AMD Hawaii-based graphics cards to mass ship in October

Well going by history and economics, when a company has a faster chip, they just say, our chip is going to be faster period.

it's going to be faster at price points, because that's just logical. it's the 4000 series again. not the absolute fastest but an amazing bargain.

My only concern would be that it would be a disaster if maxwell turned out to be a core2duo or 8800GTX paradigm game changer. Last thing we need is nvidia becoming an intel, with no serious competetion upper levels

The issue is that it might fall between the 780 and the Titan. Maybe they can't say it's outright the fastest because of the Titan.
 
Difference between the Titan and the 780 is <10% though, so it would be kind of awkward for it to be faster than the 780 but slower than the Titan, basically well within margins of error of both.
 
Yeah... an overclocked 780 is faster than a stock Titan (isn't 780 Lightening such, iirc)... so if Hawaii is faster than 780, there's a good chance that it'll be faster than a Titan.

On the flip side, if Hawaii can't beat a Titan, there's a good chance that it can't beat a 780, either.
 
If its faster than the 7970 GHz edition then it will be competitive. And it should be...pretty easy to understand.
 
Last edited:
Well going by history and economics, when a company has a faster chip, they just say, our chip is going to be faster period.

it's going to be faster at price points, because that's just logical. it's the 4000 series again. not the absolute fastest but an amazing bargain.

My only concern would be that it would be a disaster if maxwell turned out to be a core2duo or 8800GTX paradigm game changer. Last thing we need is nvidia becoming an intel, with no serious competetion upper levels

Point stands :p They never outright said it wouldn't be faster! just saying I agree that it probably won't be but it was never stated.
 
Or they don't want to tip their hand before they have to regarding performance.

Nvidia is waiting in the wings with a 15SMX Titan Ultra or GTX785 and the longer AMD can keep the exact performance of Hawaii a secret the better off they will be.
 
Or they don't want to tip their hand before they have to regarding performance.

Nvidia is waiting in the wings with a 15SMX Titan Ultra or GTX785 and the longer AMD can keep the exact performance of Hawaii a secret the better off they will be.

That too is a possibility. AMD has been known to keep its cards pretty damn close.
 
Or they don't want to tip their hand before they have to regarding performance.

Nvidia is waiting in the wings with a 15SMX Titan Ultra or GTX785 and the longer AMD can keep the exact performance of Hawaii a secret the better off they will be.

this doesnt make sense, because such a card would be in the ulra enthusist $1000 range which AMD stated flat out they aren't going to compete in.
 
this doesnt make sense, because such a card would be in the ulra enthusist $1000 range which AMD stated flat out they aren't going to compete in.

Correct, they won't be pricing their consumer cards in that range...

AMD is trying to rebuild/recreate their GPU brand, i.e. new naming/numbering scheme, you don't do that by raping the consumer for a few months with outrageous price/performance and then, when the competition shows up, compete with slightly cutdown SKUs or pricedrops to 65% your original MSRP.
 
Last edited:
Correct, they won't be pricing their consumer cards in that range...

because they don't have a chip in that range, if they could they would. companies are interesting in making money. Nvidia made alot of money with the Titan. if AMD could they would.
 
because they don't have a chip in that range, if they could they would. companies are interesting in making money. Nvidia made alot of money with the Titan. if AMD could they would.

You should search out the revenue breakdown for different market segments, since you clearly haven't before.
 
because they don't have a chip in that range, if they could they would. companies are interesting in making money. Nvidia made alot of money with the Titan. if AMD could they would.

Nvidia makes money with Titan because it is a pseudo-Quadro/Tesla, and for those purposes it was stupid cheap. For those gamers that can afford them, they are the most powerful GPU on the market and also one of only two GPUs released with >4GB of RAM, both of which are appealing, especially since Nvidia didn't put 8GB (4GB/GPU) on the GTX690 for whatever inane reason.

Remember that a GPU the size of GK110 was once sold for ~$200. It's not like Nvidia can't lower the price on these things and still make money, even if the Titan does cost them quite a bit more to make.

The real reason Nvidia can sell Titans for $1000 is that AMD totally flopped on top-end performance with the HD7970. That let Nvidia get away with not shipping a large GPU, and instead ship their mid-range 'half-Kepler' GK104 chips as their high-end SKUs. What once was relegated to the mid-range (previously GTX460/GTX560) now became high-end (GTX670/GTX680), thanks to AMD.

And hopefully, this is a chance for AMD to get back in that game; they need a high-end product that can go toe-to-toe with the GTX780, which shouldn't be hard considering the shear amount of compute resources Nvidia puts into their big chips, and they need a mid-range product that can outshine the GK104 in the GTX670/680/770.

