HD 2900XT review

[HurleyBird]

Originally posted by: BFG10K
Performance only drops under the static path. Under dynamic there is no performance drop because the slider does nothing.

In any case the slider doesn't do AA like was presented in the article.

Actually look at the results people. Stalker is the only benchmark that doesn't really change FPS when AA modes are changed. Sure, it was stupid of them to test stalker with AA (yet alone two AA modes), but it doesn't falsify the benchmark results. The picture with the R600's is a good indication that the results aren't made up.


As for the actual benchmark results, wow, very weird and interesting. If this is a good indication of the final product I have no idea which card I would want to buy. While most of the results were crap, the AA + high res results were impressive. And then there's the DX10 question...

 

[tuteja1986]

Originally posted by: swtethan

Originally posted by: Nightmare225

Originally posted by: tuteja1986

Originally posted by: ronnn
Yep that review really slammed the 2900xt, if accurate amd is out of the graphic business.

You make it sound like 2900XT is priced arround 8800GTX level. it priced at $400 and is a real compititor to 8800GTS 640MB. IT also has loads more feature and better driver for vista. Also pre early DX10 benchmark show it beating up 8800GTX in crysis. Your just blined by your love for nvidia that you start to talk all the crap nvidia throws at you. Nothing more than a fanatical fanboy.

Certain others are blinded for their love of ATI and take whatever positive "review" they can find as godly truth.

You know what I mean.

EDIT: Love how all the positive benchmarks are put at the top in order to make ATI fans want to jump on it. But then they get to the bottom, and... Who cares, these aren't even real... :disgust:

lol i didnt even see that

IDIOT I LISTED THEM IN ORDER HOW THEY WERE LISTED ON THE WEBPAGE !

 

[CaiNaM]

Originally posted by: Nightmare225
BTW, STALKER can't do AA.

AA works fine if you turn dynamic lighting off

 

[apoppin]

Originally posted by: BFG10K
Bad news if those results are true.

ATI has a 8x AA mode?

Yes, according to the slides they have 8xMSAA like nVidia.

you can move the slider all the way to the right and performance will definitely drop

Performance only drops under the static path. Under dynamic there is no performance drop because the slider does nothing.

In any case the slider doesn't do AA like was presented in the article.

well i'll be damned ... you ARE right ... i was totally under the wrong impression that AA slowed perf even though it doesn't 'do much' in the visuals

--that shoots one of my 'theories' all to hell
*zero idea* why my x1950p runs it so well


i just went ahead and tried it ... along a railing ... there wasn't any noticeable difference with AA off or maxed

unfortunately there wasn't any difference in the visuals ... maybe ...
.... but there really isn't much *need* for AA ... the 'jaggies' are not noticeable, anyway

have you tried "forcing" AA in the CCC? ... i know it doesn't enable AA, but is there any perf difference?

 

[Raider1284]

Originally posted by: Fox5
This benchies could be true, they're run at such resolutions and AA settings that the memory bus of the R600 would really come into play over the g80.

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period! anyone that believes the 2900xt can somehow gain fps while the 8800 series is losing 15+ fps is out of their mind.

we are still waiting patiently for a real benchmark.

 

[ShadowOfMyself]

Originally posted by: Raider1284

Originally posted by: Fox5
This benchies could be true, they're run at such resolutions and AA settings that the memory bus of the R600 would really come into play over the g80.

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period! anyone that believes the 2900xt can somehow gain fps while the 8800 series is losing 15+ fps is out of their mind.

we are still waiting patiently for a real benchmark.

QFT! Its not about HD2900XT underperforming, its about the numbers making ZERO sense... Its physically IMPOSSIBLE for the HD2900XT to gain fps while the mighty GTX drops THAT bad

I refuse to believe that 512 bit bus suddendly makes the card gain performance instead of dropping 20-30 fps like the GTX

 

[coldpower27]

Originally posted by: ShadowOfMyself

Originally posted by: Raider1284

Originally posted by: Fox5
This benchies could be true, they're run at such resolutions and AA settings that the memory bus of the R600 would really come into play over the g80.

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period! anyone that believes the 2900xt can somehow gain fps while the 8800 series is losing 15+ fps is out of their mind.

we are still waiting patiently for a real benchmark.

QFT! Its not about HD2900XT underperforming, its about the numbers making ZERO sense... Its physically IMPOSSIBLE for the HD2900XT to gain fps while the mighty GTX drops THAT bad

I refuse to believe that 512 bit bus suddendly makes the card gain performance instead of dropping 20-30 fps like the GTX

it might not be applying AA in the high resolution or something, who knows...

 

[palindrome]

Originally posted by: coldpower27

Originally posted by: ShadowOfMyself

Originally posted by: Raider1284

Originally posted by: Fox5
This benchies could be true, they're run at such resolutions and AA settings that the memory bus of the R600 would really come into play over the g80.

