[ TweakTown ]GeForce GTX 800 series GPUs to be cheaper, faster than 700 series

[blastingcap]

The "source" isn't TweakTown, it's WCCFTech, as it says at the bottom of the Tweaktown article.

http://wccftech.com/nvidia-geforce-gtx-880-800-series-powerful-cheaper-700-series/

THEIR source is EXPReview
http://www.expreview.com/34003.html

THEIR source is "KDramaStars"
http://www.kdramastars.com/articles/24370/20140609/gtx-800-series-news.htm

THEIR source is "Cyberland PC"

I have no idea what that is as there is no link, but I assume it's just a reliable as anyone else, i.e. not at all.
So yes. Tweaktown having nothing to say is because it's a 5 website deep rumour. GG.

We should really just close terrible threads like these where it's a rumour from a rumour from a rumour from a rumour from a rumour with no one actually caring who originally said anything.

Also I should set up a terrible website where I make up crap, then set up some more websites that link to them, and then get progressively less terrible websites to link to those so that I can make up dumb rumours and pass them off as having some basis because no one every bothers to check what the actual underlying "source" is.

Yup. Rumors of rumors of rumors... who cares?

 

[PhIlLy ChEeSe]

OK, in that rumor of a rumor of a rumor of a rumor do they hint at a date? Or does anyone know the projected date, feel free to stomp on me if it was already posted..................

 

[el etro]

Want to see too the effects of Stacked-vRAM on consumption of new-gen cards.

 

[Techhog]

Quotes from the "source"

You know you have nothing to say when you repeat the same sentence five times in an 8 sentence article. And that includes the title.

~Harder, better, faster, cheaper~

 

[blackened23]

It actually makes sense why GTX880 could end up cheaper. If it is GM204, it is a real replacement for 680/770 GK104 even if it ends up faster than 780Ti by 20-30%. There have been plenty of instances when NV's next generation mid-range either tied or beat the previous generation flagship. As a mid-range Maxwell card, with 2x performance/watt, it could easily beat a 780Ti in a 200W power envelope and sub-400mm2 die size, and be priced at $499-579.

This strategy would leave room for NV to release GM210 in 2H of 2015 for $699+.

It is a welcome rumor if NV won't continue to raise prices as it will put additional pressure on AMD as well. In the past, new generation of NV GPUs got 35-70% faster at more or less the $499-$649 level. Let's hope NV stays in this band, and closer to $499 would be even nicer for the 880.

I would say that the 880 being cheaper isn't an unbelievable rumor. People apparently have short memories of when AMD tried to charge a premium (eg, HD7970 launch at 550$ and the FX 9590 CPU) and then the market spoke that they weren't having it, they had to lower prices. The 7970 sat idle on shelves for months and it was very obvious especially after the 680 launched, and the 7970 launch software situation was just embarrassing and it took them 1-2 years to fix those software issues. And 6 months to release proper performance drivers. And people didn't want to pay 550$ for that? I was one of the stupid ones who did, but anyway. I know i'm not the only that remembers the HD 7970 launch price/software situation. That happened. Yet AMD gets praised up and down for being the bargain brand constantly. I would say they charge less for a reason, IMO mainly because of the lackluster software quality they had in prior years (although it has undoubtedly improved). And then on the flip side, the GTX 680 came in at a cheaper price than AMD's premium part. Bottom line is premium pricing has been done by both sides at various points and time, the difference is the market clearly told AMD they weren't having it. And NV has had a premium because they had the single GPU performance crown most of the time while AMD has had it very rarely.

So the performance deficit in the past years in conjunction with AMD's poor software (while it has improved in the past year) has led to a situation to where AMD was required to charge a lower price, which is what the market would bear. Of course, as a consumer I like better prices. So AMD has usually been close enough in performance to be a viable competitor in past years to drive nvidia prices,( excluding the 8800 ultra ) Which I can obviously appreciate. And like I said their software, while not where it needs to be, has improved. Anyway........the statement that AMD has always been the budget brand isn't true. They would charge a premium if they could, and they have tried in the past, and the market wouldn't let them do it. It is what it is. AMD is seen as the budget cheaper brand. AMD tried to charge premium prices in the past. And when they did, the market wouldn't let them do that.