And they need a shit-ton more RAM available at every level. I'm shooting for 2x8GB cards here, for next-gen games and for 4k output, and I need at least twice the performance of my 2GB GTX670s to do that.
 
Nvidia makes money with Titan because it is a pseudo-Quadro/Tesla, and for those purposes it was stupid cheap. For those gamers that can afford them, they are the most powerful GPU on the market and also one of only two GPUs released with >4GB of RAM, both of which are appealing, especially since Nvidia didn't put 8GB (4GB/GPU) on the GTX690 for whatever inane reason.

Remember that a GPU the size of GK110 was once sold for ~$200. It's not like Nvidia can't lower the price on these things and still make money, even if the Titan does cost them quite a bit more to make.

The real reason Nvidia can sell Titans for $1000 is that AMD totally flopped on top-end performance with the HD7970. That let Nvidia get away with not shipping a large GPU, and instead ship their mid-range 'half-Kepler' GK104 chips as their high-end SKUs. What once was relegated to the mid-range (previously GTX460/GTX560) now became high-end (GTX670/GTX680), thanks to AMD.

And hopefully, this is a chance for AMD to get back in that game; they need a high-end product that can go toe-to-toe with the GTX780, which shouldn't be hard considering the shear amount of compute resources Nvidia puts into their big chips, and they need a mid-range product that can outshine the GK104 in the GTX670/680/770.

And they need a shit-ton more RAM available at every level. I'm shooting for 2x8GB cards here, for next-gen games and for 4k output, and I need at least twice the performance of my 2GB GTX670s to do that.

That "the 680 is really a 660" argument is pure speculation. the 7970 was a HUGE jump over both the 6970 AND the 580, moreso than the 6970 over the 5870. Its ridiculous to think that that was 'dropping the ball'. The truth is that Nvidia was feeling TONS of pressure with AMD's 4,5,and 6K series posing a SERIOUS threat to Nvidia's market domination (and what do you know, AMD outsold Nvidia last quarter). They tried to rush out the GK110, but yield levels were too low to sell it for less than 2k per low-bin, let alone the $700 the 7970 was flying off the shelves for. So they beefed up the smaller, more stable GK104 core and used that instead. Chances are the GK104 was meant to be the 660, and what do you know: it still is. But changes were made to it in order to allow it to sit in the top-range.
 
That "the 680 is really a 660" argument is pure speculation. the 7970 was a HUGE jump over both the 6970 AND the 580, moreso than the 6970 over the 5870. Its ridiculous to think that that was 'dropping the ball'. The truth is that Nvidia was feeling TONS of pressure with AMD's 4,5,and 6K series posing a SERIOUS threat to Nvidia's market domination (and what do you know, AMD outsold Nvidia last quarter). They tried to rush out the GK110, but yield levels were too low to sell it for less than 2k per low-bin, let alone the $700 the 7970 was flying off the shelves for. So they beefed up the smaller, more stable GK104 core and used that instead. Chances are the GK104 was meant to be the 660, and what do you know: it still is. But changes were made to it in order to allow it to sit in the top-range.

They tried to rush out GK100- that didn't work. That's why we have GK110 instead, and it came late. But they also did something else- they took their mid-range part, GK104, with it's neutered compute capabilities, and massaged it to put out significantly more gaming performance- a large part of that being an increase in efficiency in the memory controller.

But make no mistake- GK104 is a mid-range part. It's the same size as the part that went into the GTX460 and GTX560, while the GK110 in the GTX780 and Titan is the same size as the parts that went into the GTX280/480/580. This has been going on for a while.

So yes, the GTX680 would have competed against the HD7870 in market position (as the GTX660), and the never-released GK100-based 'actual GTX680' would have competed against the HD7970, but AMD's performance jump just wasn't there while Nvidia's was- and Nvidia's comments on the HD7970's performance were made known (somewhere). They laughed.

Now, a large part of this is due to AMD's drivers; Nvidia was able to execute a new architecture with new drivers, but AMD failed quite miserably. Everyone noted how much more resources the HD7970 had over the GTX680 yet at launch the GTX680 was faster. Sure, the AMD parts came out six months earlier, but that means that six months after launch AMD couldn't get their drivers together enough to compete with a mid-range Nvidia part! Nvidia is still laughing all the way to the bank. GTX260's, which had the same size GPU as the Titan does now, once sold for $200. They could now ask $1000 for such a GPU ($700 later with the cut-down consumer version), and people would pay it, and then ask for $500 for their mid-range part with same size GPU as a GTX560!
 
It wasn't a mid-range part: was a 560 ti 30% faster than a 480? The usual jump in speeds in GPUs is 20-30%, which is just where AMDs 7 series card hit, and right were the 680 hit. There is no way Nvidia was hoping to manufacture a '660' that was 30% faster than a 580. Thats just bad business.
 
