Yes, it will definitely be announced at GTC, but nobody (outside NDA) knows if it will be a real or paper launch.
There were some shrouds leaked showing that they are indeed going to be called the 1080 and 1070, rather than the X80 and X70. I preferred the latter, particularly because 1080 looks like 1080p, and the 1080 will be able to handle considerably more than 1080p!
Anyway, if all proceeds like past generations expect the 1080 to be 20-30% faster than the 980ti and the 1070 to be 5% faster than the 980ti.
Correct if I’m wrong, but these parts are going to be highly derivative of NV’s current architecture, right? Most of the performance gains are from transition from .28 to .16? And no HBM 2 yet.
They’re under NDA, so nobody knows for sure who can talk about it, but nVidia’s briefings at GTC last year said it was an entirely new architecture with double the performance per watt. So not just a die shrink. HBM2 probably isn’t happening this generation.
At that same presentation their CEO said it would be 10x faster than Maxwell so you know… grain of salt. I’d be happy with the usual 30% jump at the GTX1080.
10x faster in very specific situations running certain code that will never actually happen in a game. But, yeah, a 25-30% across-the-board jump is enough to entice me opening up the wallet.
Could’ve sworn the next gen arch is coming next year, but I don’t follow this stuff like I used to. And we’ll find out here soon enough.
To his credit, he did say “10x was ‘CEO math’”. He was referring specifically to highly-parallelized AI code. But double the power efficiency (some very possibly fake leaks showed a X70 card with a single 6-pin connector!) and a ~25% performance boost would be very welcome indeed. That would be enough for the 1080 to comfortably handle 4k gaming, no problem, at high quality, refresh rates, and framerates. The 980ti can’t quiiiiite do that.
And who knows, maybe nVidia will suck it up and agree to support adaptive sync. G-sync still technically offers an advantage over freesync in that it can hold an unchanged image for multiple screen refreshes. Now that intel said they’d support adaptive sync, and it’s a VESA standard, everybody knows g-sync is dead. Nvidia needs to do the right thing already, and increasing the bill of lading for every monitor by ~$100 isn’t that.
Agree on the adaptive sync. One of the reasons I went with my Dell 34" ultra wide over a 4k panel is that my 970 could handle 5m vs 8m pixels. Though I did try a 28" 4k with G-sync and despite being an AA junkie I didn’t need it at that resolution. And the G-sync feature was very nice.
Nvidia did not announce any consumer products today. They did announce a pascal Tesla for AI work, but it uses HBM2. So either they create a whole separate Pascal chip with GDDR or they’re going to wait for HBM2 costs to come down before introducing the big Pascal at all. My guess is the latter-- no new NV consumer GPUs until Q4 2016 at least.
Huge window of opportunity for AMD here.
Edit: Some people are saying the pascal P104 will use GDDR. That’s the equivalent of the 980/970, as opposed to the titan and 980ti. But if they were announcing that, you’d think it would be done at the keynote.
Listened on and off to the keynote yesterday, and the comments after it was done were hilarious, including this image:
Honestly I think Q4 2016 is not going to happen either. All of the hints in the keynote were pointing at Q1 2017. I completely agree that AMD should be working their asses off right now to show something in the next couple months targeting early Q4 2016 or even Q3 2016. They could really make progress recovering their rep in the space.
Crap. Well I’m going to get a 980Ti then. Aside: does anyone know if I can gift the free copy of the Division they’re currently bundled with? I already own it.
You’ve probably got at least six months of waiting to get something better than the 980Ti. I’ll be surprised if AMD manages to release something before that. Just go for it!
I’ve been telling everyone: This is a tock coming up, not some incremental BS. Don’t waste your money right now and regret it in 6 months. That is such a shitty feeling.
Edit: I have friends upgrading from a 780Ti to a 970. Just because the card is the “next generation” doesn’t make it worth it. They’re getting 1 to 5 more frames where it counts and paying out the nose to do it. If you just buy every tock you save tons of money and don’t lose all that much performance.
Obviously they’ll battle it out at the high-end as usual, but in many ways AMD’s polaris is more interesting-- they demonstrated some insanely low power draws-- we could see tablets with the GPU muscle of a current-gen console in six months!
And combined with thunderbolt3… imagine a little dongle that you plug into your ultrabook and get a GTX750ti-level GPU capable of 1080p gaming. No external power needed.
Expect the PS4.5 and the Xb1.5 to be dramatically slimmer/lighter with much lower power consumption as well.
Skedastic – I don’t know for certain, but in the past the bundles have consisted of a code, and you can give that code to someone who can activate it in steam. The games that came out around the time the 970 was released were exclusively codes, IIRC, and were gifted/sold/resold pretty freely.
So the consensus, given that Nvidia isn’t coming out with a new card for at least six months and possibly a year, is that now isn’t a bad time to upgrade after all! Crazy. The 970 and 980 came out ~18 months ago. My recollection is week, but aren’t Nvidia’s new product release cycles/upgrade cycles more typically an annual thing rather than every other year?
AMD said “before the back to school season”, which starts in mid-July.
Rumors are that AMD is far ahead of Nvidia this cycle, so that isn’t so surprising. But I still think we’ll see GTX 1070/1080s with GDDR5x in the summer too-- just not the full-fat GP100 with HBM2 (Titan/1080ti).
Polaris in summer won’t have HBM2 yet though (that’s called Vega), so the apples to apples showdown is going to be next year. But the summer offering from AMD is likely to be very nice.
Edit: I think this is what you were saying, just trying to clarify for any who might think AMD was beating nVidia by 6 months to HBM2.
Someone at a roundtable discussion three months ago said the target is “before the back to school season.” Given no more official statement has been made in the meantime, I wouldn’t bet the house on that target being met.
