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

[Fasterthannothing]

On 12/4/2022 at 6:27 AM, leadeater said:

And any of that makes this ok how?

 

Back to the old but the number at the end actually means something to the uninformed. No, no it doesn't. Like it didn't for the RTX 4080.

 

People know the first part, rely on the first part, talk about the first part. How many people ever really talk about more than that, see GTX 1060 3GB vs 6GB. That was already painful enough and at least that was less of a kick in the shins.

 

There is zero excuse for not calling it an RTX 3050 Ti, could have been a very positively received product if so. Nvidia didn't do that, too bad for them.

Why?? We have already established that even going way way back different ram amounts mean different performance. Your treating the average consumer like their the dumbest person on the planet which in that case they wouldn't even know or care they got a slower card because they would never check or benchmark it. You can't have your cake and eat it too.

[HenrySalayne]

2 hours ago, Stahlmann said:

I'm in a similar situation. But my main problem is that i play a lot of ray traced games nowadays. I don't want to "upgrade" to a 7900 XT or 8900 XT for example while they can't get their ray tracing performance to the same level as several year old Nvidia cards. And ray tracing will only get more support going forward, being baked into popular next-gen game engines and all that.

Let's assume AMD is 30% less expensive than Nvidia for the same rasterization performance. How much raytracing performance would you personally sacrifice if you pay significantly less? 20%?  30% (the price difference)? Even 40%?

AMD's upcoming GPUs might just be that. 

[WereCat]

52 minutes ago, Fasterthannothing said:

Why?? We have already established that even going way way back different ram amounts mean different performance. Your treating the average consumer like their the dumbest person on the planet which in that case they wouldn't even know or care they got a slower card because they would never check or benchmark it. You can't have your cake and eat it too.

But... that's the entire point of why this is bad. It's misleading for people who don't know any better. 

And just because NVIDIA keeps getting away with this we should now give up and give them a pass? 

 

What's this logic behind siding with the brand while they are bending you over? 

 

Like... I can't wrap my head around why you would think that. 

[DANK_AS_gay]

On 12/2/2022 at 9:02 PM, Middcore said:

Nvidia and fucking their customers with VRAM fuckery, name a more iconic duo. 

GTX 970 INTENSIFIES

[DANK_AS_gay]

1 hour ago, Stahlmann said:

I'm in a similar situation. But my main problem is that i play a lot of ray traced games nowadays. I don't want to "upgrade" to a 7900 XT or 8900 XT for example while they can't get their ray tracing performance to the same level as several year old Nvidia cards. And ray tracing will only get more support going forward, being baked into popular next-gen game engines and all that.

This video explains some things, I think AMD's implementation is going to have more support going into the future due to the Series X and the PS5, as well as its ease of implementation compared to nVidia's implementation. He explains better than I do, I would watch the video.

[Fasterthannothing]

23 minutes ago, WereCat said:

But... that's the entire point of why this is bad. It's misleading for people who don't know any better. 

And just because NVIDIA keeps getting away with this we should now give up and give them a pass? 

 

What's this logic behind siding with the brand while they are bending you over? 

 

Like... I can't wrap my head around why you would think that. 

I didn't side with anyone all I pointed out was that the argument was invalid because you can't say consumers know enough to benchmark their cards while also saying they can't read numbers on a box! And let's not pretend anyone cared about this until like 5 minutes ago because Nvidia (and in some cases AMD) have done name differentiations without level increases like this since the 556 ti 448 edition for crying out loud. At the exact same time as AMD released their 6950 1gb and their 6950 2gb where the 2gb was actually the 6970 and could be flashed as such making it a totally different card with the only name different being wait for it THE RAM!

 

https://forums.tomshardware.com/threads/radeon-hd-6950-1gb-vs-2gb.767470/

[HenrySalayne]

Just now, Fasterthannothing said:

I didn't side with anyone all I pointed out was that the argument was invalid because you can't say consumers know enough to benchmark their cards while also saying they can't read numbers on a box!

If you can't wrap your head around the problem of manufacturers releasing similar or identical named products which are not even close in performance, you are part of the problem. This level of ignorance is just astonishing. A clearly named product stack does not take anything away from you.

[xAcid9]

1 hour ago, Fasterthannothing said:

We have already established that even going way way back different ram amounts mean different performance.

Ah yes. That's why the 2060 12GB exist. It perform much faster than the 6GB model. Oh my god, that's so easy to understand. Thank you. 