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period! anyone that believes the 2900xt can somehow gain fps while the 8800 series is losing 15+ fps is out of their mind.

we are still waiting patiently for a real benchmark.

QFT! Its not about HD2900XT underperforming, its about the numbers making ZERO sense... Its physically IMPOSSIBLE for the HD2900XT to gain fps while the mighty GTX drops THAT bad

I refuse to believe that 512 bit bus suddendly makes the card gain performance instead of dropping 20-30 fps like the GTX

it might not be applying AA in the high resolution or something, who knows...

I know it is crap

/end thread

 

[Griswold]

Originally posted by: Acanthus

Originally posted by: tuteja1986

Originally posted by: ronnn
Yep that review really slammed the 2900xt, if accurate amd is out of the graphic business.

You make it sound like 2900XT is priced arround 8800GTX level. it priced at $400 and is a real compititor to 8800GTS 640MB. IT also has loads more feature and better driver for vista. Also pre early DX10 benchmark show it beating up 8800GTX in crysis. Your just blined by your love for nvidia that you start to talk all the crap nvidia throws at you. Nothing more than a fanatical fanboy.

I strongly disagree with your assertion that the X2900 will have better drivers. You dont know that.

ATi is in the same situation nvidia was, new OS (actually 2 if you count x86-64), new API, new driver model, new hardware.

AMD may be in the same boat but their boat seems to have an engine that works. AMD is actually releasing drivers on a monthly basis, whereas nvidia is feeding us with beta garbage and WHQL drivers that - at least for me - function worse than any previous version (I am still using 100.41 because any follow-up driver breaks what worked before without fixing what didnt work). I cant say that for my other vista box with a humble x1600 in it.

Because of that, I cant wait to kick my 7800GT to the curb and replace it with a nice HD2600whatever card as soon as they're available.

 

[BFG10K]

AA works fine if you turn dynamic lighting off

In-game AA works but driver AA does not. Also in-game AA is a shader implementation and not comparable to driver AA modes.

have you tried "forcing" AA in the CCC? ... i know it doesn't enable AA, but is there any perf difference?

On my 8800 GTS driver AA can have a performance impact but it never impacts visuals.

DXTweaker also has a performance impact but no impact on image quality except for a possible marginal improvement in tree leaves but even zoomed screenshots side by side cannot confirm for sure.

Thus for all intents and purposes the only way to get AA in Stalker is static lighting combined with in-game AA.

 

[Nightmare225]

Originally posted by: tuteja1986

Originally posted by: swtethan

Originally posted by: Nightmare225

Originally posted by: tuteja1986

Originally posted by: ronnn
Yep that review really slammed the 2900xt, if accurate amd is out of the graphic business.

You make it sound like 2900XT is priced arround 8800GTX level. it priced at $400 and is a real compititor to 8800GTS 640MB. IT also has loads more feature and better driver for vista. Also pre early DX10 benchmark show it beating up 8800GTX in crysis. Your just blined by your love for nvidia that you start to talk all the crap nvidia throws at you. Nothing more than a fanatical fanboy.

Certain others are blinded for their love of ATI and take whatever positive "review" they can find as godly truth.

You know what I mean.

EDIT: Love how all the positive benchmarks are put at the top in order to make ATI fans want to jump on it. But then they get to the bottom, and... Who cares, these aren't even real... :disgust:

lol i didnt even see that

IDIOT I LISTED THEM IN ORDER HOW THEY WERE LISTED ON THE WEBPAGE !

Do you think my comment was directed at your post? Don't be so self-centered...

 

[v8envy]

Originally posted by: Raider1284

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period!

Remember the ti4200? It gave better frame rates at 1024x768 than at 800x600 -- the actual silicon was optimized to render faster at 10x7.

Software engineers will tell you that it's not possible to optimize for every possible case, but if you can restrict your domain to a small subset of possibilities you can sometimes come up with shortcuts not applicable in the general case. So in the optimal case you will run MUCH faster than in an intuitively easier case which falls into the general category.

So I'm just saying: it is absolutely possible to run faster at one single resolution than all the others, even though it's got more pixels which require pushing.

Besides, the HD2900 has a huge pile of memory bandwidth. Presumably it's got ROP power to match. One theme all leaked reviews have in common is better preformance at ludicrous resolution.

 

[nullpointerus]

Originally posted by: v8envy

Originally posted by: Raider1284

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period!

Remember the ti4200? It gave better frame rates at 1024x768 than at 800x600 -- the actual silicon was optimized to render faster at 10x7.

Software engineers will tell you that it's not possible to optimize for every possible case, but if you can restrict your domain to a small subset of possibilities you can sometimes come up with shortcuts not applicable in the general case. So in the optimal case you will run MUCH faster than in an intuitively easier case which falls into the general category.

