Nvidia Korea CEO Expands on RTX 4080 12GB Cancellation

By Aaron Klotz
published

The details trickle out.

GeForce RTX 4080 rebranded
(Image credit: Nvidia)

According to a report by Bodnara, Nvidia's Korea CEO Kim Seung-gyu further explained why the company canceled the GeForce RTX 4080 12GB. In short, Kim Seung-gyu said that the name of the RTX 4080 12GB was "confusing" compared to the RTX 4080 16GB, because the latter offered significantly more performance. Seung-gyu explained that Nvidia normally names the GPUs so that the model name can be used to predict performance, but the large difference in performance between the two 4080 models would confuse customers.  

Nvidia's official USA announcement wasn't quite so clear, boiling down to a single sentence that said the designation was confusing without any explanation of the rationale behind the decision.

It's noteworthy that the cancellation could be self-serving: The 4080 cancellation apparently raises the entry-level pricing of the 4000-series to $1,200, and there is no indication when a lower-priced alternative will come to market. That ensures a high-priced point of entry, and will also push many customers to Ampere cards, helping rid the company of its oversupply of those models. Nvidia recently wrote off $1.32 billion for excess Ampere inventory and related reserves.

To recap, Nvidia permanently killed off the RTX 4080 12GB SKU seven days ago, admitting that the card's name was confusing. There were also severe differences in specifications and performance between the 12GB and 16GB models. Nvidia's own testing confirmed this, with the 4080 12GB and 16GB featuring a whopping 30% performance disparity between the two SKUs.

Obviously, there was far more than just a memory capacity difference. The "unlaunched" 4080 12GB used a different chip, AD104, with 7680 CUDA cores and 60 Streaming Multiprocessors (SMs), compared to the 4080 16GB using AD103 with 9728 CUDA cores and 76 SMs. The memory was also clocked higher on the 16GB card, indicating it's almost certainly using Micron's 24Gbps GDDR6X chips.

Nvidia hasn't announced a replacement plan for the RTX 4080 12GB, but we do know that AIB partner versions of the RTX 4080 12GB were already created and Nvidia will be "refunding" those partners — presumably for the packaging, which will have to change. Nvidia will eventually create a new name for these "ex-RTX 4080 12GB" cards, likely sooner than later. We expect the core design will eventually show up as an RTX 4070 or RTX 4070 Ti, though Nvidia may change the specs slightly to avoid accusations of simply rebranding what was always supposed to be a 4070 GPU.

You can read more about the Nvidia Ada Architecture, where we still have full specs on the "unlaunched" RTX 4080 12GB. The GeForce RTX 4090 meanwhile has taken the top spot on our GPU benchmarks hierarchy and now ranks as the fastest of the best graphics cards. Hopefully the replacement for the 4080 12GB will get a price cut commensurate with the loss in performance and features when it reappears.

Aaron Klotz
Aaron Klotz
Freelance News Writer

Aaron Klotz is a freelance writer for Tom’s Hardware US, covering news topics related to computer hardware such as CPUs, and graphics cards.

31 Comments Comment from the forums
  • PlaneInTheSky
    "We at Nvidia deny accusations that we got caught trying to mislead consumers. We are just clarifying a misunderstanding consumers had about our products."
    Reply
  • Kamen Rider Blade
    Who in nVIDIA marketing thought this "4080 16 GB" & "4080 12 GB" naming scheme was a good idea?

    Jensen Huang should fire them for this debacle.

    It was a unnecessary stain on the companies reputation and 100% self inflicted wound.

    The worst part is, if nVIDIA let this confusing naming scheme go on, it could've led to a long term class-action law suit for deceptive business practices.

    Something nVIDIA wants to avoid.

    We all remember the 3.5 GB VRAM controversy from back in the day and how there was a class action lawsuit over that 0.5 GB of useless VRAM.
    Reply
  • Giroro
    It makes me wonder where TechPowerUp gets the info for their GPU database. They've had a "RTX 4070" listed with the same specs as the "4080 12GB" for a long time.

