Nvidia: Unlaunching the 12gb 4080

[ZetZet]

7 minutes ago, LAwLz said:

And you don't think that consumer will assume a difference in VRAM will result in a difference in performance?

 

If you ask the average consumer that doesn't follow hardware "which card is faster, the one with 12GB of RAM or the one with 16GB of RAM and costs 300 dollars more" chances are they will say the latter.

What is an average consumer in this sense, this is all enthusiast level hardware. The naming is mostly a phycological issue for people since it's not fast enough for most to consider it a next-gen 80 class card. It sits in a similar price and performance bracket to last gen 3080. So people are mad.

[LAwLz]

6 minutes ago, ZetZet said:

What is an average consumer in this sense, this is all enthusiast level hardware. The naming is mostly a phycological issue for people since it's not fast enough for most to consider it a next-gen 80 class card. It sits in a similar price and performance bracket to last gen 3080. So people are mad.

When I say the average consumer I am talking about people who do not follow hardware news and barely knows what the difference is between a CPU and GPU.

The people who buy prebuilts because they are scared of putting it together themselves. The masses.

 

If we are talking about the "average consumer" as someone who follows tech news, read reviews, and actually are considering buying a separate graphics card for like 1000 dollars then I think the risk of someone getting ripped off is even lower, since the people interested in this card will probably look up reviews and from very obvious contextual clues will be able to figure out that the two cards will perform differently.

 

 

Anyway, according to the few benchmarks we got it seems to be trading blows with an RTX 3090 Ti. Is that not fast enough for a new 80 series card? 

[ZetZet]

2 minutes ago, LAwLz said:

When I say the average consumer I am talking about people who do not follow hardware news and barely knows what the difference is between a CPU and GPU.

The people who buy prebuilts because they are scared of putting it together themselves. The masses.

 

If we are talking about the "average consumer" as someone who follows tech news, read reviews, and actually are considering buying a separate graphics card for like 1000 dollars then I think the risk of someone getting ripped off is even lower, since the people interested in this card will probably look up reviews and from very obvious contextual clues will be able to figure out that the two cards will perform differently.

 

 

Anyway, according to the few benchmarks we got it seems to be trading blows with an RTX 3090 Ti. Is that not fast enough for a new 80 series card? 

I don't think the masses buy 4080s and 4090s at all. I mean these are crazy expensive systems. 

 

It's not trading blows with 3090 Ti, loses in all 3 games Nvidia posted, not to mention the insane difference in VRAM. I think it's close to a 3090, but then last gen 3090 and 3080 were super close. So 4080 being close again just seems weird. 

[Tan3l6]

As easy as CTRL-Z...

[Kinda Bottlenecked]

It's misleading for sure. AMD and Nvidia have made this move before by supplying OEM only cards with gimped performance and same branding and I don't think the sentiments were ever positive. Unless the RTX 4080 explicitly has the VRAM advertised on the front as "RTX 4080 12GB" but from what we've seen that is not the case.

[tikker]

34 minutes ago, LAwLz said:

And you don't think that consumer will assume a difference in VRAM will result in a difference in performance?

To a different extent I think than knowing it's actually a different chip that may also perform differently. Without knowing how different the chips are exactly (maybe I've missed that and we know) it's hard to say, but it leans to the side of misleading marketing for me. If performance between the chips themselves is the same and the name merely reflects a change in architectural design or whatever to accommodate the difference in memory, then it would be okayish with me, but even as an enthusiast that would look into performance differences to see what actually changed, I'm not thrilled with this confusion.

 

For me, the problem is that I see articles mentioning the 12 GB variant is 20-30% slower than the 16 GB model. By changing two variables at once, Nvidia muddies the waters a lot. How much of that performance drop is simply a consqeuence of 25% less VRAM and how much of it is the chip itself being less powerful? Does the 25% less VRAM lead to 25% less FPS? I didn't see any benchmarks with the 12 GB model on graphically light games that don't rely on VRAM a lot, so I think an interesting benchmark would be to artificially limit the 4080 16GB to only 12 GB of its memory, for example, and see how it performs then. If there's little loss of performance, or at least significantly less compared to the 4080 12GB model, then I don't think it should be called a 4080.

