erockerYou have a TMSC insider informant?

Every one of TSMC's customers(ALtera, Quallcomm, AMD, and Nvidia) using the 28nm process that the GPUs are on wants MORE wafers. Stated above in my post with source.


You can also hear the same form any of TSMC's customers in thier quarterly financial reports. AMD stated that they managed to have "just enough" wafers, but they could have used more as well.

Both Qualcomm and Altera have stated that there are no yield issues, and they want more wafers.


Nvidia claims yield issues, but also wants more wafers.

ERgo, nVidia was unable to secure the supply they needed.


No "insider" needed; this is all public information.

Until the GTX680 has wide availability, it doesn't matter if it beats the 7970 or not (and there really isn't a clear winner, both side trade punches although nVidia has the upper hand on efficiency).

Right now, if you want a high end GPU there's the 7970 or waiting for another batch of GTX680s.

Not that I mind, I'm still chugging along with my HD3850 AGP :p

N3M3515Why would amd cut the price of 7970 if gtx 680 isn't available?

I don't think AMD would cut the price. If you read my post that you quoted, you'd see that I was in fact putting forward an counter-argument against iCookie's (post #115) assertion that AMD needed to cut prices.

johnnyfiiveSo, back to the topic.

" Do you think HD 7970 GHz Edition can make HD 7970 attractive again? "

The 7970 was never unattractive in my opinion, it was just priced wrong from the get go.
I think the GHz edition will be a nice welcome if its $449. I still think $469 is too much for a plain 7970. Fact is, the 680 does beat the 7970 in the majority of the games that actually matter (BF3, ding ding ding), so asking $30 less isn't enough. I think $50 less would make the 7970 the perfectly priced card. When the 670's launch, I have a feeling they will perform RIGHT under the 7970 but cost a good $50-60 less than the cheapest 7970. It would be in AMD's best interest to not bother releasing a "GHz" SKU and focus on strategic pricing, like they do with their processors.

I had/have both a 7970 and a GTX 680. Both are fantastic cards. However, I do like the 680 better. On the contrary, I like AMD's driver interface MUCH more than Nvidia's. They each have their pros and cons, but this time around, the 680 is the superior card. This is FACT, not fascinated fiction. It does indeed beat a reference 7970 in almost all benchmarks. It may not be a huge amount, but its enough to matter. Factor in the $549 MSRP vs 680 $499 MSRP, and its pretty obvious what the better card is. 7970 GHz and 7970 need to be $449 and $429 to make most people give a turd.

In the world i live in the 7970 costs $479 and sells in abundance, and GTX 680 is almost unexistent and where it is it costs $600 - $680

So, right now it really is $479 vs $600, amd does not need to lower the price if the situation continues this way.

Right now which is the better card? HD 7970 hands down. When GTX 680 is available and at the MSRP, then we'll talk. But for now it isn't.

what the companies need to do is make hybrid crossfire sli like they tried to do awhile back, i wouldnt mind a gtx680 sli-fired with hd7970 the hd7970 for folding and 680 for gaming and since sli-fired both for gaming ^^

HumanSmokeHate to burst your bubble, but the "GHz Edition" is 1050M core not 1000M. Secondly, here's a selection of OC'ed 7970's against the stock clocked card. Now assuming that the slightly OC'ed memory adds exactly zero to the added performance, you'll note that the gain for 1050MHz is 6-6.8% over stock

Cute story...pity it's a work of fiction

Note the first review. Four factory overclocked cards...three failed to break 1200M core. Or are you of the opinion that AIB's are happy to accept average bins five months after launch, while AMD stockpile the all-new all-dancing "SuperTahiti" for the next round of vanilla (reference) cards ?

Gunning to be Rory Reads gopher ?