R9 290x 2816 stream processors oh...my

Public webcast of AMD's new GPU products on Sept. 25th.:
http://www.anandtech.com/show/7344/amd-announces-2014-gpu-product-showcase-webcast-sept-25-3pm-edt

Also, from 3DCenter.org:
We are looking at roughly a 430mm2 chip, some 18 percent bigger than Tahiti. This corresponds with AMD&#8217;s official claim that Hawaii is about 15 percent bigger than its predecessor and 30 percent smaller than the GK110. The GPU includes 2.0 GCN architecture, while Tahiti was a 1.0 GCN part. It has four raster engines, twice as many as Tahiti.

However, the shader count and clocks are not pinned down yet. The GPU is expected to feature up to 2816 in 44 clusters, 37 percent more than Tahiti and more than Nvidia&#8217;s GTX 780, although comparing AMD and Nvidia shader counts is a case of apples and oranges. However, 3Dcenter does not have the exact number and it could be anywhere between 2560 and 2816.


The card is said to have a 384-bit memory bus and we would expect nothing less, or more for that matter.
Source: http://www.fudzilla.com/home/item/32553-hawaii-gpu-spec-leaked-sort-of
Original: 3D Center.org
 
680 was definately a mid range part, that got bumped to high end because of AMD lack of performance.
 

It's only funny because it's true- a GTX680 being faster than an HD7970 on release is like a GTX560 (not the Ti Core 448 model) being faster than an HD6970. The idea of Nvidia's mid-range part outrunning AMD's high end-part was ridiculous until the GTX680 came out.
 
It's only funny because it's true- a GTX680 being faster than an HD7970 on release is like a GTX560 (not the Ti Core 448 model) being faster than an HD6970. The idea of Nvidia's mid-range part outrunning AMD's high end-part was ridiculous until the GTX680 came out.

It's more of a case of nvidia overreaching with their design for their whole line-up. If you look at the issues they faced trying to fab GK100, they shifted their entire chip line up for this generation too far and ended with a top end that was unmanufacturable and a mid range that looked (after some tweaking) like a top end card.
 
It's more of a case of nvidia overreaching with their design for their whole line-up. If you look at the issues they faced trying to fab GK100, they shifted their entire chip line up for this generation too far and ended with a top end that was unmanufacturable and a mid range that looked (after some tweaking) like a top end card.

That was the fastest card they could put out at the time so to call it mid level is just plain stupid and quite laughable. Throwing that old argument out in a thread like this is just plain retarded.
 
It's only funny because it's true- a GTX680 being faster than an HD7970 on release is like a GTX560 (not the Ti Core 448 model) being faster than an HD6970. The idea of Nvidia's mid-range part outrunning AMD's high end-part was ridiculous until the GTX680 came out.

it's just semantics man. they can call it their 'mid-range' part if they want but if it's the fastest card they offer at that time then it's their current flagship. their 'actual high-end' part the Titan came out nearly a year after the 680. AMD could follow the same line of semantics if they wanted to and release the R9-290X calling it a mid-range part; "Our high-end gpu will be released under the name R9-295X in 2014."
 
that whole "mid range 680 argument' is pretty stupid IMO.

It was only slightly faster than the 7970 @ launch, so people still purchased the 7970 regardless. Do people really think that if NV had the chance to completely obliterate AMDs top offering with the "GK 110" that they wouldn't have? that would have taken a lot more sales away from AMD and put more money into nVs pocket.

The only logical point would have been to say they wanted to sell 680s, and then 780s, but then they could have designed a card between the 680 and 780 to give them more of an edge over AMD and a decisive lead vs the neck to neck fight it ended up being.
 
Any guesses so far on the value proposition of the new 7970 replacement vs $200 7950 and $300 7970's?
 
Yeah I don't see AMD charging equal to 780 pricing unless it seriously outperforms. At or slightly less than 780 for $499 will prob do just fine though.
 
it's just semantics man. they can call it their 'mid-range' part if they want but if it's the fastest card they offer at that time then it's their current flagship. their 'actual high-end' part the Titan came out nearly a year after the 680. AMD could follow the same line of semantics if they wanted to and release the R9-290X calling it a mid-range part; "Our high-end gpu will be released under the name R9-295X in 2014."

The real point is that what went into the GTX680 was an evolution of the part that went into the GTX460/560- and that it was more than competitive with AMD's top end. The real failure is on AMD for not improving performance, both from a hardware standpoint and from a driver standpoint.

