• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD beliefs: DirectX 11 Radeons pleasantly fast

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Originally posted by: evolucion8
Originally posted by: BFG10K
Yes, there have been changes to caches, memory and shaders, but the fundamental reason why the 4xxx absolutely destroys the 2xxx/3xxx is because of 800 vs 320 SPs, 40 vs 16 TMUs, and ROPs that are twice as fast and resolve the box modes in hardware.

That is to say, it?s largely the existing design, but massively beefed up.

So if ATi can remain competitive with next generation by doing exactly the same thing they did with the 4xxx series, why risk another 2900 XT?

But aren't the shaders/stream processors layout identical? I acknowledge the changes of the caches, the ROP, the new TMU which are much smaller and much more efficient, but didn't see a significant change in the shader layout.

Originally posted by: SirPauly
But he didn't say pleasantly fast -- he said, "really pleasant surprise" and the context was aimed at consumers/customers. In others words consumers will be really pleasantly surprised on the performance of this family to me. Not just surprised -- not just a pleasant surprise but a really pleasant surprise, hehe!🙂 He was adamant about offering awareness for the performance point based on the perception that this family is just added DirectX 11 compliance to some mind-sets -- but more so DirectX 11 compliance with strong performance to boot. That's what I did get out of his message.

Redundancy department of redundancy 🙂

It's not redundancy and a clear difference between "pleasantly fast" and "really pleasant surprise"

 
Originally posted by: BFG10K

Exactly what I said: a new architecture is not necessarily required to be competitive.

That depends on if your competitor has a new architecture or not.

The 4xxx series was mostly competition for the older G92 not so much with the newer GTX2xx series.

I'm sure if NVIDIA wanted to they could just patch DX11 onto the GTX200 series, add a die shrink and still remain comfortably ahead of ATI.
 
Originally posted by: evolucion8
[
But aren't the shaders/stream processors layout identical? I acknowledge the changes of the caches, the ROP, the new TMU which are much smaller and much more efficient, but didn't see a significant change in the shader layout.

Redundancy department of redundancy 🙂

the shaders themselves and the "SIMD cores" that encapsulate them are largely unchanged, but the sheer number of them has been significantly increased and with the special case of RV790, a spatially altered layout as well. back in the day of pipelining GPUs, doubling the number of your functional units and altering the compute:texture ratio was more than enough to call it a "new" architecture. why? because it sure as hell wasn't the old architecture. I'm not sure BFG10K could recognize a new architecture unless we were to get away from stream processing altogether, certainly if AMD were to not alter its shaders for this next generation. It depends on how fuzzy you want to be with the word. One could argue that the HD2400 and HD4890 are on the same architecture because their shaders are identical and they both support dx 10.1, which is often referred to as an architecture itself rather than an instruction set, but clearer arguments could be made for why they are not the same architecture (memory, cache, pin out, shaders per cluster per TMU per ROP). Can you put a RV790 into a RV630 board as a drop in upgrade? Where does one plug the GDDR2 bus into the GDDR5 controller? Similar compute regions do not constitute architectural unity, but rather an architectural sisterhood.


But when someone says a new architecture is not necessary for competition, it sounds like you're saying that a shrink to 40nm would get the 2900 XT back in the game, when the truth is that an underclocked passively cooled 4770 with a tiny GDDR5 bus would murder it.
 
Originally posted by: Wreckage
Originally posted by: BFG10K

Exactly what I said: a new architecture is not necessarily required to be competitive.

That depends on if your competitor has a new architecture or not.

The 4xxx series was mostly competition for the older G92 not so much with the newer GTX2xx series.

I'm sure if NVIDIA wanted to they could just patch DX11 onto the GTX200 series, add a die shrink and still remain comfortably ahead of ATI.

Yeah right, thank to that, nVidia had to dropped prices to remain competitive and had to launch a sandwich card to compete against the almighty HD 4870X2, you don't even believe your own lies.

http://www.anandtech.com/showdoc.aspx?i=3341&p=23

"but in our testing we definitely saw this $300 part perform at the level of NVIDIA's $400 GT200 variant, the GTX 260. This fact clearly sets the 4870 in a performance class beyond its price."

http://www.anandtech.com/showdoc.aspx?i=3408&p=10
Originally posted by: SirPauly
It's not redundancy and a clear difference between "pleasantly fast" and "really pleasant surprise"

But is pleasant to know that it's pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such pleasantly fast card!!
 
Originally posted by: evolucion8

It was a work that took 3 years, if something isn't broken, don't fix it, nVidia did the same with the GTX 280 and GTX 260 which has more similarities with their G92 counterparts than the RV770 against the RV670.

yes, i agree.
 
