AMD Radeon 7000-Series 28nm (Southern Islands) | 7990 7970 7870 7770 | Discussion

[Vesku]

I could see the rumored $550 if they think demand will outstrip supply and that the odds of NVIDIA launching a competitor within a ~8-12 week timeframe is low. $450 would be great though and would really test the loyalty of many NVIDIA only types. I don't think pricing will have much to do with Xfire mid-tier though, GTX 460 SLI didn't seem to phase GTX high end much neither has 5770 and 6870 Xfire performance caused price shifts in the top tier. If anything it just seems to effect pricing within their own tiers.

 

[ArchAngel777]

Edit: If HD7870 will be at $200 it will cannibalize the sales of the high End part because two of them in CF at $400 will be faster than a single HD7970 and AMD will not want that to happen

After reading Tom's hardware review on Crossfire and microstutter, I am going to keep very, very, very far away from running two cards in cross fire.

So far SLI has been great. But then again, Tom's hardware substantiated that for the most part. Basically saying 'nVidia provides better consistency at the cost of scaling' AMD, on the other hand, provides better scaling by the numbers, but shows a lot of microstutter. It was an awesome article on Tom's... So... No xfire for me, unless I pickup three cards.

 

[tigersty1e]

No, perhaps we are all talking about different things. But JAG87 came in this thread and was pissing all over the price because the performance didn't warrant it in his mind. He then tried to make a correlation between architectures, etc... My point is. If you product is the current KING. You can price it accordingly. Just because AMD priced it at $369 last generation doesn't mean they should continue to do so. What logic is there for that?

The fact of the matter is (we will know this for SURE in a few days) that the 7970 is believed to be at least 30% faster than the 580 GTX. It doesn't matter if this is AMD 15th generation or the 40th. If they currently have the faster card, then it stands to reason they should charge more. Even so, it would only be a premium of $50 over their competition while dominating it by 30%.

I guess my main issue is why is there an allowed grace for nVidia to rip off customers and not AMD? Why can nVidia charge so much? Do they get a pass?

When Kepler comes out, AMD will either adjust MSRP or come out with a new product... But until such a time, they are free to pretty much price this card anywhere from $550+ if it truly have a 30% performance advantage.

To further illullstrate the issue with his logic is using nVidia's own lineup. The 580 GTX is 15% or so faster than the 570 GTX. Yet it carries a 43% price premium. So what? There is a price premium to own the fastest single GPU graphics card. That is how it has always been. If you want price/performance, you go with a mid-range card. So if you don't like the 7970's price, go buy a 6970.

I agree. AMD should be allowed to charge a premium. There's always been a price premium for the fastest card above the performance ratio of the next highest card.

AMD priced the 5800 series so relatively low because they were going after market share. And they acheived their goals. BFG went under and XFX switched sides. They haven't had the fasted single gpu on single card many times. Going back to the 9800pro days, only the 5800 series (and possibly 1900 series, but can't remember) was it.

But word was solid that the nvidia was going to be waaaay late for the dx11 party, so amd did what they did for 5800. Now circumstances are different. Nvidia has dx11 products that can compete and has a new gen coming 2-3 quarters later.

And the parent company of amd graphics is not doing too well. They are strapped for cash. This 7900 series is going to be priced very high.

 

[lavaheadache]

People forget that HD7970 will replace HD6970 which was released at ~$400 price point.

HD7870 will have almost the same performance as HD6970 and replace the HD6870 at the $200 price point.

Do you actually believe that there will be that much of a gap ($500 to $200) in price between the two ???

I believe HD7970 will be introduced close to $399-449

Edit: If HD7870 will be at $200 it will cannibalize the sales of the high End part because two of them in CF at $400 will be faster than a single HD7970 and AMD will not want that to happen

We need to stop hearing about 2 mid range cards being faster than the top dog.

45 fps on a sli/crossfire setup =! 45 fps single gpu

 

[Zed03]

We need to stop hearing about 2 mid range cards being faster than the top dog.

