AMD Cayman info (or rumor)

[zerazax]

actually he's been benching with the 480 now:

http://translate.google.com/translat...768017-27.html

That's nice but he's comparing it in synthetics still. I'm saying real games

None the less, AMD has lost a vary valuable marketing tool. If AMD cards become a nonfactor on HWbot, then alot of people won't buy them(benchers).

I benched just as much as I gamed before so it is a real significant factor for me.

I don' think it's doom and gloom if they are priced right then it can be a decent product. It's just this being xtremesystems, people expectations for this card were sky high, when they don't come off as close to matching expectations. The dreams of 30-40% faster than a gtx 580 are gone and all that is being left is a sobering reality. It doesn't have to be bad if we control our expectations before hand. My prior performance expectations were realistic compared to most people on this board, I got caught in the hype a bit. But I still feel this card can be a good card if priced right.

If this card is priced at around 400 dollars, we can thank the gtx 570 and 580. Maybe spurs will get to stay on this board too, but because the 6970 performed well below expectations .

Yes, a lot of people bench equal if not more than they game, but that's where unrealistic expectations come in - a 1920 SP card would absolutely kill performance wise, but a 1536 SP card means that for benching, where raw SP counts do matter, it's not going to provide as big a boost over Cypress - whereas for games, it can matter more

But I think the entire "zomg Cayman = fail because it is lower in 3dMark Vantage" is ridiculous - what if it sucks in 3dMark but is comparable to the 580 in gaming?

We're only getting 3dMark benching which isn't giving us anything about gaming

And yes, HWBot is big, but again... for gamers, how many follow it based off that? They're going to see the benches for teh games they care about. If enthusiasts are a small enough niche, then the hardcore benching crowd is even smaller. I don't think that hurts AMD as much as you think - indeed, look at how many people ridicule Fermi for being a benching queen w/r/t Heaven

[migunani]

This fight has become less interesting. The crowds are really dissapointed and are ready to go home. Looks like the RED CORNER is going to throw in the towel. hopefully second wind coming to the RED Corner and give a good fight. A good fight = everyone wins. If not, See you in the NEXT FIGHT!

[elzhi]

so many drama queens in this thread, lol. every rumour is a reason enough for this card to be "fail", people making their minds up already even though no actual review has been done and no one is 100% sure on what the spec are...

only 3 days left, chill the f*** out

[zerazax]

This fight has become less interesting. The crowds are really dissapointed and are ready to go home. Looks like the RED CORNER is going to throw in the towel. hopefully second wind coming to the RED Corner and give a good fight. A good fight = everyone wins. If not, See you in the NEXT FIGHT!

.... ooooooooookkkaaayyy

[RaZz!]

who is that guy and what is he trying to tell ? lol i missed the point is he pissed at amd cause of the leak or something ?

that's what i asked myself as well.

and why is he mad at amd? the guys over at the hardwareluxx forums bought the card from a regular shop (for a reasonable 395€ btw) - atleast that's what they said.
how is it amd's fault when some random retailer sends out these cards before the official launch date

[tajoh111]

I think AMD didn't expect the gtx 580 so soon. None of us did, look at the initial gtx 580 rumors and we all thought it was a joke and FUAD was just trying to get attention. I think this card was meant to go against the gtx 480. Performance was a chief concern but AMD also knew it didn't need sky high performance and would just need to beat the gtx 480 in heat and power consumption, while taking it in performance to some extent. The gtx 480 was a disaster when its cons were taken into account. Basically AMD just needed to make a decent card to make an excellent card when compared to the flawed gtx 480. I think this was a far more realistic interpretation of what was to come.

Some people were expecting an 80% performance jump on the same node when the prior generation of chips were already 333mm. That's completely unrealistic(unless AMD wanted to double the size of their chip). I know the card isn't released but leaks and specs are indicated something far less than what a lot of people anticipated.

This card probably isn't a fail card itself. It just fail for the fanboys because they don't get to huff that that NV is screwed and last gen. I think it's should be a lesson for them, don't be so crazy optimistic and be so vocal about it. In addition, don't attack anyone else if they don't share your supremely optimistic point of view because you might look foolish in the end.

[zerazax]

I think AMD didn't expect the gtx 580 so soon. None of us did, look at the initial gtx 580 rumors and we all thought it was a joke and FUAD was just trying to get attention. I think this card was meant to go against the gtx 480. Performance was a chief concern but AMD also knew it didn't need sky high performance and would just need to beat the gtx 480 in heat and power consumption, while taking it in performance to some extent. The gtx 480 was a disaster when its cons were taken into account. Basically AMD just needed to make a decent card to make an excellent card when compared to the flawed gtx 480. I think this was a far more realistic interpretation of what was to come.