So I'm just saying: it is absolutely possible to run faster at one single resolution than all the others, even though it's got more pixels which require pushing.

Besides, the HD2900 has a huge pile of memory bandwidth. Presumably it's got ROP power to match. One theme all leaked reviews have in common is better preformance at ludicrous resolution.

Yeah, but we're talking 3fps faster at 2560x1600 than 1900x1200--huge gap in pixel count there, and not too many people run at the ultra high resolution. If we assume that the article isn't just FUD, then the most likely explanations are: (a) there were driver bugs, or (b) the reviewer screwed up by mixing up a few of the resolution inputs when he generated the chart graphics.

 

[apoppin]

Originally posted by: nullpointerus

Originally posted by: v8envy

Originally posted by: Raider1284

NO THESE BENCHMARKS ARE NOT TRUE!!!!! fps does not go up as you increase resolution, it doesnt happen period!

Remember the ti4200? It gave better frame rates at 1024x768 than at 800x600 -- the actual silicon was optimized to render faster at 10x7.

Software engineers will tell you that it's not possible to optimize for every possible case, but if you can restrict your domain to a small subset of possibilities you can sometimes come up with shortcuts not applicable in the general case. So in the optimal case you will run MUCH faster than in an intuitively easier case which falls into the general category.

So I'm just saying: it is absolutely possible to run faster at one single resolution than all the others, even though it's got more pixels which require pushing.

Besides, the HD2900 has a huge pile of memory bandwidth. Presumably it's got ROP power to match. One theme all leaked reviews have in common is better preformance at ludicrous resolution.

Yeah, but we're talking 3fps faster at 2560x1600 than 1900x1200--huge gap in pixel count there, and not too many people run at the ultra high resolution. If we assume that the article isn't just FUD, then the most likely explanations are: (a) there were driver bugs, or (b) the reviewer screwed up by mixing up a few of the resolution inputs when he generated the chart graphics.

the point is" *they will* ... that IS what DX10 brings .... super-high resolutions impossible with DX9 and with less performance hit

the reviewer himself mentions there are likely driver bugs and they tested continuously for 100 hours

all these leaks 'hint' to *one thing* r600's DX10 possible performance advantages

what is it? 4 more days?

isn't the reviewer formerly associated with theInq ..

 

[Pabster]

Originally posted by: tuteja1986
You make it sound like 2900XT is priced arround 8800GTX level. it priced at $400 and is a real compititor to 8800GTS 640MB. IT also has loads more feature and better driver for vista. Also pre early DX10 benchmark show it beating up 8800GTX in crysis. Your just blined by your love for nvidia that you start to talk all the crap nvidia throws at you. Nothing more than a fanatical fanboy.

Assuming $399 is the launch price, it will be a direct competitor to the 8800GTS 640MB.

However, your point about ATI drivers is laughable. ATI has never been superb in the driver department and their latest Vista effort is hardly a slam dunk. Both camps have a lot of work to do.

PS pointing another out as a "fanboy" is going to make the mirror shatter.

 

[yacoub]

8800GTS 640MB already down to $375 w/o rebate. By the time ATi finally launches it'll be closer to $350.

That means if the performance is similar you will see MIRs on the 2900 in short order though you won't see an actual price drop for a while.

 

[apoppin]

think this one is *FaKE*


How about *PART TWO*

http://it-review.net/index.php?option=com_content&task=view&id=1325&Itemid=91

Well, some goblins kicked in and we were forced to take some more time before posting this one - and do some additional testing. It's time to check how R600 performs on Intel's 975 chipset. We actually picked these four platforms to broaden the range of possibilities for our readers so they can know what to expect from this card on their own computer. We received quite a few e-mails from people asking us to test R600 on some other platforms as well and we will, as soon as the fourth dimension allows us to do so. So, here we go again, R600-land...

[and their conclusion for those who want to skip the charts]

Page 11 of 11

As you can see for yourself, R600 performs quite a bit better on Intel's 975 chipset then on NVIDIA's NF590 SLI, to a point where we are really pleasantly surprised. We can only see one pretty strange result here, in high-res/high-AA/AF FEAR test for 8800GTS 640, but that's just about everything strange. We feel that, although we know that there are new drivers coming out, AMD/ATi really still must make some serious efforts on the driver side, where they were traditionally good. If latest news we heard are even remotely close to the truth, things are looking pretty good then. We also noticed that at lower resolutions R600's seem "a bit more CPU limited" then 8800, which is very interesting. When you crank R600 up with higher resolutions (1920x1200 and 2560x1600) and 8x/16xAA settings, this thing starts breathing and doing the job right. We also removed Company of Heroes because it really behaves strangely so it's out until we figure out what's going on.

