Nvidia Pascal Announcement

Meanwhile over at Anandtech, The Fury X would only be 66% faster...

Yeah, let's nitpick the benches that show what we want, and disregard everything else...

If we just take it as "cheaper and faster" than minspec VR as he says, he is very likely to be talking about the cut-down Polaris 10 as historically the full midrange GPU would cost more than the 970/390's current prices.

Now let's interpret their words in a way that favors our argumenting line...

280X launch price was $300.
285 launch price was %250.
380X launch price was $230.

I don't think I can remember any instance in which mid-range cards costed more than $350.

And market reality now is quite different, than when the 280X launched too. AMD needs to recover a lot of marketshare, for which more competitive prices are a given.

I guess really it comes down to if you believe small Polaris can beat the 960 by 10% or more. If not, for me AMD might as well pack it in and go home.

Then they might as well start packing it in an go home, they way it's looking... Or not, because I don't see how the future of AMD rests on a comparison of their lowest-end (100-120 mm2-ish?) chip vs Nvidia's midrange (240 mm2?) chip... Have low-end chips always bested, past midrange chips?? I have the impresion that's not been the case in 99% of cases, but admitedly that's not usually the chips I pay more attention to.
 
Yeah, let's nitpick the benches that show what we want, and disregard everything else...

What "everything else"? You were the one providing one data point based on year-old review, I just linked the most recent TPU and Anandtech benches for both cards.

Now let's interpret their words in a way that favors our argumenting line...

280X launch price was $300.
285 launch price was %250.
380X launch price was $230.

I don't think I can remember any instance in which mid-range cards costed more than $350.

The 280X was rebranded 28nm silicon based on GPUs that cost a lot more a year earlier, that's why it was cheaper. Pitcairn launched at the start of a new process at $350 and $250 respectively.

Or not, because I don't see how the future of AMD rests on a comparison of their lowest-end (100-120 mm2-ish?) chip vs Nvidia's midrange (240 mm2?) chip... Have low-end chips always bested, past midrange chips?? I have the impresion that's not been the case in 99% of cases, but admitedly that's not usually the chips I pay more attention to.

I can't remember the last time when the new mid-range didn't at least match the previous high-end after a node transition.

At 28nm the 7870 was faster than the 6970 and at 40nm the 5770 was roughly equal to 4870.

From what I can remember, Nvidia was the same between 40 and 28 with the 680 blowing away the 580...so much so you got to pay the full price for it. The 460 was roughly equal to the GTX 285.

This time around there are obvious differences, for example both companies really pushed the 28nm node to its limit so it should be more difficult.

Do you expect Nvidia to beat the Titan X with their fastest GP104 card?
 
Do you expect Nvidia to beat the Titan X with their fastest GP104 card?

It's not meant to either way. Other than that since leaks are starting to surface, would you want to place any bets what the full SKU really can replace?
 
What "everything else"? You were the one providing one data point based on year-old review, I just linked the most recent TPU and Anandtech benches for both cards.



The 280X was rebranded 28nm silicon based on GPUs that cost a lot more a year earlier, that's why it was cheaper. Pitcairn launched at the start of a new process at $350 and $250 respectively.



I can't remember the last time when the new mid-range didn't at least match the previous high-end after a node transition.

At 28nm the 7870 was faster than the 6970 and at 40nm the 5770 was roughly equal to 4870.

From what I can remember, Nvidia was the same between 40 and 28 with the 680 blowing away the 580...so much so you got to pay the full price for it. The 460 was roughly equal to the GTX 285.

This time around there are obvious differences, for example both companies really pushed the 28nm node to its limit so it should be more difficult.

Do you expect Nvidia to beat the Titan X with their fastest GP104 card?

The problem is that you're taking examples from when there were only 3 tiers of GPUs, and trying to extrapolate it to todays AMD 4 chip strategy.

7870 was a 212 mm2, beating a 389mm2 chip.
5770 170mm2 vs 4870 256mm2.

Now just because in your head "It MUST be so!!" you're pretending that a 232mm2 chip should be equal to a 600mm2 chip, with no evidence to prove such a ludicrous claim. And I'm telling you, you must be on some hard drugs, mate. Just like in the past (see the examples for reference). the 232mm2 chip is likely to be equal to the previous node 440mm2 chip, or maybe it's slightly faster. It won't touch 600mm2 parts, pal.
 
I will bet one internets that both Nvidia and AMD beat their current flagships with their incoming mid-range cards. ;)

Nvidia for sure, they have to. If AMD doesn't then it will be less of a surprise obviously and they may decide to not even try, going for perf/Watt instead. But that's how I see it.
 
OK this is interesting, more so because of the photo but is it a mock-up by them???.....
Invites also seem to be going on for Pascal consumer launch.
http://pctuning.tyden.cz/component/...p100-pujde-i-do-hernich-grafik-nejen-do-tesel
Photo in link anyway but main news is the launch invites going out:
titanusl7ja0.png


NVM looking closely does not seem right, excited as it aligned with some early rumours doh :)
Cheers
Upon closer inspection that sticker appears to say "GTX TITAN #######" and the die is ~50% of the area of a fiji. Also shares roughly the same power requirements as a GTX950. Not sure how reliable that picture is...
 