Anyway, if the 880 is cheaper than the 780ti and faster....sounds good to me. That would kick up the competition on both sides for sure. I just want new toys , hardware has been getting a bit boring lately. I'd be most interested in seeing what both NV and AMD have in store for us because GPUs have just been getting long in the tooth. That's partially due to TSMC and the slowing of node transitions, so hopefully we'll get some exciting new tech late this year. It's certainly about time.

 

[tviceman]

Blackend, I agree with all of what you said except for gtx680's performance. It was only about 25-30% faster than gtx580.

 

[tg2708]

Its seems that people (enthusiast) just brush off 25% performance bump as not really much of an improvement (even though it still is). Can anyone tell me why as I have read similar statements before?

 

[sao123]

With Both major video card chip manufacturers basically also employing intels tic-toc strategy (which I very much hate for reasons I will state here) when is the appropriate time to buy into a product cycle?

To me it seems no matter when you buy there is always something better immediately around the corner...

When a new architecture debuts, it is usually a X% margin faster and better than the previous generation. At architecture introduction, there are generally 3-4 bins: High, Mainstream, and Budget, and likewise at the architecture refresh, likewise there are also 3-4 bins.

[I am specifically talking about the GPU itself...Granted I am greatly simplifying things, because I am excluding specialty products (dual GPU cards), overclocked varieties, and version B's with a higher amount of memory in a repeated bin.]

Lets assume you generally buy the flagship video gpu for gaming.

At the tic point in the cycle of the architecture, you have the early adopters, which comes with new product pains... If you buy at this point... there are initial problems that must always be resolved with the initial launch of a new generation of hardware.

or

You could wait for the toc part of the architecture, which means you now have a more mature product, and yet you know if you buy at this point, a new architecture is simply around the corner, with new performance gains.

Furthermore, once you have hit the toc cycle, you now realize that the flagship product released in the tic, was really a stripped down version, still sold at premium flagship pricing, and yet if you wait for the toc flagship, your mature product is technically outdated at this point.

Lets examine an example of what I am saying.
In 2012, Nvidia released the Kepler architecture with the GK104 chip, released in the GTX 680 flagship card.
GK104 had 8 Streaming Multiprocessors, 4 Graphic Processing Clusters. 1536 Pixel Shaders, 128 Texture Units, and 32 Render Pipes for a hefty price of $500+...

The full matured Kepler GK110 chip released In the GTX780 / GTX780ti flagship cards had 12/15 Streaming Multiprocessors, 5 Graphics Processing Clusters, 2688/2880 Pixel Shaders, 224/240 Texture Units, and 48 Render Pipes for a hefty price of $649/699...

BUT ITS THE SAME CHIP! Those processing units were always there, they are simply turned off...

This is why I hate the tic-toc cycle...I want to be able to buy the FULLY ENABLED gpu during the tic cycle... not the one with half the gpu disabled. It sucks because by the time the FUL ENABLED GPU I want is released, the architecture is nearing EOL and its time for a new stripped down version of the next architecture to be released.

I believe the toc cycle should be for smaller performance gains due to die shrinkage, increased vram, etc. But it severly annoys me that a stripped down version of the GPU is being sold as a flagship product during Generation 1.

I see this exact decision before me again... (not that I wont consider a R9 290X... I am... but likewise it has the SAME decision problem...) but when considering the GTX 780 TI... is it mature? or outdated? Then there is the gtx 880 to consider... GM 204... flagship new architecture or overpriced half disabled GPU?

So... I want to know...when do you believe is the best time to buy into (upgrade into) a new architecture?