45 fps on a sli/crossfire setup =! 45 fps single gpu

yup.

My 4870x2 crushes even the latest single gpu cards but gameplay experience is no where as close.

Micro stutter, sli quirks, and frame spikes (more like dips).

FPS isn't the only metric that needs to be measured - the quality and consistency of FPS plays a huge role.

 

[GotNoRice]

There is no reason to upgrade your card over WoW. You can max the game out on a current gen mid-range card. If that is your concern, there is no need to upgrade over it.

You could buy a 560 or 6950 and be able to run the game on its highest settings.

I can understand why you would think that based on the existing reviews that are out but the game is still extremely demanding when run at Ultra settings and high levels of AA. Most of the reviews on the internet that include numbers for World of Warcraft run their benchmarks by measuring FPS while on a wind-rider taking a flight-path from one point to another. That's because that is about the closest you can get to something that is repeatable in an MMO like WoW. There might be a few spots where you fly through a thick Forrest or something and your numbers dip down a bit but for the most part taking a flight-path on a wind-rider is not very graphically strenuous and is going to produce very high average FPS numbers.

There are numerous examples during real gameplay where FPS can dip quite a bit, like when things are getting rough in 25-man raid or even a 40-man Alterac Valley. That's also when performance counts most.

 

[JAG87]

This is way off target. The 480 is about 40 to 50% faster than the 285 on average. Nowhere close to 100%.



So the performance jump would be basically the same, if you believe one set of supposed benches. There are other benches that show it 50% faster than a gtx 580, making it a good 70% faster than a 6970 if true.

Either way, even in the worst case scenario of these leaked benches, the performance leap is consistent with past new node releases from nvidia and AMD. 100% has not happened in a very, very long time. I think the last time close to that was probably the 8800GTX and before that the 9800PRO.

No, it's definitely 100% faster, even more than 100% after driver revisions since launch.

You just have to create the scenario for this to show. Try 2560x1600 with MSAA+TRSSAA. GT200 struggled with TRSSAA, and it wasn't a frame buffer limit (1GB vs 1.5GB), it was just a GPU horsepower limit. Just because the benchmarks on websites don't show it, because they like to benchmark without the actual candy turned on (which is fair because they need to keep the testing environment as consistent as possible), it doesn't mean the performance is not there. I have first hand experience of the GTX480 putting out 2x the frames of my GTX280.

No, perhaps we are all talking about different things. But JAG87 came in this thread and was pissing all over the price because the performance didn't warrant it in his mind. He then tried to make a correlation between architectures, etc... My point is. If you product is the current KING. You can price it accordingly. Just because AMD priced it at $369 last generation doesn't mean they should continue to do so. What logic is there for that?

The fact of the matter is (we will know this for SURE in a few days) that the 7970 is believed to be at least 30% faster than the 580 GTX. It doesn't matter if this is AMD 15th generation or the 40th. If they currently have the faster card, then it stands to reason they should charge more. Even so, it would only be a premium of $50 over their competition while dominating it by 30%.

I guess my main issue is why is there an allowed grace for nVidia to rip off customers and not AMD? Why can nVidia charge so much? Do they get a pass?

When Kepler comes out, AMD will either adjust MSRP or come out with a new product... But until such a time, they are free to pretty much price this card anywhere from $550+ if it truly have a 30% performance advantage.

To further illullstrate the issue with his logic is using nVidia's own lineup. The 580 GTX is 15% or so faster than the 570 GTX. Yet it carries a 43% price premium. So what? There is a price premium to own the fastest single GPU graphics card. That is how it has always been. If you want price/performance, you go with a mid-range card. So if you don't like the 7970's price, go buy a 6970.

I didn't mean to criticize AMD's pricing. You are certainly correct that as long as they have the performance lead they can price it however they want, but that doesn't mean that it has to be a good value for us.