 

On 12/3/2022 at 11:40 PM, Vishera said:

It will completely break my purchasing model,

With your suggestion i will probably have to upgrade a year or two sooner because of the performance.

What? I didn't suggest you to upgrade anything. I'm just saying you can get RX 6600 today for less than $250 instead of the much slower GTX1660.

 

It's a much better card anyway, I don't know why anyone want to purchase 1660 over 6600 for similar price unless they had bad experience with Radeon driver/hardware.

[Fasterthannothing]

13 minutes ago, xAcid9 said:

Ah yes. That's why the 2060 12GB exist. It perform much faster than the 6GB model. Oh my god, that's so easy to understand. Thank you. 

 

16 minutes ago, HenrySalayne said:

If you can't wrap your head around the problem of manufacturers releasing similar or identical named products which are not even close in performance, you are part of the problem. This level of ignorance is just astonishing. A clearly named product stack does not take anything away from you.

I'm not defending them all I am pointing out is that this is has been industry standard for literally a decade at this point (seriously go look at the 6950 1gb and 2gb cards). Do I agree with the practice absolutely not however if I'm shopping for a graphics cards as an average consumer I'm going to look at a card and go the one with the bigger number is faster plain and simple. Seriously I remember this was the exact reason AMD stuffed tons of RAM into their GPU I know people who bought slower GPUs because it had more RAM because the number on the box was bigger. It's exactly the thinking that drives average consumers so it's not out of the realm of possibilities that a consumer will think a item with less ram is slower. It seems people are more worried about hating Nvidia than looking at how the industry has actually structured their product stack with how the RAM numbers show on a box. It's a problem with the entire industry everywhere not just GPUs look at any electronics (resolution and ram and large amounts of storage or even a slow laptop i7 being promoted over a desktop i5 because 7> 5, HP numbers in sports cars, or a million other things heck bedding is market based on high thread count.

[xAcid9]

4 minutes ago, Fasterthannothing said:

I'm not defending them all I am pointing out is that this is has been industry standard for literally a decade

It's not an industry standard when only 1 company been repeating this "tradition" in the past 10 years, more like Nvidia standard.
HD6000 series was like 12 years ago. Try to come up with something more recent.

[Stahlmann]

1 hour ago, HenrySalayne said:

Let's assume AMD is 30% less expensive than Nvidia for the same rasterization performance. How much raytracing performance would you personally sacrifice if you pay significantly less? 20%?  30% (the price difference)? Even 40%?

AMD's upcoming GPUs might just be that. 

I wouldn't want to sacrafice ANY ray tracing performance. Why would i spend money to downgrade, even if it's just in ray tracing performance?

 

Until AMD catches up in ray tracing performance i won't buy their GPUs, even though i despise Nvidia's tonedeaf and anti consumer mentality. They simply have the superior product for me.

[WereCat]

1 hour ago, Fasterthannothing said:

I didn't side with anyone all I pointed out was that the argument was invalid because you can't say consumers know enough to benchmark their cards while also saying they can't read numbers on a box! And let's not pretend anyone cared about this until like 5 minutes ago because Nvidia (and in some cases AMD) have done name differentiations without level increases like this since the 556 ti 448 edition for crying out loud. At the exact same time as AMD released their 6950 1gb and their 6950 2gb where the 2gb was actually the 6970 and could be flashed as such making it a totally different card with the only name different being wait for it THE RAM!

 

https://forums.tomshardware.com/threads/radeon-hd-6950-1gb-vs-2gb.767470/

 

56 minutes ago, Fasterthannothing said:

 

I'm not defending them all I am pointing out is that this is has been industry standard for literally a decade at this point (seriously go look at the 6950 1gb and 2gb cards). Do I agree with the practice absolutely not however if I'm shopping for a graphics cards as an average consumer I'm going to look at a card and go the one with the bigger number is faster plain and simple. Seriously I remember this was the exact reason AMD stuffed tons of RAM into their GPU I know people who bought slower GPUs because it had more RAM because the number on the box was bigger. It's exactly the thinking that drives average consumers so it's not out of the realm of possibilities that a consumer will think a item with less ram is slower. It seems people are more worried about hating Nvidia than looking at how the industry has actually structured their product stack with how the RAM numbers show on a box. It's a problem with the entire industry everywhere not just GPUs look at any electronics (resolution and ram and large amounts of storage or even a slow laptop i7 being promoted over a desktop i5 because 7> 5, HP numbers in sports cars, or a million other things heck bedding is market based on high thread count.

