So, it's what I've been saying..

[Trackr]

When HD 7970 came out, I noticed the lack of extra horsepower and wrote about it.

It was the first time in history that the shader count (and thus performance) was not doubled.

Here's how it used to go:

X1800/X1900
320 ALU's (Doubled) -> HD 2900/3870
800 ALU's (Doubled) -> HD 4870
1600 ALU's (Doubled) -> HD 5870/6970
2048 ALU's (Increase of 25%) -> HD 7970

Now yes, because of the minute increase in shaders, the card overclocks like a champ, but it still doesn't manage more than 50%, unlike every other single top-end card.

So, OF COURSE nVidia looks at this and goes "Gosh, why have AMD done this? Oh well, guess there's no point in releasing our high-end card yet. We can just as easily rename our mid-range GPU and call it our high-end."

And so they did - both AMD and nVidia have given us mid-range (25-50% increase) GPUs under the high-end (100% increase) moniker.

This is as close to price-fixing we're going to see in this industry, so I hope ya'll are mad.

 

[gerardfraser]

I sold my 2 580GTX SLI 3GB cards(1 week ago) to buy these mid-range cards. Some people just do not care like me as long as the new cards are faster.

 

[Tsumi]

Yet another performance bashing thread...

Do we seriously need more of these?

 

[Trackr]

I sold my 2 580GTX SLI 3GB cards(1 week ago) to buy these mid-range cards. Some people just do not care like me as long as the new cards are faster.

Yet another performance bashing thread...

Do we seriously need more of these?

Alright. Well, I guess we'll be getting 25% increases every year when we used to get 100%, from now on.

 

[Tsumi]

Alright. Well, I guess we'll be getting 25% increases every year when we used to get 100%, from now on.

And what does complaining about it on a forum do besides fill up space? As long as nVidia and AMD are getting their money, and shareholders are happy, nothing will change.

 

[Trackr]

And what does complaining about it on a forum do besides fill up space? As long as nVidia and AMD are getting their money, and shareholders are happy, nothing will change.

This isn't "a forum". This is [H]. We're the enthusiasts. If we can't/don't critic the industry - who will?

Besides, all I truly want is for someone to agree with me.

 

[Hornet]

There were early rumor that nVidia had to do a bottom-up release of Kepler for whatever reason.

Now it seems nVidia can do a top-bottom. So perhaps there's some truth there, that due to yield issues or whatever, the top end Kepler is going to be delayed. But AMD being the first to release current gen cards, set the performance bar, and suddenly nVidia found themself with an opportunity to jump into the high end segment with the GK104.

The jump we saw with HD 7970 was disappointing for me as in the past we used to see something close to 2x performance with new generations (not including refresh). But it looks like that's what we'll be having for awhile now.

 

[Trackr]

There were early rumor that nVidia had to do a bottom-up release of Kepler for whatever reason.

Now it seems nVidia can do a top-bottom. So perhaps there's some truth there, that due to yield issues or whatever, the top end Kepler is going to be delayed. But AMD being the first to release current gen cards, set the performance bar, and suddenly nVidia found themself with an opportunity to jump into the high end segment with the GK104.

The jump we saw with HD 7970 was disappointing for me as in the past we used to see something close to 2x performance with new generations (not including refresh). But it looks like that's what we'll be having for awhile now.

Don't excuse them and say that it's because of yield issues.

It's at best a favorable situation for nVidia, created by AMD, which they took advantage of.

And at worst, it's AMD and nVidia agreeing behind the scenes to screw us consumers, complete with the story of nVidia "being disappointed" in AMD's performance and using it at a reason to gip us. It's like a false-flag attack.

 

[sirmonkey1985]

There were early rumor that nVidia had to do a bottom-up release of Kepler for whatever reason.

Now it seems nVidia can do a top-bottom. So perhaps there's some truth there, that due to yield issues or whatever, the top end Kepler is going to be delayed. But AMD being the first to release current gen cards, set the performance bar, and suddenly nVidia found themself with an opportunity to jump into the high end segment with the GK104.

