Home / Component / Graphics / Nvidia’s ‘Big Pascal’ GPU reportedly taped-out, on-track for 2016 launch – rumour

Nvidia’s ‘Big Pascal’ GPU reportedly taped-out, on-track for 2016 launch – rumour

Nvidia Corp. has reportedly taped out its next-generation high-performance graphics processing unit that belongs to the “Pascal” family, according to a market rumour. If the information is correct, then Nvidia is on-track to release its new GPU around mid-2016. The company needs its “Big Pascal” graphics processor to build next-generation Tesla accelerators for high-performance computing applications and better compete against AMD on the market of consumer GPUs.

An anonymous person presumably with access to confidential information in the semiconductor industry revealed in a post over at Beyond3D forums that Nvidia had already taped out its next-generation code-named GP100 graphics processing unit. Nowadays, a tape-out means that the design of an integrated circuit has been finalized, but the first actual chips materialize only months after their tape-out.

Tape-out is the final stage of the design cycle of an integrated circuit, the point at which the artwork of the IC is sent to a maker of photomasks. Once the set of photolithographic masks is ready and verified, it is sent to a contract manufacturer of the chip, which produces the first working samples of the IC. Today’s mask sets contain 50 – 70 (up to 100) photomasks and it takes 15 – 20 hours to write a typical mask. It may take several weeks to prepare a mask-set. Production cycle of a complex FinFET processor is around 90 days from wafer start to chip delivery. As a result, it takes several months to prepare a set of photomasks and build an actual chip nowadays. Hence, if Nvidia taped-out of the GP100 in May, then the company will get the first samples of its GP100 in August. Nowadays high-volume production of chips starts between nine and twelve months after the initial tape-out.
nvidia_artworklatesteyeweb_mid
The world’s No. 1 producer of discrete graphics processors will reportedly use one of Taiwan Semiconductor Manufacturing Co.’s 16nm FinFET fabrication technology to make its “Big Pascal” GPU. Given the timeframe of the tape-out, it is highly likely that Nvidia uses TSMC’s advanced 16nm FinFET+ (CLN16FF+) manufacturing technology. According to the post, the BP100 is Nvidia’s first 16nm FinFET chip and the company has changed its approach to roll-out of new architectures. Instead of starting from simple GPUs and introducing biggest processors quarters after the initial chips, Nvidia will begin to roll-out “Pascal” with the largest chip in the family.

Nvidia’s “Pascal” architecture represents a big leap for the company. Thanks to all-new architecture, the Nvidia’s next-gen GPUs will support many new features introduced by DirectX 12+, Vulkan and OpenCL application programming interfaces. The 16nm FinFET process technology will let Nvidia engineers to integrate considerably more stream processors and other execution units compared to today’s GPUs, significantly increasing overall performance. In addition, next-generation graphics processing units from Nvidia will support second-generation stacked high-bandwidth memory (HBM2). The HBM2 will let Nvidia and its partners build graphics boards with 16GB – 32GB of onboard memory and 820GB/s – 1TB/s of bandwidth. For high-performance computing (HPC) applications, the “Big Pascal” chip will integrate NVLink interconnection tech with 80GB/s or higher bandwidth, which will significantly increase performance of “Pascal”-based Tesla accelerators in supercomputers. Moreover, NVLink could bring major improvements to multi-GPU technologies thanks to massive bandwidth for inter-GPU communications. According to Nvidia’s estimates, graphics adapters based on “Pascal” architecture should deliver two to ten times higher performance than comparable graphics processors today in peak scenarios.

NVIDIA_Tesla_K80_Dual-GPU_Accelerator_Front

Nvidia needs GP100 chip in order to build next-generation Tesla accelerators for supercomputers. Since “Maxwell” architecture (even the GM200) lacks native support for double precision (DP) FP64 computing, it cannot be used for Tesla cards. As a result, Nvidia currently offers Tesla accelerators featuring GK110 and GK210 chips, which are basically three years old. The release of the “Big Pascal” will help Nvidia to boost sales of Tesla cards for HPC applications.

The accuracy of predictions of the Beyond3D forum member could not be verified, but some of his previous posts indicate that he has access to information that is not yet public. The post in the forum on Friday was republished by 3DCenter, a web-site known for predictions in the field of graphics processing units.