EDIT: You can even see the chart detailing the comparison right here:
http://www.bit-tech.net/hardware/graphics/2013/11/07/nvidia-gtx-780-ti-3gb-review/1

 

[Wild Thing]

I would say that the 880 being cheaper isn't an unbelievable rumor. People apparently have short memories of when AMD tried to charge a premium (eg, HD7970 launch at 550$ and the FX 9590 CPU) and then the market spoke that they weren't having it, they had to lower prices. The 7970 sat idle on shelves for months and it was very obvious especially after the 680 launched, and the 7970 launch software situation was just embarrassing and it took them 1-2 years to fix those software issues. And 6 months to release proper performance drivers. And people didn't want to pay 550$ for that? I was one of the stupid ones who did, but anyway. I know i'm not the only that remembers the HD 7970 launch price/software situation. That happened. Yet AMD gets praised up and down for being the bargain brand constantly. I would say they charge less for a reason, IMO mainly because of the lackluster software quality they had in prior years (although it has undoubtedly improved). And then on the flip side, the GTX 680 came in at a cheaper price than AMD's premium part. Bottom line is premium pricing has been done by both sides at various points and time, the difference is the market clearly told AMD they weren't having it.

Anyway, if the 880 is cheaper than the 780ti and faster....sounds good to me. That would kick up the competition on both sides for sure. I just want new toys , hardware has been getting a bit boring lately. I'd be most interested in seeing what both NV and AMD have in store for us because GPUs have just been getting long in the tooth. That's partially due to TSMC and the slowing of node transitions, so hopefully we'll get some exciting new tech late this year. It's certainly about time.

I'd say the 295x2 being half the price(and faster) than TitanZ just shot holes in your "bargain brand" lament.

 

[shaynoa]

So Nvidia charge an extra $200.00 to $300.00 dollars more for there video cards just because they got the drivers right. yep that sounds fair doesn't it?

 

[rgallant]

so card maker A releases a card 3 months after card maker B for $50.00 less to $500.00 and that is a brilliant marketing move ,[could have used the $450 check box maybe]
but then releases the real big chips at 1k and $650.00 [later dropped to 500.00 something came along that was faster and cheaper] and that to is a brilliant marketing move and good for the customers.[maybe not those that paid 650.00]

then release same chip again [new names but not crippled much] at only $700.00 this time and that is a good deal for the customers.[maybe not the 1k past buyers]

-then release more of the same with different names at 1k +/-$200.00

-then releases 2 $700 chips on one board for 3k to 3.35k [with wc block ] and that is a deal.

not sure that when card maker A saved peeps $50.00 they should be given any praise looking back in history.

 

[ocre]

<snip>
Furthermore, once you have hit the toc cycle, you now realize that the flagship product released in the tic, was really a stripped down version, still sold at premium flagship pricing, and yet if you wait for the toc flagship, your mature product is technically outdated at this point.

Lets examine an example of what I am saying.
In 2012, Nvidia released the Kepler architecture with the GK104 chip, released in the GTX 680 flagship card.
GK104 had 8 Streaming Multiprocessors, 4 Graphic Processing Clusters. 1536 Pixel Shaders, 128 Texture Units, and 32 Render Pipes for a hefty price of $500+...

The full matured Kepler GK110 chip released In the GTX780 / GTX780ti flagship cards had 12/15 Streaming Multiprocessors, 5 Graphics Processing Clusters, 2688/2880 Pixel Shaders, 224/240 Texture Units, and 48 Render Pipes for a hefty price of $649/699...

BUT ITS THE SAME CHIP! Those processing units were always there, they are simply turned off...

This is why I hate the tic-toc cycle...I want to be able to buy the FULLY ENABLED gpu during the tic cycle... not the one with half the gpu disabled. It sucks because by the time the FUL ENABLED GPU I want is released, the architecture is nearing EOL and its time for a new stripped down version of the next architecture to be released.


I see this exact decision before me again... (not that I wont consider a R9 290X... I am... but likewise it has the SAME decision problem...) but when considering the GTX 780 TI... is it mature? or outdated? Then there is the gtx 880 to consider... GM 204... flagship new architecture or overpriced half disabled GPU?

I can kind of see what your talking about but you have some of your facts mixed up.