As I've said previously, I tend to skip "ticks" and instead go for the "tocks", and this tock sucks. This "tock" isn't even that good if it was priced at $369, 50% improvement over last architecture is not that impressive. Of course by my same logic above that 50% will probably come closer to 100% improvement if we actually use all the features the card gives us in CCC, but nothing warrants this %40 markup from last gen, except the fact that AMD wants to milk it until Kepler comes out.

Kind of the same thing NV did with GT200 by launching a GTX280 at $650, and then when AMD put out the 4870 at $299 MSRP two weeks later, NV was handing out rebate checks. AMD might be a bit luckier this time because NV is not quite 2 weeks late, but nonetheless, as an outlook towards 2012 performance Tahiti is not a good value. This is of course based on my assumption that if history repeats itself, Kepler will double performance at the same price point of Fermi, if NV botches that chip then my whole premise falls apart and we just end up paying lots of money for not very much performance improvement.

mmmkay?

 

[aphelion02]

JAG87: Your ticks and tocks only make sense in Intel's world. The 6xxx series was a tock as they moved from VLIW5 to VLIW4, and this 7xxx series is both a tick and a tock with the drop to 28nm and move to GCN. If anything, most of the performance boost should come from the tick as GCN is focused on compute.

 

[maddie]

No, it's definitely 100% faster, even more than 100% after driver revisions since launch.

You just have to create the scenario for this to show. Try 2560x1600 with MSAA+TRSSAA. GT200 struggled with TRSSAA, and it wasn't a frame buffer limit (1GB vs 1.5GB), it was just a GPU horsepower limit. Just because the benchmarks on websites don't show it, because they like to benchmark without the actual candy turned on (which is fair because they need to keep the testing environment as consistent as possible), it doesn't mean the performance is not there. I have first hand experience of the GTX480 putting out 2x the frames of my GTX280.





I didn't mean to criticize AMD's pricing. You are certainly correct that as long as they have the performance lead they can price it however they want, but that doesn't mean that it has to be a good value for us.

As I've said previously, I tend to skip "ticks" and instead go for the "tocks", and this tock sucks. This "tock" isn't even that good if it was priced at $369, 50% improvement over last architecture is not that impressive. Of course by my same logic above that 50% will probably come closer to 100% improvement if we actually use all the features the card gives us in CCC, but nothing warrants this %40 markup from last gen, except the fact that AMD wants to milk it until Kepler comes out.

Kind of the same thing NV did with GT200 by launching a GTX280 at $650, and then when AMD put out the 4870 at $299 MSRP two weeks later, NV was handing out rebate checks. AMD might be a bit luckier this time because NV is not quite 2 weeks late, but nonetheless, as an outlook towards 2012 performance Tahiti is not a good value. This is of course based on my assumption that if history repeats itself, Kepler will double performance at the same price point of Fermi, if NV botches that chip then my whole premise falls apart and we just end up paying lots of money for not very much performance improvement.

mmmkay?

Where it's positioned in the price/performance spectrum is all that matters. Arguing about ticks, tocks, new gen, old gen, etc, etc, is ridiculous.

 

[exar333]

Perhaps we're on different wavelengths.

I am talking architectures, not generations. 480 to 580 is not an architectural step, it's still Fermi. 6970 is not an architectural step from 5870 (even though Cayman and Evergreen have different names) Cayman is just a tweak of Evergreen. Tahiti is a new architecture compared with Cayman and at $549 it's price/performance sucks.

Are you really saying this isn't a good value when you have a GTx 480SLI rig? The 480 was $499 vs. 5870 at $379 and was only marginally faster, while using way more heat and arriving 6 months late.

Does not compute.

 

[jackstar7]

I can understand why you would think that based on the existing reviews that are out but the game is still extremely demanding when run at Ultra settings and high levels of AA. Most of the reviews on the internet that include numbers for World of Warcraft run their benchmarks by measuring FPS while on a wind-rider taking a flight-path from one point to another. That's because that is about the closest you can get to something that is repeatable in an MMO like WoW. There might be a few spots where you fly through a thick Forrest or something and your numbers dip down a bit but for the most part taking a flight-path on a wind-rider is not very graphically strenuous and is going to produce very high average FPS numbers.

