ATi 4870 X2 (R700) previews thread

[AzN]

It seems 4870x2 behaves like 1gb vram card and not a 512mb ones we've seen in the past. Looking at GRID in anandtech's articles tells me memory is better managed than Crossfire. @ 2560 and 4xAA x2 is getting 84.2fps while the crossfire 4870 is tanked to 30.4fps.

 

[ArchAngel777]

Originally posted by: 5150Joker
Minimum fps is what counts to me and the R700 falls behind the GTX 280 in several instances. Oh well I doubt my damn shuttle could even handle a GTX 280 even if I wanted one.

Long time no see Joker! I hope all is going well.

I am in agreement with you. I purchased a 4870 only to cancel my order after looking closely at the minimum FPS. I know this revolves around the x2, but after checking some of these reviews, it confirms my thoughts. Many times the minimum frames per second isn't much better than the single 4870.

I wish more reviews would focus on minimum frame rates, that is what really matters when it comes to gaming, not '500fps while looking at the sky' thus skewing the average frame rate results.

Anyway, I will say this looks very promising though... I am rather impressed with the R700 to be honest. It isn't for me until it can scale properly with all games, and confirm that micro stutter is gone, in addition to minimum frame rates going up.

 

[Rebel44]

Originally posted by: BFG10K

Originally posted by: lopri

AMD shipped out ten R700 cards worldwide[/b], attempting to capitalize on the success of the 4800 and showcase the strength of AMD's small-GPU strategy.

Hmm, ten cards but only seven reviews. Maybe we should have a competition to find the other three.

HardOCP has 2

 

[Extelleron]

It looks good but not as good as I thought it would be. Crysis performance is just not that impressive at all.... Crossfire scaling needs to be seriously improved.

That's not to say that there aren't games where the 4870 X2 demolishes the GTX 280, but mostly those are games where more performance is not needed. In Crysis, where we need more performance, the 4870 X2 is pretty much equal with the GTX 280.

The power consumption is also pretty insane - normally I don't care, as you can see I bought a GTX 280 and it is very power hungry. But the 4870 X2 is off the charts in terms of power consumption; it would not even work with my power supply. Consuming 100W more than the GTX 280 which is already a power hog is not a good thing.

Overall, considering the price I paid for the 280, I don't regret the purchase. The 4870 X2 performs better on average, but not where the performance is actually needed, and Crossfire scaling cannot always be counted on. And the power consumption means that even if I wanted to buy one, I could not without upgrading the PSU as well.

nVidia could get the crown back with a 55nm GT200 clocked at 700MHz+, for me at least. In certain games the 4870 X2 is and always will be untouchable by a GT200 part, but in other games (Crysis or other games with bad CF scaling) all the GTX 280 needs is a slight boost in performance to be back on top. If nVidia were able to set up a new SKU like the Leadtek Extreme, which is clocked @ 738/1566/2520, that would be enough to be better than R700 where it really matters. What AMD needs to do is improve their Crossfire drivers. The scaling across the board is just not as consistantly good as with SLI. You can't rely on CF to scale well.

 

[ShadowOfMyself]

Originally posted by: ArchAngel777

Originally posted by: 5150Joker
Minimum fps is what counts to me and the R700 falls behind the GTX 280 in several instances. Oh well I doubt my damn shuttle could even handle a GTX 280 even if I wanted one.

Long time no see Joker! I hope all is going well.

I am in agreement with you. I purchased a 4870 only to cancel my order after looking closely at the minimum FPS. I know this revolves around the x2, but after checking some of these reviews, it confirms my thoughts. Many times the minimum frames per second isn't much better than the single 4870.

I wish more reviews would focus on minimum frame rates, that is what really matters when it comes to gaming, not '500fps while looking at the sky' thus skewing the average frame rate results.

Anyway, I will say this looks very promising though... I am rather impressed with the R700 to be honest. It isn't for me until it can scale properly with all games, and confirm that micro stutter is gone, in addition to minimum frame rates going up.

