bottlenecks as the synthetic tests? You think these bottlenecks don't exist because a 'game' is something different than a 'benchmark' to the GPU or something?
Ofcourse it can. The key is the tessellation factor. Most games are designed to run within the 'comfort zone' of AMD's hardware, which is very limited. As soon as you step outside it, as demonstrated by tons of synthetic tests, even AMD's own DX11 samples, performance tanks.

 

[Genx87]

Unless someone pays them to put in a specific amount of tessellation.

Poor AMD. They would get a fair shake of things if only Nvidia didnt pay developers to put in higher factor Tesselation, PhysX, Deferred AA for DX9 titles, and people writing apps for Cuda.

If only.......................

 

[Aristotelian]

You're not bothered by the fact that GPU vendors are fooling us all, and are manipulating games and benchmarks at the cost of image quality, just so they can pretend to have the better product?

If it were true, you'd expect there to be claims in these reviews that the 6800 series necessarily commits the user to a worse IQ experience than a 4XX or 5XXX series, right? Yet I haven't seen a single claim that expresses this. All I have seen are claims that AMD is now providing the option to choose a 'performance' mode that drops IQ for better framerates. How is that a bad thing?

 

[badb0y]

Scali, so the newer games should have a higher level of tessellation implemented right? Then all we can do is wait and check back later but simply going on about AMD tessellator not being up to snuff is worthless since 80% would disagree with you.

 

[AnandThenMan]

Poor AMD. They would get a fair shake of things if only Nvidia didnt pay developers to put in higher factor Tesselation, PhysX, Deferred AA for DX9 titles, and people writing apps for Cuda.

If only.......................

10:1 market share says AMD is doing just fine without have to worry about any of that.

 

[Genx87]

10:1 market share says AMD is doing just fine without have to worry about any of that.

Then they have it all figured out. Good to know. So why is Huddy whining about those topics again?

 

[Scali]

Scali, so the newer games should have a higher level of tessellation implemented right? Then all we can do is wait and check back later but simply going on about AMD tessellator not being up to snuff is worthless since 80% would disagree with you.

And that 80% would be ignoring factual proof provided by a number of synthetic tests, and would probably not understand what they're talking about anyway.
In short: they're simply wrong.

How many of you really understand how tessellation works anyway? How many of you have ever implemented a tessellation scheme in DX11? So how many of you have hands-on experience of what the hardware does, and how it performs?

 

[Gloomy]

Then they have it all figured out. Good to know. So why is Huddy whining about those topics again?

There's still 10% to go :thumbsup:

 

[AnandThenMan]

How many of you really understand how tessellation works anyway? How many of you have ever implemented a tessellation scheme in DX11? So how many of you have hands-on experience of what the hardware does, and how it performs?

You're just better than all of us dolts.

 

[Aristotelian]

Then they have it all figured out. Good to know. So why is Huddy whining about those topics again?

Where is the whining, exactly?

 

[Scali]

Where is the whining, exactly?

See the link in the opening post.

 

[Skurge]

In a very distant past perhaps (GeForce FX?).
If nVidia doesn't have the same issues in the same game, then apparently AMD has driver issues... or has to cover up hardware issues in the drivers.



You're not bothered by the fact that GPU vendors are fooling us all, and are manipulating games and benchmarks at the cost of image quality, just so they can pretend to have the better product?



Think man, think!
Why do most current games not have problems with tessellation? Because most of them were developed for Radeons (lack of nVidia DX11 cards, remember?).

ALL synthetic benchmarks have shown huge performance differences in tessellation. Why are you in denial?
You think a game can not possibly run into the same bottlenecks as the synthetic tests? You think these bottlenecks don't exist because a 'game' is something different than a 'benchmark' to the GPU or something?
Ofcourse it can. The key is the tessellation factor. Most games are designed to run within the 'comfort zone' of AMD's hardware, which is very limited. As soon as you step outside it, as demonstrated by tons of synthetic tests, even AMD's own DX11 samples, performance tanks.

No, It doesn't bother me If I don't see worse image quality, that's why I have catalyst AI on advanced at all times. If it did I would switch cards. Like when I replaced my FX5500 with a 9600 cause HL2 looked like crap on it.

Even on synthetic benchmarks the 6870 is at least as fast as a 460, why in this benchmark (that is heavily sponsored by nvidia), is the gap is so huge? A game that AMD said they could optimize the tessellation to be even faster on BOTH set of cards?

 

[Scali]

Like when I replaced my FX5500 with a 9600 cause HL2 looked like crap on it.

Lol, you bought an FX5500 card? What on earth were you thinking?

Even on synthetic benchmarks the 6870 is at least as fast as a 460

No it isn't. Did you actually bother to look at any of the benchmarks I mentioned?
The gap in TessMark for example is far larger even than in HAWX 2. And TessMark is not sponsored by nVidia. So HAWX 2 could be perfectly valid results.

