NVIDIA Will Unveil Pascal GeForce GTX 1080 and GTX 1070 Graphics Cards at a Special Livestream Event May 6 – Cooler Shrouds Confirmed

[Masada02]

Soooooo paper launch, manufacturer "benchmarks", and no 3rd party reviews? Yet the hype is unbelievably strong with this one. Hmmmmmmmmmm, seems hypocritical. 

 

People complain about AMD cherry-picking numbers, yet they let Nvidia tell them whatever they want?

 

Wait for 3rd party reviews people, then get hyped. Not that the presentation wasn't impressive, but I'll believe it when I see it. 

[ivan134]

Is a block diagram of Pascal available yet?

[DeathRtH]

12 minutes ago, ivan134 said:

That's funny. I guess I missed the part where he said how fast the 1070 was or the part where AMD showed the performance of their cards. Damn, I missed a lot.

https://www.twitch.tv/nvidia/v/64989878 about 1:39:50 "Faster than a Titan X" Right before he says the MSRP. 

[i_build_nanosuits]

Just now, super_skank said:

3X the compute power im guessing they mean....which in gaming will be the usual 5-10% faster in games.

no it's because it can do some stuff with VR like there is no performance hit from running two seperate images at once or something...some voodoo magic to minimise the impact on performance for VR basically...within the GPU...and some 3d positional sound stuff as well added to the cards.

[handymanshandle]

4 minutes ago, RagnarokDel said:

That's actually a pretty steep price increase for the 1080. A 980 is selling for 460/530$ USD That's in the best case scenario a 13% to 30% increase

Hmmm, thought I said around 980 prices. Still, it's less than $1000. 

[ivan134]

Just now, DeathRtH said:

https://www.twitch.tv/nvidia/v/64989878 about 1:39:50 "Faster than a Titan X" Right before he says the MSRP. 

Yea, it was already pointed out to me. I somehow missed it.

[SteveGrabowski0]

2 minutes ago, i_build_nanosuits said:

i think, honestly...i think the 1070 will perform similar to an overclocked 980ti...and the 1080 will blow it out of the water.

Nah, the stock 1080 looks like it performs a little better than an overclocked 1080ti. Nvidia definitely had to give up some IPC in Pascal. The numbers given in the presentation looked pretty consistent with the leaked Firestrike score.

[i_build_nanosuits]

1 minute ago, ivan134 said:

Is a block diagram of Pascal available yet?

http://www.pcworld.com/article/3052222/components-graphics/nvidias-pascal-gpu-tech-specs-revealed-full-cuda-count-clock-speeds-and-more.html

[ivan134]

3 minutes ago, i_build_nanosuits said:

http://www.pcworld.com/article/3052222/components-graphics/nvidias-pascal-gpu-tech-specs-revealed-full-cuda-count-clock-speeds-and-more.html

That's Tesla and I hope GeForce does not look like that because I dont see asynchronous shaders.

[grayperview]

until real world benchmarks are out id take everything they say with a pinch of salt, my money is on it scraping past an overclocked 980 ti by the skin of its teeth.  

[i_build_nanosuits]

Just now, ivan134 said:

That's Tesla and I hope GeForce does not look like that because I dont see asynchronous shaders.

don,t worry about that...faster than a titan X...379$ and 599$...that's ALL you have to get out of this.

[Kinda Bottlenecked]

Oh, hopefully the high idle clock speed on those cards are gone now 

 

http://www.pcper.com/news/Graphics-Cards/Testing-GPU-Power-Draw-Increased-Refresh-Rates-using-ASUS-PG279Q

 

That incident was so embarrassing. 

[i_build_nanosuits]

2 minutes ago, Pohernori said:

Oh, hopefully the high idle clock speed on those cards are gone now 

 

http://www.pcper.com/news/Graphics-Cards/Testing-GPU-Power-Draw-Increased-Refresh-Rates-using-ASUS-PG279Q

 

That incident was so embarrassing. 

yeah but it's been fixed within a week when they discovered that...my card now idling at 135mhz

[mariushm]

AMD is expected to reveal more details about Polaris at Computex 2016, which is between 31st of May and 4th of June this year.  That's why nVIdia's 1080 is "paper launched" on the 27th, just so they could say they were the first to officially release new generation video cards.

I suspect the stock will be extremely limited, and GTX 1070 will be more common in 10th of June, maybe 1-2 weeks later.

 

In other words, I encourage everyone to watch the nVidia presentation again, the devil's in the details, in the fineprint. Yes, the 1080 card in theory is faster than Titan X but the fineprint should say "in specific benchmarks or specific conditions".  I suspect the card will be about 10-20% faster than 980 Ti in general.  Note that nVidia boss guy said the card is faster than two 980 in SLI ... they're not 980 Ti in SLI, just plain 980,

 

He also showed some graphs with performance compared to Titan and other cards... pay attention to the scales and what the curve is really about. It's not brute performance, it's something in which the reduced power consumption is also a factor, but this reduced power consumption is mainly a side effect of making chips at 16nm, it's not some revolutionary feature.

[Daegun]

2 minutes ago, SteveGrabowski0 said:

Nah, the stock 1080 looks like it performs a little better than an overclocked 1080ti. Nvidia definitely had to give up some IPC in Pascal. The numbers given in the presentation looked pretty consistent with the leaked Firestrike score.