Tuteja just posted tons of benchmarks which show the X2 having higher mins than the GTX 280 all the time

 

[Kuzi]

Wow great performance. Hope it's ok tuteja1986, I stole your post and added percentages

http://www.anandtech.com/video/showdoc.aspx?i=3354&p=4

Anandtech's Benchamrk

Age of Conan 2560x1600 4xAA/16xAF
4870 X2 : 44.8 --> 97.3% faster
GT280 : 22.7

Crysis 1920x1200
4870 X2 : 39.1 --> 13.9% faster
GT280 : 34.3

Oblivion 2560x1600 4xAA 16xAF
4870 X2 : 50.3 --> 36.6% faster
GT280 : 36.8

Grid 2560x1600 4xAA
4870 X2 : 84.2 --> 100.7% faster
GT280 : 40.6

========================================

http://techreport.com/articles.x/15105/3

Techreport's Benchmark

Half life 2 episode 2 2560x1600 4xAA/16xAF
4870 X2 : 84.6 --> 29.1% faster
GT280 : 65.5

Quake War 2560x1600 4xAA/16xAF
4870 X2 : 100.2 --> 35% faster
GT280 : 74.2 /

Crysis 1920x1200
4870 X2 : 24.8 average /17 min -> 22.1% faster avg / 0% min
GT280 : 20.3 average /17 min

Race Driver 4xAA 1920x1200
4870 X2 : 111.9 average / 77.0 min -> 65.5% faster avg / 37.5% faster min
GT280 : 67.6 average / 56.0 min

========================================

http://www.driverheaven.net/re...?reviewid=588&pageid=1

Driver Heaven's Benchmark

Call of Duty 4 - 2560x1200 4xAA 16xAF
4870 X2 : 90 average / 49 min -> 38.4% faster avg / 48.4% faster min
XFX GT280 XXX : 65 average / 33 min

World in Conflict - 2560x1200 4xAA 16xAF
4870 X2 : 61 average / 47 min -> 38.6% faster avg / 23.6% faster min
XFX GT280 XXX : 44 average / 38 min

HL 2 Episode 2 2560x1600 8xAA
4870 X2 : 110 average / 30 min -> 59.4% faster avg / 3.4% faster min
XFX GT280 XXX : 69 average / 29 min

========================================

Average: 48.7% faster than GTX280

The R700 is pretty solid with the games tested, maybe could be a little faster in Crysis. The card seems to consume a little less power than two 4870 cards in CF, and will be cheaper while offering similar performance. So 4870X2 > 4870CF.

Nvidia needs 55nm GT200 GPUs to compete with this card, 4870X2 is simply much faster than the GTX280. But lets see how it performs with other games.

 

[ArchAngel777]

Originally posted by: ShadowOfMyself

Originally posted by: ArchAngel777

Originally posted by: 5150Joker
Minimum fps is what counts to me and the R700 falls behind the GTX 280 in several instances. Oh well I doubt my damn shuttle could even handle a GTX 280 even if I wanted one.

Long time no see Joker! I hope all is going well.

I am in agreement with you. I purchased a 4870 only to cancel my order after looking closely at the minimum FPS. I know this revolves around the x2, but after checking some of these reviews, it confirms my thoughts. Many times the minimum frames per second isn't much better than the single 4870.

I wish more reviews would focus on minimum frame rates, that is what really matters when it comes to gaming, not '500fps while looking at the sky' thus skewing the average frame rate results.

Anyway, I will say this looks very promising though... I am rather impressed with the R700 to be honest. It isn't for me until it can scale properly with all games, and confirm that micro stutter is gone, in addition to minimum frame rates going up.

Tuteja just posted tons of benchmarks which show the X2 having higher mins than the GTX 280 all the time

Depends on the benchmark and site that reported it. There were several situations where it hadn't improved more than 5% over the single 4870, while the average framerate nearly doubled. I am sure there are many benchmarks to prove either way. Until more hard numbers are tested and cross references, I am not willing to to say confidently one way or the other.

But I will tell you that the Xbit labs review with the 4870 has the 9800GTX beating it, or on par with it in the minimum frame rate department. If the 4870 struggles with mins as a single card, I can only imagine that it will struggle the likewise with the x2. Apop seems to think it is the drivers and he might be right, in fact, he probably is. But that brings us back to the point that the card is only as good as the drivers...

 

[tuteja1986]

Originally posted by: Extelleron
It looks good but not as good as I thought it would be. Crysis performance is just not that impressive at all.... Crossfire scaling needs to be seriously improved.

That's not to say that there aren't games where the 4870 X2 demolishes the GTX 280, but mostly those are games where more performance is not needed. In Crysis, where we need more performance, the 4870 X2 is pretty much equal with the GTX 280.

The power consumption is also pretty insane - normally I don't care, as you can see I bought a GTX 280 and it is very power hungry. But the 4870 X2 is off the charts in terms of power consumption; it would not even work with my power supply. Consuming 100W more than the GTX 280 which is already a power hog is not a good thing.

