It’s quite obvious at this point that NVIDIA or more rather their AIB partners have quite a lot of excess supply for various RTX 3000 series graphics cards that they are desperate to offload.
NVIDIA may have just recently launched the RTX 4090, which despite offering a 60% increase in relative rasterized performance it has several issues which we’ll discus shortly, but I suppose the rumors of NVIDIA having stockpiled RTX 4090 units months in advance may have been true considering how NVIDIA are once again facing severe oversupply issues.
These new GPUs are obviously a direct result of NVIDIA getting negative reception for trying to peddle an RTX 4070 as the RTX “4080 12GB”, leading to the company having to cancel its launch, but at least they’ve gotten away with their $2000 RTX 4090 graphics cards and the “$1200” RTX 4080.
Gotta make your shekels somehow.
Specifically, majority of GPUs sold every generation are your average run of the mill mid-range componentry which has seen a relative increase from the ~$300 price tag over the past several generations, though manufacturers make much greater profits from high-end GPU sales it makes the most amount of sense that if NVIDIA were facing oversupply problems that it would generally be the cause of excess X60 series GPUs.
AIB partners are now announcing new variants of the RTX 3060 and RTX 3060 Ti graphics cards, despite Ampere being phased out entirely these AIBs sure are desperate to relinquish their stockpile of GPUs possibly bought during the mining epidemic considering slim profit margins for AIB partners is the very reasoning behind NVIDIA losing its bottom bitch, EVGA.
ASUS for instance have announced a new revised RTX 3060 Ti graphics card, identical to the original GPU in every single way except for the memory, whereas the standard GPU offers 8GB of GDDR6 memory operating at 14Gbps effectively, across its 256-bit memory interface the RTX 3060 Ti provides 448GB/s of bandwidth.
To rid themselves of excess GA104 dies and possibly memory modules as well, NVIDIA have given the go-ahead for partners to release revised variants of the RTX 3060 Ti equipped with GDDR6X memory chips instead.
Operating at 19Gbps effectively, an RTX 3060 Ti can now produce as much as 608 GB/s in raw bandwidth across its 256-bit memory bus, an increase of over 35%, the very same as the RTX 3070 Ti.
I for one am unsure whether the intentions of this would be to call it the RTX 3060 Ti Super or perhaps 3060 Ti Tie.
Nevertheless an RTX 3060 Ti with GDDR6X memory will undoubtedly provide an exceptional performance increase over its standard GDDR6 model but you can’t help but feel as if this is probably something that should’ve been available on release if not at some point in the two whole fucking years the RTX 3000 series has been on the market, not when the series as a whole is all but End of Life.
You may also take note of the increased power requirements needed to operate the RTX 3060 Ti GDDR6X, with ASUS equipping them with two 8-pin PCIe power connectors whereas the standard GDDR6 model only requires a single power connector.
Given that the standard RTX 3060 Ti has a TDP rating of just 200 watts it makes sense that the much more power hungry GDDR6X memory would plague its efficiency a tad.
For NVIDIA to offload excess RTX 3060 Ti units with GDDR6X memory truly does show that this is a sort of rush effort to rid themselves of their own supply, but it doesn’t end with excess RTX 3060 Ti units though I’m afraid.
It’s certainly not the first time NVIDIA had come up with creative ways of reusing excess chips, such as rebranding various older generation hardware as “CMP” Cryptocurrency Mining Processors, and who could forget the six gorillion specific cutdown variants of the GTX 1060?
Manli, one of those AIB partners who are quite prevalent amongst South East Asian nations have announced a brand new RTX 3060 graphics card.
Whereas the original RTX 3060 is a rather slow lemon, a lemon equipped with 12GB of GDDR6 memory no less, moreso than each RTX 3000 series GPU until you reach the RTX 3080 Ti, this specific variant however comes with a pitiful 8GB of VRAM instead.
The real killer here isn’t just a third reduction on its overall memory capacity but also a reduction to its memory bus interface, with 15Gbps GDDR6 memory modules, this cucked variant now has an 128-bit bus compared to the 192-bit memory interface found on the standard model, this brings the overall effective bandwidth down from 360GB/s to an embarrassing 240GB/s.
This certainly will have an overall effect on performance, but on the plus side, at least for NVIDIA and AIB partners such as ASUS who are also selling this 8GB turd, is that consumers can buy into this garbage for a lower price tag, speculated to be an $80 drop from $380 to $300.
Exactly what the market needed at this point in time, as consumers are not opting to buy into new tech, possibly because they cannot afford to heat their homes for the winter, NVIDIA have instead provided consumers a quazzy RTX 3050 Ti purpose built for the purpose of offloading excess GA106 dies at a hefty profit, with its VRAM capacity and effective bandwidth being cucked by a third it’s obvious that the performance drop expected from this will be greater than that of the savings in terms of pricing versus the 12GB variant.
The standard card was already extremely limited in terms of memory bandwidth, to further decrease its ability will certainly hinder the performance by a significant margin with performance possibly dropping below that of the RX 6600.