Overclocked 980Ti? Looks about the same as a regular 980Ti. At least from what they showed, it was 1.2x the performance of the Titan X but at 400Mhz above the boost clock, or a 600Mhz total overclock. The regular cards should be right around the 980Ti/Titan X. For only $50 less than a 980Ti, and an efficiency increase. Basically the only thing exciting about this card atm is that full pascal can have 50% more cores.

[i_build_nanosuits]

2 minutes ago, mariushm said:

 Note that nVidia boss guy said the card is faster than two 980 in SLI ... they're not 980 Ti in SLI, just plain 980,

yeah but GTX 980 in SLI is quite a bit faster than a titan X!

[handymanshandle]

I agree with @super_skank that it's possible that the current numbers aren't completely accurate. Though I do have to say that, it's still impressive regardless. It still means AMD has a lot of ground to make up with Polaris.

[ivan134]

2 minutes ago, mariushm said:

AMD is expected to reveal more details about Polaris at Computex 2016, which is between 31st of May and 4th of June this year.  That's why nVIdia's 1080 is "paper launched" on the 27th, just so they could say they were the first to officially release new generation video cards.

I suspect the stock will be extremely limited, and GTX 1070 will be more common in 10th of June, maybe 1-2 weeks later.

 

In other words, I encourage everyone to watch the nVidia presentation again, the devil's in the details, in the fineprint. Yes, the 1080 card in theory is faster than Titan X but the fineprint should say "in specific benchmarks or specific conditions".  I suspect the card will be about 10-20% faster than 980 Ti in general.  Note that nVidia boss guy said the card is faster than two 980 in SLI ... they're not 980 Ti in SLI, just plain 980,

 

He also showed some graphs with performance compared to Titan and other cards... pay attention to the scales and what the curve is really about. It's not brute performance, it's something in which the reduced power consumption is also a factor, but this reduced power consumption is mainly a side effect of making chips at 16nm, it's not some revolutionary feature.

I'm seriously not sure why this performance jump from a node shrink is surprising to anyone here. This was expected going by history. I predicted Polaris 10 is going to give R9 Fury performance for about $300 and judging by this, I'm on track.

[i_build_nanosuits]

2 minutes ago, ivan134 said:

I'm seriously not sure why this performance jump from a node shrink is surprising to anyone here. This was expected going by history. I predicted Polaris 10 is going to give R9 Fury performance for about $300 and judging by this, I'm on track.

it better be...remember...faster than titan X, GTX 1070 cards 8GB -- 379$ MSRP...FuryX is nowhere near as fast as a Titan X...so asking 300$ for a card with the performance of a Fury will already be quite a stretch.

[ivan134]

1 minute ago, wcreek said:

I agree with @super_skank that it's possible that the current numbers aren't completely accurate. Though I do have to say that, it's still impressive regardless. It still means AMD has a lot of ground to make up with Polaris.

What ground to make up? Do you seriously think AMD is not going to get these performance increases? The only thing is we most likely won't get a product from them that will compete with the 1080 at 1st. I seriously don't understand how anyone is worried about the performance of AMD's upcoming cards. It's like history means absolutely nothing.

[grayperview]

1 minute ago, wcreek said:

I agree with @super_skank that it's possible that the current numbers aren't completely accurate. Though I do have to say that, it's still impressive regardless. It still means AMD has a lot of ground to make up with Polaris.

AMD are not going to bother competing with these cards, going off the numbers they are already within 25% with their current cards....Polaris is going to offer fury level (maybe even better) performance at probably the price of a 390....If this happen, you would have to be insane to buy an Nvidia card.

[NumLock21]

GTX 1080

Architecture: Pascal

Effiency: 2x Maxwell (I thought they say 3x efficiency)

16nm FinFet

2560 CUDA Cores

7.2B Transistors

8GB GDDR5X

10GB/s QDR

Over 2GHz OC capability

DP 1.4: HDR 4K at 120Hz

HDMI 2.0B: 4K at 60Hz

 

[ivan134]

1 minute ago, i_build_nanosuits said:

it better be...remember...faster than titan X, GTX 1070 cards 8GB -- 379$ MSRP...FuryX is nowhere near as fast as a Titan X...so asking 300$ for a card with the performance of a FuryX will already be quite a stretch.

Nowhere near as fast as a Titan X? Lol?

[Kinda Bottlenecked]

Just now, i_build_nanosuits said:

yeah but it's been fixed within a week when they discovered that...my card now idling at 140mhz

 

Years* 

 

780 Ti already had that problem. The high idle clock speed is just that one last problem that's stopping nvidia's architecture from being damn good. 

 

Every now and then, a question pops up and asks "Why is my so and so card idleing at 800/1000mhz?" Very unfortunate that I always have to tell them that the 144hz Gsync panel they bought is the cause. 

 

If this gets fixed then there will no longer for me to tell peeps that 144hz panel has this big flaw that comes with it  

The issue is still prevailing and some peeps even think that idling 900mhz is "normal" LOL 

[i_build_nanosuits]

Just now, ivan134 said:

Nowhere near as fast as a Titan X? Lol?

for the love of god pull your head out of your ass for once and go look at performance benchmarks...the TitanX is much faster...across the board.