While it is true that Nvidia and AMD sell products with parts fused off, you do seem to have some things confused.

The gtx680 was a full GK104. It had no parts fused off. It is from the kepler design and it wasnt the largest kepler chip. Nvidia didnt make the 680 from cutting down a gk110. The gk104 was its own chip and differs slightly from the gk110.

So if we look at the gk110, Nvidia made many GPUs by fusing off various parts in one way or the other. Some examples are Titan, GTX 780, GTX 780ti, and titan black.

The gtx680 wasnt one of them. It came from a completely different chip in the kepler family: The GK104.
When compared to the GK110 it is not "THE SAME CHIP! Those processing units were always there, they are simply turned off..."

It has a slightly different arrangement than the gk110 and a full one has a maximum 1536 cuda cores built from an arrangement of 2 SMX per GPC (4 total) where the GK110 has 3 SMX per GPC (5 total). Kepler design has 192 cores per SMX so here is how the math breaks down.

gk104
4 GPC with a 2 SMX arrangement.
4 X 2 = 8
there is 192 cores per SMX so 8 x 192 = 1536

The gk110
5 GPC with 3 SMX per GPC
5 X 3 = 15
The same 192 cores per SMX so 15 X 192 = 2880

Both chips are from the same kepler family but the gk104 and the chips cut out of it are not cut down from the gk110. They are cut down from the gk104.

GPUs using the full GK104 are the gtx680 and gtx 770. They have no parts fused off. They are full chips. There are plenty that were cut down from a gk104, such as: GTX 670, GTX 760, GTX 660ti

The GK104 is much smaller than the GK110. Every chip made from fusing off parts will still be the same physical size if you measure it. All the GK104 variants measure the same, they all have 3.5billion transistors. Even if the fused off segments dont do anything anymore, they are still physically there. The Gk110 is massive in comparison. Every chip salvaged from the gk110 is still massive in size. With 7.1 billion transistors, it is almost double the size of a gk104.

 

[tviceman]

I'd say the 295x2 being half the price(and faster) than TitanZ just shot holes in your "bargain brand" lament.

295x only looks good because of TitanZ. Both cards are overpriced rip offs, Nvidia's more so but to defend 295X pricing for ANY reason is filled with as much fallacy as anyone who justifies buying TitanZ over cheaper and faster solutions.

 

[sao123]

I can kind of see what your talking about but you have some of your facts mixed up.

While it is true that Nvidia and AMD sell products with parts fused off, you do seem to have some things confused.

The gtx680 was a full GK104. It had no parts fused off. It is from the kepler design and it wasnt the largest kepler chip. Nvidia didnt make the 680 from cutting down a gk110. The gk104 was its own chip and differs slightly from the gk110.

So if we look at the gk110, Nvidia made many GPUs by fusing off various parts in one way or the other. Some examples are Titan, GTX 780, GTX 780ti, and titan black.

The gtx680 wasnt one of them. It came from a completely different chip in the kepler family: The GK104.
When compared to the GK110 it is not "THE SAME CHIP! Those processing units were always there, they are simply turned off..."

It has a slightly different arrangement than the gk110 and a full one has a maximum 1536 cuda cores built from an arrangement of 2 SMX per GPC (4 total) where the GK110 has 3 SMX per GPC (5 total). Kepler design has 192 cores per SMX so here is how the math breaks down.

gk104
4 GPC with a 2 SMX arrangement.
4 X 2 = 8
there is 192 cores per SMX so 8 x 192 = 1536

The gk110
5 GPC with 3 SMX per GPC
5 X 3 = 15
The same 192 cores per SMX so 15 X 192 = 2880

Both chips are from the same kepler family but the gk104 and the chips cut out of it are not cut down from the gk110. They are cut down from the gk104.

