• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

NVIDIA Pascal GPU to feature 17B transistors and 32GB HBM2, coming Q2 2016 (earliest)

http://wccftech.com/nvidia-pascal-gpu-17-billion-transistors-32-gb-hbm2-vram-arrives-in-2016/

In an exclusive report published by Fudzilla, the site reveals that NVIDIA’s next generation Pascal GPU will feature 17 billion transistors crammed inside its core. Currently, the flagship GM200 core found on the GeForce GTX Titan X comes with 8.0 Billion transistors while the competitor, the Radeon R9 Fury X has a total of 8.9 Billion transistors inside its Fiji GPU. The 17 Billion transistors on the Pascal GPU are twice the transistors found on the GM200 Maxwell and the Fiji XT GPU core which is literally insane. Pascal is meant to be NVIDIA’s next high performance, compute focused graphics architecture which will be found on all market segments that will include GeForce, Quadro and even Tesla. Based on TSMC’s 16nm process node, NVIDIA’s Pascal GPU will not only feature the best performance in graphics but also the most power efficient architecture ever made by a GPU manufacturer.

It was revealed a few days ago that NVIDIA’s Pascal GP100 chip has already been taped out on TSMC’s 16nm FinFET process, last month. This means that we can see a launch of these chips as early as Q2 2016. Given that the transistor count is correct, we can expect a incremental performance increase from Pascal across the range of graphics cards that will be introduced.

With HBM2, NVIDIA gets access to more denser chips that will result in cards with 16 GB and up to 32 GB of HBM memory across a massive 4096bit memory interface which will dominate the next high-resolution 4K and 8K gaming panels.Although they may have to wait a little bit longer thanks to AMD’s priority access to HBM2 with SK Hynix, the makers of HBM. With 8Gb per DRAM die and 2 Gbps speed per pin, we get approximately 256 GB/s bandwidth per HBM2 stack. With four stacks in total, we will get 1 TB/s bandwidth on NVIDIA’s GP100 flagship Pascal which is twice compared to the 512 GB/s on AMD’s Fiji cards and three times that of the 980 Ti’s 334GB/s.

This is going to be a massive leap if it does ship early/mid next year. Does AMD have anything in the pipeline to counter something like this?

Lock it up if old.
 

Siphorus

Member
Was reading about this earlier, the leap with new architecture, new die shrink, new graphics API's, new WDDM, coupled with HBM2 ram bandwidth and density, the jump should be nuts. These things will sell like crazy if the 1070 or whatever its called is around 400$.
 

Kabouter

Member
That's freakin' insane. I wonder what we can expect from the normal-priced cards rather than this (which I'm assuming is a Titan thing, if it's accurate at all)
 

Kaako

Felium Defensor
17B transistors & 32 fucking GB of HBM2!? Ahahaha yaasssssssssssss give it to me now. This my next card for my next build. Freaking glorious!
 

twobear

sputum-flecked apoplexy
on the plus side, if it costs a lot of money you'll be able to take shelter under it when you can no longer afford your rent
 

Rizific

Member
Was reading about this earlier, the leap with new architecture, new die shrink, new graphics API's, new WDDM, coupled with HBM2 ram bandwidth and density, the jump should be nuts. These things will sell like crazy if the 1070 or whatever its called is around 400$.

UGHHHHHHHH. just when i had my sights on a 980ti. i dont know how much longer i can hold out on my 7950.
 

Kadin

Member
I just returned a 970 because I'm gonna wait it out a bit as I don't really need the power right now. Looks like I might have to try and wait a bit longer but good god that thing looks insane.
 

WolvenOne

Member
Okay, in all seriousness, what the heck are developers going to do with 32GB of video?! Even for 4k that sounds like overkill.
 
TSMC’s 16FF+ (FinFET Plus) technology can provide above 65 percent higher speed, around 2 times the density, or 70 percent less power than its 28HPM technology. Comparing with 20SoC technology, 16FF+ provides extra 40% higher speed and 60% power saving. By leveraging the experience of 20SoC technology, TSMC 16FF+ shares the same metal backend process in order to quickly improve yield and demonstrate process maturity for time-to-market value.

for those curious about 16nm FF

The Pascal GPU would also introduce NVLINK which is the next generation Unified Virtual Memory link with Gen 2.0 Cache coherency features and 5 – 12 times the bandwidth of a regular PCIe connection. This will solve many of the bandwidth issues that high performance GPUs currently face. One of the latest things we learned about NVLINK is that it will allow several GPUs to be connected in parallel, whether in SLI for gaming or for professional usage. Jen-Hsun specifically mentioned that instead of 4 cards, users will be able to use 8 GPUs in their PCs for gaming and professional purposes.

yeah, but what about consumer models? Because this will assuredly be for HPCs only at first.
 

b0bbyJ03

Member
holy fucking shit.. sounds like these will be the first cards that offer a true single GPU solution for current 4K games.
 

VariantX

Member
I honestly have no need of that much power as I currently only game at 1080 p. It does make me excited for their lower end stuff that comes from this architecture as it'll likely be a huge jump from what i currently use.
 

Makareu

Member
Even if it is delayed to Q3 or Q4 2016 it will probably be my trigger to dive into VR, right on time to replace my 970.
That being said 32 GO seem overkill, and I hope AMD can keep up.
 

tuxfool

Banned
Q2 is being very very optimistic.

HBM2 only enters into production in Q2, so there has to be a delay between their integration into graphics cards, as they will require the GPU, HBM2 and Interposer dies all be ready and available.

This of course presumes the Hynix slides are still accurate.
 

AJLma

Member
Sounds like another $1,000 video card. Not expecting this or it's smaller version to be cheap. If it really releases under a year from now a 980 Ti will probably still be beyond most people's needs.

Also, I'm not seeing how AMD can catch up to nVidia performance wise right now.
 
Doubt these will be for the GTX cards btw. Maybe the next Titan. HPCs and professional cards at first most likely. I'm mean just thinking about the 980 Pascal equivalent to ship with 16/32 is just insane lol
 

tuxfool

Banned
yeah, but what about consumer models? Because this will assuredly be for HPCs only at first.

Yes. Nvlink requires a completely different physical connector and as such I'm not certain how they will comply with ATX specs. Not to mention that it could only be used potentially for GPU-GPU communication.
 

mrklaw

MrArseFace
17 billion GPU transistors, or 17 billion including the ram (as everything is essentially on the same chip now)?

17 billion GPU transistors seems a big jump, although with the process node shrink the overall area could be similar or even smaller than a Titan X?
 

malyce

Member
This is beautiful. Next gen consoles shouldn't be gimped or half stepped like the current ones are. At the very least they should be able to do 4k @ 60fps by 2019.
 
Top Bottom