Nvidia did not comment on the news-story.

Discuss on our Facebook page, HERE.

KitGuru Says: On the one hand, the rumour comes from one source without a track record and should be taken with a huge grain of salt. On the other hand, Nvidia needs “Big Pascal” to update Tesla accelerators as soon as possible. If Nvidia wants to release its GP100-based products in mid-2016, then the chip has been taped-out by now.

Check Also

Nvidia’s Titan V benchmarks show big performance increase

Last week, Nvidia surprised us with the launch of its first Volta graphics card, the …

  • Mac DeGuzman Abergas

    again lol nvidia u troll people

  • Gary ‘Gazza’ Keen

    My next rig milestone is the release of the pascal GTX1080, with everything else beiung whatevers viable and available at the time (i./e4790k vs high end broadwell CPU). I’ve been holding out on the summer 2016 prediction and don’t actually mind if it dips into 2017. NVidia are finally creating an actual next generation in Pascal as oppose to refreshes seen in Kepler and Maxwell and I’m sure it’ll be worth the wait

  • Redwoodz

    How do you tape-out on a 16nm process that has not yet had successful yields? Is this even possible?

  • Patrick Proctor

    This poster Kitguru is quoting is smoking something awful. Nvidia is waiting for 16nmFF+, and 16nmFF hasn’t taped out for anyone yet.

  • Etienne Boutet boucher

    Well if that is true, it’s will maybe make me give up my wish to buy a gtx 970 and make me wait 1 year at least…

  • Pubicbeard

    im doing the same thing, still rocking it with my EVGA 580 😀 (5 years old now i believe)

  • Etienne Boutet boucher

    Nice, me idk why but im getting hard time with my gigabyte r9 270x when oc to get decent fps

  • OOP

    They are tapping it out because they don’t need it right now. If AMD releases a GPU that is faster than TItan X and has nice features, then nVidia probably will show Pascal.

  • VNPlayer 92

    And maxwell gpu will get poor optimization like kepler when pascal release..just sayin..

  • Smigel

    WTF are you babbling about?

  • Smigel

    I’ll get a GTX 1080 too. But I’ll be enjoying the 980 Ti until then.

  • Smigel

    “and better compete against AMD on the market of consumer GPUs.”

    LMAO. Nvidia isn’t the company that needs to “better compete” at the moment. That would be AMD.

  • Jimmy Q

    tape out = chip design that is complete and ready to manufacture.
    if the design has flaws, that’s what respins (revisions) are for – to make changes. ie: A1 vs A2 silicon. A1 meaning no changes/respins had to be made. ie: GTX 970 is A1 silicon. A tape out (masks) is done every time you have a new GPU otherwise it can’t be made. coming on a new node and/or yields have nothing to do with it. you need a tape out to manufacture the chip.

  • Jimmy Q

    refreshes? lol no, more like limited paths that can be taken with 28nm. for nvidia’s 3rd gen 28nm parts to beat the 290X by 50% (with 980Ti) is nothing short of impressive at the very least. hardly a “refresh”, especially considering AMD had to start watercooling their 2nd gen 28nm dual GPU card (295×2), and single GPU parts (Fiji XT and Pro) with their 3rd gen 28nm parts. and don’t forget the incredible power savings nvidia was able to accomplish their 3rd time around.

    it almost makes you wonder why with how close we are getting to the end of using traditional silicon, that nvidia and amd don’t stick with each node for at least two GPU’s.

  • Jimmy Q

    ~50% faster than a 290X is poor optimization? damn, in that case i can’t wait for Pascals poor optimizations!

  • Jimmy Q

    I think you’re confused. a tape out is the chip design on a node.

  • Patrick Proctor

    It’s the first trial runs to see how the chip yields are as well as tune for performance/power draw. It’s the part where masks are adjusted, light frequencies are tweaked, and deposition solutions are rebalanced.

  • Jimmy Q

    that is correct.

    This is not:
    Nvidia is waiting for 16nmFF+, and 16nmFF hasn’t taped out for anyone yet.

  • Patrick Proctor

    That is correct. Not a single 16nmFF tapeout has occurred yet, and TSMC loves to brag. Some poster on kitguru known for spreading BS is not a credible source when the biggest loudmouth of all the foundries isn’t even bragging yet.

