Question for those who have managed to avoid a 980Ti impulse buy

[xthetenth]

If you want immersion, curve screen are nothing more than gimmick. It won't make use of your peripheral vision.

However, 3440x1440 is the best resolution for things other than gaming, a very nice resolution for gaming, and the curve is great because it helps the screen look dead on uniform (if you get too close your viewing angle will shift enough that even though the color is right the brightness is slightly lower on the edges with a flat 34").

 

[RussianSensation]

They are popping up on craigslist now though. Since Titan Xs are already down to $750 on craigslist this should be a good pickup around $480

I would be in for dual Titan Xs for $480 per unit.

It is all waste and tell me which game requires a GTX 980 or a R9 290X for a 1080p

Most of the games are 1:1 ports and more than 90% of users are stuck at 1080p resolution.

For 60fps gaming experience at 1080p a GTX 960, R9 285 or R9 280X is more than enough.

I think I misunderstood you then. What you meant is that since he was still on 1920x1200 he doesn't need to upgrade from 680 SLI anyway? Ok, no I understand your point. Sorry, I didn't understand you the first time. :thumbsup:

1. 970 was priced very aggressive.

How? A GTX560/560Ti replacement for $330 is about right. This is like when GTX460 came out and wrecked GTX275/280/285/HD4890.

2. If Kepler wasn't neutered, SLI 780Ti > SLI 970.

Every next gen mid-range card from NV beat the previous gen flagship. You can check if you want:

GeForce 3 Ti 500 < GeForce 4200
GeForce 4 4600/4800 < GeForce 5600U/5700U
GeForce 5950U < GeForce 6600GT
....
etc.
etc.
GeForce 580 < GTX670/680
GeForce 780Ti < 980

What makes you think Pascal will be different?

3. The better normal comparison would be 980 SLI.

Sure, but 980 = new architecture, same node, same GDDR5
Pascal 980 successor = new architecture, new node, HBM2

Should be a greater increase in performance over 980Ti than 980 had over 780Ti.

4. 14nm HBM2 is going to be cheap like 28nm? You dreaming. Expect mid-range Pascal to go for $500/800 instead of $330/600.

You think NV will raise GP204 to $500 for 970 successor and $800 for 980 successor?

5. When do you think a 300-400mm2 Pascal with HBM2 will be launched? Q3 2016 is HBM2 ramping. Q4 would be building the GPU/HBM2 package.

September-October 2016.

I expect GM200/Fury will not be exceeded until 2017.

No way. AMD already committed to not waiting as long between generations as they acknowledged their mistake with having such a major gap between 290X and Fiji. I expect they will have a faster card in 2016. Even if it means just 20-25% faster than Fiji XT. Doesn't have to be 50-75% faster. I think they are going to realize sooner or later that with NV bifurcating a generation into 2, they can't wait 1.5 years to release cards 50% faster when NV is doing 30-40% increments every 9-10 months. AMD needs to also bifircuate their generation to keep up. It's the only strategy that makes sense now to split generations into parts. In the past we got 70-100% increase in 1 shot like 9700Pro vs. 8500 or X800XT vs. 9800XT or X1800XT vs. X800XT. It won't be like this anymore.

It'll be more like 680->780/780Ti or 980->980Ti style where newer cards should drop every 9-15 months with 30-40% higher performance. I expect an NV/AMD card to beat 980Ti/Fiji XT next year.

However, 3440x1440 is the best resolution for things other than gaming, a very nice resolution for gaming, and the curve is great because it helps the screen look dead on uniform (if you get too close your viewing angle will shift enough that even though the color is right the brightness is slightly lower on the edges with a flat 34").

Linus makes a strong case why for him 3440x1440 is actually better than 4K. Many of his points make a lot of sense since it strikes a better balance as a productivity and a gaming monitor. 4K is crazy demanding on modern GPUs and 32-34" 4K FreeSync/GSync monitors are MIA.

https://www.youtube.com/watch?v=KnrxNfxRK_4

 

[raghu78]

4. 14nm HBM2 is going to be cheap like 28nm? You dreaming. Expect mid-range Pascal to go for $500/800 instead of $330/600.

