[Rumour] Nvidia's Pascal on track for 2016

[MrRavens]

Awesome, can't wait.

[MEC-777]

Starting to save my pennies already.

[Thony]

Still wonder what number they use for that gen.

Can't wait to buy X70 or X70ti (maybe?!).

@MEC-777 for a second I though u said something else

[Valentyn]

My 980 Ti is already outdated.

 

True, but I can't play 60fps as a minimum in the Witcher 3 at 1440p Ultra with all Gameworks features on right now on Pascal.

[spartaman64]

Everyone hold their horses. TSMC isn't even shipping 16nm FF parts yet, and I've heard nothing from the industry saying their yields are yet up and volume production is either in swing or getting ready. Furthermore, Nvidia said Pascal will be 16nm FF+, or another quarter node behind 16nmFF. Even further, we have no news that HBM 2.0 is ready to enter production, and Nvidia's promising 8 and 16GB cards this time around. It's a clickbait article and nothing more.

 

 

Pascal is still looking to be a Q2/Q3 affair in 2016. You're not remotely outdated yet.

so they have to get it from samsung whom they are suing right now or get it from intel an competitor 

[LukaP]

so they have to get it from samsung whom they are suing right now or get it from intel an competitor 

Or GloFo. my bet is on a deal with intel. Even though they kinda arent talking to eachother right now, Intel does desperatly need some graphics IP, and i havent heard them ever oppose the idea of leasing fab space. Plus if a merger is to happen sometime in the future (which it will) then Its good for both parties to be working together on the same equipment before

[Commander Llama]

My 980 Ti is already outdated.

 

That old rusty soup can was outdated on June 2nd.

[patrickjp93]

so they have to get it from samsung whom they are suing right now or get it from intel an competitor

Nvidia would never let AMD have the satisfaction of seeing it move to GloFo, and Intel would want access to its fundamental GPU IP in exchange. Nvidia's too prideful and sees the writing on the wall as it is.

[Briggsy]

Everyone hold their horses. TSMC isn't even shipping 16nm FF parts yet, and I've heard nothing from the industry saying their yields are yet up and volume production is either in swing or getting ready. Furthermore, Nvidia said Pascal will be 16nm FF+, or another quarter node behind 16nmFF. Even further, we have no news that HBM 2.0 is ready to enter production, and Nvidia's promising 8 and 16GB cards this time around. It's a clickbait article and nothing more.

 

 

Pascal is still looking to be a Q2/Q3 affair in 2016. You're not remotely outdated yet.

 

I agree, Q3 2016 at the very least, probably later. Nvidia will want to make their coin back from workstation cards first, so probably Q1/Q2 before the first Pascal cards even appear, but like you say it also depends on HBM2 production. I'm guessing they want to get Pascal and HBM2 into the hands of businesses and professionals as soon as they possibly can (cha-ching), hence the decision to start with the largest chip first.

[Maslofski]

dont we need motherboards with that weird nvidia link to sli pascal then?

[ACatWithThumbs]

I agree, Q3 2016 at the very least, probably later. Nvidia will want to make their coin back from workstation cards first, so probably Q1/Q2 before the first Pascal cards even appear, but like you say it also depends on HBM2 production. I'm guessing they want to get Pascal and HBM2 into the hands of businesses and professionals as soon as they possibly can (cha-ching), hence the decision to start with the largest chip first.

I think it will be actually Q1/Q2 2016 otherwise Nvidia wouldn't have skipped the 1000 series.

They had to be very confident to release a GTX980 TI instead of of a new series which closed up any chances of releasing several cut down version of GM200 chip like they did with Kepler.

Or AMD's next series is just not going to be good enough so that they simply didn't see the need for a new series.

 

[patrickjp93]

I think it will be actually Q1/Q2 2016 otherwise Nvidia wouldn't have skipped the 1000 series.

They had to be very confident to release a GTX980 TI instead of of a new series which closed up any chances of releasing several cut down version of GM200 chip like they did with Kepler.

Or AMD's next series is just not going to be good enough so that they simply didn't see the need for a new series.

 

The reason Q1/2 just isn't realistic is TSMC itself. 16nm FF tapeouts haven't happened for anyone yet, and Pascal is planned for 16nmFF+, a quarter node smaller to accommodate HP chips like GPUs. After a tapeout, things can go wrong causing delays. In the best case a tapeout happens and volume production is a quarter later. TSMC hasn't managed that in years.

[Ketcchup]

So it is the time for a new GPU cycle.

Who remembers 7800gtx sli owners crying about a single gtx8800 destroying them?

[swifty7]

I guess I'll be keeping by gtx 670 SC 4GB edition a bit longer until pascal.

[Pomfinator]

Doesn't this just mean they've finalized the design of Pascal? Theoretically at least.  Because didn't they finalize Maxwell a long time ago just TSMC was too incompetent (hyperbole) to produce enough 20nm and they had to go back to the drawing board?

[TechCzech]

Hardware companies have their new versions of chips ready to roll out based on a long term schedule and would decide to shorten the development/test/tune period only in some special cases. There would have to be significant reason to do it.

After all...they have to get their money for the development of previous generation.

And no company on stock market will just decide they can forget about the money.

I would expect the Pascal based cards to be ready for the Christmas market the earliest.

