+1 GB of VRAM will make such a massive difference on the -25% decreased memory bus! What's the point, really? :kookoo:

Makes sense for the 5050, 1GB more capacity and slightly higher bandwidth. Feels like a downgrade to a 5060Ti though, they'd be better off axing the 8GB model and having 12GB and 16GB models, using the new 24 Gbit memory chips for the 12GB and the older 16 Gbit chips for the 16GB.

Better yet discontinue the 8GB 5060Ti and don't even replace it. Or, discontinue both and make only a 12GB, priced like the 8GB.

AusWolf+1 GB of VRAM will make such a massive difference on the -25% decreased memory bus! What's the point, really? :kookoo:

There’s no point. Which is why this won’t happen.

I should start flinging stuff on X, maybe I’ll get quoted on TPU.

We got 9GB RTX 5060 before GTAVI

If there is a cost saving using 3 GDDR7 chips instead of 4 GDDR6 chips, than nVidia will absolutely do this. Maybe the narrower bus also lets them use more silicon that might otherwise be deemed defective?

Visible NoiseI should start flinging stuff on X, maybe I’ll get quoted on TPU.

Not even X, just forums it seems! Or start a YT channel where you can basically make it up and reach a lot of people too. Bonus points if you're voice is like hearing nails on a chalkboard.

wolfMakes sense for the 5050, 1GB more capacity and slightly higher bandwidth. Feels like a downgrade to a 5060Ti though, they'd be better off axing the 8GB model and having 12GB and 16GB models, using the new 24 Gbit memory chips for the 12GB and the older 16 Gbit chips for the 16GB.

The Blackwell cards in general aren't very bandwidth limited, 5060 Ti 448GB/s performs the same as 9060 XT 320GB/s. It's kind of ridiculous nonetheless, the 9GB card will do better in capacity limited scenarios, the 8GB one will do better in bandwidth limited but not capacity limited scenarios. I think that means the 9GB one will be better for new games and the 8GB one will do better in older games that are still demanding enough to tax it but don't come close to maxing out the VRAM.

Both options are comically bad really, 8GB is not acceptable on any GPU costing $250+ anymore, mind you this was the introductory price of the RX 480 8GB ten years ago.

vbq7qK68eyYAH4iRIf there is a cost saving using 3 GDDR7 chips instead of 4 GDDR6 chips, than nVidia will absolutely do this. Maybe the narrower bus also lets them use more silicon that might otherwise be deemed defective?

With the current state of memory supplies, there’s no cost savings to be had. They will use every single component they can get their hands on.

If this does turn out to be true I think they should call it the 5060 SSuper and 5060Ti SSuper. 5060 SuperStingy has a nice ring to it. I sure hope this isn't true, what a slap to anyone that needs a gpu and can't shell out thousands. It is like giving a hungry dog one more piece of kibble instead of doing it right and giving him a meal. Just wait though, if this is true, AMD will make sure to do it worse, and Intel will continue to release pro cards with the big die and not a B590 or whatever.

ghaziThe Blackwell cards in general aren't very bandwidth limited, 5060 Ti 448GB/s performs the same as 9060 XT 320GB/s.

That could also be for other reasons that aren't memory bandwidth too, they're completely different cards, architectures and memory generations. As resolution increases, it could well be the extra bandwidth that accounts for the performance gap widening, given the 5060Ti has +4.3% 1080p, +6% 1440p, +7.7% 2160p relative to the 9060XT. My 5080 Also benefits from memory OC alone, more so than my 3080 does.

I place a lot of doubt in a 96-bit 5060Ti 9GB coming to fruition based on this 'information' alone right now, but in the meantime someone with a 5060Ti dropping their clocks by 25% and running some A and B benchmarks might shed some light on what could be expected - but even then it wouldn't be completely accurate as dropping clocks isn't necessarily identical to narrowing the bus.

They should name them then 5060 SE and 5060 Ti SE, as the SE monicker is always used on the cut-down cards with both teams.