There's one other "trick" here, and that's image quality, also "to be covered" in a separate article. In all honesty, 8 and 16xAA modes look a bit better on HD2900XT then on 8800 series. This is one of those things that make HD2900XT what it is - very interesting product. Stay tuned for CrossFire vs SLI article tomorrow, done on 975 chipset for CrossFire and 680i chipset for 8800's.

flame on ...

 

[ShadowOfMyself]

Originally posted by: apoppin
think this one is *FaKE*


How about *PART TWO*

http://it-review.net/index.php?option=com_content&task=view&id=1325&Itemid=91

Well, some goblins kicked in and we were forced to take some more time before posting this one - and do some additional testing. It's time to check how R600 performs on Intel's 975 chipset. We actually picked these four platforms to broaden the range of possibilities for our readers so they can know what to expect from this card on their own computer. We received quite a few e-mails from people asking us to test R600 on some other platforms as well and we will, as soon as the fourth dimension allows us to do so. So, here we go again, R600-land...

[and their conclusion for those who want to skip the charts]

Page 11 of 11

As you can see for yourself, R600 performs quite a bit better on Intel's 975 chipset then on NVIDIA's NF590 SLI, to a point where we are really pleasantly surprised. We can only see one pretty strange result here, in high-res/high-AA/AF FEAR test for 8800GTS 640, but that's just about everything strange. We feel that, although we know that there are new drivers coming out, AMD/ATi really still must make some serious efforts on the driver side, where they were traditionally good. If latest news we heard are even remotely close to the truth, things are looking pretty good then. We also noticed that at lower resolutions R600's seem "a bit more CPU limited" then 8800, which is very interesting. When you crank R600 up with higher resolutions (1920x1200 and 2560x1600) and 8x/16xAA settings, this thing starts breathing and doing the job right. We also removed Company of Heroes because it really behaves strangely so it's out until we figure out what's going on.

There's one other "trick" here, and that's image quality, also "to be covered" in a separate article. In all honesty, 8 and 16xAA modes look a bit better on HD2900XT then on 8800 series. This is one of those things that make HD2900XT what it is - very interesting product. Stay tuned for CrossFire vs SLI article tomorrow, done on 975 chipset for CrossFire and 680i chipset for 8800's.

flame on ...

One of the most ridiculous things about all these "benchmarks" is how the HD2900XT gets stomped by the 8800GTX in a game, and then stomps back in another... WTF?

Example:

STALKER

HD2900XT - 7
8800GTX - 28

Prey

HD2900XT - 34
8800GTX - 15

 

[apoppin]

i think it is the drivers

i tend to think this review is 'real' ... you would not "manufacture" benches that look like this

my x1950p probably DOES beat HD2900xt in STALKER ...
-for now

 

[apoppin]

i found this review originally and i also found this

http://uk.theinquirer.net/?article=39502

A SITE RUN BY a former colleague of mine just broke out with a first review of R600, now know as Radeon HD2900XT. The site does not go in depth about the cards themselves, and some results turned out weirdly, as it was mentioned in their conclusion. Since they invested 100 or so hours in reviewing these monsters, it seems to us that they have the boards for quite some time.

so it is kinda like theInq's DT

 

[Drayvn]

Originally posted by: apoppin
i found this review originally and i also found this

http://uk.theinquirer.net/?article=39502

A SITE RUN BY a former colleague of mine just broke out with a first review of R600, now know as Radeon HD2900XT. The site does not go in depth about the cards themselves, and some results turned out weirdly, as it was mentioned in their conclusion. Since they invested 100 or so hours in reviewing these monsters, it seems to us that they have the boards for quite some time.

so it is kinda like theInq's DT

They have also had the boards for quite some time with over 100 hours put into it, so it might not even be a retail version but an engineering sample or not the right version of the board thats supposed to be released.

But again its all speculation until the NDA is opened

 

[Wreckage]

Originally posted by: apoppin
i think it is the drivers

That's impossible.Their drivers can't suck, they just can't. I won't listen....noooooooooo!

 

[apoppin]

Originally posted by: Wreckage

Originally posted by: apoppin
i think it is the drivers

That's impossible.Their drivers can't suck, they just can't. I won't listen....noooooooooo!

come on now

don't be coy ...

i know you want them to suck :lips:


you don't need to listen ... open your eyes
:Q

 

[apoppin]

http://it-review.net/index.php?option=com_content&task=view&id=1325&Itemid=91

up for Part 2

fake?

i think not

 

[tuteja1986]

Originally posted by: apoppin

Originally posted by: Wreckage

Originally posted by: apoppin
i think it is the drivers

That's impossible.Their drivers can't suck, they just can't. I won't listen....noooooooooo!

come on now

don't be coy ...

i know you want them to suck :lips:


you don't need to listen ... open your eyes
:Q

Your forgetting the important rule of fanboy, they are not def but blind.