The amount of mental gymnastics I see in what you wrote is astounding. 

If you're trying to be devils advocate... fine but try to make better arguments. 

 

 

Comparing the 6950 1GB/2GB to 6970 and to this is something completely different. 

 

If you bought 6950 2GB then you've got what you've paid for. 

If you were tech save enough to flash it to 6970 and risk bricking it and voiding the warranty that was up to you. 

 

With 3060 you MAY think you know what you've paid for. Except you may actually get worse product and you can't do anything about it unless you can return it. 

 

More VRAM = BETTER 

Yes, that mindset still exist to a way lesser amount than it used to. There were times when we had almost daily discussions on this forum and explaining advantages and disadvantages to less savy people, you almost never see those posts now. 

 

And with that logic AMD should be the market leader because their product stack consistently has more VRAM than the NVIDIA counterpart. 

[porina]

14 minutes ago, Stahlmann said:

I wouldn't want to sacrafice ANY ray tracing performance. Why would i spend money to downgrade, even if it's just in ray tracing performance?

On these forums and places like this people value different aspects differently. To me raster is "done". More raster perf is neither here nor there once we hit good enough. We have not quite hit good enough on RT, so all the perf there is where the value lies. Features not direct to gaming like video encoders are another aspect for example, as is compute support. Right now I'm not even sure I'd rate AMD above Intel in GPU tech. Sure, the overall end user experience and offerings of Intel are not quite there yet, but they have hit harder in RT and matrix operations than AMD up to RDNA2, thus I feel Intel are closer to nvidia tech than AMD are.

 

AMD price their GPUs because they have decided that is what's optimal for them, considering all the factors. If all you care about is raster, they may give better value in that limited scenario. In general it costs less because you get less. Not because Lisa Su likes you. Look at CPUs for where that situation has flipped. They know they can command well in the upper regions now and are not afraid to charge for it.

 

5 minutes ago, WereCat said:

And with that logic AMD should be the market leader because their product stack consistently has more VRAM than the NVIDIA counterpart. 

It is a marketing trick that works to an extent. Look at the posts over the last couple of years where people were picking up AMD GPU with more VRAM, not because they had any specific need for that much, but because it was there. Of course some genuinely need more VRAM, but they are probably not the majority of posters here.

[HenrySalayne]

1 hour ago, Fasterthannothing said:

I'm not defending them all I am pointing out is that this is has been industry standard for literally a decade at this point

Oh - then it is completely okay. What a great point. And the spine of your argument is totally not the fallacy of relative privation.

 

I shall just diddly-do watch a review to make an informed buying decision....

 

 

So after watching three independent reviews I really think I should buy a 3060, that's exactly the performance I aimed for.. Oh look, the 8 GB variant is $20 bucks cheaper. I only game at 1080p so that is just perfect for me.

[HenrySalayne]

35 minutes ago, Stahlmann said:

I wouldn't want to sacrafice ANY ray tracing performance. Why would i spend money to downgrade, even if it's just in ray tracing performance?

That's not at all what I asked.

[leadeater]

3 hours ago, Fasterthannothing said:

Why?? We have already established that even going way way back different ram amounts mean different performance. Your treating the average consumer like their the dumbest person on the planet which in that case they wouldn't even know or care they got a slower card because they would never check or benchmark it. You can't have your cake and eat it too.

No we have established that it's both, it's always been misleading and confusing and it should never be acceptable. Ever. You can change VRAM capacity and not effect performance and in that situation and that situation only is it acceptable to carry the same model number/name i.e. RX 480 4GB vs RX 480 8GB.

 

I am treating the situation exactly as it should be treated. The RTX 3060 8GB was released to market without any public announcement, no notification to reviewers, no review samples and NO UPDATE TO OFFCIAL PRODUCT PAGES. Right now none of the official RTX 3060 product pages and performance information show nor differentiate these model variants dispite the significant difference in performance.

 

How were you, or anyone else, supposed to know what the difference actually is? Unacceptable.

 

And what on earth are you arguing, it's never ever acceptable for a consumer to receive a worse product unknowingly. If a consumer is buying an RTX 3060 and expecting an RTX 3060 then they MUST receive an RTX 3060 and nothing less. Not some weird off cast product that carries the same model name but isn't actually the same product offering the same performance.