Originally posted by: Wreckage
Originally posted by: BFG10K

Exactly what I said: a new architecture is not necessarily required to be competitive.

That depends on if your competitor has a new architecture or not.

The 4xxx series was mostly competition for the older G92 not so much with the newer GTX2xx series.

I'm sure if NVIDIA wanted to they could just patch DX11 onto the GTX200 series, add a die shrink and still remain comfortably ahead of ATI.

How do you figure that?

There was a lack of competition from nVidia when the 4870x2 ruled the higher-end price-points. The 4870's strong performance and feature set dramatically redefined price-points and forced nVidia to lower pricing their newer generational sku's. How on Earth was the 4XXX series mostly for the G-92?

The 4850 Sku and eventually the 4830 Sku's were targeted for the G-92b sku's of the 9800GT and 9800GTX+/GTS250.

A brand new architecture doesn't instantly translate into success and depends on competitional elements. When the 5800 was released it offered more performance than nVidia's 4600 but had trouble competiting with ATI's offerings for that time-line and the same can be said with the ATI's 2900 when compared to nVidia's offerings.





 
Originally posted by: evolucion8
Originally posted by: Wreckage
Originally posted by: BFG10K

Exactly what I said: a new architecture is not necessarily required to be competitive.

That depends on if your competitor has a new architecture or not.

The 4xxx series was mostly competition for the older G92 not so much with the newer GTX2xx series.

I'm sure if NVIDIA wanted to they could just patch DX11 onto the GTX200 series, add a die shrink and still remain comfortably ahead of ATI.

Yeah right, thank to that, nVidia had to dropped prices to remain competitive and had to launch a sandwich card to compete against the almighty HD 4870X2, you don't even believe your own lies.

http://www.anandtech.com/showdoc.aspx?i=3341&p=23

"but in our testing we definitely saw this $300 part perform at the level of NVIDIA's $400 GT200 variant, the GTX 260. This fact clearly sets the 4870 in a performance class beyond its price."

http://www.anandtech.com/showdoc.aspx?i=3408&p=10
Originally posted by: SirPauly
It's not redundancy and a clear difference between "pleasantly fast" and "really pleasant surprise"

But is pleasant to know that it's pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such pleasantly fast card!!

Thank you for being so pleasant!🙂
 
Originally posted by: Wreckage
Originally posted by: BFG10K

Exactly what I said: a new architecture is not necessarily required to be competitive.

That depends on if your competitor has a new architecture or not.

The 4xxx series was mostly competition for the older G92 not so much with the newer GTX2xx series.

I'm sure if NVIDIA wanted to they could just patch DX11 onto the GTX200 series, add a die shrink and still remain comfortably ahead of ATI.

Why do you post such bs?
 
http://forums.anandtech.com/me...id=28&threadid=2311476
- here us an example from my (today's) last post to shed light on "pleasant":

But is pleasant to know that it's pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such pleasantly fast card!!


It is pleasant to know that Athlon II 250 X2 is pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such a pleasantly fast CPU!!
- cheap too

Even though it cannot hold a candle to the Phenom IIs and intel Quads
😛
:roll:

that is the "UNpleasant *part*
rose.gif

 
Originally posted by: evolucion8

Yeah right, thank to that, nVidia had to dropped prices to remain competitive and had to launch a sandwich card to compete against the almighty HD 4870X2, you don't even believe your own lies.

😕

You could say ATi had to launch an X2 just to have a high end card to compete with 280 at the time.


Are you Wreckage's ATi alter-ego?
 
Originally posted by: apoppin
http://forums.anandtech.com/me...id=28&threadid=2311476
- here us an example from my (today's) last post to shed light on "pleasant":

But is pleasant to know that it's pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such pleasantly fast card!!


It is pleasant to know that Athlon II 250 X2 is pleasantly fast and is a pleasant surprise to please the buyers to pleasantly spend their money on such a pleasantly fast CPU!!
- cheap too

Even though it cannot hold a candle to the Phenom IIs and intel Quads
😛
:roll:

that is the "UNpleasant *part*
rose.gif

Hehe, this topic has been so very pleasant!🙂
 
Originally posted by: OCguy
😕

You could say ATi had to launch an X2 just to have a high end card to compete with 280 at the time.


Are you Wreckage's ATi alter-ego?

Unlike Wreckage, I post links with real information backing up my posts, most games today the HD 4870 comes dangerously close GTX 280 and in some is even faster.

