Nvidia is at it again: 3060 8GB released, up to 35% slower than its 12GB counterpart

[porina]

12 minutes ago, leadeater said:

There is exactly zero information from Nvidia that this product performs any different and without this Hardware Unboxed review nobody here would even really know about it.

I do wonder if it is more of an OEM targeted product, thus the absence of a regular product launch. It is early days and remains to be seen how this evolves. I'd expect pricing to stabilise and it'll find its place between its performance neighbours. I see one major UK seller have it listed (no marketplace) at £320, compared to £339 cheapest 12GB model at same site, with many 12GB models costing more.

 

I see mention of its existence on nvidia's web site, but like you said, they didn't break out perf numbers for the two models. I get your point that is lacking, but it is probably more a case of the graphics were created when 12GB was the only model and hadn't been updated yet.

 

I haven't, and wont watch HUB content, but I see their Techspot written equivalent is now up. Still it seems to fit my thinking, perf sits between 3050 and 3060 12GB, and assuming true price matches it is a logical progression to fit in.

 

10 minutes ago, leadeater said:

The RTX 30 series mobile GPUs already do not follow the desktop models. Look higher up the mobile product stack, they do not follow the desktop at all.

Ok, I see at 3080 tier they diverge, but I only looked to 3070 since that was the point I was buying at previously.

[porina]

10 minutes ago, LAwLz said:

It seems weird that I am only able to find a single OEM selling it, in a single store. I would file this under "some board partner doing something weird", like we have seen several times in the past. That of course can change if the situation changes. But right now I don't see any issue.

Not doing a massive search by any means, in UK I can see some major sellers with some models:

OCUK: Gainward

ebuyer: Palit

 

AWD-IT: Gigabyte (I don't know this seller)

 

Need to be extra careful not to look at Ti results as I search for 3060 8GB.

[leadeater]

8 minutes ago, porina said:

I do wonder if it is more of an OEM targeted product, thus the absence of a regular product launch.

Probably is, or a South East Asia target product since the video features a Galax branded card and I believe that is their major market. Wouldn't be the first time Nvidia has released a off cast weird VRAM model for that market, problem is they don't only stay in that market and neither do I think any market or OEM system buyer should be subjected to this types of obfuscation and lack of transparency.

 

Nvidia not wanting to call it an RTX 3050 Ti because they didn't actually want a global release of the product is all well and good for them and fits within their reasoning, if that is the situation. I just do not find it acceptable as per above reasons.

 

8 minutes ago, porina said:

I see mention of its existence on nvidia's web site, but like you said, they didn't break out perf numbers for the two models. I get your point that is lacking, but it is probably more a case of the graphics were created when 12GB was the only model and hadn't been updated yet.

It never will be updated, like they never did in the past for other similar model variants. And it is not acceptable if it's updated later either. Product information needs to be available when the product itself is.

[leadeater]

So other than Galax featured in the video both Asus and Gigabyte also sell these RTX 3060 8GB from what I can find.

 

https://www.pbtech.co.nz/product/VGAGBV331617/Gigabyte-GeForce-RTX-3060-Gaming-OC-Graphics-Card

https://www.pbtech.co.nz/product/VGAASU33060/ASUS-DUAL-RTX-3060-OC-Graphics-Card-8GB-GDDR6-PCIE

 

I can't find a Gigabyte product page for it though, Asus at least has one: https://www.asus.com/motherboards-components/graphics-cards/dual/dual-rtx3060-o8g/

[LAwLz]

15 minutes ago, leadeater said:

There is an RTX 3060 product page yes, obviously there is. I suggest you go take a look at it. Look all over it at all the marketing information and around the performance of the RTX 3060 and tell me if you can see two different RTX 3060's on it showing different performances.

 

The product page does have the specifications for both variants, yet only shows the information and performance for the original 12GB

I see.

There is a product page for the "RTX 3060 family" and it includes the 3060 Ti, the 3060 12GB and 3060 8GB.

I agree that it is bad that Nvidia haven't updated the first party benchmark charts to show the difference between the 8GB and 12GB model, but realistically someone with next to no clue about hardware will assume the 12GB model is faster than the 8GB model since the number is larger. 