 

Cake and eat it too? Yes, yes I can. Name it RTX 3050 Ti and then there is no problem, no issue, nothing to debate and it would even be a fantastic value product.

 

Why must you or anyone else make such bad arguments and support such bad business practices?

 

Quote

The RTX 4080 12GB is a fantastic graphics card, but it’s not named right. Having two GPUs with the 4080 designation is confusing.

 

So, we’re pressing the “unlaunch” button on the 4080 12GB. The RTX 4080 16GB is amazing and on track to delight gamers everywhere on November 16th.

https://www.nvidia.com/en-us/geforce/news/12gb-4080-unlaunch/

[Fasterthannothing]

58 minutes ago, xAcid9 said:

It's not an industry standard when only 1 company been repeating this "tradition" in the past 10 years, more like Nvidia standard.
HD6000 series was like 12 years ago. Try to come up with something more recent.

O you mean like the R7 250??

we could go even more modern like the R7 350 gddr3 and gddr5 models with the same name 

[Fasterthannothing]

7 minutes ago, HenrySalayne said:

Oh - then it is completely okay. What a great point. And the spine of your argument is totally not the fallacy of relative privation.

 

I shall just diddly-do watch a review to make an informed buying decision....

 

 

So after watching three independent reviews I really think I should buy a 3060, that's exactly the performance I aimed for.. Oh look, the 8 GB variant is $20 bucks cheaper. I only game at 1080p so that is just perfect for me.

By that same picture your forgetting the 3080 12gb and so your know if your buying the 10gb well that's cheaper but hey literally no one said s word about that card it's scummy practice that no one cares about until Nvidia decided to raise prices and instead of not buying the card and moving on with their lives people just want to become a keyboard warrior.

[HenrySalayne]

1 minute ago, Fasterthannothing said:

By that same picture your forgetting the 3080 12gb and so your know if your buying the 10gb well that's cheaper but hey literally no one said s word about that card it's scummy practice that no one cares about until Nvidia decided to raise prices and instead of not buying the card and moving on with their lives people just want to become a keyboard warrior.

You're arguments could literally replace Alfie Allen in the Game of Thrones series.

Please explain to me how anybody watching a 3080 review (original 10 GB variant) would get anything less than they expect?

"Ohh no, what a bad deal. My GPU is 5% faster than I thought. How could you do this to me?!? "

[leadeater]

1 hour ago, xAcid9 said:

It's not an industry standard when only 1 company been repeating this "tradition" in the past 10 years, more like Nvidia standard.
HD6000 series was like 12 years ago. Try to come up with something more recent.

AMD 6950 1GB vs 2GB was identical GPU die, identical memory bus, identical performance unless VRAM buffer is exceeded. It's the situation where having the same model number with different VRAM capacity at the end is perfectly fine.

 

Quote

As its name might suggest, the Radeon HD 6950 1GB is the exact same card as its more heavily endowed Radeon HD 6950 2GB. The only difference is that the 1GB card uses eight half-density 128MB GDDR5 memory chips rather than the eight 256MB modules found on the 2GB variant.

https://bit-tech.net/reviews/tech/amd-radeon-hd-6950-1gb-review/4/

 

AMD has a tradition of doing it the right way, the way that is clear, transparent and not misleading. Nvidia has a tradition of doing it the wrong way, in secret as much as possible and does nothing to avoid market confusion.

[Fasterthannothing]

1 minute ago, HenrySalayne said:

You're arguments could literally replace Alfie Allen in the Game of Thrones series.

Please explain to me how anybody watching a 3080 review (original 10 GB variant) would get anything less than they expect?

"Ohh no, what a bad deal. My GPU is 5% faster than I thought. How could you do this to me?!? "

Because it's literally the exact same argument word for word. If it has less ram it's slower end of story 

[Fasterthannothing]

1 minute ago, leadeater said:

AMD 6950 1GB vs 2GB was identical GPU die, identical memory bus, identical performance unless VRAM buffer is exceeded. It's the situation where having the same model number with different VRAM capacity at the end is perfectly fine.

 

https://bit-tech.net/reviews/tech/amd-radeon-hd-6950-1gb-review/4/

 

AMD has a tradition of doing it the right way, the way that is clear, transparent and not misleading. Nvidia has a tradition of doing it the wrong way, in secret as much as possible and does nothing to avoid market confusion.