It would be a surprise if they jump ahead and release them in the fall.

However one never knows of course

 

To tease you some more here are some rumored specs…

 

Architecture Pascal

Possible names of the actual chips...GP200, GP300, GP304 ... GP204

16nm FinFET+

Chip size 500-550mm square

CUDA up to 4500-5000

Memory HBM2 with capacity up to 32GB (for PRO cards)

NVLink

 

and finally... the expected performance all the way up to 60-80 percent above Titan X (Hmmmm, really!?)

 

[MrDynamicMan]

And this is a surprise how?

[Opcode]

And this is a surprise how?

IKR we'll see Pascal and Arctic Islands next year for sure. No need to start the PR now, I mean... confirm it.  :rolleyes:

[sgloux3470]

I think everyone is missing the point of this article, the point of this article is not that Pascal is due for 2016. It's the fact that they are eliminating the small die and going directly to the big die. So just to refresh your memory, with recent events, Maxwell launched with GM204, which is a small die ("mid range"). Then GM200 launched as the big die. The big die typically being 20-30% faster than the small die.

 

Here in this article they are saying that NVIDIA is completely skipping the small die and going directly to the big die for Pascal launch. So there will be no mid-range graphics card launched as a premier product like in the past (eg. GTX 680 GK104, GTX 980 GM204). Instead they will be launching their GP100 right from the get go (big die Pascal). For those who are wondering what GP100 is, it will be the equivalent of GM200 (Titan X / 980 Ti) or GK110 [originally called GK100] (Titan Black / 780 Ti); except with the Pascal architecture. As opposed to launching GP104 (eg. GM204) first, then launching GP100 (eg. GM200) later. 

 

 

This seems really unlikely to me.

 

That would mean that their low-mid range lineup is just going to be rebrands and they're going to have yet another 780Ti/980Ti scenario where their Titan is utterly invalidated by a much cheaper, equivalent slightly cut down version of the chip.

[Jcowley]

The reason Q1/2 just isn't realistic is TSMC itself. 16nm FF tapeouts haven't happened for anyone yet, and Pascal is planned for 16nmFF+, a quarter node smaller to accommodate HP chips like GPUs. After a tapeout, things can go wrong causing delays. In the best case a tapeout happens and volume production is a quarter later. TSMC hasn't managed that in years.

 

For anyone you say? I'll let you in on a little secret - I'm holding a 16nmFF piece of silicon in my hand right now my company got back from TSMC 2 months ago - we'll have 2 designs on it by the end of August.

[patrickjp93]

For anyone you say? I'll let you in on a little secret - I'm holding a 16nmFF piece of silicon in my hand right now my company got back from TSMC 2 months ago - we'll have 2 designs on it by the end of August.

Where's the TSMC loudmouth headline? Clearly yields are still an issue then, whether or not you're getting your designs. Who do you work for?

[Guest Strangerbob]

Who cares...  it will just be a 1500$ card.

 

And you will still pay 200$ for a 660ti equivalent performance and 300$ for a 680 equvalent performance.

 

 

Basically. If AMD doesnt pull a miracle and steal 40% of nvidias market share, be ready to have marginal price/performance gpu upgrades from now on... just like we have for CPU`s.

 

GG pc gaming master race. You get what you deserve.

[patrickjp93]

Who cares... it will just be a 1500$ card.

And you will still pay 200$ for a 660ti equivalent performance and 300$ for a 680 equvalent performance.

Basically. If AMD doesnt pull a miracle and steal 40% of nvidias market share, be ready to have marginal price/performance gpu upgrades from now on... just like we have for CPU`s.

GG pc gaming master race. You get what you deserve.

If AMD dies before Intel knocks Nvidia out of HPC (and thus out of direct competition), it will be good for the entire industry. Nvidia would buy up the x86 and ARM portions, and Intel would get ATI. You'd reignite competition on both fronts within a year. The longer AMD holds on, the more likely it becomes Intel walks away a legal monopoly with no one having the resources to challenge it ever again.

[Thony]

Hardware companies have their new versions of chips ready to roll out based on a long term schedule and would decide to shorten the development/test/tune period only in some special cases. There would have to be significant reason to do it.

After all...they have to get their money for the development of previous generation.

And no company on stock market will just decide they can forget about the money.

I would expect the Pascal based cards to be ready for the Christmas market the earliest.

It would be a surprise if they jump ahead and release them in the fall.

They said pascal will be out no earlier than 2016.

So surprise would be Christmass. The fall would be insanity.

Im expecting Q1 or Q2 at worst.

[BiG StroOnZ]

This seems really unlikely to me.

 

That would mean that their low-mid range lineup is just going to be rebrands and they're going to have yet another 780Ti/980Ti scenario where their Titan is utterly invalidated by a much cheaper, equivalent slightly cut down version of the chip.

 

Speculation has it that in order for this to be possible GP100 would have to be much smaller than GK110 and GM200 were, people are saying no more than 500mm² and most likely under 500mm². Some are even saying it wouldn't even break the 400mm² mark. 

 

Which means they could easily offer many versions of this chip at different levels and price points. 

 

Personally though, based on the rumor's reliability (from the source it comes from) I don't see this as likely either because it just hasn't followed NVIDIA's recent pattern. But really who knows.