 

[Lonbjerg]

You're just better than all of us dolts.

I am willing to bet than Scali writes better DX9/DX10/DX11 code than you do...wanna bet? :sneaky:

 

[Skurge]

Lol, you bought an FX5500 card? What on earth were you thinking?

Hey, no need to rub it in. I bought it on release ok. I also didn't do my research.

No it isn't. Did you actually bother to look at any of the benchmarks I mentioned?
The gap in TessMark for example is far larger even than in HAWX 2. And TessMark is not sponsored by nVidia. So HAWX 2 could be perfectly valid results.

Are there tessMark benches for the 6870?

The Problem I have is that tessmark ONLY have tessellation in it, nothing else. HAWX2 has other stuff that has to be rendered, which would bring the cards closer. Heaven and StoneGiant are the heaviest use of tesselation I have seen yet and the cards are equal. Is HAWX2 made up of 95% tessellation or what?

 

[Aristotelian]

See the link in the opening post.

If you think there is whining in the link in the opening post, you do not have an adequate grasp of the English language for us to have a discussion. He isn't whining. Perhaps you should see the link in the opening post again: you've posted so much in this thread it seems as if you've lost the plot.

You keep talking about 'AMD's problems' but I do not see them. Did you read Anandtech's 6850/6870 review? Do you think that because a synthetic benchmark can highlight the tessellation differences between AMD's and Nvidia's offerings that this renders AMD's competitive products irrelevant? I'd bet Anandtech's review itself will generate thousands of sales of the 6850/70 over a 460 1GB. Does that frustrate you, realizing that the point that you are arguing for (that nobody debates: Nvidia's cards have better tessellators) accounts for next to nothing in the real world? Of course, it makes for great fodder on message boards.

Finally, where is AMD's enormous problem given how they do with gaming framerates? You realize that people buy these cards to play...GAMES, right? Not to discuss theoretical impacts on what might be the case if more tessellation is used than less in the future. They buy cards to play games. Gaming benchmarks aren't a problem for AMD's cards. There are games where Nvidia's cards perform better and there are games where AMD's cards perform better. It just turns out to be the case that:

"As for the Radeon HD 6850 however, things are much more lopsided in AMD’s favor. It’s give and take depending on the benchmark, but ultimately it’s just as fast as the GTX 460 1GB on average, even though it’s officially $20 cheaper. And at the same time it draws less power and produces less noise than the GTX 460 1GB. In fact unless the GTX 460 1GB was cheaper than the 6850, we really can’t come up with a reason to buy it. For all the advantage of an overclock when going up against the 6870, the stock clocked card has nothing on the 6850. Even the GTX 460 768MB, while $10-$20 cheaper than the 6850, still has to contend with the fact that the 6850 is almost 10% faster and only marginally louder."

That's from the Anandtech 6850 review. Why isn't he saying that there is a glaring defect in the 6850 because it has a "weak" tessellator? These are important questions because your point seems to rest on an identity claim that gaming==tessellation, and that's empirically false.

 

[Scali]

Hey, no need to rub it in. I bought it on release ok. I also didn't do my research.

You basically just fell for the vendor telling you that the card is okay?
Just like AMD is now trying to tell us that their tessellation is okay?

Are there tessMark benches for the 6870?

Not that I know of...
But we know that the best-case scenario for 6870 is 2x the tessellation of 5870.
So doubling the 5870 scores would be best-case, and that still is slower than the 460.
I think it's pretty obvious that the 6870 isn't going to beat the 460 in practice.

The Problem I have is that tessmark ONLY have tessellation in it, nothing else.

I don't see the problem. It demonstrates a huge bottleneck.
While less pronounced in actual games (and the HAWX 2 benchmark), it's not going to disappear.
AMD's own slides already indicate the problem: above tessellation factor 11, the 6870 tanks back down to 5870 levels, and we already know that 5870 has been struggling when tessellation is involved.
Now you just need to convince yourself that the tessellation factor can go all the way up to 64x for a good reason. They didn't make it like that just so they could stop at a mere 11.
I already mentioned the example of using a 12-triangle cube as a basis for creating a perfect sphere by tessellating it into infinity... Try to work with that idea.

 

[Red Storm]

I'm not really in the business of trusting a benchmark from an incomplete, NVIDIA sponsered game that is recommended by NVIDIA on the eve of their competitor's new product launch. Call me crazy, that sounds a bit fishy. It's a heavy dose of NVIDIA PR, just like Huddy's remarks are a heavy dose of AMD PR. Thankfully most review sites seemed to agree.

When the game is actually out and playable, you can benchmark it to your heart's content and then if your claims hold true, it'll be interesting to see.