Which is real cool for AMD assuming that Nvidia don't actually improve their arch. It's thinking like that that got AMD in the position where Kepler came as a complete surprise to them-not to mention a fully functional 512 shader GTX 580 before that. Do you think that if AMD had clue one about GK104's ability they would have released a 925M core part in the first place?
BTW: How do you know GK110 is going to be 20% faster than GK104?....or is this a story that starts out "In a perfect AMD world..."

what part of a new revision can you not understand? factory overclocked cards are regular tahitis with higher frequency, a bit higher voltage, and higher power consumption, those cards barely max out at 1180mhz

amd is now talking about a new revision, able to clock 1ghz or more at the same voltage as the first tahiti(925) and at the same consumption and these chips are said to be easily getting 1250mhz overclocks, meaning pro overclockers might even get more out of it
that being said, kepler is nomore efficient than GCN, it never was, efficiency is performance per watt, and an hd7750,7770, and 7850,7870 all do better in performance per watt and tahiti now will follow with this new revision, now lets see if the rest of the kepler line up will beat that.

also one note about the power consumption figures you mentioned in ur other post, i hope u are well aware than tahiti destroys gk104 in computer right? meaning a gk104 tesla or quadro or what not is hopeless against a firepro tahiti.

as for fermi yes they were hot, but they beat amd hands down in compute and gpgpu, so while they werent as efficient as amd in gaming, they were beasts in gpgpu and compute, amd did the best of both worlds
as for gk110 being 20% thats what i heard from the rumors online, the chip will be 40% bigger, but because of that it will have lower clocks, and a bigger memory controller so the performance will not perfectly scale with the size of the chip, look at tahiti vs pitcairn, tahiti has 40% more cores but is only 25-30% or so faster clock-clock

videocardz.com/31650/geforce-gtx-685-gk110-features-4gb-512bit-memory

I think Charlies maths is more correct this time (without his overly dramatic rendition of it).

NV were 'rumoured' to release the smaller parts first (to test the process)
AMD released Tahiti in Dec, realistically in Jan.
NV were not ready yet for whatever reason.
A full quarter later, NV paper launched GK104.
So they didn't release their small models first as rumoured (no big deal - was just rumour).
But they didn't release GK100 first either. GK100 rumoured to be broken.
So GK104 released a month and a half ago is still scarce in the US and on limited supply in UK (and no matter what websites people link to to say, hey here it is - it's still limited supply from a marketing perspective - and also pushing up prices).
GK110 is rumoured to be out soon but as Crap Daddy is linking to, is a pure gpgpu part - not for gaming.
Why GK110? Because the initial Kepler design wouldn't work on TSMC, ironically probably from a power consumption point. HPC parts need to be 'reasonably' power efficient. When you have thousands in a super computer they tend to gobble up Watts.

So i think 680 is in respin mode and will probably improve but until then, the stocks will suck.

The 7970 GHz edition might also be a respin (maybe even a tweak a la the GTX 580 from GTX 480).

Either way people are losing track of the qualities of the cards. It looks like Nvidia is splitting the Kepler line in to gaming OR HPC. So maybe GK104 is as good as it gets - which makes it their least 'winning' winning card for ages.
This would also explain why they spent so much attention to detail on the super star 690, using GK104 parts. This will be the single 'card' ruler for this round.

And if AMD hobbled the compute performance of the 7970, it's power use would be a lot less. Remember this is why Fermi drew so many Watts, it was a gpgpu card. This is also why GK104 draws relatively few Watts it's - not a gpgpu card.

+1 to 54ths

So the 7970 1ghz edition is basically going to be like the 4870 tweaked and re-released as 4890?
That was pretty epic back in the day, huge performance for value.
But the 7970 prices are just plain silly, even if they manage to squeeze some more performance out of it.

techtardBut the 7970 prices are just plain silly, even if they manage to squeeze some more performance out of it.

I'm not so sure. On a popular swedish site (Komplett.se) they can be had for €390 and there are loads and loads of them in stock. The 680 is about €440 and are pretty much non existent in the market with a preliminary delivery day of May 31. Some of them will not come until july.

techtardSo the 7970 1ghz edition is basically going to be like the 4870 tweaked and re-released as 4890?
That was pretty epic back in the day, huge performance for value.
But the 7970 prices are just plain silly, even if they manage to squeeze some more performance out of it.

i doubt that 4890 had a new gpu in it, this 7970 its just cherry picked 7970 at best at least that's my guess.

FrickI'm not so sure. On a popular swedish site (Komplett.se) they can be had for €390 and there are loads and loads of them in stock. The 680 is about €440 and are pretty much non existent in the market with a preliminary delivery day of May 31. Some of them will not come until july.

