Wait for GTX1080 (pascal)?

Being new to PC gaming I've made some poor choices with building my current rig. I bought 2x 780Ti cards a year and a half ago and have realised it struggles to run many recent games at max settings. Couple that with lack of SLI support, I know games like Quantum Break have very poor optimisation but I like to run everything at max otherwise I may as well play on consoles.

I didn't see the benefit of upgrading from 2x 780Ti to a single 980Ti since the performance is only slightly higher, would the next gen Nvidia card be twice as powerful as my current GPU setup? From what I can see, 1x 980Ti has much better performance (approx 20fps more) than 2x 780 (non Ti) cards so given that 780Ti to 1080 has 2 releases between them, is it likely to be twice as powerful?

I have to play games at 1440p, maxed, 60fps.

yes. its better to wait for pascal and see what the performance will be.

780Ti to 1080Ti or whatever will NOT be 2x faster, it will probably be more like 50% faster, the 980Ti being 25-30% faster.

They are also still using GDDR5.

The rumours around AMD Polaris is they will be nearly as fast as Nvidia GTX980ti for around $300, it's definitely best to wait.

Things are really wierd at the time we speak.

According to this graph we need reference GTX1080 to be only 20% faster than a very good factory o/ced GTX980Ti in order to be twice as fast as reference GTX780Ti:

Now that's not impossible. The numbers we have from GP100 and the GP104 rumors (40SMs with 2560 cores with 1600MHz clock), going to this way.

So Nvidia can do it. The big question is: Will they do it?

Now this is why I said its wierd.

I expected (as I posted in a previous conversation) GP104 to be a better performer than Polaris 10 and Nvidia to dominate the $350 and higher market, while AMD to dominate $250 and lower market with polaris 10 and 11. (And probably a full polaris at $300 will fight with a heavy crippled GP104-150)

What I didn't expect is that the same time I was posting this, AMD's marketing department with Roy Taylor, was confirming my thoughts, saying that Pascal will be very powerful and will be a high end product while Polaris 10 will be a mainstream $300 product.

AMD’s Polaris will be a mainstream GPU, not high-end

"The reason Polaris is a big deal, is because I believe we will be able to grow that TAM [total addressable market] significantly," said Taylor. "I don't think Nvidia is going to do anything to increase the TAM, because according to everything we've seen around Pascal, it's a high-end part. I don't know what the price is gonna be, but let's say it's as low as £500/$600 and as high as £800/$1000. That price range is not going to expand the TAM for VR. We're going on the record right now to say Polaris will expand the TAM. Full stop."

...

With this, more questions came to my mind.

GTX980Ti performance will be mainstream???

GP104 is so far ahead in performance that AMD considers this to be a high end product while their Polaris 10 to be a mainstream???

And the final question thats coming from those two questions is... What the hell did they do to GP104 to be so far ahead of GTX980Ti???

@Coseniath: I think it was said that Polaris would be the mainstream, and VEGA would be high end someplace...

@horgen: Yes. We knew that.

But today, Roy Taylor revealed that that GP104 would have way more performance...

Now that's something AMD never done before (accepting that the opponent has superior performance) in GPUs (since they admit that they can't follow Intel in CPUs).

And when Vega will come, Nvidia will use GP100 as Titan for sure and none of these will happen soon (AMD is pointing early 2017 for Vega and Nvidia are not stupid to use a card that they sell for $13K each to gaming without serious competition).

ps: Since clock is something you can adjust even in the last moment, I think Nvidia will hold their horses and will come with less aggresive clock...

2x780Ti's shouldn't struggle at 1440p

AMD has been somewhat behind ever since GTX680 managed to compete with 7970.

If nVidia waits with their GP100 chip then I guess I will get more out of GM200 chip then :P

Though seriously I hoped AMD would deliver something excellent this time. I want to try their route as well, but their offerings doesn't seem as good =/

@GhostHawk196: Definitely wait for Pascal. Everyone makes choices that don't pan out. I previously bought 970 SLI and ended up swapping them for a 980 Ti. SLI/Crossfire is garbage now. Stay away from it at all costs. My 970 SLI was a bad choice, but swapping them for a 980 Ti was a great choice. Always buy the top tier card if you need beastly performance.

@horgen: Yeah I think people with GTX980Ti wouldn't have a reason to upgrade (neither performance nor VRAM would be an issue), unless they want a probably 30-40%ish upgrade.