[BuckGup]

22 hours ago, GOTSpectrum said:

And yet again we are shown any EVGA made the right decision. Think of the man hours that have gone into designing the cards, coolers, boards ect.

 

Not even thinking about any potential stock they have our the design ready to ship 

Internally NVIDIA is even worse to deal with. It's truly a monopoly. Everyone hates dealing with Nvidia but they provide the best product so you have to let them take your balls

[leadeater]

1 hour ago, tikker said:

so I think an interesting benchmark would be to artificially limit the 4080 16GB to only 12 GB of its memory, for example, and see how it performs then. If there's little loss of performance, or at least significantly less compared to the 4080 12GB model, then I don't think it should be called a 4080.

The performance would be identical. Limiting a RTX 4080 16GB to any capacity amount above what a game needs will result in no performance change. That's the issue, VRAM capacity doesn't effect performance, not enough vram does. But you might say then it does, which points to the problem of correctness of statements, the importance of conveying information.

 

Any amount above what you need is identical regardless of how much more.

 

The problem is how VRAM capacity is added and that can cause confusion. Adding more, but not doubling, GDDR modules means you have to widen the memory controller to actually connect that many resulting in an increase to memory bandwidth. You can also double the number of GDDR modules and have two per phy so double the capacity and the same bandwidth. The final way is by using higher capacity memory modules, also resulting in the same bandwidth.

 

Above all assuming modules are all the same transfer rate.

 

Adding unnecessary confusion and reinforcing a misconception is simply not good for consumers. It's a tricky thing but also easily avoidable. 

[CerealExperimentsLain]

EVGA must be laughing their butts off, having not to recall any factory shipped 4080 12GB, reflash their vBios, print new boxes, rebrand the cards, return money to retailers...

[leadeater]

8 hours ago, LAwLz said:

But if we are being realistic, how many people would actually have been mislead?

Actually does not matter, the potential is there, it's obvious it's there. It's so easily avoidable, how many numbers exist in history so what's the need to reuse the same ones?

 

And yes plenty could easily get mislead. The potential is very high for the subset of buyers who do not build their own systems and buy a prebuilt from an OEM or SI, so you know the largest segment of the PC gaming community...

 

If we are being realistic why bother to try and not criticize this, why bother to try and make this seem like not an issue. There are just so many why's I could add to this sentence but then... why.. haha

 

So basically in short, the situation is pathetic from the outset and I see no reason at all to do anything other than point out the issue. Simply do not cede your rights and protections as a consumer just because the issue is small as that is the historic path of eroding those protections.

 

8 hours ago, LAwLz said:

Also, I don't appreciate the attempt at silencing me for having a different opinion. You're basically telling me to fuck off if I don't jump on the bandwagon.

I made the comment because you are trying to do exactly this. You are trying to silence what is by basically the majority of opinion of the community and media a valid issue. You can have a different opinion yet you keep coming here and other topics throwing shade at people for voicing their own opinion. Now not to start more of an argument but reflect on yourself and your conduct, have you really acted in the best way?

 

I'm not at all saying you should jump on the bandwagon, that comment was "stop actively trying to kick people off it" `because you don't agree with their opinion. Also calling it a "bandwagon" is itself kind of a problem...

 

I mean how is showing your indifference to the situation the same thing as jumping on the bandwagon 

 

Your actions speak far louder than your words do and they do not align with each other.

[tikker]

23 minutes ago, leadeater said:

The performance would be identical. Limiting a RTX 4080 16GB to any capacity amount above what a game needs will result in no performance change. That's the issue, VRAM capacity doesn't effect performance, not enough vram does. But you might say then it does, which points to the problem of correctness of statements, the importance of conveying information.

Any amount above what you need is identical regardless of how much more.

Yeah that was my thought. I figured VRAM wouldn't really relate to FPS directly, but since the benchmarks were at 2160p and Flight Sim 2020 (which if I remember was a title that could make actual use of a lot of VRAM) I was wondering if the performance decrease could be from a "lack" of VRAM if those titles would actually have use for more. Thought reading it back I guess it shouldn't really affect it in the sense that being able to load more into memory doesn't mean it can render it any faster.