I just miss being able to get the second fastest AMD single gpu card for about $275. My 5850 was a bargain compared to what both camps are offering these days.

It wouldn't surprise me if there's some good old fashioned price fixing going on. LCD producers were semi-busted recently, Ati and nVidia were also caught before.

Maybe I'm just turning into an old, cheap asshole. :)

they should just name it HD7980, its good for sales.. nvidia has done this with success

techtardI just miss being able to get the second fastest AMD single gpu card for about $275. My 5850 was a bargain compared to what both camps are offering these days.

It wouldn't surprise me if there's some good old fashioned price fixing going on. LCD producers were semi-busted recently, Ati and nVidia were also caught before.

Maybe I'm just turning into an old, cheap asshole. :)

I rather think it was extremely cheap. Go back more and the prices were what they are now, if not higher. And I still think we get a lot for our money. A 7950 is about €350.

FrickI rather think it was extremely cheap. Go back more and the prices were what they are now, if not higher. And I still think we get a lot for our money. A 7950 is about €350.

mmm. . . . . . going back....
hd4850 $180 - hd4870 $299
hd3850 $180 - hd3870 $230

Hey, actually we're experiencing the highest prices ever :)

I don't recall the second fastest card from amd ever being at $450 at launch, not even $380.

N3M3515In the world i live in the 7970 costs $479 and sells in abundance, and GTX 680 is almost unexistent and where it is it costs $600 - $680

So, right now it really is $479 vs $600, amd does not need to lower the price if the situation continues this way.

Right now which is the better card? HD 7970 hands down. When GTX 680 is available and at the MSRP, then we'll talk. But for now it isn't.

The better card in performance is the gtx680.. Price means nothing, to people with money. It's not the chepest card, for sure. but it beats the competition in gaming, hands down. Idk how people, can even say things like this.

They don't need to lower the price I agree. But for people who solely choose, based on performance, in games.. Answer is pretty obvious to me.. Made me go from ATI to Nvidia, "no fan boy needed"..
I'll pay 100 extra, for a better card, anyday. This thread is just a flame war apparently. Most of the statements here, are pure, opinionated, dribble..

If cost is your issue, go AMD this time. If performance is your issue, go Nvidia this time. That's how I see it.. You're going to end up, with a BA card, either way..

TRWOV(and there really isn't a clear winner, both side trade punches although nVidia has the upper hand on efficiency).

Performance wise yes, they trade blow's, with more going to Nv side. It bother's me that people have this impression of efficiency when in reality it isn't as good as it look's because everyone seem's to keep forgetting that gk104 is supposed to be MID RANGE chip so when you look at it in that light your perspective change's.

I mean seriously a gpgpu crippled "mid range" chip that consumes nearly as much as the competition's full fledged top tier chip is hardly what i'd call efficient. Apologies if i offended anyone :o

Hopefully Nvidia get's their shit sorted so we can see some real progress

N3M3515mmm. . . . . . going back....
hd4850 $180 - hd4870 $299
hd3850 $180 - hd3870 $230

Hey, actually we're experiencing the highest prices ever :)

I don't recall the second fastest card from amd ever being at $450 at launch, not even $380.

I meant further back. x1900xtx was about $600 iirc, x850xtpe about the same. 8800GTX was about $500 and so on. 2007 to last year was very cheap compared to what it was before that and now prices have gone normal. My x1950 PRO was about €190 and it was far from the performance king.

But then early 2000's it was a bit cheaper iirc.. So it does move up and down. Price/performance figures are way up though. Now it's most of the time pointless in buying anything over €300 if you're on a 1080 monitor. That will change the next couple of years though.

FrickI meant further back. x1900xtx was about $600 iirc, x850xtpe about the same. 8800GTX was about $500 and so on. 2007 to last year was very cheap compared to what it was before that and now prices have gone normal. My x1950 PRO was about €190 and it was far from the performance king.

But then early 2000's it was a bit cheaper iirc.. So it does move up and down. Price/performance figures are way up though. Now it's most of the time pointless in buying anything over €300 if you're on a 1080 monitor. That will change the next couple of years though.

This. I remember when some GPU's were 700 bones for the top tier cards. But I think that was more to do with the RAM costs then anything. I honestly dunno. I just remember the 4800 series being awesome from a price performance perspective and at the time it wasn't the norm.