GPUs using the full GK104 are the gtx680 and gtx 770. They have no parts fused off. They are full chips. There are plenty that were cut down from a gk104, such as: GTX 670, GTX 760, GTX 660ti

The GK104 is much smaller than the GK110. Every chip made from fusing off parts will still be the same physical size if you measure it. All the GK104 variants measure the same, they all have 3.5billion transistors. Even if the fused off segments dont do anything anymore, they are still physically there. The Gk110 is massive in comparison. Every chip salvaged from the gk110 is still massive in size. With 7.1 billion transistors, it is almost double the size of a gk104.

I get what you are saying and I see my mistake... but I still wish they would go straight to the GM 210 variant in Gen 1, and then there is still room to improve it in Gen 2.

And... you still ignored the question.
When is the proper time to upgrade... Tic or Toc?

 

[blastingcap]

295x only looks good because of TitanZ. Both cards are overpriced rip offs, Nvidia's more so but to defend 295X pricing for ANY reason is filled with as much fallacy as anyone who justifies buying TitanZ over cheaper and faster solutions.

I agree with the above, with the exception of those who are limited to one slot for some reason. And even then, multi-GPU rigs are to be avoided unless absolutely necessary (3D/120Hz/multi-monitor/4K may need more than one GPU).

 

[ocre]

So Nvidia charge an extra $200.00 to $300.00 dollars more for there video cards just because they got the drivers right. yep that sounds fair doesn't it?

Fair?

haha

That is funny.
Buy what you want to buy. If you dont think it is "fair", than dont buy it.

There are many here that bash Nvidia all the time. Funny more is the fact that Nvidia is still very popular. If you personally have an issue with their pricing, there just so happens to be an alternative for you. Issue solved.

so card maker A releases a card 3 months after card maker B for $50.00 less to $500.00 and that is a brilliant marketing move ,[could have used the $450 check box maybe]
but then releases the real big chips at 1k and $650.00 [later dropped to 500.00 something came along that was faster and cheaper] and that to is a brillant marketing move and good for the customers.[maybe not those that paid 650.00]

than release same chip again [new names but not crippled much] at only $700.00 this time and that is a good deal for the customers.[maybe not the 1k past buyers]


-then release more of the same with different names at 1k +/-$200.00

-than releases 2 $700 chips on one board for 3k to 3.35k [with wc block ] and that is a deal.

not sure that when card maker A saved peeps $50.00 they should be given any praise looking back in history.

HUH?

Titan was way over priced in my book. I just wouldnt pay that much for a GPU. Not for my needs. I fully believe that but it doesnt make everything else go away.

I think the gtx680 was a great deal when it launched but the gtx670 as well. see

I think the GTX 760 also launched at a great price
-And-
I think Titan was priced way too high for me. No way would I pay that much for a card just to play a few games on. But that is me. Now if I was rich............I dont know....
Doesnt matter. The titan was sooo high I wouldnt even consider it.

Doesnt change the good buys they have had though.

 

[3DVagabond]

295x only looks good because of TitanZ. Both cards are overpriced rip offs, Nvidia's more so but to defend 295X pricing for ANY reason is filled with as much fallacy as anyone who justifies buying TitanZ over cheaper and faster solutions.

I agree that both cards are overpriced even in today's market. Nobody should buy either one, IMO. At 1/2 the price though, you can hardly call them comparable. Someone might justify the 295X price because it's "a halo product". It's the fastest and that's the price of admission to that club. It also improves over the 2 cheaper cards you could get in it's place. Titan-Z, on the other hand isn't either a halo product, it isn't the fastest, nor does it improve performance, noise and heat in this case, like the 295X2, and to add insult to injury... it's actually slower. The Titan-Z is just an insult and slap in the face to anyone who might like to own one. With the exception of maybe those who hand $100 bills to valet attendants for parking their Ferrari. Even those people though tend to like to be getting something extra for their money. Certainly not slower, louder, and hotter.

While the 295X2 is overpriced it's not some sort of bad joke played on their customers like the Titan-Z is.

 

[rgallant]

Fair?

haha

That is funny.
Buy what you want to buy. If you dont think it is "fair", than dont buy it.

There are many here that bash Nvidia all the time. Funny more is the fact that Nvidia is still very popular. If you personally have an issue with their pricing, there just so happens to be an alternative for you. Issue solved.