There are numerous examples during real gameplay where FPS can dip quite a bit, like when things are getting rough in 25-man raid or even a 40-man Alterac Valley. That's also when performance counts most.

Just asking, because I don't know, but isn't WoW more CPU bound? And has it been optimized for quad+ CPUs? Because I thought the lag in points wasn't bound to effects and such... but lots of moving pieces and the like.

 

[ArchAngel777]

No, it's definitely 100% faster, even more than 100% after driver revisions since launch.

You just have to create the scenario for this to show. Try 2560x1600 with MSAA+TRSSAA. GT200 struggled with TRSSAA, and it wasn't a frame buffer limit (1GB vs 1.5GB), it was just a GPU horsepower limit. Just because the benchmarks on websites don't show it, because they like to benchmark without the actual candy turned on (which is fair because they need to keep the testing environment as consistent as possible), it doesn't mean the performance is not there. I have first hand experience of the GTX480 putting out 2x the frames of my GTX280.





I didn't mean to criticize AMD's pricing. You are certainly correct that as long as they have the performance lead they can price it however they want, but that doesn't mean that it has to be a good value for us.

As I've said previously, I tend to skip "ticks" and instead go for the "tocks", and this tock sucks. This "tock" isn't even that good if it was priced at $369, 50% improvement over last architecture is not that impressive. Of course by my same logic above that 50% will probably come closer to 100% improvement if we actually use all the features the card gives us in CCC, but nothing warrants this %40 markup from last gen, except the fact that AMD wants to milk it until Kepler comes out.

Kind of the same thing NV did with GT200 by launching a GTX280 at $650, and then when AMD put out the 4870 at $299 MSRP two weeks later, NV was handing out rebate checks. AMD might be a bit luckier this time because NV is not quite 2 weeks late, but nonetheless, as an outlook towards 2012 performance Tahiti is not a good value. This is of course based on my assumption that if history repeats itself, Kepler will double performance at the same price point of Fermi, if NV botches that chip then my whole premise falls apart and we just end up paying lots of money for not very much performance improvement.

mmmkay?

Ok, now with that response I can see where you are coming from. Now, I don't particular agree with all of it, but now at least I can understand. Basically, since you already have two high performning GPUs, the idea of spending $1100 to upgrade seems like a bad bargain. I'd fully agree with you. If I had two 480's in SLI I wouldn't dream of dropping $1100 for a simply 50% increase. But that doesn't have to do with the 7970 being a bad value so much as you already sitting with a high-end setup. For someone who has a GTX 280, or a 4870, the upgrade to a 7970 is going to be a steal compared to someone that already has a high end card.

With that said the biggest jump in performance ever, in my opinion, was the 7800GTX to the 8800GTX. It was often 3-4 times the performance... The 8800GTX was probably the best stomping, ever. nVidia really kicked butt with that design. Since that time, the gains have been sub 100%. Generally around 50-70%, or even less.

8800 GTX -> Ultra = 25%
8800 GTX Ultra -> 280 GTX = 65%
280 GTX -> 480 = 65%

I'd grant you that the 8800 to 9800 series was negligable, that 280 to 285 was negligable and that 480 to 580 was negligable... So, I guess, I'd agree with you about the 'tock' so to speak, but quite honestly, I can care less whether a tick or tock is what brings the performance boost. If it is the highest performing part at the time, it deserves to have a price premium.

Again, as a BUYER, I'd love to see these things priced at $1. But, that isn't going to happen, is it?

 

[JAG87]

JAG87: Your ticks and tocks only make sense in Intel's world. The 6xxx series was a tock as they moved from VLIW5 to VLIW4, and this 7xxx series is both a tick and a tock with the drop to 28nm and move to GCN. If anything, most of the performance boost should come from the tick as GCN is focused on compute.

No it doesn't. This type of strategy was in place long before Intel brought it up.