D007The better card in performance is the gtx680.. Price means nothing, to people with money. It's not the chepest card, for sure. but it beats the competition in gaming, hands down. Idk how people, can even say things like this.

They don't need to lower the price I agree. But for people who solely choose, based on performance, in games.. Answer is pretty obvious to me.. Made me go from ATI to Nvidia, "no fan boy needed"..
I'll pay 100 extra, for a better card, anyday. This thread is just a flame war apparently. Most of the statements here, are pure, opinionated, dribble..

If cost is your issue, go AMD this time. If performance is your issue, go Nvidia this time. That's how I see it.. You're going to end up, with a BA card, either way..

I have to disagree with you in something, performance means nothing for me if it is 7% for A HUNDRED bucks more.... :)

And 7% is not hands down, it isn't a beating, etc. They are equals when you're playing with any of the two.

I would accept if you tell me you go for NV for TXAA, and stuff, but $100 more for 7% perf that you won't even realize?, are you for real? NO, you just love nvidia, just admit it. there's nothing wrong with it.

For me, i'm a bang-for-bucker, be nv or be dammit, whichever gives me the best bang for my buck. Hell, i still have my $150 HD 4870 and nothing can replace it at the moment at the same price.

N3M3515mmm. . . . . . going back....
hd4850 $180 - hd4870 $299
hd3850 $180 - hd3870 $230

Hey, actually we're experiencing the highest prices ever :)

I don't recall the second fastest card from amd ever being at $450 at launch, not even $380.

1. note that there is something called inflation

2. at the days of ati they never made bigger chips and would never compete with nvidia on the fastest gpu, but were all about efficiency, the hd3870 was a tiny chip with 320radeon cores that had a die size(192mm2) smaller of that of the hd7870 pitcairn(212mm2)

the hd5870 was the first time amd made the card bigger around 320mm2, but then for the 6000 series amd introduced the 6900 series line and started using a different methodology in building chips in each family , so instead of using 2 chip designs one for x700 and one for x800, now they use 3 chip designs, one for 7700(cape verde) one for 7800(pitcairn) and one for 7900(tahiti)
with tahiti being 360mm2
the 6000 series also had 3 chips one using the old hd5700 series die, one using the new improved vliw5 chip called barts (hd6800) and then the vliw4 cayman (hd6900) which measured 389mm2


that being said, you must stop comparing the 4800 series to the 7900 series in terms of price as the whole methodology has changed, as for hd5800 series it was a step in between so that is were it gets tricky with hd5800 being bigger than hd6800 but smaller than hd6900. but that is why the 5870 had a smaller price premium

so while the hd 4870 cost way less than amds high end today, its simply because ati never used to make high end chips to compete with nvidias bigger cards and would only counter with dual gpu cards
now in the 299$ bracket is the hd7870(i know its 350 but once competition shows up it will go down) and that is the gpu class that the 4870 filled in a few years ago which you
should be looking at

sergionographywhat part of a new revision can you not understand? factory overclocked cards are regular tahitis with higher frequency, a bit higher voltage, and higher power consumption, those cards barely max out at 1180mhz. amd is now talking about a new revision, able to clock 1ghz or more at the same voltage as the first tahiti(925) and at the same consumption

The article says no such thing about the GPU being a revision. All it is saying is that as the process is refined there is less voltage leakage and a bit more performance headroom (i.e. the standard deviation curve is moving to higher freqs. If you're expecting Tahiti XTX to be a foundry respin you're in a waking dream.

sergionographythat being said, kepler is nomore efficient than GCN, it never was, efficiency is performance per watt, and an hd7750,7770, and 7850,7870 all do better in performance per watt

So your idea of comparison is to take the highest efficiency mainstream (and lower) card and measure it against a card a considerable step up in market segment. What next? comparing power consumption of the GTX 680 against that of the HD 6450 ?, acoustics of the GTX 680 against a passive cooled card ?

sergionographyand tahiti now will follow with this new revision, now letsCsee if the rest of the kepler line up will beat that.

Sorry, not convinced that a binned Tahiti is the next messiah. Don't save me a pew at the Church of Redfanboyism

sergionographyi hope u are well aware than tahiti destroys gk104 in computer right? meaning a gk104 tesla or quadro or what not is hopeless against a firepro tahiti.

