Saturday, February 9th 2008

ATI RV770 'On Par' With Expectations

With the launch of the GeForce 9 series getting closer and closer, AMD is hard pressed to find something to keep themselves competitive. While the RV670 and R680 are regaining some much needed market share, they will both pale when the GeForce 9 series is released to the public. Thankfully, AMD is not going down without a fight. About the same time as the GeForce 9 series is released, AMD is releasing a little something called the RV770. At this point, it appears that the RV770 is about 50% faster than the current HD3870, which is certainly respectable. How this compares to the GeForce 9 series is still a mystery. The release of CrossFire X technology ought to really help benchmark numbers, assuming AMD can make buying four AMD GPUs cost about as much as two from NVIDIA.Source: Nordic Hardware
Add your own comment

57 Comments on ATI RV770 'On Par' With Expectations

#1
btarunr
Editor & Senior Moderator
50% faster than R670? That might not help with the performance increments people are talking about GeForce 9 having over the current line-up. If 4 ATI cards cost as much as two NVidia cards, assuming the two setups perform equally, I'd choose the NVidia setup over ATI for obvious heat/power/space/motherboard price/game-optimisation issues.
Posted on Reply
#2
Ketxxx
Heedless Psychic
Card by card basis ATi offers far better price-performance ratio, and better drivers. Power consumption issues can be curbed by reducing GPU voltage and clocks in 2D mode.
Posted on Reply
#3
springs113
by: btarunr
50% faster than R670? That might not help with the performance increments people are talking about GeForce 9 having over the current line-up. If 4 ATI cards cost as much as two NVidia cards, assuming the two setups perform equally, I'd choose the NVidia setup over ATI for obvious heat/power/space/motherboard price/game-optimisation issues.
I believe that this article is not stating everything...as like the r670 core the r770 is basically the lower end of the spectrum...the r670 aka 3870 x2 is the highend...
the r700 is the high end chip...eventually made up of a couple of r770s...i have heard multiple..with a possibility of upto four...so in actuallitythe r700 will be the high end while the 770 the low end or core component of the 700.
ati' see multiple gpus as a future so as long as the 3870x2 does well the r700 will continue to grow...and we are all waiting on drivers.
Posted on Reply
#4
[I.R.A]_FBi
by: btarunr
50% faster than R670? That might not help with the performance increments people are talking about GeForce 9 having over the current line-up. If 4 ATI cards cost as much as two NVidia cards, assuming the two setups perform equally, I'd choose the NVidia setup over ATI for obvious heat/power/space/motherboard price/game-optimisation issues.
what process will either manufacturers gpu be done on?
Posted on Reply
#5
btarunr
Editor & Senior Moderator
R770 = 45nm ; G100 = 55nm (on the 9800 GTX), G94 = 65nm (on the 9600 GT)
Posted on Reply
#6
imperialreign
R700 is supposed to be 45nm, also, and IIRC, they're supposed to be fairly energy efficient, too.

Still, though, dual core GPU . . . and if ATI decide to make a dual R770/R700 GPU PCB - that's 4 cores on one card

Rumor has it the R770 and R700 will also support GDDR5, and we all know how ATI loves to jump on new tech like that

brace yourselves . . . the GPU wars are about to become extremelly interesting again :toast:
Posted on Reply
#7
phanbuey
by: Ketxxx
Card by card basis ATi offers far better price-performance ratio, and better drivers. Power consumption issues can be curbed by reducing GPU voltage and clocks in 2D mode.
Actually... any review site will show that from any price-performance consideration the 8800GT leads... even ahead of the 2600XT, and even ahead of HD3870 and 3850. -

Source:
http://www.techpowerup.com/reviews/HIS/HD_3870_X2/23.html << techpowerup has best reviews!!
Posted on Reply
#8
[I.R.A]_FBi
by: btarunr
R770 = 45nm ; G100 = 55nm (on the 9800 GTX), G94 = 65nm (on the 9600 GT)
45 nm + powerplay = much cooler card
Posted on Reply
#10
mandelore
sweet, maybe just what is needed to replace my 2900xt..
Posted on Reply
#11