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=15 <<3-4fps slower average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=16 <<9fps slower average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=17 << 3fps slower average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=18 << 3fps faster average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=19 << 3fps slower average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=20 << 9fps slower average

http://www.anandtech.com/video/showdoc.aspx?i=3539&p=21 <<9fps faster average

so ATi launched an X2 to have the performance crown, and then nVidia launched the GTX 295 to have it back, simply as that. And now,

are you Wreckage's Lawyer or Bodyguard? Or some kind of Rollo's incarnation?
 
I think everyone should be able to agree that AMD redeemed themselves from the R600s. There should be no argument there. Speaking to the original 295 as some sandwiched abomination doesn't really work for me though, since it is a fast card and still more power efficient than a 4870 X2, let alone a 3870 X2. Also, speaking to the 4870 as if it was competing against G80/G92 is ludicrous. If that were the case, they would have stopped at the 4850 and not gone any farther. But I don't see how splitting hairs over the very, very well documented performance of these cards with misinformed individuals will allow us to extrapolate any valuable speculation about the new architectures... or am i the only one interested in RV870/G300?
 
I'm pleasantly waiting for the GT300/RV8X0 to see a more pleasant landscape of the pleasantly future of the pleasant DX11 and their pleasantly fast performance.
 
people don't buy cards for the instruction set. no one bought the 8800 to play dx10. they bought it because it was 200% faster than whatever 7 series or 500% faster than whatever 6-series they were upgrading from.
 
Originally posted by: evolucion8

Unlike Wreckage, I post links with real information backing up my posts, most games today the HD 4870 comes dangerously close GTX 280 and in some is even faster.

Oh really.

http://www.techreport.com/articles.x/16681/1

4870 vs GTX280 2560x1600

FarCry 2 4.7 fps slower
Left4Dead 19.7 fps slower
CoD 14.5 fps slower
Fallout 3 3.8 fps slower
Crysis Warhead 4.7 fps slower

The 4870 is slower in every game. Mostly breaking even or falling behind the GTX260. Which meant the rest of the 4xxx lineup was competing against the old G92 series.

In several games it falls below 30fps which many consider unplayable.
 
wreckage, even if your data were not hilariously shamelessly gratuitous by selecting the unconventional, even superfluous resolution of 2560x1600 as your basis for comparison, you fail to understand that simply because a given radeon performs slower than a given geforce by 2 or 3 percent does not imply that the slower card was intended to compete with hardware from previous years. The Radeon 4890 in the article is well within the margin of error compared to the GTX 285. The differences at conventional resolutions such as 1920x1200 are vanishingly small, yet the 4890 can be had for under $200. most people understand that a 3% performance disadvantage can be negated entirely by 50% or greater reduced costs and effortlessly mitigated by a 3% overclock.
 
Originally posted by: alyarb
wreckage, even if your data were not hilariously shamelessly gratuitous by selecting the unconventional, even superfluous resolution of 2560x1600 as your basis for comparison, you fail to understand that simply because a given radeon performs slower than a given geforce by 2 or 3 percent does not imply that the slower card was intended to compete with hardware from previous years. The Radeon 4890 in the article is well within the margin of error compared to the GTX 285. The differences at conventional resolutions such as 1920x1200 are vanishingly small, yet the 4890 can be had for under $200. most people understand that a 3% performance disadvantage can be negated entirely by 50% or greater reduced costs and effortlessly mitigated by a 3% overclock.

2 or 3%?

Left4Dead 19.7 fps slower
CoD 14.5 fps slower


Hardly.

How about another review. This time using a slower GTX260
http://www.bjorn3d.com/read.php?cID=1414

We weren't expecting it, but the Nvidia BFG GTX-260 Core 216 (MaxCore) made it a clean sweep. With every game at every resolution, it took the lead and kept it. If this were a Shootout at the OK Corral the ATI GPU would be sprawled in a pool of blood with five holes in its core and the Nvidia GPU would be left standing unscathed.

In fact most review sites put the 4870 behind the GTX260 and nowhere near the 280.
http://www.hardocp.com/article...wxLCxoZW50aHVzaWFzdA==
 
Originally posted by: Wreckage
Originally posted by: evolucion8

Unlike Wreckage, I post links with real information backing up my posts, most games today the HD 4870 comes dangerously close GTX 280 and in some is even faster.

Oh really.

http://www.techreport.com/articles.x/16681/1

4870 vs GTX280 2560x1600

FarCry 2 4.7 fps slower
Left4Dead 19.7 fps slower
CoD 14.5 fps slower
Fallout 3 3.8 fps slower
Crysis Warhead 4.7 fps slower

The 4870 is slower in every game. Mostly breaking even or falling behind the GTX260. Which meant the rest of the 4xxx lineup was competing against the old G92 series.

In several games it falls below 30fps which many consider unplayable.