But it shouldn't deserve it unless you can sell your GTX980Ti for a good amount and just add something like $200ish for a GTX1080.

@Coseniath: I am probably going to swap my 980 Ti for a 1080 when it launches. My 980 Ti does not struggle, but I like to get as close to 1440p/144fps as possible on all games and then let G-Sync do its thing. I always get a good return when I sell my cards. :)

@BassMan: Yeap, this is the point.

If you can sell your old GPU for a good amount then giving some $$ to get the newer architecture is a good choice!

And the guy that buys it gets more performance for less money. Its a win-win.

I guess the question is, based on past experience, how many more GPU releases until I will be able to see a 20-30% performance of a single X80 card over my current 2x 780Ti setup?

From what I can see a single 980Ti is much more superior than a 2x 780 SLI setup, is it reasonable to presume a single 1080 pascal will provide anywhere from 20-50% increase over a 2x 780Ti setup.

I realise now that the way you PC game is buy the latest X80 card, keep upgrading it instead of doing these SLI setups, I bought 1 780Ti mid 2014 and then another one at half price when they announced the 980 and now it doesn't do what I need it to but I don't want to put too much more money into my PC at this moment but a GPU upgrade is acceptable to me. I presume the 1080 will release later this year in around Sep or Oct so the ROG OC version will release either December or early next year which is when I plan to upgrade.

@GhostHawk196: 1080 is coming out in June and third party cards with custom coolers will be available at launch. Start getting ready to sell those 780 Tis. :)

I don't expect a huge increase in performance... How was the jump from 780Ti to 980?

The jump from 780Ti to 980, didn't include a double fabrication jump. It was from 28nm to 28nm. Yet Nvidia somehow managed with architecture only upgrade to give so much performance even with less transistors.

Now we have 28nm to 16nm. An 610mm2 chip based on 16nm (GP100) manage to deliver 1,5GHz 24/7 for super computing market. Consumer GP104's clocks will be way higher.

By the way...

We will soon find the real performance:

NVIDIA Pascal Editors’ Day to take place next week

@Coseniath said:

The jump from 780Ti to 980, didn't include a double fabrication jump. It was from 28nm to 28nm. Yet Nvidia somehow managed with architecture only upgrade to give so much performance even with less transistors.

Now we have 28nm to 16nm. An 610mm2 chip based on 16nm (GP100) manage to deliver 1,5GHz 24/7 for super computing market. Consumer GP104's clocks will be way higher.

By the way...

We will soon find the real performance:

NVIDIA Pascal Editors’ Day to take place next week

I don't believe there will be a massive improvement in architecture changes this like last time. But yeah higher clock speed alone will help.

Either way I'm waiting for the GP100 chip or whatever they decide to use in 1080Ti or if they call it 1180...

Waiting for GP100 is the best since you already have a GTX980Ti.

Personally I never upgraded my GPU for less than 70% more performance.

That's why I kept my GTX570, when I saw GTX670's performance (the difference was like 30%-40%).

But if someone would said to me, give your GTX570 and $150 for a GTX670, I would probably do it.

In the end, what matters is the price you pay for the upgrade you want to make...

ps: With AMD's absence in high end GPUs, I expect lower clocks from Nvidia, so they can reuse them with higher clocks in their 11xx version (#Kepler_history_repeats)...

:P

@Coseniath: With that clock increase the TDP for 680 to 770 increased a lot. 195 or 185 to 230 or so. Besides they needed better production techniques for the memory. My 680 is not stable with a memory speed at 7GHz effective. The core on the other hand is stable up to 1329MHz or so.

7970ghz was the last time they lead which seems like ages already.

I may settle for amd's mid end polaris stuff if it turns out good. My 290 turned out to be a good investment after a bad launch. They were behind 780s, now they are competing against 970s and even 980s. Perhaps the consoles using their gcn archie is doing amd a favor with optimization.

My noggin tells me both polaris and pascal will have a big jump after such a long wait for the die shrink, but my gut tells me I'll be disappointed once again. I wish June was here already.

@BassMan: Looking forward to it, I'm still deciding whether to get the stock gpu in June or wait a few months until the ASUS ROG OC version comes out. Hopefully there are better games to use such a powerful card for other than just Doom, Homefront, Deus Ex or Mafia.