 

 

19 minutes ago, leadeater said:

.It's an official Nvidia model variant that is offered by multiple AIBs, could well be all of them.

I know Nvidia has officially acknowledged its existence, but this launch seems more like an OEM product to me than something that is going to be widely available. 

I couldn't find a single store in Sweden that sells or even lists them. I am far more lenient when it comes to lacking information and announcements regarding OEM parts than parts intended for general consumers to order.

 

 

28 minutes ago, leadeater said:

And if nobody had bothered to buy the card and review it? You know since Nvidia didn't notify reviewers nor send out review samples. Relying on the diligence of reviewers to spot these things and put in the effort to investigate is acceptable situation to you?

I addressed that in my post, in the part you cut off.

37 minutes ago, LAwLz said:

If I couldn't find any benchmarks then I would have been weary and gone with the product I know how it performs, which would have been the 12GB model.

 

I am not sure why you cut off my post only to force me to repeat myself. If we are going to have a conversation I'd prefer it if you don't do things like that. It just makes the conversation annoying and repetitive. 

[leadeater]

1 hour ago, Fasterthannothing said:

If it has less ram it's slower end of story

Please do explain the identical performance of the 6950 1GB vs 2GB then? Or the multiple other similar examples in past and future generations for GPUs that feature the exact same GPU die configuration and only different in VRAM capacity.

 

Surely you have the self awareness to admit that this statement is not correct. 

[leadeater]

32 minutes ago, LAwLz said:

I addressed that in my post, in the part you cut off.

Well I'm sorry but I will never agree that less VRAM = less performance, should mean this, be normalized or anything otherwise.

 

I won't address it because as I've pointed to much earlier, this point has been address thoroughly to death.

 

RTX 3060 8GB and RTX 3060 12GB must have the same performance otherwise change the damn model name. It's not hard.

 

32 minutes ago, LAwLz said:

I am not sure why you cut off my post only to force me to repeat myself. If we are going to have a conversation I'd prefer it if you don't do things like that. It just makes the conversation annoying and repetitive. 

Because in my post I said there would have been no benchmarks otherwise yet you went in on the I'd look at benchmarks line which would have been impossible which is a big part of the entire point and why this type of thing is a problem.

 

You cannot claim you should inform yourself if that information is being withheld from you.

 

And anyway just because you would have doesn't mean everyone would. An RTX 3060 is an RTX 3060, wait but is it? But the reviews just say RTX 3060, this is the same thing right? Just with 8GB. There is any number of situations where things would not turn out like you'd say they would and the most easy way to avoid that is a different model name altogether. It is the most safe, consumer first way to do it.

[porina]

2 minutes ago, leadeater said:

Please do explain the identical performance of the 6950 1GB vs 2GB then? Or the multiple other similar examples in past and future generations for GPUs that feature the exact same GPU die configuration and only different in VRAM capacity.

For those cases where ram bus does not change with capacity (and assuming same speeds), then If we say you keep within the lower ram limit the two would perform the same. Then there would be no point in the higher ram version existing if that did not also offer the capacity advantage. Pick a setting the upper card satisfies where the lower does not, and you will see a performance difference. On average, even if the ram bus and clocks were the same, you could argue more is faster. Then we get into arguments about performance where users would reasonably be expected to use the GPUs, rather than max everything.

[LAwLz]

11 minutes ago, leadeater said:

Well I'm sorry but I will never agree that less VRAM = less performance, should mean this, be normalized or anything otherwise.

 

I won't address it because as I've pointed to much earlier, this point has been address thoroughly to death.

 

RTX 3060 8GB and RTX 3060 12GB must have the same performance otherwise change the damn model name. It's not hard.

But I mean... The model names are different. They are just as different as for example 3080 vs 3080 Ti.

Or if we want to bring up another product from Nvidia that I think is way more confusing than this, the GTX 1660 vs GTX 1660 Super vs GTX 1660 Ti.

 

Anyway, the general public will assume that something with a higher number, be it RAM, VRAM or product name, will perform better. If the argument is that the average Joe will be fooled, then we also have to think as an average Joe.