I agree. HBM2 volumes should be low in 2016. 16/14nm yields on 300 sq mm dies are not going to be good. Heck the foundries are having trouble yielding < 100 sq mm chips and 5W TDP. whats the hope for 300 sq mm GPUs with 175-200W TDP. I see both Nvidia and AMD bumping up the price on a 300 sq mm mid range 16/14nm FINFET HBM2 GPU to USD 700-750 for the flagship and USD 550-600 for the salvage SKU.

5. When do you think a 300-400mm2 Pascal with HBM2 will be launched? Q3 2016 is HBM2 ramping. Q4 would be building the GPU/HBM2 package.

I expect GM200/Fury will not be exceeded until 2017.

I surely expect the 300 sq mm 16/14nm FINFET GPU with HBM2 to beat the Fury/GM200. Heck even a dumb shrink with some minor architectural enhancements and slightly higher clocks can get you 15- 20% improvements.
I am expecting AMD and Nvidia will aim for a GTX 580 -> GTX 680 leap. Roughly 25-35% higher performance at roughly 60% of the TDP.

btw do not expect any FINFET GPU in 2016 to be > 300 sq mm chip. Neither Nvidia nor AMD can afford that sort of risk. The yields just won't be there. Nvidia would be going for their first HBM + 16/14nm FINFET + Pascal GPU. They would be even more conservative on die size and not have a repeat of Fermi.

Heck even Intel will be shipping the first > 300 sq mm 14nm Broadwell Xeons (16 cores or higher) only in Q1 2016. Thats with 3+ years of FINFET manufacturing experience and on a 2nd gen FINFET process with more than 12 months of HVM. Good luck to the foundries trying to ramp a 300 sq mm GPU at >=75% yields and in high volume. :whiste:

Should be a greater increase in performance over 980Ti than 980 had over 780Ti.

yeah we can expect at best a 30-35% perf increase over Titan-X (stock for stock) with GP204 and HBM2. But if you look at Titan-X OC vs GP204 OC thats debatable. It depends on the 16/14nm process maturity and what kind of OC headroom we will have on the first FINFET GPUs

You think NV will raise GP204 to $500 for 970 successor and $800 for 980 successor?

why not ? Definitely possible. Heck I would say quite likely. The only way it won't happen is if AMD R9 3xx is very good and takes significant market share back. If AMD get back to better than traditional market share of 35% and get closer to 40% by H2 2016 Nvidia will get aggressive on GP204 pricing. But still they are likely to be at the mercy of TSMC/Samsung FINFET yields just as AMD will be.

September-October 2016.

yeah i would say thats the earliest for 16/14nm FINFET GPUs with HBM2. I think AMD will get to it first. Nvidia might get to it 2-3 months later. I expect AMD to have a smoother transition to 16/14nm HBM2 as they have invaluable experience with designing HBM and bringing Fiji to market on a mature and high yielding 28nm node.

 

[escrow4]

There is nothing worth playing coming in 2015 to justify a 980 Ti. And this is coming from somebody who bought a 5930K just before Inquisition came out just because. I deleted AC Unity after a bit, it was just a pile of rubbish (play Black Flag instead), Watch Dogs was nothing special, Skyrim can run on a toaster, and the rest meh. If Fallout 4 comes out this year (and it already looks like something from 2010 not 2015 HA!) by all means upgrade. Now, for what? The issue you'll hit is lack of VRAM. Is that worth it to play some old ports? AAA gaming is just piffling this year and last. 2 or three big release if that and that's it for the year. Meh.

 

[Majcric]

1. 970 was priced very aggressive.
2. If Kepler wasn't neutered, SLI 780Ti > SLI 970.
3. The better normal comparison would be 980 SLI.
4. 14nm HBM2 is going to be cheap like 28nm? You dreaming. Expect mid-range Pascal to go for $500/800 instead of $330/600.
5. When do you think a 300-400mm2 Pascal with HBM2 will be launched? Q3 2016 is HBM2 ramping. Q4 would be building the GPU/HBM2 package.

I expect GM200/Fury will not be exceeded until 2017.

I do think it will be beaten but not by a lot and even when it is the price won't be any better than the cards we have now.

2016 it will be beaten with the die shrink but as we all know that's a long time in the tech world.