Sound's interesting!
I think ATI can make it Large Again! (Rise from the Ashes:))
Sorry not ATI, AMD!
#12
HaZe303
I hope ATI/AMD can come up with some really good and intresting stuff, we know how much they need the comeback. And maybe this would get Nvidia of their lazy buttĀ“s!? We need something to get some competition again in the gfx market, and ATI getting the performance crown would be that something! Hope hope hope.... Hope they dont do another "2900xt" on us and fail miserably??
Posted on Reply
#13
imperialreign
well, if what rumors have surfaced about ATI's upcoming R7xx series are true, they've defi got what it takes to go toe-to-toe with nVidia with the next generation of hardware. The R7xx series looks like it'll be a strong contender against nVidia's G92 series.
Posted on Reply
#14
zOaib
by: imperialreign
well, if what rumors have surfaced about ATI's upcoming R7xx series are true, they've defi got what it takes to go toe-to-toe with nVidia with the next generation of hardware. The R7xx series looks like it'll be a strong contender against nVidia's G92 series.
tumhaaraay moun main ghee shakkar <-------- a saying in india

translation,

may your mouth be filled with sweets and sugar ..... lol

wub be nice to see the red team right there on the horizon =P
Posted on Reply
#15
Kreij
Senior Monkey Moderator
I'm waiting for AMD (ATI) to come out with a dual core GPU with built in crossfire and PPU.
Single chip, 45nm. Use their new FABs for GPU as well as CPU

Put two on a card and you have single-slot, quad crossfire and a pair of PPUs.

That would seriously rock.
Posted on Reply
#16
btarunr
Editor & Senior Moderator
Guys, forget dual-core / many GPU's on one PCB / X2 GX2 , etc. The architectural finesse of a company lies in its ability to make leap-improvements, not notched improvements. With a weaker X1, X2 is bound to be weak. If ATI concentrates on a strong single GPU/core architecture, it can truly call itself a leading company which it is not right now.

It's like to reach 1.00 you can use two 0.50 or four 0.25. a 0.25 is weaker than 0.50, isn't it? The idea of using many weaker units to make a strong unit is worse than fewer units to reach the same? So to attain a certain performance would you take four ATI cards or two NVidia when a lot more things go against the ATI setup like the price of the platform, power and cooling?

Jab do balwan toh chaar ka kya kaam? - zOaib can translate that.
Posted on Reply
#17
eidairaman1
I dunno 3870 Kept them from dying so to speak since the 2900XT couldnt keep in spec due to heat. 3870X2 draws about as much as the 2900XT. I think the 50% is modest performance gain, but you gotta expect, usually RV is the Mid Range sort of like RV670. who knows how R780/90 will be.
Posted on Reply
#18
Kreij
Senior Monkey Moderator
by: btarunr
Guys, forget dual-core / many GPU's on one PCB / X2 GX2 , etc. The architectural finesse of a company lies in its ability to make leap-improvements, not notched improvements. With a weaker X1, X2 is bound to be weak. If ATI concentrates on a strong single GPU/core architecture, it can truly call itself a leading company which it is not right now.
True, but if ATI could put a strong pair of GPUs on a single die with the crossfire logic built in, it would perform much better than having discreet chips on a board. Add a PPU to the mix and you would be looking at an excellent performing chip.

If they could do it in quad on a single chip, even better.
Posted on Reply
#19
btarunr
Editor & Senior Moderator
but end of it look at the design philosophy...you're clubbing many weaker pieces to make something big. There's a difference between a quad-core CPU and a quad-core GPU. The difference is that parallelism has already been achieved on a single core/die/PCB GPU with multiple pixel-pipelines/multiple shaders, etc. You can't boost parallelism by clubbing many tiny pieces and expect something big. OTOH, a quad-core CPU adds to its capabilities with stepping-up parallelism with increase in thread/core count. So clubbing four GPU cores with say 0.25x performance each would still be a no improvement design than a single GPU with 1.0x performance, it in fact becomes worse in the aspect of thermal envelope, operation costs, the price would be the same anyway as buying a powerful NVidia card.
Posted on Reply
#20
sneekypeet
Unpaid Babysitter
by: springs113
I believe that this article is not stating everything...as like the r670 core the r770 is basically the lower end of the spectrum...the r670 aka 3870 x2 is the highend...
the r700 is the high end chip...eventually made up of a couple of r770s...i have heard multiple..with a possibility of upto four...so in actuallitythe r700 will be the high end while the 770 the low end or core component of the 700.
ati' see multiple gpus as a future so as long as the 3870x2 does well the r700 will continue to grow...and we are all waiting on drivers.
Just to address this...you all are assuming R770 and the like when it is the RVVVVVVVVV770 we are speking about. The V IIRC stands for Value. It is not made to be top of the market.

So dont everyone get worried Im sure there will be a R770 released as well
Posted on Reply
#21
happita
So much interesting news today. I hope the R700 is all it can be!!!!
Posted on Reply
#22
ChillyMyst
by: btarunr
Guys, forget dual-core / many GPU's on one PCB / X2 GX2 , etc. The architectural finesse of a company lies in its ability to make leap-improvements, not notched improvements. With a weaker X1, X2 is bound to be weak. If ATI concentrates on a strong single GPU/core architecture, it can truly call itself a leading company which it is not right now.

It's like to reach 1.00 you can use two 0.50 or four 0.25. a 0.25 is weaker than 0.50, isn't it? The idea of using many weaker units to make a strong unit is worse than fewer units to reach the same? So to attain a certain performance would you take four ATI cards or two NVidia when a lot more things go against the ATI setup like the price of the platform, power and cooling?