Amazing how "computer" ( I presume you mean compute function/ GPGPU) has suddenly become of major importance with AMD followers -where was all this corncern when Fermi and Evergreen were having to go round.
On your second point, you do realise 1. that Quadro/Tesla will be based on GK110 since GK104 has no ECC 72-bit memory and is constrained of double precision FP performance, and...2. AMD have had capable workstation cards for generations- they just heaven't put much effort into a software enviroment or drivers for the pro sector. Big engine great. Not being able to figure out how to shift out of neutral bad.
Since you're all for lopsided comparisons, are you willing to bet that Tahiti will be a GPGPU match for GK110...It sounds like Cray aren't

sergionographyas for gk110 being 20% thats what i heard from the rumors online, the chip will be 40% bigger, but because of that it will have lower clocks, and a bigger memory controller so the performance will not perfectly scale with the size of the chip...videocardz.com/31650/geforce-gtx-685-gk110-features-4gb-512bit-memory

The link you posted actually quotes 20-25%. Leaving aside your lowballing. The 20-25% is gaming performance not compute. Since you have trouble distinguishing the two:
GTX 680 FLOPS 1006M core x 1536 shader x 2 OPC = 3090.432 GFlop...Double precision artificially capped at 1:24 rate
GK110 would need only a 800M core clock to have a 20% FLOP advantage ( 800 x 2304 x 2 OPC = 3686 GFlops), but here's the kicker. Quadro DP is a full 1:2 rate. Now according to this 3DCentre article (probably a bit more credible than Videocardz and OBR)single precision is estimated at 4000+ GFlops (2000+ double precision) so;
GK110 4000+ FP32 and 2000+ FP64
Tahiti XT 3788 FP32 and 947 FP64...and that making a huge assumption that an AMD pro card could be built around Tahiti XT. For AMD's last arch, they used Cayman LE...a HD 6950 with 128 shaders fused off (Firepro V7900)

HumanSmokeThe article says no such thing about the GPU being a revision. All it is saying is that as the process is refined there is less voltage leakage and a bit more performance headroom (i.e. the standard deviation curve is moving to higher freqs. If you're expecting Tahiti XTX to be a foundry respin you're in a waking dream.

So your idea of comparison is to take the highest efficiency mainstream (and lower) card and measure it against a card a considerable step up in market segment. What next? comparing power consumption of the GTX 680 against that of the HD 6450 ?, acoustics of the GTX 680 against a passive cooled card ?

Sorry, not convinced that a binned Tahiti is the next messiah. Don't save me a pew at the Church of Redfanboyism

Amazing how "computer" ( I presume you mean compute function/ GPGPU) has suddenly become of major importance with AMD followers -where was all this corncern when Fermi and Evergreen were having to go round.
On your second point, you do realise 1. that Quadro/Tesla will be based on GK110 since GK104 has no ECC 72-bit memory and is constrained of double precision FP performance, and...2. AMD have had capable workstation cards for generations- they just heaven't put much effort into a software enviroment or drivers for the pro sector. Big engine great. Not being able to figure out how to shift out of neutral bad.
Since you're all for lopsided comparisons, are you willing to bet that Tahiti will be a GPGPU match for GK110...It sounds like Cray aren't

The link you posted actually quotes 20-25%. Leaving aside your lowballing. The 20-25% is gaming performance not compute. Since you have trouble distinguishing the two:
GTX 680 FLOPS 1006M core x 1536 shader x 2 OPC = 3090.432 GFlop...Double precision artificially capped at 1:24 rate
GK110 would need only a 800M core clock to have a 20% FLOP advantage ( 800 x 2304 x 2 OPC = 3686 GFlops), but here's the kicker. Quadro DP is a full 1:2 rate. Now according to this 3DCentre article (probably a bit more credible than Videocardz and OBR)single precision is estimated at 4000+ GFlops (2000+ double precision) so;
GK110 4000+ FP32 and 2000+ FP64
Tahiti XT 3788 FP32 and 947 FP64...and that making a huge assumption that an AMD pro card could be built around Tahiti XT. For AMD's last arch, they used Cayman LE...a HD 6950 with 128 shaders fused off (Firepro V7900)

i like that super computer link, for an odd reason i was looking up all about super computers its intrigues me to no end... but after looking it up for so long...it just seems its not about the power of the chip... its the power of a nation... japan vs america and all that crap... it is definetly turning into national superiorism.... but what isn't these days.

