980ti - when?

B

Brahmzy

Supreme [H]ardness
Joined
Sep 9, 2004
Messages
5,024
If I buy an eVGA 980, or 2 980s, do you think the 980tis will drop within my 90 day step-up window?
 
Last edited:
Rumor is end of summer but if 390x is all that I'm sure they will bring it sooner. though no date set for 390x.
 
It really all depends on when the 390x comes out. The sooner the better. I personally think it will come by then becUse I can't imagine amd wants to wait till the end of the summer. No response to the 9xx really wasn't a good thing for consumer or amd
 
390X will probably launch sometime June with availability sometime in July if I had to guess. The 980 Ti will probably follow by releasing in August/early Sept. So probably a 4-5 month wait from today.
 
Jasheon said:
Might as well wait for Pascal
Click to expand...

Small Pascal will probably show up around January 2016 (or earlier) and big Pascal in the summer of 2016. The OP has 670 SLI so he would be well off by just grabbing a single Titan X right now, it would still be a big boost over what he has.
 
I'm real close to grabbing one of those 40" Sam 4Ks. A single Titan X isn't enough horsepower. 2 980s / 980tis should be.
 
2 980s won't get you too much more than an OC'd single Titan X and it costs more but limits your upgrade path unless you plan to add a third 980 which doesn't scale well.
 
Brahmzy said:
Thus my question about the step ups...
Click to expand...

Well then to answer your question, you would most likely miss out on the 980 Ti upgrade window if you bought today. That's why I think if you're set on NVIDIA, go with a single Titan X and OC it (they all pretty much do 1.4 GHz) for now.
 
I'm in your boat - bought my 980s after I got rid of a pair of 970s and now hoping the 980ti will drop within my 90 day step up. Since I bought in mid-March, however, I'd say the odds are very slim for me. I'm thinking we'll see a late summer (maybe August/Sept) release for the 980ti, but it could be even later if the 390s continue to be delayed.
 
I'm not buying the end of summer thing. I'd guess closer to June or maybe early July. With 16nm/16FF+ ramping up after this summer I don't see why they'd bother dropping a 980 Ti that late unless the 390X completely blows.
 
don't evga still have the policy of the step up that you can't step up for the same series?
 
Well, just ordered a single 980FTW for now - I couldn't take the waiting any longer, lol. Will probably pick up a 2nd one soon.
 
90 days seems too optimistic.
But, yeah the 980Ti seems to be the card to get (Unless NVIDIA screws up the pricing).
 
After the 390x comes out...

NVidia is waiting to slap down the 390x hype and price point by releasing a nearly full Titan X at an equal or lower price... Even if the 390x is better overall, NVidia can cut it off at the legs by releasing the 980ti at $50-$100 less, with less power draw...

Honestly, release timing has been 9/10ths of nVidias success imo. People get mad when they trump their own Titan cards quickly after release, but those Titan/follow-up money grabs are why NVidia makes a LOT more money than AMD, despite AMD being more innovative and typically better performance per dollar. NVidia is business savy.
 
5150Joker said:
Small Pascal will probably show up around January 2016 (or earlier) and big Pascal in the summer of 2016.
Click to expand...

Doubtful.
I would expect them to shrink GM206 and/or GM204 to 16FinFet to get a feel of the process before trying to manufacture a new architecture on a new process.

GM206 is a great candidate.
~230mm2 with a 128bit bus. Would shrink down to 130-140mm2 which is a great size for feeling out a new process.

GM204 is also a good candidate for a shrink.
~400mm2 with a 256bit bus. Would shrink wonderfully down to ~220-230mm2, depending on if they want to add anything and how much.
 
LordEC911 said:
Doubtful.
I would expect them to shrink GM206 and/or GM204 to 16FinFet to get a feel of the process before trying to manufacture a new architecture on a new process.

GM206 is a great candidate.
~230mm2 with a 128bit bus. Would shrink down to 130-140mm2 which is a great size for feeling out a new process.

GM204 is also a good candidate for a shrink.
~400mm2 with a 256bit bus. Would shrink wonderfully down to ~220-230mm2, depending on if they want to add anything and how much.
Click to expand...

NVIDIA's roadmap clearly shows Pascal in 2016 which is assumed around Q1:

Pascal1.png


They're not AMD, they don't recycle designs over and over and over for their tier 1 products. GM204 could conceivably be shrunk and optimized for notebooks and that would work fine but even then I'm doubtful.
 
Last edited:
5150Joker said:
NVIDIA's roadmap clearly shows Pascal in 2016 which is assumed around Q1:

They're not AMD, they don't recycle designs over and over and over for their tier 1 products. GM204 could conceivably be shrunk and optimized for notebooks and that would work fine but even then I'm doubtful.
Click to expand...

Lol. Good luck seeing Pascal in Q1... Only thing I have heard that seems credible is Q3 '16.

Nvidia doesn't recycle designs? The facts don't agree with you. They constantly respin their top tier GPUs into new products...
G80 -> G92 -> G92b
GT200 -> GT200b
GF100 -> GF110
GK100 -> GK110AB -> GK110A -> GK110B

Don't forget their roadmap changing constantly the past 4-5 years.
 
This was presented during the Titan X launch. You saying NVIDIA is lying and you know better?
 
Don't need to, if NVIDIA was going to release it Q3 they'd say as much on their recent slide.
 
I doubt pascal would be rolled out for consumers in 2016 q1. This is merely a roadmap that depends on so many different factors, like TSMC's production yields. Right now Nvidia is binning their flagship chips with disabled cuda cores and packaging them as 2nd tier cards. I reckon q2 or q3 in best case scenario. If not then they'll be forced to make changes to their roadmap and come up with something to satisfy customers, maybe modify maxwell and call that a new chip. There wasn't a huge difference between kepler and maxwell anyway besides better power delivery (fluctuating power rapidly to achieve a lower average tdp) and modified smm architecture to allow segmentation.

Still an improvement over previous generation - unlike amd..
 
Dang, that will be nice to see. I was waiting on the 980ti, but instead grabbed a second 780ti classi for cheap. It should do me well till a huge performance gap in GPU tech.
 
Full 3072? how and why. If true, then what about yields? I still think it's 2600 cuda cores which would allow overclocking, higher power limits but with reduced vram utilisation due to dead cores and segmentation, hence 6gb. However, they released the 780 ti after the titan black with half the vram so it's not completely impossible for the 980ti to be the full GM200. And since the titans are so expensive they've probably got a lot of chips lying around waiting to be sold. Maybe they've had excellent yields and TSMC's perfected the 28nm node, who knows.

But Mother of God. RIP amd. Preparing my wallet.
 
Last edited:
I would be surprised if the 980Ti is specified as +127 core out of the box compared to Titan X ... that's about the usable oc limit for my particular Titan X, unless they are harvesting the best cores for 980Ti 3-4 months from now. Yeah I can bench +200-250 like ocn, but it isn't game stable at those speeds. Maybe if they tighten it up such that it runs much closer to the boost spec rather than overshooting. I seriously doubt there's a respin for 980Ti, so it should be about the same in the end after manual overclocking.
 
Last edited:
LOL why is anyone surprised at this news? History has shown that Nvidia have no problem releasing faster cards than their "Titan" brand.

The 28nm yields have been fine for ages, it's a very mature process now.
 
You must log in or register to reply here.
Back
Top