The average Joe will see one card has 12GB of VRAM and the other has 8GB of VRAM, and will assume that the 12GB model is faster because it has more "stuff" (the stuff being VRAM, which I don't think the average Joe will even understand what it is).

 

You don't have to look far to find posts from me arguing that people overvalue and overstate the importance of VRAM. I 100% agree with you that if everything else is equal, a card with 8GB of VRAM will perform just as well as a card with 12GB of VRAM unless you run into a situation where more than 8GB of VRAM is required. But that is how you and I think. Both you and I will research products and would notice things like the difference in memory bus width and possibly benchmarks showing the difference. So we are not affected by this because we would have found the truth to begin with.

The people who may fall for this are the average Joe who has very little understanding of computer hardware, and people like that will see 8GB vs 12GB and assume the 12GB card is faster.

 

They will come to the correct conclusion for the wrong reasons.

 

 

The only person I can think of that would be negatively affected by this is some weird person that I don't think exists. A person who are interested in buying off the shelf hardware and understands how memory buffers work and affect performance, but isn't interested enough in knowing what his 400 dollar computer part performs like before ordering it.

 

 

 

  

11 minutes ago, leadeater said:

Because in my post I said there would have ben no benchmarks otherwise yet you went in on the I'd look at benchmarks lines which could have been impossible which one a big part of the entire point and why this type of thing is a problem.

 

You cannot claim you should inform yourself if that information is being withheld from you.

I don't understand what you mean. I explained what I would have done in that situation. I am genuinely confused by this whole conversation.

You: How would you know the difference?

Me: I would have looked at benchmarks. If benchmarks didn't exist for one product then I would have picked the product that had benchmarks. Better safe than sorry.

 

Benchmarks for the 12GB model exists, and have existed for a long time. If I was unable to find benchmarks for the 8GB model then I would have been wary and bought the 12GB model because I had third party verification of how it performs. I would not buy a 400 dollar product without seeing third party benchmarks.

If I can't inform myself then I am not buying it.

 

 

16 minutes ago, leadeater said:

And anyway just because you would have doesn't mean everyone would.

True. But I don't really see any person being in a situation where this would cause any issue. Again, the average Joe will mist likely end up with the right conclusion anyway.

 

17 minutes ago, leadeater said:

An RTX 3060 is an RTX 3060, wait but is it?

It isn't, just like an RTX 3080 isn't always an RTX 3080. It might be an RTX 3080 Ti.

 

18 minutes ago, leadeater said:

But the reviews just say RTX 3060, this is the same thing right? Just with 8GB.

Now you are once again assuming that people understand how frame buffers work and won't just assume that a card with 12GB of VRAM will be faster than one with 8GB of VRAM. I think 90 people out of 100 will just assume the RTX 3060 12GB is faster than the RTX 3060 8GB when presented with nothing but the texts "RTX 3060 12GB" and "RTX 3060 8GB".

 

20 minutes ago, leadeater said:

There is any number of situations where things would not turn out like you'd say they would

Such as? I am having a hard time imagining those situations.

 

 

20 minutes ago, leadeater said:

and the most easy way to avoid that is a different model name altogether. It is the most safe, consumer first way to do it.

Would that really be simpler? Maybe I am thinking of someone else, but I am fairly sure both you and I in the 4080 thread said that we think the product name should be determined by the GPU die. In this case, both cards have the same die (at least as far as cores and clock are concerned). The difference is in the memory configuration. Not sure if they physically altered the bus width on the die itself or if they are just using fewer memory chips and that's what caused the reduction is bus width. In either case, I think there is a fairly strong case for calling both of these the RTX 3060.

I think it would be weirder to have a card called the 3050 Ti have the same CUDA core config as the 3060, but that's probably because I think the name should be dictated by the die name rather than memory config.

[WereCat]

1 minute ago, LAwLz said:

But I mean... The model names are different. They are just as different as for example 3080 vs 3080 Ti.

Or if we want to bring up another product from Nvidia that I think is way more confusing than this, the GTX 1660 vs GTX 1660 Super vs GTX 1660 Ti.

 

 

Once again. That's wrong. 