G70>G71, G80>G92, GT200>GT200b, GF100>GF110.

And ATI is exactly the same

R520>R580, R600>R670, RV770>RV790, Evergreen>Cayman, Tahiti should be a tock.

If Cayman was supposed to be a tock, it was a pretty awful one.

Are you really saying this isn't a good value when you have a GTx 480SLI rig? The 480 was $499 vs. 5870 at $379 and was only marginally faster, while using way more heat and arriving 6 months late.

Does not compute.

Don't turn this into an NV vs AMD value discussion.

This isn't what I'm contesting. I'm contesting value against AMD's own products, value of 7970 vs 6970, at 50% more performance = 50% more price. Do you want to know where we would end up like this?

8970 = 50% more performance than 7970 = $823
9970 = 50% more performance than 8970 = $1235

This card is a HORRIBLE value if priced at $549. And the only reason it's being priced like this is because NV has no product to counter it yet. Competition shouldn't be necessary to keep the price down. Look at Intel vs AMD. AMD hasn't had a competing product in years, yet we've gone through Core 2, Nehalem, Sandy Bridge, and now Ivy Bridge, all being released at comparable price points to their predecessors and offering us more performance for the same money.

But whatever, buy one, you'll see. You'll either feel taken by AMD when NV releases their card in Q2 and the MSRP drops to $399, or we will all feel taken if NV follows the same pricing model, i.e. GK100 is 50% faster than GTX580 = $549*150% = $823. This card should be priced at no more than $399, considering GK110 has very good chances of ending up faster.

But hey, don't listen to me. Buy one for $549 on launch day. I bought two GTX280s on launch day for $650, so I've walked the road before. I'm just trying to give you a heads up.

Ok, now with that response I can see where you are coming from. Now, I don't particular agree with all of it, but now at least I can understand. Basically, since you already have two high performning GPUs, the idea of spending $1100 to upgrade seems like a bad bargain. I'd fully agree with you. If I had two 480's in SLI I wouldn't dream of dropping $1100 for a simply 50% increase. But that doesn't have to do with the 7970 being a bad value so much as you already sitting with a high-end setup. For someone who has a GTX 280, or a 4870, the upgrade to a 7970 is going to be a steal compared to someone that already has a high end card.

With that said the biggest jump in performance ever, in my opinion, was the 7800GTX to the 8800GTX. It was often 3-4 times the performance... The 8800GTX was probably the best stomping, ever. nVidia really kicked butt with that design. Since that time, the gains have been sub 100%. Generally around 50-70%, or even less.

8800 GTX -> Ultra = 25%
8800 GTX Ultra -> 280 GTX = 65%
280 GTX -> 480 = 65%

I'd grant you that the 8800 to 9800 series was negligable, that 280 to 285 was negligable and that 480 to 580 was negligable... So, I guess, I'd agree with you about the 'tock' so to speak, but quite honestly, I can care less whether a tick or tock is what brings the performance boost. If it is the highest performing part at the time, it deserves to have a price premium.

Again, as a BUYER, I'd love to see these things priced at $1. But, that isn't going to happen, is it?

No not really. I have no problem spending $1100 to upgrade. I've had these 480s for almost TWO years, so you see, my initial $1100 investment has well payed off.

All I'm saying is that if something launches at a certain price premium, it will either drop, or we will pay through the nose going forward. Hence this card launched at $549 is bad value for us.

 

[Silverforce11]

Awesome CF scaling, 3gb vram, and zero power used on extra cards on idle, completely catering to multi-GPU setups, especially for HPC.

~200W TDP, +50W from powertune for OC, giving this kind of performance is beastly. $550, it will sell like hotcakes when you consider gtx580 ~$490 and the 3gb variant ~$590.

Edit: Above JAG^, GK110 is coming soon? Talking about the theoretical competitor is iffy when by the time it arrives, 6 months late, 7970 isn't its competitor. Until NV releases a product thats faster, AMD will sell the fastest GPU for $550 no problems.

