Everyone calm down : It is amazing how short memories are

[D.P.]

You know the chart they showed with the 1080 ahead of the titan x by 22% was in terms of relative VR performance. It gains its lead mostly from Single pass rendering and Multi projection in VR.

22% was without VR

 

[D.P.]

This and this was presented by nVidia which ofc are going to try and show their card in the best light. Lets remember back when AMD had slides showing the Fury x beating the 980Ti however reviewers ended up painting a complete different picture.

Oh, I agree entirely that the real world numbers wont be nearly as good as what NVidia have used as marketing material, merely Nvidia marketing material reinforces all the other evidence that the 1080 is at least as good as any previous generally jump on to new node size. One only has to look at chip size, clock speeds, CUDA core counts, conservative architecture improvements to realize what the 1080 can do.

 

[Marine-RX179]

This x1000.

You can tell a lot of people are very angry that the 1080 is going to be as good as it is after months of ridiculing Nvidia.

Who is angry at 1080 could be "too fast"?

I mean considering the launch price of the 1080 has gone from $550 to $699 comparing to 980, for the consumers' sake I honest hope the relative performance increase would be greater than 780Ti to 980.

 

[Jono8]

I honest hope the relative performance increase would be greater than 780Ti to 980.

It will be. You only have to look at the specs to ascertain this

 

[Marine-RX179]

It will be. You only have to look at the specs to ascertain this

lol of course you do...just look at how the Nvidia rushed to launch the 980Ti at a more moderate price because of how well the spec of the Fury X looked before launch

Clearly people on the forum are better expert at guessing performance base on spec than the people that work at Nvidia

 

[Jono8]

lol of course you do...just look at how the Nvidia rushed to launch the 980Ti at a more moderate price because of how well the spec of the Fury X looked before launch

Clearly people on the forum are better expert at guessing performance base on spec than the people that work at Nvidia

? We are not comparing AMD to Nvidia. We are comparing Nvidia to Nvidia. Unless you think Nvidia have somehow REDUCED the individual performance of each cuda core

It is a 2560 cuda core card clocked at 1733mhz at stock, of course it is going to come close/match 980 sli and beat a 980ti/Titan X. Even if Nvidia have made no IPC gains, it would still be faster than a 980Ti

 

[Marine-RX179]

? We are not comparing AMD to Nvidia. We are comparing Nvidia to Nvidia. Unless you think Nvidia have somehow REDUCED the individual performance of each cuda core

It is a 2560 cuda core card clocked at 1733mhz at stock, of course it is going to come close/match 980 sli and beat a 980ti/Titan X. Even if Nvidia have made no IPC gains, it would still be faster than a 980Ti

But it is not the on the same architecture nor on the same process. You have to look at the example of the previous die-shrink: have a look back on the GTX580 with 512 cuda core on 40nm vs GTX680 with 1536 cuda core on 28nm- individual performance of each cuda core reduced was exactly what happened...otherwise the GTX680 would had been like 2-3 times the performance of the GTX580.

The new 16nm/14nm process is unknown water for all of us, and we cannot use conventional approach for comparing 1080 to current gen cards like we did for between 40nm cards.

 

[D.P.]

lol of course you do...just look at how the Nvidia rushed to launch the 980Ti

What, There was loads of stock on release, wasn't a rush job at all. Nvidia were ready to release ahead of AMD, had the cards in stock and no point waiting around when they could be selling cards. They had planned the release date well in advance, having carefully reduced the supply channel foe older cards.

 

[Marine-RX179]

What, There was loads of stock on release, wasn't a rush job at all. Nvidia were ready to release ahead of AMD, had the cards in stock and no point waiting around when they could be selling cards. They had planned the release date well in advance, having carefully reduced the supply channel foe older cards.

I didn't say Nvidia didn't have the 980Ti ready, I meant Nvidia was sitting and waiting to find out as much as they could about Fury to decide if they should make the move first, or to wait for AMD to launch the Fury first and to counter with a higher stock clocked for the 980Ti to better ensuring them or increase their odds of claiming the crown on the benchmarks.

 

[Jono8]

But it is not the on the same architecture nor on the same process. You have to look at the example of the previous die-shrink: have a look back on the GTX580 with 512 cuda core on 40nm vs GTX680 with 1536 cuda core on 28nm- individual performance of each cuda core reduced was exactly what happened...otherwise the GTX680 would had been like 2-3 times the performance of the GTX580.

The new 16nm/14nm process is unknown water for all of us, and we cannot use conventional approach for comparing 1080 to current gen cards like we did for between 40nm cards.

But that was no doubt mainly due to all the compute being torn out for the 600 series.

I do love this place, you simply can't win.

One moment everyone is telling you "Pascal is just Maxwell on a die shrink, don't expect much" then when you point out that spec wise the performance Nvidia have claimed checks out, you get "you don't know because Pascal is clearly a completely different architecture to anything before it"

Round and round the magic roundabout we go.

From the specs, all the leaked benchmarks and actually what NVidia have told publicly to millions of people, I have very little doubt that the 1080 will be around 980 sli performance (so around 1.7x the performance of a 980)

 

[Gregster]

I didn't say Nvidia didn't have the 980Ti ready, I meant Nvidia was sitting and waiting to find out as much as they could about Fury to decide if they should make the move first, or to wait for AMD to launch the Fury first and to counter with a higher stock clocked for the 980Ti to better ensuring them or increase their odds of claiming the crown on the benchmarks.

NVidia had AMD by the danglies when they launched the Fiji series. They basically **** all over AMD with the 980Ti and price cuts of the 980/70. AMD were in a rock and a hard place with sky high prices and no stock.

 

[Marine-RX179]

But that was no doubt mainly due to all the compute being torn out for the 600 series.