 

 

 

 

 

see the difference in names? 

[leadeater]

3 minutes ago, porina said:

For those cases where ram bus does not change with capacity (and assuming same speeds), then If we say you keep within the lower ram limit the two would perform the same. Then there would be no point in the higher ram version existing if that did not also offer the capacity advantage. Pick a setting the upper card satisfies where the lower does not, and you will see a performance difference. On average, even if the ram bus and clocks were the same, you could argue more is faster. Then we get into arguments about performance where users would reasonably be expected to use the GPUs, rather than max everything.

A lot of the time the higher VRAM capacity variants did indeed make no sense to exist. However there are cases where one or a few games at much higher resolutions did get a benefit from it or the other situation is much later in life, many years, the larger VRAM capacity did not suffer from insufficient ram degrading performance.

 

Like I said before within the VRAM buffer capacity performance is identical and if someone really wants to dive in to what is the difference that type of information does exist. However if they do not the likelihood of being adversely effected is very low. It's not good that someone might buy the higher one spending more getting no benefit realistically but at the very least the performance and product specification matches what can be found online in reviews as well as official product pages.

 

This detail difference matters because when GPU core configuration is different, or memory bus, the performance will always be different no matter the setting. You don't need to go to 4k or 8k or Eyefinity or some other much more rare and bespoke setting that might only apply in a few games to actually get a performance difference.

 

Normalizing incorrectness and misunderstanding is just bad all round. Less VRAM does not always equal less performance always in all situations. Making people think this is actually the case is just bad.

[leadeater]

16 minutes ago, LAwLz said:

But I mean... The model names are different. They are just as different as for example 3080 vs 3080 Ti.

Now we are just going to re-do already existing discussions. No this is not same thing, at all. You want people to disregard existing common knowledge and understanding about established naming conventions and standards always assuming something on the end they don't fully understand must always mean something like a performance difference.

 

This is reinforcing the mindset of making bad, ill-informed assumptions that I simply cannot support. Nor do I think everyone is going to treat it like you think.

 

To a lot of people that have some understanding of Nvidia model naming an RTX 3060 will always be just that irrespective of what is slapped on the end unless it fits within a known model designation and there are matching products people talk about and review i.e. your 3080 vs 3080 Ti.

 

There is no basis for 8GB or 12GB to signify a single thing in regards to Nvidia product model names, it's just something on the end that might mean something. I know what it means, you know what it means, I know it doesn't just mean VRAM capacity like I think it should and nothing else. I don't assume everyone thinks like you or me.

 

And I simply cannot agree that RTX 3050 Ti is functionally equivalent to RTX 3060 8GB and not more clarifying based on the above reasoning. I've repeated this once just for you since I know you have not read the whole topic.

 

Anyway the fundamental point is established and understood naming actually matters, more than maybe you want to believe. And it doesn't excuse how simple it would have been to avoid naming conflicts like this, that alone is a big problem to me. How easily avoidable it is.

[LAwLz]

12 minutes ago, WereCat said:

Once again. That's wrong. 

 

-images-

 

see the difference in names? 

Just because the box has a particular layout does not mean that's the official name of the product.

When I go to for example Newegg to order a GPU, the product name includes the VRAM amount. That matters way more than what the box looks like, because you only get the box after you have bought the product (when buying online, the reverse is true for buying things in physical stores).

 

The product name on stores like Newegg.com, which includes the VRAM amount, is far more important than what the box looks like.

[leadeater]

17 minutes ago, LAwLz said:

Just because the box has a particular layout does not mean that's the official name of the product.

When I go to for example Newegg to order a GPU, the product name includes the VRAM amount. That matters way more than what the box looks like, because you only get the box after you have bought the product (when buying online, the reverse is true for buying things in physical stores).

 

The product name on stores like Newegg.com, which includes the VRAM amount, is far more important than what the box looks like.

Well I'm so glad to know that ZOTAC GeForce RTX 3080 AMP Holo 12GB LHR and ZOTAC GeForce RTX 3080 Trinity OC 12GB LHR are entirely different models of Nvidia GPUs and that  AMP, Holo, Trinity and OC all each means something and all signify performance difference.