The jump we saw with HD 7970 was disappointing for me as in the past we used to see something close to 2x performance with new generations (not including refresh). But it looks like that's what we'll be having for awhile now.

i don't think we will see 2x performance jumps any longer. we have pretty much gotten to the edge of the performance curve, now its about power efficiency. either way the 680's release had nothing to do with the 7970's lack of performance and far more to do with the fact that Nvidia failed to create a working version of the GK100/110 in a timely manor that would keep the share holders happy. the GK104 shift from mid range to high end had far more to do with saving their own asses then anything else and i had a pretty good idea we were going to see this happen when Nvidia was commenting about the 7970's performance compared to their own product which is something nvidia usually NEVER does before their product releases. even during the media craze of the 400 series they never commented on the 5k series nor attempted to compare their product to it.

will we ever see the GK100/110, probably but it'll probably be a 3rd or 4th revision of the chip and released as the 700 series instead.

as far as the original post goes, try a little harder next time.. not only do you fail to mention the shift from the VLIW5 to VLIW4 architecture you even fail to mention the change from VLIW4 to GCN which are two key factors in why the ALU's haven't doubled from the 5800 series to the 6900 series to the 7900 series.

yet to fail to even mention the fact that Nvidia was only able to MATCH the 7970's performance after more then doubling their cuda cores from 512 to 1536 while also increasing their GPU clocks from 772mhz to 1050mhz along with the shader clocks from 1544mhz to 2000mhz. don't you think that would of been something important to mention?

like i said, try a little harder next time. doing your home work before making a "lets bash AMD" thread will make you sound a like you actually know what you are talking about.

 

[firas]

the high end card used to have a 100% increase over the previous gen?

 

[sirmonkey1985]

the high end card used to have a 100% increase over the previous gen?

performance or ALU's? if you are referring to performance, no they are typically 40-50% between each new architecture. but as of the last 4-5 years its been about a 20% improvement with each new generation(e.g. 8/9 series to GTX 200 to 400 to 500/HD3800 to 4800 to 5800, etc)

 

[Trackr]

the high end card used to have a 100% increase over the previous gen?

Yes.

performance or ALU's? if you are referring to performance, no they are typically 40-50% between each new architecture. but as of the last 4-5 years its been about a 20% improvement with each new generation(e.g. 8/9 series to GTX 200 to 400 to 500/HD3800 to 4800 to 5800, etc)

Okay, I am going to stifle off this feeling of motion sickness..

Performance = ALU's

Thus, every generation is twice as powerful as the last.

HD 5870 = Twice as powerful as HD 4870
GTX 480 = Twice as powerful as GTX 280

Etc.

 

[Trackr]

yet to fail to even mention the fact that Nvidia was only able to MATCH the 7970's performance after more then doubling their cuda cores from 512 to 1536 while also increasing their GPU clocks from 772mhz to 1050mhz along with the shader clocks from 1544mhz to 2000mhz. don't you think that would of been something important to mention?

like i said, try a little harder next time. doing your home work before making a "lets bash AMD" thread will make you sound a like you actually know what you are talking about.

Oh, okay.. I thought you were being serious.

 

[Phishy714]

so uh.. do you really really not think that AMD is currently, and has been for a good while, working on the 8000 series card? What if they publically announced that they were working on it? Would you then say, damn, the 7900 series are crap they are ripping us off!

I can guarantee you that they already started work on the 9000 series cards too. Should we just forgo next generation because OBVIOUSLY the 9000 will be a higher end card which makes the 8000 a middle card.

Bad logic is bad.

 

[Mr. K6]

I got a 7970 that overclocked gives me double the performance of my previous 6950 @ 1GHz + another 1GB or vRAM and some other goodies. This is the biggest performance boost I've seen in over two years. Anyway, back to enjoying my games.

 

[sirmonkey1985]

Yes.





Okay, I am going to stifle off this feeling of motion sickness..

Performance = ALU's

Thus, every generation is twice as powerful as the last.