I do love this place, you simply can't win.

One moment everyone is telling you "Pascal is just Maxwell on a die shrink, don't expect much" then when you point out that spec wise the performance Nvidia have claimed checks out, you get "you don't know because Pascal is clearly a completely different architecture to anything before it"

Round and round the magic roundabout we go.

From the specs, all the leaked benchmarks and actually what NVidia have told publicly to millions of people, I have very little doubt that the 1080 will be around 980 sli performance (so around 1.7x the performance of a 980)

Sorry, I don't know why I should be held accountable for opinions of other people, but I was merely addressing what you stated on "Unless you think Nvidia have somehow REDUCED the individual performance of each cuda core", and then I was scratching my head and the pointed out "Yes it is happened in before the pass (via example)".

 

[Marine-RX179]

NVidia had AMD by the danglies when they launched the Fiji series. They basically **** all over AMD with the 980Ti and price cuts of the 980/70. AMD were in a rock and a hard place with sky high prices and no stock.

You are missing my point...there's little doubt that Nvidia can release a card that can beat the the Fury card, but Nvidia would have want to make sure the card that would beat the AMD's offering would offer as little extra performance as possible, so that they don't need to try as hard or cost as much to bring a card the following gen that will look to be faster by enough margin.

They could have either:
a) wait for the launch of the Fury X, and launch their 980Ti afterward with whatever higher stock clock necessary
b) launch the 980Ti first, and if Fury X beat it, they will just release another card with higher core counts (like that did with 780Ti), and call it a different name like 980Z or 980X to reclaim the crown...

Option a) is probably more ideal for them as less work is involve and less hassle, but as they waiting for longer and closer to the Fury X launch, they probably thought it would be better to go with option b).

Like back on the gen with when AMD released the 7970/7950, the reason why Nvidia's 670/680 didn't come till few months later was only because they saw the 7970/7950 was not a threat at all, so they held back their big chip, and bought the release of what should had been the 660/660Ti forward, and branded it as 670/680 to complete with the 7950/7970, and then the Titan branding for the big chip early adopter was born. For Nvidia, it is all about giving just the right amount of performance- not too much, but just enough to beat the competition would do; if they carelessly released a card that's too much faster than the competition in the gen before, then it would mean they would have to push the performance new gen card to higher performance to maintain a certain % increase in performance.

*Stare up to the 2nd half of the title of this topic- "Yep it is quite right".

 

[Gregster]

I think I mostly understand what you are saying and yer, NVidia's 680 was a mid chip, the same as the 1080 is a mid chip but they were the fastest cards at their respective times (assuming the 1080 is indeed faster than a TX) and if NVidia are guilty of giving "just enough of an increase", AMD are guilty of the same. You have the same rebrands from both camps (GTX 770 = GTX 680 - 390X = 290X),

Actually looking at it the same way you are, the 680 and 7970 were low end GPUs, as the same node size (28nm) has produced the 680/Titan/Titan black/Titan X/7970/290X/390X/Fury X) so yes, both camps are indeed guilty of feeding just enough performance.

As for NVidia holding back their big chip, I lol

 

[Dave2150]

NVidia had AMD by the danglies when they launched the Fiji series. They basically **** all over AMD with the 980Ti and price cuts of the 980/70. AMD were in a rock and a hard place with sky high prices and no stock.

I don't think much changed, even after fury stock become common place.

The Fury owners thread is as barren as AMD itself.

 

[TwsT]

This is why I'm waiting for reviews, I never believe in what AMD and Nvidia claims.

Everyone is just impatient for a new card though .

Indeed AMD sponsored info graphics are equally as ridiculous as nvidias and lets not forget that the fury x was a great overclocker or 970's memory and rops issues


Both companies have a job to do expecting them not to do it is kinda retarded but that doesn't mean you have to get wrapped up in all the hype.

 

[melmac]

Like back on the gen with when AMD released the 7970/7950, the reason why Nvidia's 670/680 didn't come till few months later was only because they saw the 7970/7950 was not a threat at all, so they held back their big chip, and bought the release of what should had been the 660/660Ti forward, and branded it as 670/680 to complete with the 7950/7970, and then the Titan branding for the big chip early adopter was born. For Nvidia, it is all about giving just the right amount of performance- not too much, but just enough to beat the competition would do; if they carelessly released a card that's too much faster than the competition in the gen before, then it would mean they would have to push the performance new gen card to higher performance to maintain a certain % increase in performance.

Nvidia didn't hold back their back chip. They had problems with the GK100, not sure what, but they were due to fill a contract for the Oakridge supercomputer in November 2011. They didn't manage to fill that contract until September 2012 with the GK110. There is no way Nvidia would have held the big card back just because of the assumed performance of the 7970.

You are also wrong about the 7970, it had way more potential than the 680. If AMD had been a bit more aggressive with their pricing and clock speeds, basically, if they released the Ghz edition from the start, Nvidia would have been very hard put to match it.

As for NVidia holding back their big chip, I lol

Yeah, me too, I can't believe people still think this is true.

 

[Marine-RX179]

Yeah, me too, I can't believe people still think this is true.

Oh but they DID hold back the big chip, even if not due to that, but they did it in the form of "holding back the release time from the mainstream" by launching the Titan line, to further delay the launch of fat 80 card

At least Titan and Titan Black had the excuse of for compute purpose, but the Titan X is just glorified double-memory 980Ti.

 

[Gregster]

Marine, you are wrong on all parts!

 

[Marine-RX179]

Marine, you are wrong on all parts!

OK then if I was wrong on "all parts", please do explain why we now have the same pattern of always having to wait for the release of Titan card first, before we can get hold of the fat 80/80Ti card for the pass few gen? (and it's is quite scary how some people have now accepted it as the norm).