HUH?

Titan was way over priced in my book. I just wouldnt pay that much for a GPU. Not for my needs. I fully believe that but it doesnt make everything else go away.

I think the gtx680 was a great deal when it launched but the gtx670 as well. see

I think the GTX 760 also launched at a great price
-And-
I think Titan was priced way too high for me. No way would I pay that much for a card just to play a few games on. But that is me. Now if I was rich............I dont know....
Doesnt matter. The titan was sooo high I wouldnt even consider it.

Doesnt change the good buys they have had though.

just trying to say those who saved $50.00 buying a gtx680[as mentioned in a post] at going high end card prices [500.00] over the 550.00 card might have saved $50.00 in 2012 ,other buyers were ripped off by $150-$500.00 but did not want to start a fight.
add to the fact those that saved 50.00 are now finding that 2 gb cards might not last that long if they have 2 vs the 550.00 cards with 3gb of vram.

 

[Lepton87]

People remember that GTX680 was only cheaper then 7970 in the US and CANADA, elsewhere it was more expensive.

 

[3DVagabond]

just trying to say those who saved $50.00 buying a gtx680[as mentioned in a post] at going high end card prices [500.00] over the 550.00 card might have saved $50.00 in 2012 ,other buyers were ripped off by $150-$500.00 but did not want to start a fight.
add to the fact those that saved 50.00 are now finding that 2 gb cards might not last that long if they have 2 vs the 550.00 cards with 3gb of vram.

Yes, Tahiti is definitely worth ~10% more than GK104. AMD marketing failed miserably though and couldn't get that value across to the public. Not even to the tech press who failed to see Tahiti's advantages.

 

[Wild Thing]

295x only looks good because of TitanZ. Both cards are overpriced rip offs, Nvidia's more so but to defend 295X pricing for ANY reason is filled with as much fallacy as anyone who justifies buying TitanZ over cheaper and faster solutions.

Two full speed 290X chips will cost you over $1000.
Then add a liquid cooling system and a warranty on both.
You won't get out of that much under the price of a 295x2.

TitanZ on the other hand....:hmm:

 

[blastingcap]

Two full speed 290X chips will cost you over $1000.
Then add a liquid cooling system and a warranty on both.
You won't get out of that much under the price of a 295x2.

TitanZ on the other hand....:hmm:

Any cost-conscious person who doesn't need that much horsepower could get a pair of r9 290s instead of r9 290x's. Some even come with waterblocks already, else you can rig it up yourself. A pair of r9 290s with mild overclocks should be as fast as a stock r9 295x2 and cost either WAY less (air cooled) or still significantly less (watercooled r9 290s).

 

[3DVagabond]

Any cost-conscious person who doesn't need that much horsepower could get a pair of r9 290s instead of r9 290x's. Some even come with waterblocks already, else you can rig it up yourself. A pair of r9 290s with mild overclocks should be as fast as a stock r9 295x2 and cost either WAY less (air cooled) or still significantly less (watercooled r9 290s).

With the 290's, add AIO cooling and brackets you are at ~$1000, save ~$500. By the same token you could get a pair of 780's, over clock them (Would you even have to overclock them against Titan-Z?), and save yourself ~$2000. Still not even in the same universe as far as a rip off proposition. AMD is charging a premium for the 295X2. No doubt about it. It's a halo product, it's all in one and convenient, they are hitting you with a fairly hefty premium for that. 3X the price though? Nah! Keep in mind to that Titan-Z can't claim Halo. It's not even as good out of the box as 2X 780 ti. It's louder, hotter, and slower.

 

[SinOfLiberty]

Want to see too the effects of Stacked-vRAM on consumption of new-gen cards.

This will be only with Volta.

 

[SinOfLiberty]

OK, in that rumor of a rumor of a rumor of a rumor do they hint at a date? Or does anyone know the projected date, feel free to stomp on me if it was already posted..................

Expect Q3 release. Sometime in mid to late July/early August.