  • I wanna link to that image 🙁

  • Pubicbeard

    Most recent games are well optimized for nvidia cards, amd users are having a bad time because of that.

  • Redwoodz

    So since the node isn’t ready yet,it really doesn’t matter much,no?

  • Asasa

    Too much money?

  • Jasonelmore

    The node is ready.. 16nm already shipping at samsung and global foundries.. TSMC is not ready maybe, but they aren’t the only foundry nvidia could hire.

  • Darien

    ??? 1080p? because 4k 290X beat or close to 970 where is 50% comes from?

  • 1011101001001

    my 290x’s are powering through everything like champs. I am waiting for pascal for my next upgraDE.

  • 1011101001001

    Dont be silly. 50% is wrong.

  • Jimmy Q

    I love how you have no idea what a tilde represents….

  • Pubicbeard

    Same here, have you tried project cars when it got released? i heard it was horrendous for amd cards..

  • 1011101001001

    Oh I am sooo sorry. I ment to say around 50% is wrong. Or should I say an average of 50% is wrong. Maybe I should just be a tool and put ~50% is wrong, since it is such a well used mathematical expression on comment sections.

  • Jimmy Q

    wow. you get really grumpy when AMD doesn’t send you your check on time. relax man. have a beer or something.

  • 1011101001001

    No, I get grumpy when people like you post fanboy comments like yours and then rather than showing facts or opinions of why when somebody else says you are wrong, you instead attack grammar.
    The 980 is on average 20-25% more powerful then the 20-25% cheaper one year older r9 290x. I bet that gives you great pride in knowing you’re best friend Nvidia has a more powerful more expensive card.
    I personally will be getting pascal next year and I promise I will not come to comment sections and type thing likes “Pascal is ~50% faster” if it really is not.

  • Jimmy Q

    i attacked your grammar? are you high? Pascal? who the f*ck brought up Pascal? and finally, have you read a review saying up to 40% faster against the 290X? get help bruh. seriously.

  • 1011101001001

    wow, nice answer. got anything else you can squeeze out of my wording? or do you have something constructive to say?

  • GregoryGr

    Me too. Every year i say to myself i ll upgrade but months pass and when i can upgrade the card i want (GTX970 this time) is already about a year old so it is not worth it. I’m sitting tight with that GTX660 lol.

  • Tyrann

    What about the GTX980ti? That one is for sure ~50% faster than the 290x.

  • Tyrann

    I think Maxwell was originally made for 20nm but because that got delayed so hard, they had to put their new efficient design on 28nm. Imagine if everything went as planned, i think Maxwell would of been even more impressive than it already was. Pascal is going down 2 nodes to 16nm i think, so AMD better have something amazing to bring to the table cause next year Nvidia is going for the throat.

  • 1011101001001

    What about it. that card would compare to the r9 fury x. i can say AMD is better because the R9 290x is 50% better than a gtx 670. Doesnt make sense to compare a year newer card that is far more expensive.

  • Tsais

    My only real reason to think about upgrading is the fact that my GTX 680 doesn’t have HDMI 2.0 yet, for doing 4K @ 60Hz.

    Might still wait for Pascal though and just deal with 30Hz for now

  • Cura

    One thing is to understand people gaming with 60hz in 2015, another is 30hz, I’m shocked

  • Cura

    GTX 970 was release Sep 18, 2014
    GTX 980 Ti was released May 31, 2015

    My point being, that mid 2016 seems as the latest release time for Pascal series.
    Would not complain about a winter release, and its not entirely impossible. But there is a higher chance of it being a spring release. Late summer or early autumn is almost out of the question, because 970 came out in sept 2014. Two years with Maxwell doesn’t make sense.

    Somewhere between Q4 2015 Q2 2016 makes sense.

  • Cura

    GTX 970 was released Sep 18, 2014
    GTX 980 Ti was released May 31, 2015

    My point being, that mid 2016 seems as the latest release time for Pascal series.
    Would not complain about a winter release, and its not entirely impossible. But there is a higher chance of it being a spring release. Late summer or early autumn is almost out of the question, because 970 came out in sept 2014. Two years with Maxwell doesn’t make sense.

    Somewhere between Q4 2015 Q2 2016 makes sense.