Overall, considering the price I paid for the 280, I don't regret the purchase. The 4870 X2 performs better on average, but not where the performance is actually needed, and Crossfire scaling cannot always be counted on. And the power consumption means that even if I wanted to buy one, I could not without upgrading the PSU as well.

nVidia could get the crown back with a 55nm GT200 clocked at 700MHz+, for me at least. In certain games the 4870 X2 is and always will be untouchable by a GT200 part, but in other games (Crysis or other games with bad CF scaling) all the GTX 280 needs is a slight boost in performance to be back on top. If nVidia were able to set up a new SKU like the Leadtek Extreme, which is clocked @ 738/1566/2520, that would be enough to be better than R700 where it really matters. What AMD needs to do is improve their Crossfire drivers. The scaling across the board is just not as consistantly good as with SLI. You can't rely on CF to scale well.

Crysis was suppose to get a octomisation patch but then developer never end up releasing it because they want to their crysis warhead.

Mini frame rate :
I still have to see a creditable benchmark where the mini frame rate of GTX 280 is higher than 4870X2.


GAMES :
Now GTX280 and 4870X2 are priced the same so your arugment of crysis running a little better than GTX280 is debunked as both card cost the same.

Games 4870X2 wins in :

Age of conan : Twice as fast as GTX280
Crysis : equal
Oblivion : 37%
Grid : Twice as fast
HL2 episode 2 : 30% faster
Quake War : 25% faster
COD 4 : 40% faster
World in conflict : 38% faster

Power consumption
Power consumption taken from techpowerup

4870 : 274
GTX280 : 318
4870X2 : 436
GTX280 SLI : 564

Now , IF anyone was buying a powersupply for a GTX280 SLI , they really would need a really good 750W + PSU. A Crosair 750W which comes with 4x 8pin PCIE connector is really good for $110.

Micro shutter

Well according to some result microstuttering problem has been solved but i need to look at more result before i will go with that line.


Overclocking
Well Driver Heaven used a 670Mhz core clocked GTX280 and it couldn't beat a 4870X2.

Depends on the benchmark and site that reported it. There were several situations where it hadn't improved more than 5% over the single 4870, while the average framerate nearly doubled. I am sure there are many benchmarks to prove either way. Until more hard numbers are tested and cross references, I am not willing to to say confidently one way or the other.

But I will tell you that the Xbit labs review with the 4870 has the 9800GTX beating it, or on par with it in the minimum frame rate department. If the 4870 struggles with mins as a single card, I can only imagine that it will struggle the likewise with the x2. Apop seems to think it is the drivers and he might be right, in fact, he probably is. But that brings us back to the point that the card is only as good as the drivers...

You have absolutely nothing to back your statements.

Sound more like a buyer trying to justify his purchase. GTX280 is excellent single card but ain't match for a 4870X2.

 

[hooflung]

Where are the Nvidia representatives at? The guys who said the 4000 series didn't scale. The ones where they said the SLI 280 was going to whip whatever AMD could put out. Yeah. C'mon. Waiting for the, 'well... well it uses a lot of power' and 'well... crysis is the way forward...' posts.

 

[AmberClad]

Originally posted by: tuteja1986

Originally posted by: BFG10K

Originally posted by: lopri

AMD shipped out ten R700 cards worldwide[/b], attempting to capitalize on the success of the 4800 and showcase the strength of AMD's small-GPU strategy.

Hmm, ten cards but only seven reviews. Maybe we should have a competition to find the other three.

Rage3d has one : )

LegitReviews has another. So how many left unaccounted for is that?

 

[tuteja1986]

http://www.legitreviews.com/article/745/5/

Amazing , ATI fixed stalker scaling issue :!

 

[ArchAngel777]

Originally posted by: tuteja1986
You have absolutely nothing to back your statements.

Sound more like a buyer trying to justify his purchase. GTX280 is excellent single card but ain't match for a 4870X2.

Well, since I don't own any of the new cards, I don't think I have any justification to make. In fact, I went through and ordered a 4870 only to cancel it when I seen 'this review. If you look closely, there are several situations where the 9800GTX has better minimum frame rates. Not in every situation, but enough to cause me to pause on my purchase decision. Also, if you look too, you will find that the 4870 often ties the 4850 for minimum frame rate, while the 4870 has a much higher average frame rate (goes to show that averages mislead).

Because I know how you typically post, I don't really plan on responding to you further. The information is there for you to review in the Xbit labs review. You can take whatever you want from that review and do with it as you please.

 

[tuteja1986]

Originally posted by: ArchAngel777

Originally posted by: tuteja1986
You have absolutely nothing to back your statements.

Sound more like a buyer trying to justify his purchase. GTX280 is excellent single card but ain't match for a 4870X2.