    Not that the name matters much. The problem is that Nvidia thinks they can get away with charging $900 for a midrange card that is worth half that.
    Reply
  • Giroro
    Kamen Rider Blade said:
    Who in nVIDIA marketing thought this "4080 16 GB" & "4080 12 GB" naming scheme was a good idea?

    Jensen Huang should fire them for this debacle.

    It was a unnecessary stain on the companies reputation and 100% self inflicted wound.

    The worst part is, if nVIDIA let this confusing naming scheme go on, it could've led to a long term class-action law suit for deceptive business practices.

    Something nVIDIA wants to avoid.

    We all remember the 3.5 GB VRAM controversy from back in the day and how there was a class action lawsuit over that 0.5 GB of useless VRAM.

    Why would Jensen Huang fire himself? He's making money for the company and himself hand over fist. Investors love the never-ending price hikes. If the card comes back as a 4070 with a $700 price tag, then they've made it "look" like a major discount, despite a $200 price hike over last gen.

    And here's the thing. It won't come back with a significant discount. Monopolies don't work that way. They'll come back and say "well the card is the same, and now it costs even more to produce because of the rebrand. We still need to make those unheard of 75 points of profit, so be grateful we didn't raise the price even higher."
    If customers say "the RTX 4070 is a bad value, I'll buy a 4080 instead" They think that's good. They want that. From the beginning, they've designed their prices to trick you into being upsold, on purpose. A sticker on the box doesn't change the math.
    Reply
  • JarredWaltonGPU
    Giroro said:
    Why would Jensen Huang fire himself? He's making money for the company and himself hand over fist. Investors love the never-ending price hikes. If the card comes back as a 4070 with a $700 price tag, then they've made it "look" like a major discount, despite a $200 price hike over last gen.

    And here's the thing. It won't come back with a significant discount. Monopolies don't work that way. They'll come back and say "well the card is the same, and now it costs even more to produce because of the rebrand. We still need to make those unheard of 75 points of profit, so be grateful we didn't raise the price even higher."
    If customers say "the RTX 4070 is a bad value, I'll buy a 4080 instead" They think that's good. They want that. From the beginning, they've designed their prices to trick you into being upsold, on purpose. A sticker on the box doesn't change the math.
    I don't think it's coincidence that AMD is also launching RDNA 3 GPUs next month. We'll see what those bring to the table, but it's possible Nvidia got enough additional information about its main competitors plans that it knows the $899 RTX 4080 12GB wasn't going to fly.

    All indications are that AMD will have an RX 7950 XT with 24GB of GDDR6 18Gbps memory as the top SKU, with an RX 7900 XT as well that may or may not have 24GB. With significantly more shader cores, and without using a bunch of transistor space on tensor core equivalents, don't be surprised if AMD's native performance ends up being very competitive with RTX 40-series. I would be surprised if AMD goes higher than $1199 on the RX 7950 XT, which means it could already have a better price to performance ratio. Take a small step down to a $999 RX 7900 XT that still has 20GB of VRAM and the RTX 4080 12GB would look pathetic. Probably the 7950 XT will also have 192MB of L3 cache, while the 7900 XT 'only' gets 80MB.

    Maybe that doesn't happen, but right now it's absolutely possible that it does. Nvidia isn't stupid, and it might just be better to unlaunch a product that it knows will need a major price adjustment by the time it arrives. Bring out the popcorn, because next month is going to be fun!
    Reply
  • USAFRet
    As my eyes scrolled down the list of threads, I saw:
    "North Korean CEO..."

    wait, what???
    Reply
  • Giroro
    JarredWaltonGPU said:
    I don't think it's coincidence that AMD is also launching RDNA 3 GPUs next month. We'll see what those bring to the table, but it's possible Nvidia got enough additional information about its main competitors plans that it knows the $899 RTX 4080 12GB wasn't going to fly.