HD 5870 = Twice as powerful as HD 4870
GTX 480 = Twice as powerful as GTX 280

Etc.

um no performance is not equal to ALU's, if that was the case then the 6970 should of been slower than the 5870 since they went from 1600 to 1536. in the case of VLIW5 yes you are right performance was equal to the ALU count. but that idea disappeared when AMD decided to fix their broken VLIW5 architecture and moving to VLIW4 which improved its performance by taking out an ALU per shader core which in turn allowed them to push more bandwidth through the remaining 4 ALU's thus giving them the better performance with less ALU's.

no the GTX 480 was not twice as fast as the GTX 285(yes the 285 was the refresh replacement of the 280 and should be the card used to compare against the 480). now the GTX 580 which is what the 480 should of been from the beginning is but thats a whole different story and different generation which what do you know ISN'T 100% faster than the GTX 480.. oh snap! (generation =/= architecture)

case in point
http://www.anandtech.com/show/2977/...tx-470-6-months-late-was-it-worth-the-wait-/9
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/10
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/12 at least in this one it was close to 100% improvement..
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/13


in the end there is a heck of a lot more to do with performance gains then just adding more ALU's/cuda cores. but i'm done with the nit picking match, because i know we'll end up getting our posts deleted and i don't want that happening since we both do have good points that hopefully forces people to actually do their own research when they buy products instead of just going by straight performance numbers they see in a review.

 

[Druneau]

I'm with Mr. K6.

Went from 5870 to 7970. In the games I play I saw my average fps increase by 2.2x times.

 

[Trackr]

um no performance is not equal to ALU's, if that was the case then the 6970 should of been slower than the 5870 since they went from 1600 to 1536. in the case of VLIW5 yes you are right performance was equal to the ALU count. but that idea disappeared when AMD decided to fix their broken VLIW5 architecture and moving to VLIW4 which improved by taking out an ALU per shader core which in turn allowed them to push more bandwidth through the remaining 4 ALU's thus giving them the better performance with less ALU's.

no the GTX 480 was not twice as fast as the GTX 285(yes the 285 was the refresh replacement of the 280 and should be the card used to compare against the 480). now the GTX 580 which is what the 480 should of been from the beginning is but thats a whole different story and different generation which what do you know ISN'T 100% faster than the GTX 480.. oh snap! (generation =/= architecture)

case in point
http://www.anandtech.com/show/2977/...tx-470-6-months-late-was-it-worth-the-wait-/9
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/10
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/12 at least in this one it was close to 100% improvement..
http://www.anandtech.com/show/2977/...x-470-6-months-late-was-it-worth-the-wait-/13


in the end there is a heck of a lot more to do with performance gains then just adding more ALU's/cuda cores. but i'm done with the nit picking match, because i know we'll end up getting our posts deleted and i don't want that happening since we both do have good points that hopefully forces people to actually do their own research when they buy products instead of just going by straight performance numbers they see in a review.

This is how it goes.

You have Tick - Tock.

GTX 280 = Tick
GTX 285 = Tock

GTX 480 = Tick
GTX 580 = Tock

Is GTX 480 twice as powerful as GTX 280? Yes.
Is GTX 580 twice as powerful as GTX 285? Yes.

Done.

 

[Trackr]

I'm with Mr. K6.

Went from 5870 to 7970. In the games I play I saw my average fps increase by 2.2x times.

Well, let's see how that works out.

HD 5870 = 1600 V5 @ 850Mhz.
HD 7970 = 2048 V4 @ 950Mhz

Since V4 is 25% more efficient, it's equivalent to 2560 ALU's.

HD 5870 = 1,360,000.
HD 7970 = 2,432,000.

Yeah, I suppose it could get to 2x if you overclock it.

But my problem is, the HD 7970 loads at 55C. They could have given it 3,072 ALU's.

They chose not to, and to name it the HD 7970 so that the next iteration has to compete with nVidia's GTX 780 - and here in the conspiracy begins.

 

[sk3tch]

Are you changing your argument as you go here?

GTX 480 -> GTX 580 = ~20% improvement
GTX 580 -> GTX 680 = ~20% improvement

This release is not really that different than recent ones.

 

[Mr. K6]

Well, let's see how that works out.

HD 5870 = 1600 V5 @ 850Mhz.
HD 7970 = 2048 V4 @ 950Mhz

Since V4 is 25% more efficient, it's equivalent to 2560 ALU's.

HD 5870 = 1,360,000.
HD 7970 = 2,432,000.

Yeah, I suppose it could get to 2x if you overclock it.

But my problem is, the HD 7970 loads at 55C. They could have given it 3,072 ALU's.

They chose not to, and to name it the HD 7970 so that the next iteration has to compete with nVidia's GTX 780 - and here in the conspiracy begins.

You clearly have no idea what you're talking about and understand very little about GPU's. First, the 7970 is GCN, not V4. Secondly, load temperatures have nothing to do with how many shaders they can fit on a die.

