Nvidia GeForce GTX 1080 rumor mill just blew up.

prava said:
Where did you see that the x80 will use GDDR5X?
Click to expand...

From the article.

NVIDIA GeForce GTX 1080 Features GDDR5X and GeForce GTX 1070 Features GDDR5 DRAM
Rest of the board was not pictured but the specific SKU was pictured with a total of eight 8 Gbps memory chips featuring 8 Gb density. The board uses the latest Samsung K4G80325FB-HC25 chips which use the 8 Gb DRAM and run at 8 Gbps speeds. This will allow NVIDIA to boost not only the bandwidth but also the VRAM size on their upcoming GeForce cards.
...
We know that the GP104-200-A1 SKU will be featured on the GeForce GTX 1070 graphics card featuring GDDR5 memory while the GeForce GTX 1080 graphics card will feature GDDR5X memory. It is pointed that both GDDR5 and GDDR5X memory are not pin-to-pin compatible since GDDR5 uses a 170 pin ball grid while GDDR5X uses a 190 pin ball grid layout. Both cards feature the same amount of VRAM of 8 GB but 1080 will have higher total bandwidth available up to 384 GB/s while retaining a 256-bit memory bus.
Click to expand...
 
Any guess on prices? I saw a foreign leak yesterday claim $650 and $500 but they just converted currencies.

YzoyqaH.png
 
xorbe said:
So the x70 is returning to a full 256-bit vram bus?
Click to expand...

We'll see. I have to imagine Nvidia won't make the same mistake as last time, and they'll cut things on an even boundary rather than maximize their spec sheets for consumers.

I mean, the GTX 950 was on an even boundary (1/4 of the chip), and it still has 85% of the performance of the 960, so I can't imagine that tiny extra slice of die space buying the GTX 970 anything. It was a whole lot of bullshit for absolutely no gain.

Also, I'm happy to see another generation of HD 4850 (GDDR3) versus HD 4870 (GDDR5) supported by the same chip (but they were using different PCBs as well). I'm sure this wasn't hard to add GDDR5x support.

Hopefully will mean a more massive difference between the price of the 1070 versus 1080 (more akin to MAxwell than Kepler).
 
Last edited:
10 Gb/s GDDR5X is available now (how much is debatable but possibly enough for one or two product models if willing to pay for it), only the 11Gb/s and 12Gb/s are in sample status.
Worth noting that one of the leaks for AMD showed a Polaris model using 10Gb/s GDDR5X (maybe early revision that is not going into production just yet), and this was photographed in early-mid March this year.
Cheers
 
Yeah, I wasn't sure I believed the bullshit rumor I hear passed around about GDDR5X not being available at all this year. That went double when they officially announced P100 sampling this year, after everyone said zero HBM2 availability until next year.

It just means it's limited quantities at a a higher price.

And GDDR5X should go down in price enough so they can drop the price o fthe flagship down next year (to make room for the 1080 Ti part).
 
defaultluser said:
We'll see. I have to imagine Nvidia won't make the same mistake as last time, and they'll cut things on an even boundary rather than maximize their spec sheets for consumers.
Click to expand...

Because what I'm half expecting is x70 to continue to shrink to 192-bit, fulfilling its x60 destiny. But then it would have to be 6GB, or 6GB + hobbled 2GB to reach the current x70 8GB claim ...
 
Today's articles show the GP104-200 die and refer to it as the "GTX 1080" but isn't there also the GP104-400? Shouldn't that be the 1080?

The two chips launching in June are GP104-200 and 400.
 
TaintedSquirrel said:
Today's articles show the GP104-200 die and refer to it as the "GTX 1080" but isn't there also the GP104-400? Shouldn't that be the 1080?

The two chips launching in June are GP104-200 and 400.
Click to expand...

Expecting anything besides the GTX 1070 and 1080 to be announced is really just wishful thinking.
 
TaintedSquirrel said:
Today's articles show the GP104-200 die and refer to it as the "GTX 1080" but isn't there also the GP104-400? Shouldn't that be the 1080?

The two chips launching in June are GP104-200 and 400.
Click to expand...

There are also Quadro SKUs to be launched as well I think.
 
TaintedSquirrel said:
Guru3D is calling this the 1080 Ti.

New Pascal GP104 (GeForce GTX 1070 and 1080) GPU Die Photos Surface
Click to expand...

If NVIDIA did that I think they would not realise the backlash it will create even from those that love their hardware.
The ti model as far as enthusiasts are concerned must be from the reference die and not the 104.
But with no competition it is also my concern they may consider this when reading some of the rumours, sigh.
Fingers crossed those rumours are mistaken.
Anyway interesting that picture shows the 12Gb/s GDDR5X, so earliest I think it will be released is June, and probably will cost even if sold as a 1080 (replacement for 980).

I still think more likely it is the 1080, and a ti model when released will have HBM and a 102 die.
This gives a broad differentiation of; 1070 standard 256-bit 8Gb/s, 1080 12Gb/s (equivalent to 512-bit 6Gb/s memory), and ti as HBM.

Cheers
 
Last edited:
Even amongst enthusiasts, most people don't care if it's the "reference die". Most people have no idea what that even means, they just know relative performance numbers.

All that matters is performance per dollar. If the GP104 GDDR5X 980ti is fast and worth the money, enthusiasts will embrace it.

NV can call their Vega competitor the Ultra instead. It's fine.
 
schizo said:
Even amongst enthusiasts, most people don't care if it's the "reference die". Most people have no idea what that even means, they just know relative performance numbers.

