Because they can sell these defective dies to unknowing consumers for the price of a fully enabled 3060.
Nvidia's 3000 Series RTX GPU [3090s with different memory capacity]
- Thread starter Shortbread
- Start date
-
- Tags
- nvidia
arandomguy
Regular
I actually wouldn't say that they reason they avoid doing that is down to wanting obfuscate models (or at least not entirely), but could just strongly be driven by branding reasons from market research. This has the by product of limiting how many model numbers are available.
We can argue whether or not it actually matters, but companies due put in research and even make it part of their branding how they label products.
From a purely functional stand point however having 4 digits but not using 2, and not even using the entire numerical range (eg. no xx1x, xx2x, xx3x, xx4x anymore at least) doesn't make any sense.
So we should pay attention to what's plastered on the box but not base our purchase decision based on what's plastered on the box? You could argue that the price is a clue, but given the prominent memory size "plastered all over the box" why would you fault the imperfectly prepared consumer for assuming the price difference is due to the memory size and not performance?
Let's talk car analogies. Mercedes doesn't sell a "Mercedes C" with different engines for different prices. They sell the C300 and C350 or whatever. The difference is in the name as well as the price. The difference in video card memory size is more akin to a difference in vehicle interior trim or tire. Cloth might be less pleasant than leather but it doesn't stop your Merc from performing the same as another with the same engine. The default long life tire may be less performant than a sport tire but you can always swap tires. In general, you can just lower texture quality to get an 8GB card to perform the same as a 12/16GB card, ceteris paribus.
Your argument that it's the customer's responsibility to know what they're buying both doesn't negate that what Nvidia is doing here is deceptive and is negated by Nvidia themselves having to
Nvidia does a lot of things to be proud of. Some of their product naming isn't worth defending.
Last edited:
It makes perfect sense if you are trying to bamboozle your customers.
So we have MX, FX, GT, GTX and RTX as prefixes, where MX is inferior („Mediocre eXperience“, and originally a suffix), FX the absolute worst, GT a denominator for not-quite-gaming, GTX is standard (and also a suffix) and RTX considered preferable.
And we have a plethora of suffixes: ZX (double the memory), M64, Vanta, Vanta-16, Vanta LT (cut down memory bus), Pro (higher clocks, except when introduced after Ultra, then lower clocks than Ultra), SDR, DDR (signifying memory type), Ultra (highest clocks in a model), GTS (Giga Texel Shader, otherwise bog standard model), GTS Pro (higher memory clock than GTS), Ti (somehow faster model, occasionally not a different chip, only clock bin, sometimes followed by an arbitrary number), SE (lower clocks), A (denominator for AGP-variants), 8X (AGP-revision), GS (slower than GT), GTO (lower clocked than standard), LE (I think, cut down), GT (higher clocks than non-GT, also a prefix for non-gaming), XE, XL, GX2, Dual, Z (for a 2nd GPU on the same card or PCB), [digits] (mostly signifying a certain amount of memory), M (mostly mobile, thus cut-down), Go (mobile, mostly cut-down), GSO (a one-off I think, don't get me started on this), OEM (mostly signifying a diversely cut-down model, occasionally a different GPU with some areas of advantage as in GF GTX 560 OEM with GF110), GTX+ (better than GTX!), Mac Edition (for scots-loving folks), Core [digits] (signifying a certain amount of cores, without telling if this is better or worse than the model without "Core", rarely there were versions with different core-counts without the denominator as in the 8800 GTS with 112 Cuda cores), Black (used in the 2nd Titan model I think, signifying more cores and higher clocks).
And that's not even including the Quadro Plex, Tesla and Quadros …
YES, I definitely do think Nvidia should be much more clear and consistent in naming their cards.
And we have a plethora of suffixes: ZX (double the memory), M64, Vanta, Vanta-16, Vanta LT (cut down memory bus), Pro (higher clocks, except when introduced after Ultra, then lower clocks than Ultra), SDR, DDR (signifying memory type), Ultra (highest clocks in a model), GTS (Giga Texel Shader, otherwise bog standard model), GTS Pro (higher memory clock than GTS), Ti (somehow faster model, occasionally not a different chip, only clock bin, sometimes followed by an arbitrary number), SE (lower clocks), A (denominator for AGP-variants), 8X (AGP-revision), GS (slower than GT), GTO (lower clocked than standard), LE (I think, cut down), GT (higher clocks than non-GT, also a prefix for non-gaming), XE, XL, GX2, Dual, Z (for a 2nd GPU on the same card or PCB), [digits] (mostly signifying a certain amount of memory), M (mostly mobile, thus cut-down), Go (mobile, mostly cut-down), GSO (a one-off I think, don't get me started on this), OEM (mostly signifying a diversely cut-down model, occasionally a different GPU with some areas of advantage as in GF GTX 560 OEM with GF110), GTX+ (better than GTX!), Mac Edition (for scots-loving folks), Core [digits] (signifying a certain amount of cores, without telling if this is better or worse than the model without "Core", rarely there were versions with different core-counts without the denominator as in the 8800 GTS with 112 Cuda cores), Black (used in the 2nd Titan model I think, signifying more cores and higher clocks).