Jab do balwan toh chaar ka kya kaam? - zOaib can translate that.
duno, the japnese have pretty much stomped the rest of the worlds eletronics markets into only buying stuff they designed by doing notched updates.

please lets be clear, if i cay "japs" that is not an insult, i love the japs, great people!!!

well their way of looking at tech advancement is why they can consistnatly bring out stuff that drives other countrys products off the market.

unlike americans and many western based culturs the japs dont go for the big score/touchdown/home run every time, they are happy hitting singles all day and just burrying you with runs(baseball anniligy) American companys alwase go for the big advancement, where as japs and some other culruts will just make small improvements and bring them out more regularly, eventuly they BURRY you in small improvements AND they by using this method they have managed to keep production costs VERY LOW, so dont go by the "home runs are alwase better" assumption,

your annaligy would lead one to belive it would be better if amd and intel had stuck with making singel core chips and had just made them drastickly more powerfull and higher clocked, when thats just not true, for SMP/threded apps its better to have more cpu's/cores to spred the load over, and slowly but surely the world is headed threded!!!!

and it really dosnt matter how you get there if the results back it up, if amd uses 45nm vs nvidias 55 and 65nm chips then amd will almost certenly have the more energy efficent chip, if u tag 2 dual core versions of that on 1 pcb you have quadfire in 1 card!!!

as to the guy who said he wanted to see a PPU onboard as well, if game developters get off their ass they could do PPU work on the x1k cards, but they havent bothered, mostly because intel bought havoc and havoc was the main company working on gpu as ppu emplimentations.

a "True" ppu is just a very powerfull math unit, guess what, thats also what a modern shader based GPU is, an x1300pro is ruffly equivlant to a first gen agena ppu card, an x1300xt kills it, if you look at the HD seirse the 2400 kills agena's current products for raw prosessing power, its just that nobodys bothered emplimenting it YET.

we need somebody like MS to step in and say "heres a new standred for phsyics thru a common API, make your drivers fit the API calls and you cans upport phsyics"
Posted on Reply
#23
btarunr
Editor & Senior Moderator
2x 0.50 > 4x 0.25. --- while the baseball example did sound good, let me give a more natural one. Let's say tomorrow a game that's unoptimised for a multi-GPU setup comes up, what happens? The game would only exploit one of the four cards and it's 0.25. But, had the unit been 0.50, at least an unoptimised game would run better so it's important that ATI works on a strong single GPU than look at making powerful solutions based on multiple weak single units. We've pretty-much seen that happen with the Crysis benches for the HD 3870 X2:


^Look at what an unoptimised game can do. So, it's important to have stronger units. Let's face it, the G92 is a superior core to the R670, we've seen enough people attesting that, just that when in an SLI array, it slightly falters but that's the platform to blame, not the cards. In a 32-lane SLI chipset like the NForce 590 SLI or the 680i SLI, the northbridge and southbridge give out 16 lanes each to a card while in a 32 lane CrossFire compliant chipset like the AMD 790 FX or the Intel X38, the northbridge supplies all the 32 lanes. So well, a SLI of two G92 based cards falters slightly to a Crossfire of two R670 but if NVidia works well with its internal SLI on the 9800 GX2, we'll have a card waay more powerful than a HD 3870 X2, reason: the use of a powerful unit core. And supposing the 9800 GX2 does end up facing an unoptimised game like the HD3870 X2, it will perform better due to its stronger unit core.
Posted on Reply
#24
xfire
[quote]2x 0.50 > 4x 0.25[quote]
Agreed. but
2x1.0>2x5.0
Which means not only should each of the company concentrate on producing faster single GPU but also on dual.
Earlier you said that the increments are not enough considering th geforce 9 series, correct me if I am wrong but isn't the fastest in the 9x series a dual GPU based on the 8800GTX GPU?
Also see the heading, its on par with expectations. Didn't a thread posted few days back say it was as good as nvidia's next gen.
Also as chillymist said what about dual core/quad core cpu's. Shouldn't they have made a faster single core cpu.
Posted on Reply
#25
btarunr
Editor & Senior Moderator
by: xfire

Agreed. but
2x1.0>2x5.0
utterly confused :confused:

by: xfire

Which means not only should each of the company concentrate on producing faster single GPU but also on dual.
Earlier you said that the increments are not enough considering th geforce 9 series, correct me if I am wrong but isn't the fastest in the 9x series a dual GPU based on the 8800GTX GPU?
Also see the heading, its on par with expectations. Didn't a thread posted few days back say it was as good as nvidia's next gen.
Also as chillymist said what about dual core/quad core cpu's. Shouldn't they have made a faster single core cpu.
The 9800 GTX isn't ?
Posted on Reply
Add your own comment