NVIDIA GeForce header image

NVIDIA’s Ampere GPUs will be based on TSMC’s 7nm node, next-gen Hopper lineup might use Samsung’s 5nm process

Written by Metal Messiah

In a new report posted by ChinaTimes it has been outlined that Nvidia’s upcoming “Ampere” GPUs will indeed use TSMC’s 7nm process node technology, while the next-generation “Hopper GPU family would likely be based on Samsung’s 5nm EUV process node.

There have been some rumors circulating before as well as to whether Nvidia will tap TSMC’s 7nm process node to manufacture the AMPERE lineup of cards.

The report suggests that Nvidia’s Ampere GPUs for the HPC segment, data center, and most probably even for Gaming, will be based on TSMC’s advanced  7nm manufacturing process.

However, there’s also a chance that “Ampere” is just a logical successor to VOLTA and will only be used and implemented on the company’s professional-grade enterprise GPUs, and not the gaming consumer cards which are expected later this year. But this has not being fully confirmed yet.

As soon as this news came out, Samsung also released a statement in its Q1 results report confirming that it will begin production of its 5nm EUV process (5LPE) most probably by Q2 2020.

According to this new “ChinaTimes” report, Nvidia will be the biggest customer of TSMC’s 7nm process node. The article further alleges that the new Ampere lineup from Team Green will be announced in may itself, with a sooner than expected launch date.

Meanwhile, AMD is going to leverage a more mature form of the “7nm process” for its upcoming Ryzen 4000 desktop CPUs, as well as the Navi 2x GPUs.

ChinaTimes also alluded to another upcoming GPU from Nvidia. Ampere’s successor, HOPPER, was also mentioned. Hopper is rumored to use multi-chip modules (MCM).  Based on previous rumors it seems likely that the next-generation Hopper GPUs from nvidia are going to use Samsung’s 5nm EUV process node.

Samsung has already received orders for processors, graphics cards, and several other products based on its “5nm” process node. According to the report, most of these products are going to be used in consumer and end user applications.

Nvidia is slowly preparing to launch the RTX 3000 series in third quarter, as AIBs also quickly try to clear the current inventory of cards for Ampere.  For now though, it seems that TSMC has won another round.

However, it should be noted that TSMC also has its own EUV process. In october, TSMC began mass production of its new N7+ line. One takeaway from all this is that “EUV lithography” is surely going to play a major role in the upcoming node fabrication prcoess.

45 thoughts on “NVIDIA’s Ampere GPUs will be based on TSMC’s 7nm node, next-gen Hopper lineup might use Samsung’s 5nm process”

      1. I knew it must be a scientist or mathematician. That is what Nvidia likes to name their generations for a long time now. The name just struck my funny bone.

        Didn’t mean any disrespect to her or her significant accomplishments.

        1. “” Didn’t mean any disrespect to her or her significant accomplishments.””

          Hey, obviously I know you didn’t mean to disrespect her in the first place. Just curious about the name.

          Hopper does actually sound a bit strange when it comes to GPUs. I could be wrong though.

          1. Metal Messiah, your helpfulness and heartful, sincere comments makes me feel good, man. Thanks for this amazing article and thank you for helping us tech simpletons with helpful comments without contempt or humblebrag. Much appreciated.

      2. “It is named after a lady scientist, similar to Turing…”

        Are you suggesting that Alan Turing was a “lady” due to his sexual orientation? What filthy bigotry is this??

          1. For someone named “TrollHunter” I’d expect you’d have caught on that this preposterous statement of mine was a troll.

          1. Just trolling you man 🙂

            I’m new to this trolling thing and clearly I suck at it

    1. The S20 can come with two chips, depending on what region you are in. The Exynos does fall behind and should be avoided if possible, but it isn’t horrible. But, it should not be the same price. The 865 version is rather good.

    1. One could also assume it is a dual GPU graphics card, or at least one with a GPU of twice the size.
      Should be a beast if it scales properly.

    2. Just a name for the upcoming next-gen RDNA 2 based GPUs.

      It’s just a successor to the current RDNA architecture, and it has been labelled as NAVI 2X, aka Big Navi for the time being.

      1. Did they say when will they release it because it seems to me that Raja”s damage is still in effect and they still didn”t release a reliable driver or a card for a while

        1. No. There is no such word though. Previously RDNA2 was scheduled for Q2/Q3 2020, but because of this COVID-19 pandemic things might get delayed even more.

          It’s kind of hard to predict anything right now given the current state of global economy worldwide.

  1. I’ll be excited about Nvidia once they stop wringing their customers of more and more money every generation.

    1. I doubt nvidia will stop milking customers even with their RTX 3000 series of cards. I will be damned, if they really lower down the prices and make it more competitive tho.

          1. I am saying that the cost of admission goes down with AMD every gen, especially with their CPUs.
            You can game on a 100 dollar CPU now. No issues. The same can be said for their GPUs. 100 bucks and you can play, 1080p/60 mostly.
            Their cards continue to bring down Nvidias costs. Remember Nvidia dumping their prices when the 5700 and 5700xt? They could have launched higher and matched Nvidias price/performance, but they chose to go more affordable.
            You can thank AMD for current Nvidia prices. Sadly, they don’t have anything to compete with the 2080 up, which is why those prices are absolute theft.

          2. AMD’s recent move with their Ryzen CPUs has also led INTEL to re-think on their pricing strategy, and as we can see Intel was forced to lower down the price of its 9/10’th GEN CPU lineup.

          3. The shift we have witnessed in Intel’s pricing structure has been nothing short of amazing.

          4. “They could have launched higher and matched Nvidias price/performance, but they chose to go more affordable.”

            They don’t choose to go more affordable. If they go with prices to NVIDIA level, they would sold much less GPUs. Even now NVIDIA GPUs with higher prices are selling better. Do you really think, that AMD would choose to be paid less if they have another choice? These are commercial companies which want to earn money. 🙂

          5. I think you’re forgetting that AMD did launch their prices first. They could match their pricing and still have the same performance/dollar ratio, assuming Nvidia doesn’t drop (not their style).

          6. HD 5700 series. Damn those cards were good but that was the last time AMD gave Nvidia some competition. Wish they would do something and knock those prices down from Nvidia

    2. I am sure the company you work at reduces its prices every year and does its very best to ensure that its revenue and profits are lower every year.

        1. There is – sadly- no competition in the high end market almost since the ATI 7xxx series.

      1. You may have a hard time getting your head around this, but we don’t charge for anything. Completely not-for-profit.

  2. Man I am so ready for a new 3080 Ti + the 3000 series, My 1080 Ti has served Me well and has been / is amazing.

    I really hope on May 14th out minds will be blown and everything will be lined up for all of use to benefit. Let’s Go! + fingers crossed.

  3. Nvidia will be the biggest TSMC 7nm node customer? I thought they said there is nothing left for nvidia since all the 7nm capacity left by big player like apple/qualcomm/huawei goes to AMD.

    personally i still think TSMC will be nvidia major foundry going forward. Samsung probably will continue to be nvidia second. IMO To make nvidia do the complete switch samsung needs to prove they can at least make 800mm2 gpu with with satisfactory yield according to nvidia standards. TSMC and nvidia has been doing this “pushing the limit” for over a decade.

  4. “there’s also a chance that “Ampere” is just a logical successor to VOLTA ”

    Nope. 0% chance.

Leave a Reply

Your email address will not be published. Required fields are marked *