23 minutes ago, leadeater said:

The problem is how VRAM capacity is added and that can cause confusion. Adding more, but not doubling, GDDR modules means you have to widen the memory controller to actually connect that many resulting in an increase to memory bandwidth. You can also double the number of GDDR modules and have two per phy so double the capacity and the same bandwidth. The final way is by using higher capacity memory modules, also resulting in the same bandwidth.

 

Above all assuming modules are all the same transfer rate.

Interesting. Didn't know you could have double the chips on the same bandwidth.

23 minutes ago, leadeater said:

Adding unnecessary confusion and reinforcing a misconception is simply not good for consumers. It's a tricky thing but also easily avoidable. 

Agreed. Different chip, different amount of memory, worse performance, this thing should not be called a 4080.

[leadeater]

17 minutes ago, tikker said:

Interesting. Didn't know you could have double the chips on the same bandwidth.

Good example of this is RTX 3090 vs RTX 3090 Ti both with 24GB VRAM, lets just put aside the different transfer rates of the modules used as they are very close anyway. Same memory bus width, just 2 modules per channel vs 1 module per channel.

 

RTX 3090 had back side GDDR modules

 

RTX 3090 Ti does not have back side GDDR modules

[Fasterthannothing]

Everyone in this entire thread is over thinking how they will fix their problem. If a relauch was that hard or cost prohibitive they would have never unlaunched the card in the first place. No all they will do is relabel them 4070 and move on with it. They get to use the chips, sell the 4070 in a higher price bracket while also delivering more performance then usal and most likely launching early and then have the yeilds for lower tier cards.

[leadeater]

2 minutes ago, Fasterthannothing said:

If a relauch was that hard or cost prohibitive they would have never unlaunched the card in the first place. No all they will do is relabel them 4070 and move on with it.

Not for Nvidia 

 

All the AIBs have already produced coolers with RTX 4080 12GB on them so those have to be changed. At least from what I heard in GN's video Nvidia is helping with the costs though.

[LAwLz]

6 hours ago, ZetZet said:

I don't think the masses buy 4080s and 4090s at all. I mean these are crazy expensive systems. 

 

It's not trading blows with 3090 Ti, loses in all 3 games Nvidia posted, not to mention the insane difference in VRAM. I think it's close to a 3090, but then last gen 3090 and 3080 were super close. So 4080 being close again just seems weird. 

Yes it lost, by 1% in Flight Sim and 7% in F1 22. I would say that is pretty damn close.

A Plague Tale seems to be an outlier. Not sure why but those results seemed very low on the 4080 12GB. 

 

It's not a good idea to draw conclusions based on 3 results so I'd suggest we wait for more benchmarks. But according to the data we got right now, it seems to only be slightly slower than a 3090 Ti. That is very good if you ask me.

 

 

4 hours ago, leadeater said:

Actually does not matter, the potential is there, it's obvious it's there. It's so easily avoidable, how many numbers exist in history so what's the need to reuse the same ones?

It absolutely matters if we want to gauge how bad this was and what reaction it warrants.

You can't just start talking about consumer protection laws and then when asked how many people it would actually protect you go "that's irrelevant". Laws exist to protect people. If nobody is being protected then the law is meaningless. 

 

I agree that it was a bad name. I don't dispute that. I think it would have been better if they called it the RTX 4070 or something along those lines. In my first post in the announcement thread I said they should have called it the 4070.

My issue right now, and what I want to discuss, is how much it actually matters in practice. That is what I want to discuss, and how many people would actually have been mislead by the naming is the only thing that matters in that conversation. I want to talk about the practical effects of the naming, not what someone feels would have been the right thing to do.

 

 

 

4 hours ago, leadeater said:

And yes plenty could easily get mislead. The potential is very high for the subset of buyers who do not build their own systems and buy a prebuilt from an OEM or SI, so you know the largest segment of the PC gaming community...