Case in 'point': if you bought one of those HD7970's with the hope of getting a second for a Crossfire/Eyefinity setup, or you did get the second (or third) card, you're still screwed because AMD hasn't managed to fix their drivers yet. They didn't even know they were broke (or will never admit that they knew). But if you bought Nvidia cards, you were set from the get-go; you had a faster, quieter solution that actually worked and delivered value, and a large part of that was Nvidia delivering an efficient, mid-range part with a decent cooling, that could actually deliver the performance without sounding like a dustbuster. Hell, you could put two or three in a smaller enclosure using a positive-pressure setup with stock coolers. Just getting the AMD cards to get that quiet meant using at least a large third-party open-air cooler and a larger enclosure with a higher flowing, carefully balanced cooling setup.

Nvidia won the last generation on so many levels; in retrospect, they won it even more than it really seemed at the time, by delivering more value for both their customers and their shareholders.

And really, this isn't a good thing for gaming- AMD refocused on compute this last generation, likely trying to get into the 'Tesla' market of compute products, throwing a lot of transistors at stuff that doesn't make games faster, while Nvidia's mid-range product (unlike their top-end product that failed at the foundry) didn't have any of that stuff. Essentially, AMD made smaller versions of the same products that Nvidia did, with a larger compute-focused GPU and a smaller (HD7870) GPU lacking the strong compute resources focused on gaming. And that meant them being slower at every performance level die-for-die, which meant that they made far less than Nvidia did- they could have shipped three times the product and they still would have took home less pay.

So here's hoping that this generation is different- here's hoping AMD refocuses their product line to better compete with their rival, and they get all of the other important stuff like drivers and cooling right too.

My personal interest here, of course, is in trying to power next-generation games on a 4k display- I'm going to need more than one card, and lots of VRAM, and damn good driver support. Nvidia has that, but they're going to charge an arm and a leg- or at least they will, if AMD doesn't show up to this fight.
 
The real point is that what went into the GTX680 was an evolution of the part that went into the GTX460/560- and that it was more than competitive with AMD's top end. The real failure is on AMD for not improving performance, both from a hardware standpoint and from a driver standpoint.

The 7970 brought a pretty solid performance increase. I think the real take away from this is that Nvidia designed their GTX6xx series line up with a top end chip that was damn near unmanufacturable.
 
The 7970 brought a pretty solid performance increase. I think the real take away from this is that Nvidia designed their GTX6xx series line up with a top end chip that was damn near unmanufacturable.

That they did- and lucky for them, they were still able to 'hold the line', again partly because they did so well optimizing their mid-range chip and partly because AMD's top-end performance was less than they were expecting.

The HD7970 series did bring a realized performance increase, but as we're seeing now (and understood then) there was both a lot of untapped performance and a lot of performance potential that wasn't taken advantage of for gaming, but rather for compute purposes. That shift to compute for them came at entirely the wrong time for their gaming market performance; not because they weren't competitive, but because they weren't able to make the kind of money that Nvidia was.

At least this console win will be a boon to PC gaming- a big one, really. It's not an advantage for AMD over Nvidia, of course, but rather a shift towards modern GPU architectures as a development baseline, which means that we're about to see a whole new battle; and one that AMD really has a chance at competing heavily in, given their status as the 'underdog' and their issues with drivers. If they can prove the driver stuff is behind them and put out products that are outcompeting Nvidia's aging products at 28nm, they have a very good opportunity to regain momentum in the PC market and really take home some profits, while giving us the products we need to handle the incoming PC releases at reasonable prices!
 
And really, this isn't a good thing for gaming- AMD refocused on compute this last generation, likely trying to get into the 'Tesla' market of compute products, throwing a lot of transistors at stuff that doesn't make games faster,

the compute makes games better like dirt and latest Lara tombraider with hair.
you could want them to go game mode only but adding compute will add layers of possibility to add functions that actually allows the developer to do nice things.

since the majority of gamers play at 1920x1080 todays and previous generation cards are fast enough. even my set up runs fine on a single 7970 using compute with 80+ fps in BF3 at 5040x1050 sure I dont use max settings as I care for the fps there but other games runs fine maxed out, so not sure why people would need a gaming fully transistor card as current 7900 series do just fine.

20nm is around the corner, which is likely to be a gamechanger in videocards as single cards will run eyefinity like todays 7990 or so performance wise.
that shrink will do a lot.
 
I saw techspot had the unofficial spec's (along with a couple other websites) for the upcoming flagship card and it does look impressive. If what is posted is true or near it then how can you not think it will compete with the 780's?
 
I saw techspot had the unofficial spec's (along with a couple other websites) for the upcoming flagship card and it does look impressive. If what is posted is true or near it then how can you not think it will compete with the 780's?

I expect slightly performance above a 780 with price to macth the 780.
 
I'd rather have something around the price range of the 7970, but my bank account is ready for the reamin' if these cards look good.

Oh, and they need to be well below 33cm and have a leaf blower-design. I have a small case.

Bonus points for 6GB model.
 
Back
Top