 

[Digital Viper-X-]

You clearly have no idea what you're talking about and understand very little about GPU's. First, the 7970 is GCN, not V4. Secondly, load temperatures have nothing to do with how many shaders they can fit on a die.

I'm with this guy.....

 

[Trackr]

Are you changing your argument as you go here?

GTX 480 -> GTX 580 = ~20% improvement
GTX 580 -> GTX 680 = ~20% improvement

This release is not really that different than recent ones.

8800 GTX -> GTX 280 = 75%
GTX 280 -> GTX 285 = ~15%
GTX 285 -> GTX 480 = ~75%
GTX 480 -> GTX 580 = ~15%
GTX 580 -> GTX 680 = ~25%

Guess which one is odd man out.

You clearly have no idea what you're talking about and understand very little about GPU's. First, the 7970 is GCN, not V4. Secondly, load temperatures have nothing to do with how many shaders they can fit on a die.

Oh wow, Ad Hominem. I'm so offended.

GCN isn't any bit more efficient than V4, so say the benchmarks. It's designed to have other advantages.

Second, load temperatures have EVERYTHING to do with ALU/SP amount.

8800 GTX - 90nm - 128 SPs
GTX 280 - 65nm - 240 SPs
GTX 480 - 45nm - 480 SPs

They are all basically equivalent in power consumption and heat output.

I'm with this guy.....

You mean because of his confident, assured tone?

Hopefully you don't make important life decisions thusly.

 

[sirmonkey1985]

Well, let's see how that works out.

HD 5870 = 1600 V5 @ 850Mhz.
HD 7970 = 2048 V4 @ 950Mhz

Since V4 is 25% more efficient, it's equivalent to 2560 ALU's.

HD 5870 = 1,360,000.
HD 7970 = 2,432,000.

Yeah, I suppose it could get to 2x if you overclock it.

But my problem is, the HD 7970 loads at 55C. They could have given it 3,072 ALU's.

They chose not to, and to name it the HD 7970 so that the next iteration has to compete with nVidia's GTX 780 - and here in the conspiracy begins.

its three fold when you add shader cores. not only do you create heat you also increase the power usage. along with that you also take up more space on the gpu its self. so its balancing act of do you add more shader cores or do you use less but increase the clocks to offset it. the biggest down side to AMD's shader architecture is that they still for some retarded reason use a 1:1 locked ratio between the gpu and shaders, why they still do this we'll probably never know since its quite obvious the shaders can handle the higher clocks. if they were to switch to a 1:1.5 or even 1:2 ratio i bet we would actually see that 100% improvement we have all been waiting to see between gpu generations for the last 5 years.

i don't know why you keep thinking this is all a conspiracy.. we all know the 7k series is a stepping stone into the GCN + 28nm architecture, the true successor will be the 8k series. its going to be the same way for Nvidia as well. both AMD and Nvidia will learn a lot with this generation and learn what didn't work and what needs to be improved for the next version. the move from the old architecture to the GCN architecture is huge for AMD. kepler as also a big step for Nvidia since they have spent the last 4 years pushing the heck out of GPGPU where as with kepler its taking a back seat. we all knew the biggest problem with fermi was the fact that Nvidia tried to make GPGPU fermi's biggest selling point, which is great in the enterprise market but useless in the consumer market given that applications are still years behind from actually using its full potential.

heck if anything i find it quite interesting watching Nvidia switching to what AMD did with the last 3 generations where as now AMD is doing what Nvidia has been doing for years. AMD went brute force and Nvidia went with efficiency. i was hoping we would finally see this day again where nvidia didn't release some giant power hog gpu. its been a long road since the G92 which in my opinion is still the best GPU Nvidia's ever release to this point.

btw you can't just say well this architecture is 25% more efficient then that architecture so its the equivalent to this number of ALU's. GCN was built from the ground up. it shares some similarities to northern island but other then those small similarities they are so different from each other you can't compare them in that way. you can on the other hand say well VLIW4 is 25% more efficient then VLIW4 but that would only be comparing the shader architecture, not the gpu as a whole.

well scratch that i was wrong on my original statement of the 2Ghz cuda core clocks(reading [H]'s review now) i guess they followed AMD and switched to a 1:1 ratio.

 

[astroboy]

It is true we should expect a larger perf improvement.
Die size tells the story.