 

Now throw in every other AIB bollocks in to this. Somehow I think the most important part comes at the start and is what a lot of people mainly look at, just a suspicion I have.

 

If you want to go down the path you keep trying then you cannot ignore all the other crap slapped on graphics cards names.

[Fasterthannothing]

1 hour ago, leadeater said:

Please do explain the identical performance of the 6950 1GB vs 2GB then? Or the multiple other similar examples in past and future generations for GPUs that feature the exact same GPU die configuration and only different in VRAM capacity.

 

Surely you have the self awareness to admit that this statement is not correct. 

I mean I can point out high resolution loads where more RAM gives more performance?? I'm not understanding what your trying to prove. Not to mention it's not just about the capacity in this case it the ram speed the same thing that AMD did with the R7 250 ddr3 and ddr5 additions 

[leadeater]

14 minutes ago, Fasterthannothing said:

I mean I can point out high resolution loads where more RAM gives more performance??

So can I, I already have. But this is not what you said was it.

 

3 hours ago, Fasterthannothing said:

If it has less ram it's slower end of story 

You know very well the story is much longer and more complicated than this. You'd be more accurate in such a statement if you said smaller memory bus then it's slower, but even that depends on other things like cache sizes.

 

14 minutes ago, Fasterthannothing said:

I'm not understanding what your trying to prove

That equating performance to VRAM capacity is wrong and will always be wrong. The actual reason these products are slower is not the VRAM capacity. Signifying them via VRAM capacity designations will forever be reinforcing misinformation. 

 

14 minutes ago, Fasterthannothing said:

Not to mention it's not just about the capacity in this case it the ram speed the same thing that AMD did with the R7 250 ddr3 and ddr5 additions 

Well it's a good thing I already address that with my first reply to you today about it. I've told you already when it's acceptable to only designate the capacity and no other situation.

[ouroesa]

On 12/3/2022 at 2:04 PM, starsmine said:

The names for the products are different. Dont perpetuate the false idea that the name of the product stops after 4 digits. 

3060 8GB is a different name then 3060. more then half the time, its always been more then just a vram ammount difference.
AMD does this to. its literally different names for different products.

like the original 3060 was never called just a 3060 now was it. It was called a 3060 12GB, it literally always was. 
 

I mean, its no different then prebuilts that just say i7 or i5 with no model number. 
I can brag about my i7 till the cows come home, the lowest end 12th gen i3 beats my i7 in literally everything.


Honestly, I really think Nvidia should have just made a 3060 6GB with same bus myself. the 12GB thing was so stupid.
And it would have been more clear to call a card like this a 3050ti and take out an SM since its not like it can use it anyways with the bandwith limitations.

'Then' and 'than' is different. So is 'too' and 'to', just like RTX 3060 and RTX 3060

[WereCat]

8 hours ago, LAwLz said:

Just because the box has a particular layout does not mean that's the official name of the product.

NVIDIA has full control over the branding of their cards even on OEM boxes. What's on the box is the official name. If they did not wanted to be misleading they would put the VRAM amount directly next to the card name. The box and branding on the box is made so that your eyes are immediately drawn to the name of the card and it's branding. There is no VRAM amount in that area on the box. They do this because it works. If they can put (Super / Ti) next to the name they can as well put in the VRAM amount at the very least.

 

8 hours ago, LAwLz said:

When I go to for example Newegg to order a GPU, the product name includes the VRAM amount.

Yes, and what do you see when you go directly to Microcenter and walk between shelves of RTX 3060 cards?

 

Even if the VRAM amount is there on the box next to the name of the card and the difference between them in performance is this big, it's still wrong! At the very least it's a bit less elusive but it's still wrong! I can somewhat understand choosing less VRAM that's clocked higher or lower and keeping the branding but cutting down memory bus that affects performance this much is just ridiculous.

[Poinkachu]

18 minutes ago, WereCat said:

NVIDIA has full control over the branding of their cards even on OEM boxes. What's on the box is the official name. If they did not wanted to be misleading they would put the VRAM amount directly next to the card name. The box and branding on the box is made so that your eyes are immediately drawn to the name of the card and it's branding. There is no VRAM amount in that area on the box. They do this because it works. If they can put (Super / Ti) next to the name they can as well put in the VRAM amount at the very least.

