Nvidia Unveils Pascal Gpu: 16gb Of Memory, 1tb/s Bandwidth

[Guest]

Im upgrading my GPU when pascal comes out.

I'll be fine for another 2-3 years I hope

[CatCloud]

Look at how big a bump GTX 680 at 28 nm was over GTX 580 at 40 nm.

http://www.anandtech.com/bench/product/1350?vs=1348

And the 680 wasn't even supposed to be the flagship, it was originally supposed to be the 670 Ti. You gotta think they were sitting on the 780/Titan since they could keep selling the 680 for $500, like they did with the 980 this generation. A true flagship to flagship comparison might be 580 vs the original Titan:

http://www.anandtech.com/bench/product/1350?vs=1060

So if past trends hold, Pascal could really be something huge.

We shall See, all I see is flare and bling on paper.

AMD didn't exactly blow minds with the fury cards.

[AlwaysFSX]

At this point, I'm most interested in nVidia's naming scheme for Pascal and Volta. Oh and benchmarks too. Do they go 1000 series, or something totally different from before.

Nobody knows, hopefully they change it.

[VagabondWraith]

http://www.theregister.co.uk/2015/11/16/nvidia_noaa/

 

"The US government's NOAA agency will build a new computer system boosted by 760 Nvidia Pascal GPUs, according to Ian Buck, the chip company's veep of accelerated computing and CUDA inventor. When will this machine go online? Well, to give you a clue, Pascal is not due to arrive until 2016."

 

http://www.anandtech.com/show/9791/nvidia-sc15-national-oceanic-and-atmospheric-administration-to-build-tesla-weather-research-cluster

 

"Along with touting the number of major HPC applications that are now GPU accelerated and the performance impact of that process, NVIDIA’s other major focus at SC15 is to announce their next US government contract win. This time the National Oceanic and Atmospheric Administration (NOAA) is tapping NVIDIA to build a next-gen research cluster. The system, which doesn’t currently have a name, is on a smaller scale than the likes of Summit & Sierra, and will be comprised of 760 GPUs. The cluster will be operational next year, and giving the timing and the wording as a “next-generation” cluster, it’s reasonable to assume that this will be Pascal powered."

 

Thought that was kinda cool. nVidia tech making its way into major compute categories in tech, government, server, home use. Seems they're growing larger and larger every year.

[RagnarokDel]

pfft not even 32 gb of vram? If I cant load a game completely, it's not worth the upgrade!

[LeeWonky]

I just came back from a wank break and then I see this...

[Xorbot]

So 4k is def a thing next year?

 

4K will be "low settings" next year.

[RexinOridle]

What?

[Jerakl]

source pls

i want to see this

 

He just mentioned it on a stream once.

[handymanshandle]

I'm still holding out for Arctic Islands to be better or at least for it to be on par with Pascal. 

 

Regardless I'd still rather buy a 390X because honestly the PPI of a 1440p screen at 25" to 27" is going to be high enough not to notice pixels and the 390X should be enough to run 1440p at high to ultra settings and maybe even with a little MSAA.

[Lethal Seraph]

Ohhh mama

[That Norwegian Guy]

It's a shame Nvidia won't have much in the way of async compute capability until Volta, making some of that power largely unrealized in new APIs.

[Blake]

calling it now, no ti variant. they'll rename it to p, so the 1080ti becomes the 1080p.

[ShadowCaptain]

Oh no I will actually have to replace my 780ti haha

[Bouzoo]

But it seems this generation will be a MASSIVE performance boost, 28nm was holding the gpus back.

 

Let's wait and see, it may not be that massive in the end. I hope it will be but don't expect 50% more fps in games

[JebKerman]

Let's wait and see, it may not be that massive in the end. I hope it will be but don't expect 50% more fps in games

ha, your profile pic is surprisingly accurate (just add a 1)

[IhazHedont]

DSR is gonna be very interesting.

 

Even if you buy a solid 1440P monitor (like the PG279Q), the DSR will be your best friend with this kind of gpu.

[Citadelen]

Indeed. AMD tends to have the best 2D drivers, especially for multi monitor setups. I've never understood the hardware reviewers who has multi monitor setup using NVidia. Things like hydra vision and mixed resolution multi monitor setups just works on AMD no problems.

 

How often don't we hear these reviewers whine about their NVidia drivers sucking at just that?

 

 

Depends. 16nm is not just a node shrink, but also a completely new transistor design in FinFet. So if FF is twice as effective and the chips gets twice as many transistors it should ad up. If the transistors can run faster even, they we might see heat going down compared to performance even more. However I don't think FF is THAT effective, but we will see a huge boost none the less.

 

 

AMD's Arctic Island cards based on 16nm FF will also launch about the same time. Since NVidia is already getting rekt in almost all parts than the extreme highend, when we look at price to performance, then yes, they have a lot to fear. Why do you think they are so fixated on proprietary technology, that causes vendor lock in?

 

 

AMD's Arctic Island will match Pascal just fine, using same 16nm FF process and HBM2, which is 1024 GBps. I believe Arctic Island so far are 3 series of chips, each capable of being x and non x variant, so a max of 6 new cards (although we usually don't see x versions of all of them).

AMD's Arctic Island GPU's will be on the 14nm node and will contain around two billion more transistors.

[IMPERIUS]

Anyone has detail on pricing?

[Dabombinable]

Would still not buy because of horrible drivers.

Both have drivers that take turns at being shitty so......

[Mr_Troll]

no news here... they already presented the same info a year ago .... same shit ... HBM2, 1TB bandwidth, NVlink and thats it...... i am not to hyped about pascal... every new nvidia architecture  lasted for a year and then they killed it with drivers... same thing is going to happen with pascal...calling it ...

[MrDynamicMan]

Interesting. Take into account nvidia IS a micro transistor company so don't expect most of these numbers to be acurrate. I'm looking forward to seeing a big change in the GPU market.

[don_svetlio]

no news here... they already presented the same info a year ago .... same shit ... HBM2, 1TB bandwidth, NVlink and thats it...... i am not to hyped about pascal... every new nvidia architecture  lasted for a year and then they killed it with drivers... same thing is going to happen with pascal...calling it ...

Be careful - some people will beat you with a stick for those claims (I agree full-heartedly however)

Honestly - from the looks of it - the new Titan and 1080 will be HBM and Pascal - the 1070 likely GDDR5X and pascal and the rest will be Maxwell rebrands and Kepler rebrands (1060 = 980/970 (hopefully with 4GB ) and 1050 = 970M(the 960 Ti we needed) with the 1040 = 760)

[Exhilarare]

Am I the only one who thinks GTX 1050/1060/1070/1080 sounds a little ridiculous?

[don_svetlio]

Am I the only one who thinks GTX 1050/1060/1070/1080 sounds a little ridiculous?

Titan XYZ!