 

[JAG87]

Edit: Above JAG^, GK110 is coming soon? Talking about the theoretical competitor is iffy when by the time it arrives, 6 months late, 7970 isn't its competitor.

That's true, but AMD shouldn't be milking it.

Plus I have a feeling GK104 will be here by Q1 2012, and will probably match the 7970 in performance (768 cores vs 512 on fermi = 150% theoretical, a little less in practice).

Now imagine, GK104 is supposed to be a mid-range product with 768 shaders, and GK110 should be the high end with 1024. By this pricing logic, NV should price the GK104 at $549? $549 for a mid range product?

Or will they price it at $349, and AMD will have to cut it's card's price by $200 after only a few months?

 

[Concillian]

All I'm saying is that if something launches at a certain price premium, it will either drop, or we will pay through the nose going forward. Hence this card launched at $549 is bad value for us.

Top end cards have never had much in the way of value. You pay to play, and pay through the nose.

If these are indeed 200W TDP, then AMD is leaving TDP in it's back pocket to use on a refresh. 40-50% improvement with 80% of the power is impressive. Potential is there for a 15-25% improvement or so just by bumping up to 250-300W.

A lot of the comparisons people are drawing in this thread are for TDPs that were equal across generations or TDPs that increased in the new generation. If the TDP is going down compared to last gen, there is a good reason it's not well over +50%

It's very clear that high end gaming is not the primary market AMD is going for here, just like computer enthusiast was not the primary market for BD. Our market is small ball now. We are relegated to a checklist feature that these server oriented pieces of hardware can also do well, but it's not the primary marketed purpose anymore.

 

[Vesku]

Since when has the top GPU being sold ever been about value? AMD just had an aggressive pricing scheme starting with the 4800 series, which they only executed because the 2900 and 3800 series had compared poorly to the NVIDIA competition. They've gained whatever marketshare they were going to gain after 3 product generations of pricing under NVIDIA in almost every performance bracket. So I'd say, somewhat tongue in cheek, it's all those people buying 580s over 6970s that will be to blame for a $550 7970.

~$500 seems to be tolerable for those who want the top tier card at least compared to the rate at which they can be fabricated. Seems both GPU makers have relegated the $600+ bracket to dual gpu single card solutions. It would be great for me and every other GPU buyer if they stick to the 6000 series pricing but it would also surprise me. I don't see this as a trend towards $850 top tier graphics cards. We'd need to see NVIDIA release their next gen card and price it compared to 7970 performance and then see AMD price their gen after 7970 based on comparison to the NVIDIA card before we could all start cursing the GPU bean counters.

Don't turn this into an NV vs AMD value discussion.

This isn't what I'm contesting. I'm contesting value against AMD's own products, value of 7970 vs 6970, at 50% more performance = 50% more price. Do you want to know where we would end up like this?

8970 = 50% more performance than 7970 = $823
9970 = 50% more performance than 8970 = $1235

This card is a HORRIBLE value if priced at $549. And the only reason it's being priced like this is because NV has no product to counter it yet. Competition shouldn't be necessary to keep the price down. Look at Intel vs AMD. AMD hasn't had a competing product in years, yet we've gone through Core 2, Nehalem, Sandy Bridge, and now Ivy Bridge, all being released at comparable price points to their predecessors and offering us more performance for the same money.

But whatever, buy one, you'll see. You'll either feel taken by AMD when NV releases their card in Q2 and the MSRP drops to $399, or we will all feel taken if NV follows the same pricing model, i.e. GK100 is 50% faster than GTX580 = $549*150% = $823. This card should be priced at no more than $399, considering GK110 has very good chances of ending up faster.

But hey, don't listen to me. Buy one for $549 on launch day. I bought two GTX280s on launch day for $650, so I've walked the road before. I'm just trying to give you a heads up.