 

Yes, and what do you see when you go directly to Microcenter and walk between shelves of RTX 3060 cards?

Based on the pics I've seen, I just wanna add that it looks kinda like the small prints on a contract.
I'm a bit into designs, so I agree that they designed it so that sight will be "pulled" towards the Brand & Series name, then to model number.

[suicidalfranco]

Is this a storm in a cup type of thread

[leadeater]

2 hours ago, suicidalfranco said:

Is this a storm in a cup type of thread

Yes 

[LAwLz]

11 hours ago, leadeater said:

Well I'm so glad to know that ZOTAC GeForce RTX 3080 AMP Holo 12GB LHR and ZOTAC GeForce RTX 3080 Trinity OC 12GB LHR are entirely different models of Nvidia GPUs and that  AMP, Holo, Trinity and OC all each means something and all signify performance difference.

 

Now throw in every other AIB bollocks in to this. Somehow I think the most important part comes at the start and is what a lot of people mainly look at, just a suspicion I have.

 

If you want to go down the path you keep trying then you cannot ignore all the other crap slapped on graphics cards names.

Now you're being silly.

In your example you didn't change any of the numbers, while the real life products in this case do.

 

But you do bring up a good point. If we go by the logic that "it's not the GPU die that matters, it's performance", then should we also start changing the numbering on OC models of cards?

I think it becomes very messy very quickly if we start arguing that "3080" and "3080 OC" is misleading because they perform differently, and we can't expect customers to understand what OC means in terms of performance (just like we apparently can't expect consumers to understand that a model with 12GB after it performs better than one with 8GB after it".

 

 

Yes, GPU SKU names can be long and confusing, but if the argument is that "consumers shouldn't have to read more than exactly 4 numbers" then we run into issues where it is apparently deceptive and evil for Nvidia to call their GPUs things like "Ti" as well, or AMD calling their GPUs "XT".

 

I don't get why "RX 6800" vs "RX 6800 XT" and "RTX 3080" vs "RTX 3080 Ti" are apparently super easy to understand and not at all misleading. But "RTX 3060 8GB" vs "RTX 3060 12GB" is somehow too much for consumers to take in and we can't expect people to read the full name, just the first 4 numbers.

 

 

If the argument is that consumers only read the the first 4 numbers, why are you not mad at AMD for doing the whole "XT" thing? Or Nvidia for doing the whole "Ti" thing?

[leadeater]

27 minutes ago, LAwLz said:

Now you're being silly.

In your example you didn't change any of the numbers, while the real life products in this case do.

I didn't change any of the numbers? Those are both real life 100% legitimate products taken directly from the retail website selling them. Why would I change them? I'm giving you something that actually exists....

 

So you tell me exactly how changing 12GB to another number would make any difference at all if you didn't really know any of the mess other than RTX 3080?

 

 

Edit:

27 minutes ago, LAwLz said:

If the argument is that consumers only read the the first 4 numbers, why are you not mad at AMD for doing the whole "XT" thing? Or Nvidia for doing the whole "Ti" thing?

Fortunately I've already said such a thing is dumb in this very topic.

 

But as I already said also, at least those are established model designations that have a proper meaning and significance that goes across multiple generations. Random VRAM numbers at the end do not. They also get added after the fact after all the information and understanding about the original product has been imparted in to the market and consumers. Slapping 8GB on the end and fundamentally changing the product is equivalent to re-writing history, very hard to do and shouldn't even be done.

 

Edit 2:

Oh and if you really do want to argue VRAM capacity signifies both performance and model difference and sufficient to inform buyers then put these 3 GPUs in order of performance: GTX 1060 3GB, GTX 1060 5GB & GTX 1060 6GB. Hint they are not in performance order currently.

[leadeater]

21 minutes ago, LAwLz said:

If the argument is that consumers only read the the first 4 numbers, why are you not mad at AMD for doing the whole "XT" thing? Or Nvidia for doing the whole "Ti" thing?