sergionography1. note that there is something called inflation

2. at the days of ati they never made bigger chips and would never compete with nvidia on the fastest gpu, but were all about efficiency, the hd3870 was a tiny chip with 320radeon cores that had a die size(192mm2) smaller of that of the hd7870 pitcairn(212mm2)

the hd5870 was the first time amd made the card bigger around 320mm2, but then for the 6000 series amd introduced the 6900 series line and started using a different methodology in building chips in each family , so instead of using 2 chip designs one for x700 and one for x800, now they use 3 chip designs, one for 7700(cape verde) one for 7800(pitcairn) and one for 7900(tahiti)
with tahiti being 360mm2
the 6000 series also had 3 chips one using the old hd5700 series die, one using the new improved vliw5 chip called barts (hd6800) and then the vliw4 cayman (hd6900) which measured 389mm2


that being said, you must stop comparing the 4800 series to the 7900 series in terms of price as the whole methodology has changed, as for hd5800 series it was a step in between so that is were it gets tricky with hd5800 being bigger than hd6800 but smaller than hd6900. but that is why the 5870 had a smaller price premium

so while the hd 4870 cost way less than amds high end today, its simply because ati never used to make high end chips to compete with nvidias bigger cards and would only counter with dual gpu cards
now in the 299$ bracket is the hd7870(i know its 350 but once competition shows up it will go down) and that is the gpu class that the 4870 filled in a few years ago which you
should be looking at

So......in two years what? X970 at $700 and X950 at $600?
The way i understand gpu pricing is this:
4 ranges
DualGPU >600
Highend 300 - 500 (4870, 5850, 5870, 6970, GTX570, GTX260, GTX280, GTX480, GTX470, X850, X1900XT, X1800XT, 7800GTX, 6800ultra, 9800XT, 8800GTX, 8800GTS)
Midrange 180 - 250 (3850, 3870, 4850, 5770, 6850, 6870, gtx 560ti, gtx 560, gtx 460, 6800GS, 5700ultra, X700, 9600XT, 7600GT, X1900GTO, 8800GT, 9500pro)
Lowend 80 - 150 (...)

Right now there is a gigantic hole in the 150 - 250 range, which for the most people is the sweet spot($200 to be exact). 7870 should have started at the most, at 300 and that is still expensive but understandable for a new product and all the inflation and stuff...
.....and 7950 at 350. I have no doubt they will go down in price once nvidia put its shit together and have a complete lineup and good supply of new generation chips, BUT, i pretty much doubt 7950 will ever go down to $260 like the 6950 2GB, because the starting price is so sky high. Just like the 7870 will never be as low as the $155 HD 6870.

N3M3515So......in two years what? X970 at $700 and X950 at $600?
The way i understand gpu pricing is this:
4 ranges
DualGPU >600
Highend 300 - 500 (4870, 5850, 5870, 6970, GTX570, GTX260, GTX280, GTX480, GTX470, X850, X1900XT, X1800XT, 7800GTX, 6800ultra, 9800XT, 8800GTX, 8800GTS)
Midrange 180 - 250 (3850, 3870, 4850, 5770, 6850, 6870, gtx 560ti, gtx 560, gtx 460, 6800GS, 5700ultra, X700, 9600XT, 7600GT, X1900GTO, 8800GT, 9500pro)
Lowend 80 - 150 (...)

Right now there is a gigantic hole in the 150 - 250 range, which for the most people is the sweet spot($200 to be exact). 7870 should have started at the most, at 300 and that is still expensive but understandable for a new product and all the inflation and stuff...
.....and 7950 at 350. I have no doubt they will go down in price once nvidia put its shit together and have a complete lineup and good supply of new generation chips, BUT, i pretty much doubt 7950 will ever go down to $260 like the 6950 2GB, because the starting price is so sky high. Just like the 7870 will never be as low as the $155 HD 6870.

intel cpus hardly ever deprieciate in price, if newegg still had Pentium 4 EE's in stock from 2004 they would cost $999 easily