 

[JAG87]

Since when has the top GPU being sold ever been about value? AMD just had an aggressive pricing scheme starting with the 4800 series, which they only executed because the 2900 and 3800 series had compared poorly to the NVIDIA competition. They've gained whatever marketshare they were going to gain after 3 product generations of pricing under NVIDIA in almost every performance bracket. So I'd say, somewhat tongue in cheek, it's all those people buying 580s over 6970s that will be to blame for a $550 7970.

~$500 seems to be tolerable for those who want the top tier card at least compared to the rate at which they can be fabricated. Seems both GPU makers have relegated the $600+ bracket to dual gpu single card solutions. It would be great for me and every other GPU buyer if they stick to the 6000 series pricing but it would also surprise me. I don't see this as a trend towards $850 top tier graphics cards. We'd need to see NVIDIA release their next gen card and price it compared to 7970 performance and then see AMD price their gen after 7970 based on comparison to the NVIDIA card before we could all start cursing the GPU bean counters.

I sure hope you are right.

 

[Pantalaimon]

8970 = 50% more performance than 7970 = $823
9970 = 50% more performance than 8970 = $1235

Your calculation logic would hold only if the 8970 comes out at a time when the 7970 is still selling for 500+ USD. Some how I doubt that.

 

[maddie]

That's true, but AMD shouldn't be milking it.

Plus I have a feeling GK104 will be here by Q1 2012, and will probably match the 7970 in performance (768 cores vs 512 on fermi = 150% theoretical, a little less in practice).

Now imagine, GK104 is supposed to be a mid-range product with 768 shaders, and GK110 should be the high end with 1024. By this pricing logic, NV should price the GK104 at $549? $549 for a mid range product?

Or will they price it at $349, and AMD will have to cut it's card's price by $200 after only a few months?

How do you feel about Nvidia "milking it"? I hope you realize that everything you wrote besides the 1st sentence are pure fantasy assumptions, with no facts, but none the less arriving at a decision

 

[Will Robinson]

I sure hope you are right.

You don't know the price and you don't know the performance of HD7970.
Your arguments FAIL and the fact that you bought two GTX480 Thermis shows you barely understand what makes a good GPU bargain anyway.

 

[ShadowOfMyself]

Huh, whats up with people complaining about the price? The top performing card has ALWAYS been around 500$, and it will always be like that since it has no competition

Be glad this isnt Nvidia actually, or you might get something like the 750$ 7800 GTX 512

 

[SlowSpyder]

That's true, but AMD shouldn't be milking it.

Plus I have a feeling GK104 will be here by Q1 2012, and will probably match the 7970 in performance (768 cores vs 512 on fermi = 150% theoretical, a little less in practice).

Now imagine, GK104 is supposed to be a mid-range product with 768 shaders, and GK110 should be the high end with 1024. By this pricing logic, NV should price the GK104 at $549? $549 for a mid range product?

Or will they price it at $349, and AMD will have to cut it's card's price by $200 after only a few months?

http://www.guru3d.com/article/geforce-gtx-280-review-test/20

The GTX280 released at $650. Looking through those benches I would say that, on average, it offered ~50% more performance than the 9800GTX that it replaced as Nvidia's single GPU king. High end GPU's are not about bang for the buck. I'm with you in that I will not pay $550 for this card. But, it is also not reasonable to expect AMD to price it at $380 if it's 30-50% faster than the $470+ GTX580, is it?

 

[Zargon]

@Zargon
BF3 is the most demanding game?

I thought that title went to something like Metro 2033?

I wouldn't say it's the most demanding, just that it demands the most performance from most players, i.e. run the game at 60fps minimum with all eye candy.

its certainly close.

forgot about metro 2033, probably due it not being played much

and gen hardware and video card forums have daily 'i need better bf3 performance what should I do' threads

seems like from some stuff posted shogun is rougher, and bf3 can edge out metro

 

[Silverforce11]

With the specs of 7950 being so close to 7970, honestly unless you really are a rich gamer with nothing better to spend your money, its pointless to pay extra for 7970.

Plus, they come with 3gb vram as well, so CF 7950 with be awesome for 1600p and eyeinfinity.