If you think that is the argument and the point then you've walked off the cliff Wile  E Coyote  style and have yet looked down to start falling yet. The argument is known established naming conventions not peoples ability to read nor look at numbers. When things have a known and established basis of understanding then people can use that, when you do not use this then understanding starts to go by the wayside.

 

This is not a difficult point to understand. There is literally a reason Nvidia goes GTX 1080, RTX 2080, RTX 3080 & RTX 4080 and not GTX 1080, RTX 2078, RTX 3084 & RTX 4088

[WereCat]

49 minutes ago, LAwLz said:

 

 

But you do bring up a good point. If we go by the logic that "it's not the GPU die that matters, it's performance", then should we also start changing the numbering on OC models of cards?

I think it becomes very messy very quickly if we start arguing that "3080" and "3080 OC" is misleading because they perform differently, and we can't expect customers to understand what OC means in terms of performance (just like we apparently can't expect consumers to understand that a model with 12GB after it performs better than one with 8GB after it".

Those are completely different things.

 

A GPU with "OC" in it means it's factory overclocked and you typically pay more for it while not really getting any benefit because any amount of OC on the card done by OEM is overshadowed by the boosting tech on the card. Only thing you're getting there are slightly higher guaranteed boost clocks which are typically several 100 of MHz lower than the actual clocks you will get.

 

Not only that but by getting the non-OC card you can still OC it yourself and pay less and get more. You're still getting the same card with just barely different clocks over which you have an actual control as a consumer because you can tune the clocks yourself.

 

Is this a trap for non-savy people? Kind of yes, sometimes people get to spend more because of this. This has always been the case and sometimes these OC models approach higher tier cards with their price. But getting the lowest priced card in the same tier vs the highest priced card in the same tier never amounted to much of a difference. You typically get better build quality and cooling but not a whooping 20% to 30% performance difference. You're still buying the same card with just slight difference in clocks. But this is also something set by OEMs and not NVIDIA. NVIDIA provides the base specs and you should expect the base specs at the very least. And they still haven't even updated their website...

 

You can't increase memory bus yourself on the RTX 3080 8GB and claw back the 20% to 30% performance vs the 12GB. The issue here is not 8GB vs 12GB the issue here is that the 8GB one has cut down memory bus which makes the card a LOT slower. These cards should be the same performance tier despite having different memory amounts but they don't because they are not the same tier of cards!

 

An i5 12400f with 16GB RAM does not fare worse than the same CPU with 64GB of RAM as long as you are not using more than 16GB. Well, as long as you run them both with the same amount of channels. Reviewers keep bashing OEMs for putting single RAM stick in the PC for a reason. But in this case you still have option to fix it by adding another stick of RAM, you can't do anything about it on a GPU.

 

That is the only context that should matter in this. You get 8GB to save money because you don't have use for 12GB VRAM in which case you should have the same experience because you're supposedly buying the same card with just less VRAM. Except in this case you barely pay less, at least now these cards cost the same like the 12GB versions from what I've looked up.

1 hour ago, LAwLz said:

 

 

I don't get why "RX 6800" vs "RX 6800 XT" and "RTX 3080" vs "RTX 3080 Ti" are apparently super easy to understand and not at all misleading. But "RTX 3060 8GB" vs "RTX 3060 12GB" is somehow too much for consumers to take in and we can't expect people to read the full name, just the first 4 numbers.

 

 

If the argument is that consumers only read the the first 4 numbers, why are you not mad at AMD for doing the whole "XT" thing? Or Nvidia for doing the whole "Ti" thing?

Excuse me, what?

 

RX 6800 vs RX 6800 XT implies they are different tiers of cards

RTX 3060 8GB vs RTX 3060 12GB implies they are the same tier of card with different amount of VRAM.

 

When you got GTX 780 3GB vs GTX 780 6GB you still got the same performance with both cards unless you did something that actually needed more than 3GB of VRAM in which case the 6GB pulled ahead ... and you got it for that purpose and that performance difference in this case is acceptable.

However, if you got 780 3GB and 780 6GB and you did something that only used up to 3GB and yet the 3GB version performed 20% to 30% less then you know there is something amiss.

 

Not the same thing.