Some people were expecting an 80% performance jump on the same node when the prior generation of chips were already 333mm. That's completely unrealistic(unless AMD wanted to double the size of their chip). I know the card isn't released but leaks and specs are indicated something far less than what a lot of people anticipated.

Well FUD (not sure anyone should trust him anyways with AMD stuff) said the chip was the biggest AMD ever done - and R600 was 420mm^2

Turns out its 389mm^2, which is proof that Fud either doesn't have a source and makes $h!t up, is an idiot, or a combination of both (most likely) - and so expectations need to change.

A 430mm^2 1920SP Cayman probably takes teh crown easily - a 390mm^2 1536SP Cayman probably won't.

The 580's release was definitely earlier than people thought, although the 580 seems to be in short supply in a lot of countries, so maybe it was indeed rushed out earlier by Nvidia to steal the thunder regardless of stock, esp. since the 6900's apparently are in good supply and were stockpiled for some time

[zalbard]

GTX580 is a very minor refresh... No reason for it not to come out 6 months after GTX480. Just activate a few more shaders, sort out the TDP with a limiter chip, and do very minor performance tweaks = voila!

[Dimitriman]

GTX580 is a very minor refresh... No reason for it not to come out 6 months after GTX480. Just activate a few more shaders, sort out the TDP with a limiter chip, and do very minor performance tweaks = voila!

and apparently thats all it took to ruin amds one year project cayman.

[zerazax]

GTX580 is a very minor refresh... No reason for it not to come out 6 months after GTX480. Just activate a few more shaders, sort out the TDP with a limiter chip, and do very minor performance tweaks = voila!

GF110 = GF110b... a'la GT200 -> GT200b

Both lowered power draw, heat, improved yields (at least w/ GF110 we see it improved functional units)

I don't think it surprised AMD as much as people think though - after all, a 512SP Fermi was due last year. The timing might have surprised AMD a bit, but I don't think the overall card + performance did as much.

Again though, we don't have any gaming #'s yet - the Unigine benchmark for 6970 is below the 570 and 580, but it is VERY close to the 5970 - and we know the 5970 performs in game much better than the 570 and just above the 580. So what does that tell us about the 6970? TBA

and apparently thats all it took to ruin amds one year project cayman.

Stop being so emotional.

How did it ruin it? Do you have concrete evidence Cayman's gaming performance is going to be like its 3dMark results? Where are you getting this conclusion? Seriously?

[Nintendork]

And cut the cache and HPC/GPGPU things that made G100 a power hungry chip to make a G110 a more gamer-oriented gpu.

[tajoh111]

GTX580 is a very minor refresh... No reason for it not to come out 6 months after GTX480. Just activate a few more shaders, sort out the TDP with a limiter chip, and do very minor performance tweaks = voila!

I don't think the gtx 580 is that good of a card, it just looks fantastic compared to the gtx 480 as would the current leaks of the 6970 would. Sorting out the TDP and heat issues was a big victory for NV because that what was preventing NV from making the gf100 any faster and what was preventing them from making a dual card. The gtx 580 still isn't the best with power, but it is no longer the joke of the industry that can be poke fun of like AMD commercials have shown.

[zerazax]

who is that guy and what is he trying to tell ? lol i missed the point is he pissed at amd cause of the leak or something ?

Reviewer/editor at PCGH

No clue why he's pissed at AMD though - that sounds like a retailer ing up

[Nintendork]

I don't think the gtx 580 is that good of a card, it just looks fantastic compared to the gtx 480 as would the current leaks of the 6970 would. Sorting out the TDP and heat issues was a big victory for NV because that what was preventing NV from making the gf100 any faster and what was preventing them from making a dual card. The gtx 580 still isn't the best with power, but it is no longer the joke of the industry that can be poke fun of like AMD commercials have shown.

They used a Vapor-Chamber to make it look cooler. It certainly is, but no one would be impressed if they used the same 480 heatsink.

[FischOderAal]

anyone here has an account there ?

curious about gpu load..

I will ask

[Periander6]

And cut the cache and HPC/GPGPU things that made G100 a power hungry chip to make a G110 a more gamer-oriented gpu.

Source for this?

[zalbard]

And cut the cache and HPC/GPGPU things that made G100 a power hungry chip to make a G110 a more gamer-oriented gpu.

As far as I know it's even faster for folding than GTX480.