GTX680 = 295mm2
GTX580 = 529mm2

Considering the nice efficiency improvements, A 28nm ~500mm2 Kepler should actually be more than 2x 580 perf

But then again.. I don't think NV could price it at $499, given the inevitably crappy yields that comes with a new process.
Seems like both NV and AMD are just being smarter about not coming into a new process with giant chips and destroying their margins.

 

[teletran8]

From benches looks like the 680GTX is twice as powerful as a 560Ti I enjoy this thread

 

[Trackr]

It is true we should expect a larger perf improvement.
Die size tells the story.

GTX680 = 295mm2
GTX580 = 529mm2

Considering the nice efficiency improvements, A 28nm ~500mm2 Kepler should actually be more than 2x 580 perf

But then again.. I don't think NV could price it at $499, given the inevitably crappy yields that comes with a new process.
Seems like both NV and AMD are just being smarter about not coming into a new process with giant chips and destroying their margins.

Exactly!

Sometimes I forget how simple things are because no one appears sentient enough to understand and/or they love flame wars.

The entire problem is that they're pricing the GTX 660 at 500$ after calling it GTX 680.

We need to stop them!

 

[teletran8]

Exactly!

Sometimes I forget how simple things are because no one appears sentient enough to understand and/or they love flame wars.

The entire problem is that they're pricing the GTX 660 at 500$ after calling it GTX 680.

We need to stop them!

they're laughing all the way to the bank

 

[sirmonkey1985]

Exactly!

Sometimes I forget how simple things are because no one appears sentient enough to understand and/or they love flame wars.

The entire problem is that they're pricing the GTX 660 at 500$ after calling it GTX 680.

We need to stop them!

why should we stop them for finally making a power efficient high performing flagship card..

and i agree with teletran's post above mine and its taking me a lot of effort to say this..

 

[teletran8]

and i agree with teletran's post above mine and its taking me a lot of effort to say this..

And that was b4 I read the [H] article still on last 2 pages

 

[Cr4ckm0nk3y]

When HD 7970 came out, I noticed the lack of extra horsepower and wrote about it.

It was the first time in history that the shader count (and thus performance) was not doubled.

Here's how it used to go:

X1800/X1900
320 ALU's (Doubled) -> HD 2900/3870
800 ALU's (Doubled) -> HD 4870
1600 ALU's (Doubled) -> HD 5870/6970
2048 ALU's (Increase of 25%) -> HD 7970

Now yes, because of the minute increase in shaders, the card overclocks like a champ, but it still doesn't manage more than 50%, unlike every other single top-end card.

So, OF COURSE nVidia looks at this and goes "Gosh, why have AMD done this? Oh well, guess there's no point in releasing our high-end card yet. We can just as easily rename our mid-range GPU and call it our high-end."

And so they did - both AMD and nVidia have given us mid-range (25-50% increase) GPUs under the high-end (100% increase) moniker.

This is as close to price-fixing we're going to see in this industry, so I hope ya'll are mad.

Thank God for Trackr and his exposing this conspiracy. /sarcasm

So let me understand this, you have become accustomed to large improvements with GPUs and one generation didn't met your expectations so it is a conspiracy? Since the cards are priced similarly it is also a price matching conspiracy?

Are you dillusional or just entitled? Not every new gen of any device will yield huge performance increases. That is reality not some conspiracy or purposeful action of AMD or NV. Do you actually believe AMD engineered the cards on purpose to have a smaller than expected performance increase? How would that benefit AMD or NV? "

"Hey the cards can do X but let's slow it down to Y just so the new tech looks bad."

Makes complete sense.....

Maybe the lack of performance increase has all to do with moving to a new die size and dealing with newer tech.

Also it is not price fixing. There has been nothing but rumors that the 680 was originally a mid-range card and now NV is making it a high end to get more money. Even if that was true it still does not amount to price fixing. That is competitive pricing.

Price fixing would require both AMD and NV to meet and make an agreement at what price their products will sell for. Not what the market allows but what they want. Price fixing would be the high end cards suddenly selling for $700 with no market reason.

There is no conspiracy just a lower than expected result from a new gen, if you don't like it don't buy it. There is absolutely no proof the next set of cards won't give you huge improvements. If this was the 3rd or 4th gen in a row with low increases your argument might have some merit but that is not the case.