No the 6950 2gb has a cut down die you didn't even bother to quote the actual link proving it. https://forums.tomshardware.com/threads/radeon-hd-6950-1gb-vs-2gb.767470/post-6831109

 

Not to mention the entire r7 250 and r7 350 ddr3 ddr5 swap

[porina]

5 minutes ago, leadeater said:

And what on earth are you arguing, it's never ever acceptable for a consumer to receive a worse product unknowingly. If a consumer is buying an RTX 3060 and expecting an RTX 3060 then they MUST receive an RTX 3060 and nothing less. Not some weird off cast product that carries the same model name but isn't actually the same product offering the same performance.

 

Some more random thoughts:

I'm thinking the 1060 was the first instance of different VRAM different performance. 1060 3GB was about 10% fewer cores than 6GB, however that translated to perf. But they were based on the same chip. It was the same silicon. If it weren't for the cores, I think the VRAM performance was same.

 

Maybe 3080 is the next example? We started with the 10GB model, but then got the 12GB model as refresh. Again, same silicon, different cores, but now also different VRAM quantity and width. Pricing aside, this seemed to be accepted. Then again the adjustment was in an upwards direction.

 

Then we have the unlaunched 4080 12GB. This was the biggest deviation, since it didn't even use the same silicon as 4080 16GB and is the nearest we've seen so far in being a different product.

 

Today we're now talking 3060 8GB vs 12GB. Same silicon, so this is not a 4080 situation. Smaller VRAM on average performs less. I'd dare say it'll perform somewhere between 3050 and 3060 12GB, and once volume gets up, it'll be priced in that gap too. You get what you paid for compared to its neighbouring products.

 

5 minutes ago, leadeater said:

Cake and eat it too? Yes, yes I can, Name it RTX 3050 Ti and then there is no problem, no issue, nothing to debate and it would even be a fantastic value product.

That could have been an option if only the 3050 Ti mobile version didn't already exist. We have 3050 mobile and 3050 Ti mobile both based on GA107. The 3060 mobile and 3060 desktop are both GA106, so a GA106 8GB does make more sense as a 3060.

 

Without going through nvidia's product stack, I'm pretty sure there are examples where "tier down Ti" have used silicon from the higher tier.

[leadeater]

19 minutes ago, porina said:

I'm thinking the 1060 was the first instance of different VRAM different performance. 1060 3GB was about 10% fewer cores than 6GB, however that translated to perf. But they were based on the same chip. It was the same silicon. If it weren't for the cores, I think the VRAM performance was same.

Yea, came in at an actual ~8% difference and what's most annoying about that situation is changes to the GPU die/package simply shouldn't have been done.

 

Nvidia is quite capable of doing it correctly i.e. GTX 960 2GB vs GTX 960 4GB, the prior generation.

 

19 minutes ago, porina said:

Then we have the unlaunched 4080 12GB. This was the biggest deviation, since it didn't even use the same silicon as 4080 16GB and is the nearest we've seen so far in being a different product.

 

Today we're now talking 3060 8GB vs 12GB. Same silicon, so this is not a 4080 situation. Smaller VRAM on average performs less. I'd dare say it'll perform somewhere between 3050 and 3060 12GB, and once volume gets up, it'll be priced in that gap too. You get what you paid for compared to its neighbouring products.

It doesn't actually matter that the GPU die is different so long as the active configuration of it is the same. There is already an RTX 3060 GA104-150 and it is identical in configuration and performance to the RTX 3060 GA106-300.

 

Where you to get an RTX 3060 GA104 or RTX 3060 GA106 you got the same actual product as advertised. Now there is a GA106-302 based product that is not configured identically and does not have the performance as advertised.

 

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. And even if you were intelligent enough to figure out something was amiss you'd have zero way to know what the performance difference is or could be BEFORE purchasing it.

 

Had it been called an RTX 3050 Ti then it would necessitate it's own product page which would require all the accompanying documentation and that would have to actually match the product or it would literally be direct false advertising, zero question.

[leadeater]

16 minutes ago, porina said:

That could have been an option if only the 3050 Ti mobile version didn't already exist. We have 3050 mobile and 3050 Ti mobile both based on GA107. The 3060 mobile and 3060 desktop are both GA106, so a GA106 8GB does make more sense as a 3060.

No the mobile product has nothing to do with it at all and more often the mobile variants do not use the same GPU major die codes anyway. This has no bearing on the desktop models at all.

 

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.