    All indications are that AMD will have an RX 7950 XT with 24GB of GDDR6 18Gbps memory as the top SKU, with an RX 7900 XT as well that may or may not have 24GB. With significantly more shader cores, and without using a bunch of transistor space on tensor core equivalents, don't be surprised if AMD's native performance ends up being very competitive with RTX 40-series. I would be surprised if AMD goes higher than $1199 on the RX 7950 XT, which means it could already have a better price to performance ratio. Take a small step down to a $999 RX 7900 XT that still has 20GB of VRAM and the RTX 4080 12GB would look pathetic. Probably the 7950 XT will also have 192MB of L3 cache, while the 7900 XT 'only' gets 80MB.

    Maybe that doesn't happen, but right now it's absolutely possible that it does. Nvidia isn't stupid, and it might just be better to unlaunch a product that it knows will need a major price adjustment by the time it arrives. Bring out the popcorn, because next month is going to be fun!
    Maybe AMD will really bring it this time. Nvidia rarely acts like they care what AMD is doing, though. I don't know if it was a production issue, or what, but I would think AMD would be making a lot of noise if RDNA2 had been a big boost to their market share.
    Reply
  • hotaru251
    In short, Kim Seung-gyu said that the name of the RTX 4080 12GB was "confusing" compared to the RTX 4080 16GB, because the latter offered significantly more performance.

    becasue this is 1st time you've done it?


    done it at least twice before.
    forget 1st one but 2nd was the 1060 3gb vs 6gb.
    Reply
  • Kamen Rider Blade
    Giroro said:
    Why would Jensen Huang fire himself? He's making money for the company and himself hand over fist. Investors love the never-ending price hikes. If the card comes back as a 4070 with a $700 price tag, then they've made it "look" like a major discount, despite a $200 price hike over last gen.

    And here's the thing. It won't come back with a significant discount. Monopolies don't work that way. They'll come back and say "well the card is the same, and now it costs even more to produce because of the rebrand. We still need to make those unheard of 75 points of profit, so be grateful we didn't raise the price even higher."
    If customers say "the RTX 4070 is a bad value, I'll buy a 4080 instead" They think that's good. They want that. From the beginning, they've designed their prices to trick you into being upsold, on purpose. A sticker on the box doesn't change the math.
    Anybody who thinks there is a "Price discount" coming with the rename is delusional.
    That original MSRP price for the 4080 12GB is the price that Jensen wants to set, he can do whatever he wants, but I doubt he's going to budge.

    Doesn't matter if he changes the name of the "4080 12GB" to:
    4070
    4070 Ti
    4070 SUPER
    4070 <New Suffix>

    Jensen is literally going to wear his leather jacket, put on sun glasses and tell people to "Get Over It" and accept his new pricing.

    Pray that Jensen doesn't raise the MSRP on the new video cards again.

    The price is HIGHLY unlikely to change, especially given the glut of RTX 30 series & RDNA2 cards + RTX40 series inbound & RDNA3 card inbound.

    Don't forget the used mining cards are going to flood the used Video Card market, driving prices down.

    So it's a buyers market and Jensen isn't going to give up his SWEET PHAT Profit margins.
    Reply
  • JarredWaltonGPU
    hotaru251 said:
    becasue this is 1st time you've done it?


    done it at least twice before.
    forget 1st one but 2nd was the 1060 3gb vs 6gb.
    1060 3GB was about 90% of the 1060 6GB performance, unless you happened to exceed the 3GB VRAM capacity. Which wasn't really that big of a problem in 2016 but is now quite limiting.
    4080 12GB would be 82% of the compute offered by the 4080 16GB, but 70% of the memory bandwidth. And with a 192-bit interface compared to 256-bit, it's no surprise that bandwidth can be a limiting factor.
    Reply