Fair enough... but for 99% of us, who do not use a 30" display the single GPU 48x0 cards compete just fine with the single GPU GTX2xx cards.

And you continue to ignore the 4850x2, which is cheaper and faster than the GTX280 and competes well with the GTX285 while being much cheaper (Newegg current prices).

From your link. 4850x2 vs. GTX280 2560/4xAA.

FarCry 2 - 4850x2 is 7.7FPS faster (faster than GTX285 too)
Left4Dead - 4850x2 is 6FPS faster.
CoD:WoW - 4850x2 is 1.5FPS faster.
Fallout 3 - 4850x2 is 16.5FPS faster. (faster than GTX285 too)
Crysis Warhead - 4850x2 is 1.5FPS faster.

Looking through the benches with 1680 res and 1920 res (you know, the resolutions the vast majority of us use) it's pretty easy to see that the 4870 1GB has no problem competing with the GTX260 and sometimes the GTX280 as well.



 
Originally posted by: SlowSpyder

Fair enough... but for 99% of us, who do not use a 30" display the single GPU 48x0 cards compete just fine with the single GPU GTX2xx cards.
That's a generalization. I could also say that 99% of people don't buy a high end card to run at low resolutions.

And you continue to ignore the 4850x2, which is both cheaper and faster than the GTX280 and GTX285 (Newegg current prices).
Yes I did ignore 2 GPUs vs 1 GPU.

We are talking about the next gen GPU. That's single not plural. I could dig up SLI numbers as well I suppose. Spin it how you like, the facts are stacked against you.
 
http://www.guru3d.com/article/...n-hd-4890-pcs-review/9 <<HD 4870 faster than GTX 260
http://www.guru3d.com/article/...-hd-4890-pcs-review/10 << GTX 260 faster than HD 4870
http://www.guru3d.com/article/...-hd-4890-pcs-review/11 <<HD 4870 faster than GTX 260
http://www.guru3d.com/article/...-hd-4890-pcs-review/12 <<HD 4870 faster than GTX 260
http://www.guru3d.com/article/...-hd-4890-pcs-review/13 <<Tie
http://www.guru3d.com/article/...-hd-4890-pcs-review/14 <<Tie
http://www.guru3d.com/article/...-hd-4890-pcs-review/15 << Tie
http://www.guru3d.com/article/...-hd-4890-pcs-review/16 <<GTX 260 faster than HD 4870

It seems that the GTX 260 is competitive with the G92 chip, since it can match the HD 4870 which is mean to compete with the G92 like Wreckage posted, so much for competition loll
 
Originally posted by: Wreckage
Originally posted by: SlowSpyder

Fair enough... but for 99% of us, who do not use a 30" display the single GPU 48x0 cards compete just fine with the single GPU GTX2xx cards.
That's a generalization. I could also say that 99% of people don't buy a high end card to run at low resolutions.

And you continue to ignore the 4850x2, which is both cheaper and faster than the GTX280 and GTX285 (Newegg current prices).
Yes I did ignore 2 GPUs vs 1 GPU.

We are talking about the next gen GPU. That's single not plural. I could dig up SLI numbers as well I suppose. Spin it how you like, the facts are stacked against you.

Or you could say that Nvidia needs 2x as much silicon on a single GPU... two 4850's are the about the same size as a single GTX285 GPU. There's a lot of different ways to get to a performance goal... 470mm2 or 2 x 250mm2 are examples of that.

It could be said just as easily that it's a generalization that the 48x0 cards were not meant to compete with the GTX2x0 cards based on one monitor size that most people don't use. What size monitor do you run? What if the 4870 cleaned up at 1024 res? It's just not a res most people use. It may be important for a small minority of people, but not most of us. Most people aren't spending $1100+ on a display to game on it with a sub $200 card...
 
Tend to think that the GTX-260 216 Sku is comparable from a performance stand-point to the 4870 1 gig sku with both having advantages based on resolution, settings, title offered. Tend to think that the GTX-275 Sku is comparable from a performance stand-point of the 4890 Sku with both having advantages based on resolution, setttings, title offered. In other words, if one looks around, one may find reviews where one may be deemed superior.

I don't think that the 4870 1 gig sku is comparable to the 9800GTX+ or 9800GT over-all -- sure, in some tests, but over-all? No.

Even when we take testing based on reviews -- what about x8 multi-sampling performance from an ATI view ? Then I will use a nVidia view; why not compare performance with CFAA vs CSAA?



 
There's also some scenarios which the 9800GTX+ matches the GTX 260 in performance, but usually those are not very demanding scenarios, the latter is much more powerful.
 
Back
Top