Like I said earlier, I strongly suspect that those people would have understood that 16 > 12 so therefore there will be a difference in performance. Especially since there is also a 300 dollar price difference.

If we start pretending like people don't understand "higher number = better" then we might as well argue that calling something an i5 or i7 will mislead people, because how are consumers suppose to know an i7 is faster than an i5?

 

Higher number = better.

The 4080 16GB has a higher number than the 4080 12GB. 16 is higher than 12, so therefore it should perform better, which it does.

 

 

4 hours ago, leadeater said:

If we are being realistic why bother to try and not criticize this, why bother to try and make this seem like not an issue. There are just so many why's I could add to this sentence but then... why.. haha

I have criticized it. I am also criticizing the people who are making a hen out of a feather.

You can be slightly against something and still call out people who overreact.

 

 

  

4 hours ago, leadeater said:

I made the comment because you are trying to do exactly this. You are trying to silence what is by basically the majority of opinion of the community and media a valid issue. You can have a different opinion yet you keep coming here and other topics throwing shade at people for voicing their own opinion. Now not to start more of an argument but reflect on yourself and your conduct, have you really acted in the best way?

Asking people to take a step back and actually think if the outrage is warranted, is very different from telling someone to shut up.

It is also a very big difference to "silence" the majority vs trying to silence the minority opinion. Not that I tried to silence anyone. Can you find a post where I told people to shut up the same way you told me to shut up?

"Tone down your reaction" does not mean "shut up" by the way. It means that instead of making like 50 posts about it which I have seen some people do, maybe they should make one or two posts about it? 

[leadeater]

12 hours ago, LAwLz said:

I strongly suspect that those people would have understood that 16 > 12 so therefore there will be a difference in performance. Especially since there is also a 300 dollar price difference.

That's assuming people actually look that far. Your average gamer right now and in the past talks about things like the RTX 3080 being a good graphics card and I simply do not think it's a safe assumption that these people are going to read past or take a lot of notice about the 12GB or 16GB being there.

 

Lets take an example, Joey the The Anime Man. He got, now I forget which model, an Nvidia graphics card by name recommendation only and knew roughly what was a good model by it's name reputation and installed it in his computer. He never plugged in the GPU power cables and never used it for a good I think 3 years? It wasn't until Connor (CDawgVA) opened his computer and had a look that it was not even being used, looked due to performance complaints with his video editing and streaming.

 

I really do think there is a huge portion of people that will only look at and care about the RTX 4080 part of the name and shrug off the 12GB and 16GB because they are still getting a "RTX 4080".

 

If the OEM/SI box/spec sheet says RTX 4080 in it then that's what they think they are getting.

 

Better to be safe than make your assumption you've been making.

 

12 hours ago, LAwLz said:

I have criticized it. I am also criticizing the people who are making a hen out of a feather.

You can be slightly against something and still call out people who overreact.

Having and creating that argument is itself making the issue exist, hence reinforcing your notion. You're being the fox and walking in to the chicken coop then complaining that the chickens are making a racket. 

 

It's not a huge issue, it still shouldn't be left alone and Nvidia should change it. So I still come back to why bother at all with trying to say anything other than this including your arguments over how the name doesn't matter (even though I think it matters more than you do). It doesn't change that the name conflict should not exist. Nvidia has agreed, the issue is resolved.

 

Story and situation now over....

 

12 hours ago, LAwLz said:

Asking people to take a step back and actually think if the outrage is warranted

If that was ACTUALLY what you did and are doing then yes warranted, yet that's not actually how you come across. What I'm saying is you have a delivery issue, what you want to say is being lost in how you are going about it.

[ZetZet]

7 hours ago, LAwLz said:

Yes it lost, by 1% in Flight Sim and 7% in F1 22. I would say that is pretty damn close.

A Plague Tale seems to be an outlier. Not sure why but those results seemed very low on the 4080 12GB. 

Plague Tale is the most graphically intense there, 4080 12GB probably ran out of vram.

[ewitte]

20 hours ago, LAwLz said:

And you don't think that consumer will assume a difference in VRAM will result in a difference in performance?