Do you expect Nvidia to beat the Titan X with their fastest GP104 card?

(Sorry for double post, but I missed this part.)
As Ailuros said, I don't think it was ever meant to. But to be honest yes. I do, I do think it will be somewhat faster, because of resons.

What's the difference again? GM200 600mm2, GP104 stimates 300+ mm2.
Would I expect GP106 to beat the Titan X or even come close to it? Hell no!
 
The problem is that you're taking examples from when there were only 3 tiers of GPUs, and trying to extrapolate it to todays AMD 4 chip strategy.

7870 was a 212 mm2, beating a 389mm2 chip.
5770 170mm2 vs 4870 256mm2.

Now just because in your head "It MUST be so!!" you're pretending that a 232mm2 chip should be equal to a 600mm2 chip, with no evidence to prove such a ludicrous claim. And I'm telling you, you must be on some hard drugs, mate. Just like in the past (see the examples for reference). the 232mm2 chip is likely to be equal to the previous node 440mm2 chip, or maybe it's slightly faster. It won't touch 600mm2 parts, pal.

Yes because Fury X is an awful GPU in so many ways. It barely beats that same 440mm2 Hawaii GPU in most cases - certainly not by enough to warrant HBM and the extra area.

And on top of that, Hawaii is another unbalanced GPU. Doesn't even have compression, so the only real factor preventing it from losing to Polaris 10 (bandwidth), is basically a non-factor.

I'm coming from a simple angle here - AMD's current generation of GPUs are basically, terrible. If they can't get at least back on track vs Nvidia, given that they barely even put out a passable GPU the past 3 years, then they just need to give up altogether.
 
(Sorry for double post, but I missed this part.)
As Ailuros said, I don't think it was ever meant to. But to be honest yes. I do, I do think it will be somewhat faster, because of resons.

What's the difference again? GM200 600mm2, GP104 stimates 300+ mm2.
Would I expect GP106 to beat the Titan X or even come close to it? Hell no!

We're not talking about GP106. Polaris 10 is AMD's true midrange GPU, not some weak wannabe.

Would you also not have expected the 294mm2 680 to have blown past the 520mm2 580 by 30%? Because that's what happened.
 
I'm coming from a simple angle here - AMD's current generation of GPUs are basically, terrible. If they can't get at least back on track vs Nvidia, given that they barely even put out a passable GPU the past 3 years, then they just need to give up altogether.

Terrible, really? Have you compared them to the GPUs they were competing at launch lately? No, not the 2 year old benchmarks, latest benchmarks.
 
This time around there are obvious differences, for example both companies really pushed the 28nm node to its limit so it should be more difficult.
There we go again... You seem to think that silicon is like an engine with a throttle that you can push. It just doesn't work that way.

Do you expect Nvidia to beat the Titan X with their fastest GP104 card?
It should blow right past it.
 
There we go again... You seem to think that silicon is like an engine with a throttle that you can push. It just doesn't work that way.

What? Compare Fury X and Titan X to any previous node and you'll find they are both well ahead in performance vs the earlier GPUs on the node. The node lasted longer and the area was pushed further than ever before.

It should blow right past it.

Yep I expect 25%.
 
Terrible, really? Have you compared them to the GPUs they were competing at launch lately? No, not the 2 year old benchmarks, latest benchmarks.

Maybe terrible is unfair but they sure didn't help AMD's cratering market share over the past 3 years. There's no denying the power draw either, especially on Hawaii.
 
Neither die sizes, nor transistor counts can tell you much if you don't know the efficiency increase (ideally per cluster) for each of the upcoming architectures. There's nothing wrong with current Radeons either, except maybe that the power consumption could had been somewhat better.
 
What? Compare Fury X and Titan X to any previous node and you'll find they are both well ahead in performance vs the earlier GPUs on the node. The node lasted longer and the area was pushed further than ever before.
Yes. And none of that has anything to do with pushing a process node.
If you'd port Fiji or gm200 to 16nm or 40nm and their predecessors as well, you'd see just the same improvement.

There are no indications whatsoever that they were process related.
 
We're not talking about GP106. Polaris 10 is AMD's true midrange GPU, not some weak wannabe.

Would you also not have expected the 294mm2 680 to have blown past the 520mm2 580 by 30%? Because that's what happened.
yes the 680 does go past the 580 due to a major re haul of the architecture and process change 40nm to 28 nm two full node jumps (what we have here for this gen is node and half jump 28nm to 16nm).

Suffice to say, everything else is still up in the air, there isn't enough information out to draw any conclusions, outside of people that have already signed an NDA or people that are working for nV and AMD.
 
Just because we haven't found yet the right combination of amount of units + N% added effficiency + N% added frequency for upcoming Pascal SKUs, doesn't mean that similar stunts aren't possible even if it's just from 28HP to 16FF+. It's more of a question where does one halt to have a reasonable perf/W ratio and also leave enough headroom for vendors and final users to overclock.
 
Last edited:
from what they have stated, parts of it were redesigned and parts of it still the same. But yeah there just isn't enough info yet.
 
Back
Top