Well, since I don't own any of the new cards, I don't think I have any justification to make. In fact, I went through and ordered a 4870 only to cancel it when I seen 'this review. If you look closely, there are several situations where the 9800GTX has better minimum frame rates. Not in every situation, but enough to cause me to pause on my purchase decision. Also, if you look too, you will find that the 4870 often ties the 4850 for minimum frame rate, while the 4870 has a much higher average frame rate (goes to show that averages mislead).

Because I know how you typically post, I don't really plan on responding to you further. The information is there for you to review in the Xbit labs review. You can take whatever you want from that review and do with it as you please.

The problem with xbit review is they come out very late and they don't do testing on the latest driver. They never used Driver hot fix and 2nd it doesn't have GTX280 or GTX260 result . XBIT do good review but not the best reviewer in town.

The problem with mini frame is that sometime games have a bug , it may activate somehow and could literally bring the mini frame rate too 1FPS in some cases.

What really good review website would instead use is average of mini frame rate. Now you know why many site don't include it because it gives really no real representation. Also review from Techreport , Computerbase , Firingsquad are on par and in some cases even better than XBIT LAB. Also ask your self why is XBIT review only major review not compare the GTX2XX with 48XX.


Now ask your self why can't really defend your statements.

 

[n7]

AT review fixed.

It's a 2x1024 MB card, as we suspected.

 

[Caveman]

Originally posted by: nkdesistyle

Originally posted by: n7
I would say minimum fps is actually very important, in many, even most, cases, moreso than average fps.

Generally, if you have a well matched card for the rez/settings/etc. in the games you want to play, you'll be getting adequate fps on average.
But the minimums are what get annoying.

Ido agree they get annoying but what I was trying to say is that minimum fps is only encountered here and there with these cards, if you are playing constantly at 50+ and here and there the fps dips to 30 but still smooth than I don't have problem at all, but yea a constant dip to low fps could certainly annoy the crap out of me, but if that was the case than the average would not be high, the minimum fps is important but whats more important to me is how ofter it is encountered in a particular game.

Yeah, but when you're stressing a card with a flight simulator (the only "real" test for a GPU and CPU , the min is what kills you because "smooth" is expected at 30-50 fps and the dips can send things down into the single digits and are very apparent/annoying.

 

[OCGuy]

Originally posted by: hooflung
Where are the Nvidia representatives at? The guys who said the 4000 series didn't scale. The ones where they said the SLI 280 was going to whip whatever AMD could put out. Yeah. C'mon. Waiting for the, 'well... well it uses a lot of power' and 'well... crysis is the way forward...' posts.

Or they could bring up that it looks like at low rez and low AA/AF that it gets spanked by 2X 9800GTX+, and a 9800GX2.......

That could be just Crysis, id really like to see alot more games benched....

 

[ShadowOfMyself]

Well whatever suits you, I think thats ridiculous... In their conclusion the only bad points about the card are heat and power, and yet, you make up issues to justify canceling your order?

What youre saying about "avg frames being misleading" could also apply to min frames... The game could very well run on a consistent 60 fps for 1 whole hour, and then theres a scene where your pc loads up for whatever reason and it drops to 20 for 1 second... Min fps will be stated as 20 in the chart, while the average was clearly 60

 

[ArchAngel777]

Originally posted by: ShadowOfMyself
Well whatever suits you, I think thats ridiculous... In their conclusion the only bad points about the card are heat and power, and yet, you make up issues to justify canceling your order?

What youre saying about "avg frames being misleading" could also apply to min frames... The game could very well run on a consistent 60 fps for 1 whole hour, and then theres a scene where your pc loads up for whatever reason and it drops to 20 for 1 second... Min fps will be stated as 20 in the chart, while the average was clearly 60

I don't think I made up any 'issues'. I reported based on the results from the Xbit lab review. Reading a reviewers conclusion often doesn't even line up with their results. I learned long ago that the conclusion is pretty much subjective. So, I personally, don't really pay attention to their conclusions.

Also, your example would have merit if someone simply ran the benchmark one time and reported the results. If that is the way you think these reviewers test the card, you obviously have never taken the time to properly benchmark a card. You should run the same benchmark at the same settings several times to root out these 'anomolies' that you speak of. If I ran a test 5 times, and only one of the times experienced a 'glitch' like the ones you say happen, then I would discount that and make a note of it. I would then see if it happens in other games, etc and then decide to comment on that in the review or not. If one card experiences those 'anomolies' frequently, while another competing card does not, then I know there is an issue with the card and would report that in the review. But if it just happened once or twice, I'd remove those results from the tests and make an average on the 4 runs rather than 5 runs. Minimum Frame Rate is KING, IMO. Average frame rates are still important, but the minimum frame rate is more important.