[Nintendork]

So what makes GTX 580 faster? We’ll start with the obvious: it’s a complete chip. All the L2 cache, all the ROPs, all the SMs, it’s all enabled. When it comes to gaming this is as fast as GF110 can be, and it’s only through NVIDIA’s artificial FP64 limitations that double-precision computing isn’t equally unrestricted.

http://www.anandtech.com/Show/Index/...eforce-gtx-580

I mixed GF104, bad memory.

[freeloader]

Looks like AMD may have got caught with their pants down by Nvidia.

[-=DVS=-]

Maybe they never planned for 6970 to be a monster chip but rather address some weak points eg. "tessellation" and to remain within similar TDP of 5870, since we are still on 40nm. From synthetic benchmarks we see they compromised in some areas to increase others.

[zalbard]

They said it would be the new R300...

[Solus Corvus]

ye we can tell ourselves that but how many times before this much info has converged right before release and turned out to be wrong? almost never.
info leaks at this point are 80% true and if several sources are leaking 1600 shaders and gtx570 performance, thats whats likely to come. And talking about synthetics, if we looked at vantage alone I would ignore it but all benchmarks show similar results. When was the last time a card from any camp did mediocre in synthetics and awesome in games? There may be a difference, but it is often not so substantial that would put the card in a different level.
I think at this point wishing for magic drivers and miracle game performance is straight up denial or wishful thinking. Were looking at a competitor for the 570 and thats the truth. 580 will reign supreme as single chip and Amd will throw antilles against it but it will not be as attractive as 5970 was one year ago. Back then it had no competition so people cared less about xfire issues. It will not be the case this time. Many will still take the 580 over antilles for being single chip.

/end rant

and this coming from an Amd customer since 2007

If you are so confident about the price/performance info we have at the moment don't let me stop you from deciding to buy or not right now. I'm not saying that the info is necessarily right or wrong - only that it is probabilistic, not hard facts.

I don't believe in the magic drivers theory. Being a new architecture there may indeed be some significant driver improvements eventually. But that generally takes months to materialize. As for a differential performance increase in synthetics vs games, I hypothesized such a difference over 2000 posts ago based solely on the transition between 5D and 4D shaders. There are plenty of reasons why that analysis might be very wrong. But writing it off on the basis of not having happened before is a simple logical fallacy. As for the rest I am simply saying we don't have the whole picture. How do they clock? How do they perform in HPC apps? How do they perform in the specific games you play? These are open questions that could affect the purchasing decisions of many people here.

I must admit I would be disappointed if the benchmarks so far are indicative of overall performance in everything. But that's a flaw in my expectations, not a rational analysis.

I have been an ATI customer on and off since, IIRC, 1993. I don't see how that's relevant.

[Dimitriman]

Stop being so emotional.

How did it ruin it? Do you have concrete evidence Cayman's gaming performance is going to be like its 3dMark results? Where are you getting this conclusion? Seriously?

You are right I am being hasty about benchmarks. My frustration comes in fact from the decision of Amd to add so little smds to the architecture. I just get the feeling if they wanted to make a 1920 shader they could and it would dominate but instead they chose a cheaper smaller die that will once again compete on perf/watt mostly. that was a big letdown for me but I will try not to whine about it anymore my bad.

If you are so confident about the price/performance info we have at the moment don't let me stop you from deciding to buy or not right now. I'm not saying that the info is necessarily right or wrong - only that it is probabilistic, not hard facts.

I don't believe in the magic drivers theory. Being a new architecture there may indeed be some significant driver improvements eventually. But that generally takes months to materialize. As for a differential performance increase in synthetics vs games, I hypothesized such a difference over 2000 posts ago based solely on the transition between 5D and 4D shaders. There are plenty of reasons why that analysis might be very wrong. But writing it off on the basis of not having happened before is a simple logical fallacy. As for the rest I am simply saying we don't have the whole picture. How do they clock? How do they perform in HPC apps? How do they perform in the specific games you play? These are open questions that could affect the purchasing decisions of many people here.

I must admit I would be disappointed if the benchmarks so far are indicative of overall performance in everything. But that's a flaw in my expectations, not a rational analysis.

I have been an ATI customer on and off since, IIRC, 1993. I don't see how that's relevant.

you are right and Ill wait for a full review before making any more comments. I suppose a lot of things will be different on 15th. I mentioned that I was a customer because it seems that come dec 15 if cayman dissapoints a lot of amd customers will try to shine many bright lights on it but I think as a customer its better to criticize when criticism is due than to spin a positive side.
But anyway I wont be buying this generation as I am planned to replace my system with either bulldozer or sandy bridge and by then wel have 28nm cards finally. Just felt like voicing my opinion.

[Tomasis]

oh my good.. some decided already by "facts"

really useless thread .. most useless in all time history

[WeaponX]

Dammit! I'm out of popcorn!