Skip a gen and stop complaining. Also you might want to cut down on the Jesse Ventura conspiracy shows.

 

[teletran8]

Skip a gen and stop complaining. Also you might want to cut down on the Jesse Ventura conspiracy shows.

LOL! But the thing is if your coming from a 580 GTX to a 680GTX the performance increase is around 25 percent. You really have to be coming from a 560Ti or 6870 to be overall impressed with the 680GTX, and since I'm on a 560Ti it looks just O.K. IMHO. (Well when they put out the TF 4GB model for 500-550 my interest will be higher. The 680GTX is really the 660Ti in wolves clothing (500+ dollars not 250+)

If you have a 580, 7870, 7950, 7970 and are not uber power consumption mental then it doesn't justifie the 680GTX imho. But I'm glad there is a 680GTX because I have a 560Ti it's a slightly better 580GTX @ the same cost for me!

 

[sirmonkey1985]

LOL! But the thing is if your coming from a 580 GTX to a 680GTX the performance increase is around 25 percent. You really have to be coming from a 560Ti or 6870 to be overall impressed with the 680GTX, and since I'm on a 560Ti it looks just O.K. IMHO. (Well when they put out the TF 4GB model for 500-550 my interest will be higher. The 680GTX is really the 660Ti in wolves clothing (500+ dollars not 250+)

If you have a 580, 7870, 7950, 7970 and are not uber power consumption mental then it doesn't justifie the 680GTX imho. But I'm glad there is a 680GTX because I have a 560Ti it's a slightly better 580GTX @ the same cost for me!

i don't know.. 25% performance using roughly 60-80w less at full load and ~20w less at idle makes its a very compelling upgrade from the 580 in my opinion. doesn't sound like a lot but if you are like me and have your system running 24/7 it makes a pretty big difference.

 

[Trackr]

So let me understand this, you have become accustomed to large improvements with GPUs and one generation didn't met your expectations so it is a conspiracy? Since the cards are priced similarly it is also a price matching conspiracy?

Yes, think about it - what if from now on every generation is like this - GTX 980 will be barely twice as powerful a GTX 580.

It's a very dangerous precedent.

Are you dillusional or just entitled?

Here, let me show you where you are:

http://24.media.tumblr.com/tumblr_l6b9heXpZC1qb2qxro1_500.jpg

Not every new gen of any device will yield huge performance increases.

Up until now, they have.

"Hey the cards can do X but let's slow it down to Y just so the new tech looks bad."

Makes complete sense.....

Let me fix that for you.

"Hey, we can create a 3,072 ALU core on 28nm.. but let's sell them a 2,048 ALU one for full price first, then the 3,072 ALU one later, and make twice as much money!"

Then, "Hey nVidia, we have a plan to make twice the fiscal revenue, you interested?"

Maybe the lack of performance increase has all to do with moving to a new die size and dealing with newer tech.

But they move a die every 'Tick' generation. Are you retarded?

Also it is not price fixing. There has been nothing but rumors that the 680 was originally a mid-range card and now NV is making it a high end to get more money. Even if that was true it still does not amount to price fixing. That is competitive pricing.

Only if AMD and nVidia did not orchestrate it behind the scenes.

In that case, it's just sloppy.

Price fixing would require both AMD and NV to meet and make an agreement at what price their products will sell for. Not what the market allows but what they want. Price fixing would be the high end cards suddenly selling for $700 with no market reason.

Economics 101:

If you heighten prices, you lessen demand.

But, if you offer a product half as good a precursor to the actual product, both for an averagely high price you'll make far more.

There is no conspiracy just a lower than expected result from a new gen, if you don't like it don't buy it. There is absolutely no proof the next set of cards won't give you huge improvements. If this was the 3rd or 4th gen in a row with low increases your argument might have some merit but that is not the case.

Lower than expected? What, you think they're throwing a rock into a stream and seeing how many ALUs fit?

nVidia and AMD spend billions annually. They know EXACTLY what's going to happen and they control it all.

Skip a gen and stop complaining. Also you might want to cut down on the Jesse Ventura conspiracy shows.

Did you see the pyramid above?