 

[2is]

There is nothing worth playing coming in 2015 to justify a 980 Ti. And this is coming from somebody who bought a 5930K just before Inquisition came out just because. I deleted AC Unity after a bit, it was just a pile of rubbish (play Black Flag instead), Watch Dogs was nothing special, Skyrim can run on a toaster, and the rest meh. If Fallout 4 comes out this year (and it already looks like something from 2010 not 2015 HA!) by all means upgrade. Now, for what? The issue you'll hit is lack of VRAM. Is that worth it to play some old ports? AAA gaming is just piffling this year and last. 2 or three big release if that and that's it for the year. Meh.

I've already played through Black Flag and all the other AC's that came before it, Unity is next in like. But yeah, the whole gaming outlook for the present and near future are largely the reason I erected this thread. Trying to decide if the upgrade itch is worth $650 to scratch it considering there's only a small handful of game that I'd see an appreciable difference in, at least for a while anyway.

One thing I forgot to mention in the OP is I do intend to purchase an Oculus Rift VR headset sometime next year, which needs both high frame rates and low latency for an enjoyable experience that won't cause motion sickness, so a single, powerful GPU is certainly the way to go for that, but again, that's in 2016

 

[Mondozei]

680 in SLI is a very powerful setup even today. It beats a 780 Ti.

And yes you're right. Pascal is going to be huge. New arch+process node in of itself is a big step but then you add HBM Generation 2 on top of that as well, and you got a trifecta that we've rarely seen before.

I was also thinking about going to 980 Ti, but I had even less of an excuse than you (290s in Crossfire). I'll almost certainly going to go Pascal in 2016.

Part of what helped me not even ordering (and then cancelling) is that I want a new monitor to boot. Unfortunately, I don't like paying the G-Sync tax, but since the next monitor will be mine for the next 5+ years, I can't guarantee AMD will be around for that long. Also, the prices on the Acer IPS 144 Hz monitor right now is very high. I think by the time Pascal comes out, we'll probably have more choice and better prices. Still, Pascal flagship + new monitor will set me back at least 1200 dollars I think. So I got another incentive to wait: saving.

 

[richaron]

680 in SLI is a very powerful setup even today. It beats a 780 Ti.

And yes you're right. Pascal is going to be huge. New arch+process node in of itself is a big step but then you add HBM Generation 2 on top of that as well, and you got a trifecta that we've rarely seen before.

I was also thinking about going to 980 Ti, but I had even less of an excuse than you (290s in Crossfire). I'll almost certainly going to go Pascal in 2016.

Part of what helped me not even ordering (and then cancelling) is that I want a new monitor to boot. Unfortunately, I don't like paying the G-Sync tax, but since the next monitor will be mine for the next 5+ years, I can't guarantee AMD will be around for that long. Also, the prices on the Acer IPS 144 Hz monitor right now is very high. I think by the time Pascal comes out, we'll probably have more choice and better prices. Still, Pascal flagship + new monitor will set me back at least 1200 dollars I think. So I got another incentive to wait: saving.

You seem quite reasonable by this post, but I'm unsure as to why anyone would consider locking themselves into any AMD or nV tech when buying a monitor. As you, and many others know, a monitor lasts a long time; so if you're worried about longevity surely you'll buy a monitor which supports the displayport adaptive refresh standard?

More on topic, I'm in the market for a new GPU myself but I can't even consider buying a flagship product on an outgoing node. As far as I'm concerned it's either the sweet spot for price/perf on a mature process, or you pay the bleeding edge tax and get a flagship* on a new node.

*I'm still giving nV the death stare for gtx 680

 

[Dnicu]

Im more pissed off than ever at the current situation.

If i buy Gsync im stuck with Nvidia. If i buy Free sync im stuck with AMD, Monitors outlast GPU's and i wouldnt want to lose variable sync once i have it. This is whats stopping me buying. I feel like im being trapped into a manufacturer.

atleast Freesync CAN be adopted by Nvidia if they want to. However, it will probably never ever happen.

You could go the way of something they did at LinusTechTips. They made a video about a PC with both Nvidia and AMD, an outrageously expensive solution, but you could end up running with both a Gsync and a Freesync monitor hooked up to either one of the monitors.
I think they managed "select" the cards by just changing a different display as "master" under windows.... It is an interesting, but expensive, middleground that would stop you from ever having to wonder about which games you can play at max.....

 

[Smoblikat]

Ill buy a 980Ti when it costs $250. Its my same rule for buying any card that comes with an nvidia tax. What im really looking foward to is the 290X''s that cost $200. Im sitting on my 7970 until either of those 2 criteria are met.