RuruThey should name them then 5060 SE and 5060 Ti SE, as the SE monicker is always used on the cut-down cards with both teams.

It's a shame that dropped away over time, it was a great way to differentiate. Lately at least with Nv it seems they delineate mostly on VRAM size alone.

RuruThey should name them then 5060 SE and 5060 Ti SE, as the SE monicker is always used on the cut-down cards with both teams.

Sadly, these low-end model monicker like SE, LE Lite has been drop ages ago. Even 1060 3GB doesn't have any sort of indication that it has less CUDA cores than 6GB model.

(Meanwhile in NVIDIA marketing)…. “7's the key number here... 7-Elevens. 7 dwarves. 7, man, that's the number.”

NIEN! NIEN! NIEN! 3 x 3 = 9

ApocalypseeSadly, these low-end model monicker like SE, LE Lite has been drop ages ago. Even 1060 3GB doesn't have any sort of indication that it has less CUDA cores than 6GB model.

Which is quite a shame. I remember the storm the 1030 DDR4 vs GDDR5 fiasco stirred up among the handful of people interested in the product.

I also miss the time when x30 or even x10 existed as the low end. People say x60 is the new low end, but honestly, there's no more low end anymore.

nGreedia is just trolling the gaming plebe right now. I mean, you just cannot make this $hit out.

Late to April Fools' I assume

At this point I wouldn't be surprised if Nvidia released a slower line of cards that's cheaper for them to make but still make them more expensive, label them "Super", and tell us with tears in their eyes that this is actually a great sacrifice from them, they could make 10x more by making AI accelerators, so we should be grateful.

regardless of the news source quality and validity - the cards are nice desktop cards for basic tasks when you want a dedicated card with dedicated graphic memories. the only drawback are the binary drivers and maybe the questionable hardware design.

nvidia 970 was the first and thanks for reminding me on how nvidia fooled the customers with different hardware designs and performance characteristics over the years.

using different memory chip manufacturers which may lead to a bit different performance is acceptable. using less calculation units on purpose or using very slow memory for certain memory bars makes it very evil.

for most every processor we get a new label like i3-10100 or i3.-10105. nvidia is unable to calculate and unable to distinguish. that needs brain to label cards differently. the customer knows what he gets. those who do not care see i3. i see i3-101xx and know it is a crapshot. it is not fraudulent as 10100 or 10105 is a different number. the le, lhx or whatever crap scheme is bad. just bad.

i highly doubt there are 10 differnt types of a nvidia class. so nvidia which is stupid to calc and use brain, still could use nvidia 5060 to 5069 to distinguish clearly what it is. big letter on the box nvidia 5060 class, 1/5th font below nvidia 5062.

the wattage of a card can most likely often be researched. whats on the pcb can be not. nvidia was hardly ever customer friendly anyway.

Visible NoiseWith the current state of memory supplies, there’s no cost savings to be had. They will use every single component they can get their hands on.

I don't mean to sound like a dick, and I'm not trying to be a dick, buuuut, unless you know what nVidia is paying for 2GB GDDR6 and 3GB GDDR7, you're guessing as much as i am. Even if it saves them 5c on a SKU, over millions of units, i'm sure they're be keen on the savings.

When a 5070 Ti is the minimum for a bunch of high-end games let's make a POS version of the 5060 POS, very naaaayce. In related news Nintendo plans to launch a pair of 5060 POS-compliant pixel-perfect games, Stuper Mafio 4 and Animalympics the Rerun.

Further lowering of the bar for the enshitification Nvidia has been pushing trough its product stack for a decade now.
I can see them adding some lpddr5 on top of their lower range of products chips and selling that to,gamers

glad I did not even consider buying Ngreedia's crap. Been enjoying my ARC B580 GPU ever since I bought it for cheaps last year.

Should call them "5060MX" back from the days of the GeForce 2-4 MX models