And that's not even including the Quadro Plex, Tesla and Quadros …
YES, I definitely do think Nvidia should be much more clear and consistent in naming their cards.
He clearly wants more specs in the product name:
“RTX 3080 10GB 1.18GHz 320-bit GDDR6X 1.44GHz 8704 CUDA Cores 272 TMU’s 96 ROP’s”
Either the boxes will have to be massive or the name of the card will be written in teeny tiny font — maybe like ingredients on food labels. But that would look ridiculous, so we can write a shortened product name on the box such as “RTX 3080” while having the full name on the side or underside of the box!
Oh, wait…
Fortunately they've simplified the nomenclature in their professional lineup, so now we have the RTX 6000, RTX A6000 and RTX 6000 Ada
Rootax
Veteran
Note that being an expert and being a twat are not mutually exclusive
Why are you suddenly talking about me ?!
So wait, when Nvidia released the 3080 Ti, by your logic they should have called it the 3080 12GB instead?
Yes. Why is this surprising to you? You need to pay attention to what you're buying but you shouldn't make the buying decision based on what the manufacturer has put on the packaging. These two aren't at all mutually exclusive and they in fact have the same reasons behind them.
They actually do sell C300 and C350 with different engines and loads of everything else being very different too.
It does when it's actually a better option over the alternative which people seem to prefer for no apparent reason but tradition.
It would be a more fair name for it IMO, yes. Not only does "3080Ti" has some +5% in performance over a 3080 but it also basically was just a 12GB version of such.
A "Ti" can be all sorts more misleading on a product's name than stating the VRAM size there. VRAM size is just that, it's impossible to mislead anyone by stating it. And substituting it for something which has no technical meaning whatsoever does not improve the situation in the slightest.
Not to that extreme but if you aim at providing the most relevant information about a videocard through its name only then yes, something like this is considerably better than a "suffix".
Note that nothing which other people here stated as an advantage for a "suffix" or even an "x5" approach is actually true: the same name doesn't mean that there's always the same chip there, or a chip with the same number of processing units, or the same memory bus width, or size, or clocks (you do know that a 3060 for example have AIB models with different clocks, right?).
The biggest issue is that we're looking at a generation of a lineup back from where we are today while like 50% of products in that lineup hasn't been launched - or planned probably - back when the lineup naming was formed initially. You can do "x50"s but does that help really? AMD did this with 6000 and it's still a mess where you have 6700XT and 6750XT being essentially the same thing in practice. Any scheme will fail when you need to launch several "mid-gen" products in a place where you didn't plan to launch any initially. The only way to do it properly is to make sure that the new products can be easily identified by the customers so that they wouldn't buy something which they didn't want to buy in the first place. The rest is down to doing your own research and using these amazing GPU review websites we have - which is very much a luxury in the world of consumer goods at large or even other PC components often. It's not that hard really.
WTH? 3080 Ti is within about 1-2% of 3090 and 10%+ faster than 3080 10 GB according to TPU numbers. Sadly for 12GB model they only have OC numbers so apples to oranges.
TechSpot did run their 3080 12GB review at NVIDIA reference clocks which resulted in 5.8 % faster avg and 4.0 % faster minimum FPS than 10 GB, but it still leaves 3080 Ti closer to 3090.
I'm sorry, but what?!?
Well points for consistency at least. But can you really not get why everyone else here thinks that suffix/x5/etc. is a more consumer-friendly measure of performance than VRAM size is? Why would a consumer assume that a different amount of VRAM on the same GPU model means different shader performance, and not just literally a different amount of VRAM?
You’ve talked about the need for consumers to be informed, but an informed consumer by now knows instinctively that “Ti” means “somewhat faster” without having to fully research every chip parameter.
I’ve seen a few car analogies here, but to me this is more like nutritional information on food. What NVIDIA did here would be like the 500 ml and 1000 ml versions of a product having a different number of calories per ml.
Yes.
Why would a consumer assume the same from a suffix or a x5 model number? VRAM is at least a technical thing which actually give you relevant information.
Which can just as easily be a lie as anything else. There is no rule which says that a "Ti" is even a faster GPU than a "non-Ti".
Which is also completely possible as there are no rule preventing this from being true. You as a consumer have to do your research regardless.
That is indeed the case, so I would like to point out again that the performance gap between the 3060 8GB and the 3060Ti (8GB) is in the same ballpark as the gap between the 3070 and the 3090. NVIDIA is making it very easy for informed consumers to make a poor decision. And I do think that is intentional. Especially when there is no 3050Ti.
Is your point that it’s ok to blatantly lie and mislead consumers because widely accepted naming conventions aren’t perfect?
Subtlesnake
Regular
The established precedent is that the Ti version of a GPU is faster than the same model without any prefix (vanilla).
Similar threads