  • Cura

    GTX 970 was release Sep 18, 2014
    GTX 980 Ti was released May 31, 2015

    My point being, that mid 2016 seems as the latest release time for Pascal series.
    Would not complain about a winter release, and its not entirely impossible. But there is a higher chance of it being a spring release. Late summer or early autumn is almost out of the question, because 970 came out in sept 2014. Two years with Maxwell doesn’t make sense.

    Somewhere between Late Q4 2015 Early Q2 2016 makes sense.

  • Tsais

    Could you elaborate?

    Yeah, there’s 120Hz and 144Hz monitors. some FPS PvP fans think this gives them an advantage.

    Granted, people’s eyes and transfer speed of their nerves are not all the same. Some people can perceive flicker in fluorescent lights (I can), others can not.

    Most people’s eye’s ability to discern individual frames is below 24Hz (so long as there’s no on-off flicker – LCDs don’t flicker), but just a change in frames. But research has shown that some people can perceive flicker or image changes up to around 70Hz.

    People who claim they have an advantage from 144Hz displays make pretty arcane arguments to support their case. Maybe they can tell a difference, but I cannot. I’d much rather have a 50″ 4K display at 60 Hz than a 27″ 1080p display at 144Hz.

    For my forgiving eyes, 30Hz works fine and 60Hz will suffice for passive 3D at half the 4K resolution, which is still twice as sharp as 1080p.

  • Tsais

    All my current games run fine. The new ones from my list that need Pascal aren’t released yet.

    So its only a “wait” in terms of how soon I can cut down my computer’s electric bill.

  • shadowhedgehogz

    Yup, will sell my 980Ti just before the next card comes out.. as long as it’s not a ripoff Titan card. 1-2 years with a high end single card is best. Not worried about Fury x since 980Ti is just as fast or faster in that DX12 benchmark.

  • How_delightful

    Yup (ProjectCARS). Me too.
    Scraggy is how that was for AMD and the 290 (especially in Crossfire).
    But, luckily, I have a Haswell i7 NVidia rig also.

    Damn Steam and the freaks for mods.

  • How_delightful

    That 660 is good for dedicated PhysX.
    I`m even thinking of using my iChiLL GTX780 for dedicated PhysX.

  • How_delightful

    I just recall being a kid, and me being the Tv Remote control. I like it that progress happens; however it happens..

  • How_delightful

    Reliability is good. Will 16nm be reliable….?
    Longevity is nice.
    I still have an X1950 Sapphire that works great… and a 32Mb Matrox Millenium card when they were king of the hill…. Heck I got an 8Mb Creative Graphics card somewhere.

    I`m just wondering; When will I get to use my iChiLL GTX780 as Dedicated PhysX (QED:- Bring on Pascal for the recent & new gamez)… New is good. But I might need another Desk for yet another PC.

  • Adrian S

    This!
    Was actually planning on buying the 970 if the price took a dive because of the new announcement. As it is now, I’d rather wing it on my current setup than to wish I had waited in 6m-1y time. It will hurt their sales, so I doubt it’s a smart move on their part to brag too much about the improvments of the next series…

  • wargamer1969

    Bottom line is a single 980 ti will last you another 2 years with Win 10 installed. People with 780/Ti series cards should wait till next Summer to upgrade as those play games now just fine for 99% of folks.

  • Peter Den Gamer

    I guess the main thing why people go with Nvidia is because Amd is behind on their new design and no one wants to pay a high elektricity bill…

  • 1011101001001

    I still don’t get this electric bill argument. If i pay $800 for a gpu, do I really care about the extra $0.45 a month on power?

  • oplix

    Pascal will allow Oculus Rift to survive in the mainstream market. Kneel down Zuckerburg.

  • I am trying to read how wide the processor is in bits like the previous 384 bits. All I see is the Big Pascal gpu.

  • Sean Johansson

    770 SuperClocked Sli on 144hz BenQ, no need to upgrade yet!

  • Sean Johansson

    Cause you own AMD gpu….that is why!

  • Etienne Boutet boucher

    thank for the reply nvidia fanboy 😛

  • DougP

    Re: allow Oculus Rift to survive in the mainstream
    But Valve & HTC might not let it. 🙂

  • Matthijs Wolthuis

    Will cpus like the GTX 950 or GTX 750 Ti get cheaper when Pascal releases??

  • Etienne Boutet boucher

    you say that because you green team fan… that is why !