 

Why would it the model number is the same.  It shouldn't inherently have anything to do with speed.  If you have a customer that knows something about computers you can tell they typically shave down a few SMs by cutting out memory channels.  That should have a performance hit of less than 10%.  Having a completely different die (different die is a different GPU) for the same model and something like 30% performance difference is unacceptable.  If they had priced things accordingly for the performance targets and hardware used there probably would have been less backlash.  The card doesn't have $900 performance in it vs last generation even.  This is a less than 300mm die!  Plus the BOM from what I hear on that model is they have insane margins WAY over the typical 30%.

[ewitte]

6 hours ago, ZetZet said:

Plague Tale is the most graphically intense there, 4080 12GB probably ran out of vram.

Flight Sim has some really heavy CPU requirements.  There is barely any difference enabling DLSS2 (DLSS3 there is a huge difference).

[LAwLz]

20 hours ago, leadeater said:

That's assuming people actually look that far. Your average gamer right now and in the past talks about things like the RTX 3080 being a good graphics card and I simply do not think it's a safe assumption that these people are going to read past or take a lot of notice about the 12GB or 16GB being there.

Are you seriously making the argument that people don't read the entire product name, and therefore it is the manufacturer's fault someone might not get what they think they were buying?

By that same logic, we should be mad at AMD because the 7600X and 7950X performs differently. "Someone might hear Ryzen 7000 is good" and then buy an 7600X thinking it will perform like a 7950X!

 

The product name indicates that there is a performance difference since 16 is larger than 12, and the price should give a hint as well. If people don't read the product name then changing the name slightly will not make a difference. People like what you are describing won't be any more protected by changing the name from 4080 to 4070 either.

 

 

 

7 hours ago, ewitte said:

Why would it the model number is the same.  It shouldn't inherently have anything to do with speed.

That is not what the general consumer will think.

If you tell them "which is faster, a computer with 12GB of RAM and a computer with 16GB of RAM" they will probably pick the one with 16GB of RAM.

The people who know more, the ones you describe that will look up the die used, how many SM units are on the GPU etc, they will know that there is a performance difference other than the VRAM as well.

 

 

7 hours ago, ewitte said:

The card doesn't have $900 performance in it vs last generation even.

How did you come to that conclusion?

We only have a few benchmarks but in those it is fairly close to a card that up until a few weeks ago cost ~1500 dollars in stores.

I think it is very silly to say it doesn't have "X dollars worth of performance" when we don't know how it performs and what it costs. At this point you are mad about things that only exists in your head and haven't been confirmed.

[NumLock21]

If Nvidia were to relaunch it as a different card name, it better not be a RTX 4070 with that narrow memory bus of 192bit. Their cards ending with XX70s and XX80s, are always 256bit. 192bit belongs to their XX60s.

[Mihle]

4 hours ago, NumLock21 said:

If Nvidia were to relaunch it as a different card name, it better not be a RTX 4070 with that narrow memory bus of 192bit. Their cards ending with XX70s and XX80s, are always 256bit. 192bit belongs to their XX60s.

I don't think that matters, I think the performance and chip is what matters.

[LAwLz]

5 hours ago, NumLock21 said:

If Nvidia were to relaunch it as a different card name, it better not be a RTX 4070 with that narrow memory bus of 192bit. Their cards ending with XX70s and XX80s, are always 256bit. 192bit belongs to their XX60s.

The bus width should not determine what the name of the product is. Especially not when they have reworked the cache the way they have.

The bus width is pretty irrelevant. Just like how horse power does not necessarily determine how fast a car is around a track. 

 

 

The 3060 Ti and 3080 Super has the same memory bus width. Does that mean Nvidia should have called the 3060 Ti a 3080 class card? Of course not, because bus width is only a tiny part of the equation that determines performance. On its own it is fairly useless and not at all an indication of performance.

[Arika]

On 10/16/2022 at 9:13 AM, leadeater said:

If the OEM/SI box/spec sheet says RTX 4080 in it then that's what they think they are getting.

 

Better to be safe than make your assumption you've been making.

welp

 

 

[My_Computer_Is_Trash]

3080>12Gb 4080