 

[Paratus]

Impressive!

But is it strange I want to hear more about the architecture than it's performance?

 

[LittleNemoNES]

I learned long ago that the conclusion is pretty much subjective. So, I personally, don't really pay attention to their conclusions

This doesn't make any sense

If you don't trust a reviewer enough to take their conclusion into consideration, then why trust their data?

There's only so much a graph can tell you. Microstuttering for example. Those who have experienced it know the pangs it brings.

 

[JimiP]

This is a glorious day.

I can only hope that hard launch will be successful for AMD. If this card performs as well (if not better than) the early review demonstrates then I feel there is some serious competition in the GPU market. I can't wait to see prices fall even more in the coming months.

 

[ArchAngel777]

Originally posted by: gersson

I learned long ago that the conclusion is pretty much subjective. So, I personally, don't really pay attention to their conclusions

This doesn't make any sense

If you don't trust a reviewer enough to take their conclusion into consideration, then why trust their data?

There's only so much a graph can tell you. Microstuttering for example. Those who have experienced it know the pangs it brings.

It makes total sense. We can analyze the same numbers and come up with different conclusions base on what we consider important. If, for instance, I deemed average frame rate far more important than minimum and could care less, I could come up with a totally different conclusion than a person who held the opposite view. Same data, different conclusion.

Also, be sure not to fall into the trap that so many people here do, which is taking a statement and and making it absolute. So with that said, of course I would not 'ignore' the part of a conclusion if it talked about micro sutter. But, if the micro stutter existed, it would (or should) be mentioned in more than just the conclusion.

Not sure if anyone has a family here so they can relate, but the older I get the more tight I am with my money. If I have any doubt at all on my purchase, I will not purchased/cancel. In this case, I am doubting the 4870's ability to provide me with better minimum frame rates than my current video card. $319 is not chump change, unless you happen to have a ton of extra money and no responsabilities.

For the record here, no one that knows me can claim I have an nVidia bias. If anything, I slant towards ATi. They have an outstanding card for the money. But going from an 8800GTS 512MB to HD 4870 isn't worth $319 and as you can imagine, the GTX280 is even less worth it to me.

 

[SlowSpyder]

Originally posted by: Ocguy31

Originally posted by: hooflung
Where are the Nvidia representatives at? The guys who said the 4000 series didn't scale. The ones where they said the SLI 280 was going to whip whatever AMD could put out. Yeah. C'mon. Waiting for the, 'well... well it uses a lot of power' and 'well... crysis is the way forward...' posts.

Or they could bring up that it looks like at low rez and low AA/AF that it gets spanked by 2X 9800GTX+, and a 9800GX2.......

That could be just Crysis, id really like to see alot more games benched....

Sweet! Let's buy a 9800GX2 or a 2 9800GTX+ for SLI and run them at low res and low AA/AF, that sounds like a great idea. You know that the Nvidia guys are grasping at straws when they point how $400+ video card(s) run better at low res/low AA/low AF then the competition's cards.

Of course OCguy31 was the same person who put it into his sig to make fun of someone who suggested that the 4870 may even be as fast as the GTX280. What happened to your sig?

 

[n7]

Sampsa working on a microstuttering review: http://www.xtremesystems.org/f...howthread.php?t=194808

I think we're going to find that it's not completely gone so much as it's minimal enough to be largely unnoticeable & not an issue.

Just my guess...we'll see.

 

[CP5670]

I have to admit, this card looks pretty impressive for its price, even though I'm not a fan of multi GPU. I guess it benefits from the fact that a single 4870 is already quite a strong performer. However, there is no indication that it uses something other than AFR, which means that triple buffering may well remain broken even if they have somehow fixed microstuttering.

Originally posted by: Caveman

Originally posted by: nkdesistyle
Ido agree they get annoying but what I was trying to say is that minimum fps is only encountered here and there with these cards, if you are playing constantly at 50+ and here and there the fps dips to 30 but still smooth than I don't have problem at all, but yea a constant dip to low fps could certainly annoy the crap out of me, but if that was the case than the average would not be high, the minimum fps is important but whats more important to me is how ofter it is encountered in a particular game.

Yeah, but when you're stressing a card with a flight simulator (the only "real" test for a GPU and CPU , the min is what kills you because "smooth" is expected at 30-50 fps and the dips can send things down into the single digits and are very apparent/annoying.

Exactly, those minimum drops are the things you actually notice, even when the average is good. They also have a tendency to occur when you least want them to, like during big fights in FPSs.