 

[sirmonkey1985]

ok seriously quit throwing around the ALU crap.. you keep using it like you actually know what it is, and yet you have proven over and over you have absolutely no bloody clue what ALU's are or do. you continue to act like you know how businesses and markets work yet you don't. i mean ffs i didn't even graduate high school and i know more about marketing and business models then you have shown. why do you keep trying to push this bogus conspiracy crap, there is no conspiracy ffs. this is how how the business world works, don't like it then to bad.

ok i need to stop posting in this thread before i get in trouble, have fun everyone.

 

[Mr. K6]

GCN isn't any bit more efficient than V4, so say the benchmarks. It's designed to have other advantages.

Actually it's a good deal more efficient than V4, in the order of 15-20% in performance alone.

Second, load temperatures have EVERYTHING to do with ALU/SP amount.

8800 GTX - 90nm - 128 SPs
GTX 280 - 65nm - 240 SPs
GTX 480 - 45nm - 480 SPs

They are all basically equivalent in power consumption and heat output.

Temperatures reflect the efficiency of the cooling solution, nothing more. You really should read up on basic thermodynamics. Also, those cards don't have anywhere near the same power consumption and therefore heat output. The 8800GTX was ~155W, the GTX 280 was ~235W, and the GTX 480 was ~300W.

Honestly, if you're going to post tinfoil hat nonsense, it would help to be the least bit informed.

 

[Trackr]

Actually it's a good deal more efficient than V4, in the order of 15-20% in performance alone.

Temperatures reflect the efficiency of the cooling solution, nothing more. You really should read up on basic thermodynamics. Also, those cards don't have anywhere near the same power consumption and therefore heat output. The 8800GTX was ~155W, the GTX 280 was ~235W, and the GTX 480 was ~300W.

Honestly, if you're going to post tinfoil hat nonsense, it would help to be the least bit informed.

I'll check the tidbit about GCN vs V4 again.

Temperatures reflect the efficiency of the cooling solution, nothing more..?

Are you serious?

So, a GTX 460 with the same cooler as a GTX 480 would consume the same amount of power and dissipate the same amount of heat?

No?

Then, isn't it because the GTX 480 has 480 SPs while the GTX 460 has 336?

You want basic thermodynamics?

More SP's/ALU's = More transistors = More heat

 

[Mr. K6]

I'll check the tidbit about GCN vs V4 again.

Temperatures reflect the efficiency of the cooling solution, nothing more..?

Are you serious?

So, a GTX 460 with the same cooler as a GTX 480 would consume the same amount of power and dissipate the same amount of heat?

No?

Then, isn't it because the GTX 480 has 480 SPs while the GTX 460 has 336?

It seems to me that you have absolutely no concept of thermodynamics, so I'm going to start at a very basic level. Don't take offense to that, but you literally can't argue points against something you clearly don't understand. To begin, integrated circuits (IC) like GPU's can be thought of as large resistors. They consume power to function, and in this case almost all of it is turned into heat. Heat is a quantity, and the temperature reflects how much of that heat stays in the chip before it is whisked away by the cooling solution.

To provide an example of how your point is completely wrong, the GTX 460 has 336SP's and consumes ~160W on use, but the GTX 280, with only 240SP's, consumes ~235W. As you can see, there's no correlation, simply because they're different architectures on different processes. The GTX 460 consumes less power than a GTX 480 because it is a much smaller chip (332mm^2 vs. 529mm^2 of GF100) that not only has fewer SP's, but also a lot of the compute features cut out as well.

 

[Eisenblut]

Yet another performance bashing thread...

Do we seriously need more of these?

 

[Trackr]

To provide an example of how your point is completely wrong, the GTX 460 has 336SP's and consumes ~160W on use, but the GTX 280, with only 240SP's, consumes ~235W. As you can see, there's no correlation, simply because they're different architectures on different processes.

8800 GTX - 128 SP's @ 90nm = 490mm^2 x (575Mhz) = 225w (250w Peak)
GTX 460 - 336 SP's @ 45nm = 332mm^2 x (675Mhz) = ~160w (185w Peak)
GTX 280 - 240 SP's @ 65nm = 576mm^2 x (600Mhz) = ~275w (300w Peak)
GTX 480 - 480 SP's @ 45nm = 529mm^2 x (700Mhz) = ~275w (300w Peak)
GTX 680 - 1536 SP's @ 28nm = 294mm^2 x (1006Mhz) = 195w (215w Peak)

So again, it's - Amount of SPs / Manufacture Process = Die Size x Clock Speed = Heat Output.