All that matters is performance per dollar. If the GP104 GDDR5X 980ti is fast and worth the money, enthusiasts will embrace it.

NV can call their Vega competitor the Ultra instead. It's fine.
Click to expand...

That is true but depends upon the press-publications.
If they make it known as part of their analysis-articles-reviews that NVIDIA is now naming and also charging Ti prices (yeah assumption if it is named ti they would also keep comparable pricing) for a model that used to be x80 then customers may care, I agree it will be split between those that shrug as long as it has the performance and those that feel they are being squeezed by NVIDIA for their cash.
But the bad feeling may stick, as NVIDIA is already known for their high priced margins.

That said this may all be meaningless as we have not seen a shroud with Ti (although those are also rumour leaks), but still feel it is more probable a Ti specific die will come out later and the 400 die will be used for the 980 - anything is possible though with NVIDIA when there is no competition at the enthusiast tier.
Cheers
 
Last edited:
Eh, they already shifted the tiers once with Kepler and nothing happened. They can call it whatever they want, as long as the pricing, performance and marketing matches up.

As long as there's a performance bump and no answer from AMD, they have a monopoly.
 
trandoanhung1991 said:
As long as there's a performance bump and no answer from AMD, they have a monopoly.
Click to expand...
For 6 months, until Vega, yes. At the enthusiast high-end only. But all the money is in the mass-market, and AMD is looking very strong there with its cheap, tiny, extremely power-efficient GPUs with 390-class performance.
 
schizo said:
For 6 months, until Vega, yes. At the enthusiast high-end only. But all the money is in the mass-market, and AMD is looking very strong there with its cheap, tiny, extremely power-efficient GPUs with 390-class performance.
Click to expand...

Revenue is in the mass market. Profit comes from high end. AMD needs both.
 
schizo said:
For 6 months, until Vega, yes. At the enthusiast high-end only. But all the money is in the mass-market, and AMD is looking very strong there with its cheap, tiny, extremely power-efficient GPUs with 390-class performance.
Click to expand...


I agree with that but gp106 will come out within one quarter of the launch of Polaris, so I don't think they are going to gain much marketshare, I would say 5% at most hopefully better than my fiji/r3xx line prediction on gaining market share.
 
looks like just jpg compression artifacts. Just zoomed in with adobe, they both look aliased.
 
Hypocrisy:
If you tell someone the GTX 1080 is using GDDR5X and we have a photo as proof, they claim it can't possibly be launching in June since GDDR5X won't be in mass production by then.

If you tell someone Polaris is launching in June and yet we still have zero pics of the boards... That's totally cool.

Also remember how Pascal wasn't going to be available until Q4 because JHH held up a fake board at CES*?

So much backpedaling.
 
TaintedSquirrel said:
Hypocrisy:
If you tell someone the GTX 1080 is using GDDR5X and we have a photo as proof, they claim it can't possibly be launching in June since GDDR5X won't be in mass production by then.

If you tell someone Polaris is launching in June and yet we still have zero pics of the boards... That's totally cool.

Also remember how Pascal wasn't going to be available until Q4 because JHH held up a fake board at GDC?

So much backpedaling.
Click to expand...

Everyone in the AMD threads has already figured out the full product line, price, relative performance to Fiji, AND determined that AMD's next gen cards are going to the faster than the previous gen. Bunch of fucking Internet Nostradamus fortune tellers over there.
 
Ocellaris said:
Everyone in the AMD threads has already figured out the full product line, price, relative performance to Fiji, AND determined that AMD's next gen cards are going to the faster than the previous gen. Bunch of fucking Internet Nostradamus fortune tellers over there.
Click to expand...

My favorite one has gotta be 'pascal is becoming more gcn-like, because 64 ALUs per SM. Vega will beat pascal titan because gcn is the better architecture'

Meanwhile there's no information regarding either
 
They can call it the donkey puncher 1080 for all I care... just make it fast and awesome and release it meow!
 
Ocellaris said:
Everyone in the AMD threads has already figured out the full product line, price, relative performance to Fiji, AND determined that AMD's next gen cards are going to the faster than the previous gen. Bunch of fucking Internet Nostradamus fortune tellers over there.
Click to expand...

You noticed that too, huh?
 
KickAssCop said:
Is this 2X speed of 980 Ti?
Click to expand...

There have been rumors of 50% to 90% higher performance per watt, and some people misread that not noticing the "per watt" and went totally overboard with the optimism.

I would say a realistic expectation is around 20% faster than a 980 Ti, a happy surprise at 33% faster, and really not expecting more than that.
 
no 104 for me, learned my lesson with the 680. fool me once shame on you, fool me twice, shame on me
 
shinotenshi said:
no 104 for me, learned my lesson with the 680. fool me once shame on you, fool me twice, shame on me
Click to expand...
Which at this point means 1Q2017 at the earliest for GP100 game-edition based on the most updated soothsayer reports and crystal ball readings, right?
 
evilsofa said:
Some of you may be forgetting that the GTX 980 was, in most cases, around 8% to 11% faster than the 780 Ti. Expecting 20% for the 1080 over the 980 Ti is being generous, really.
Click to expand...

why?. GTX 680 destroyed the GTX 580 by an ample margin way bigger than 20% even the GTX 670 did it, man even the 660TI was better and traded blows with the GTX 660 so yes, its always possible.
 
I have to believe the new GPU/memory architecture is going to be impressively better (20%+) at 2K, 4K, and VR applications than Maxwell
 
You must log in or register to reply here.
Back
Top