Tuesday, October 25th 2022
NVIDIA Partners Quietly Launch GeForce RTX 3060 with 8GB (128-bit) Memory
NVIDIA's add-in board partners today began quietly launching the GeForce RTX 3060 8 GB, a variant of the RTX 3060 with a third of its memory size and memory bus-width sawed off. The RTX 3060, NVIDIA's best-selling desktop graphics SKU from the RTX 30-series "Ampere," originally launched with 12 GB of GDDR6 memory across a 192-bit wide memory bus, which at its reference speed of 15 Gbps (GDDR6-effective), makes 360 GB/s of memory bandwidth. The new variant comes with 8 GB of GDDR6 memory across a narrower 128-bit memory interface, with the same 15 Gbps data-rate, which works out to 240 GB/s memory bandwidth.
Besides memory size, bus-width, and bandwidth; NVIDIA hasn't tinkered with the core-configuration with the RTX 3060 8 GB. It still comes with 3,584 CUDA cores across 28 SM, which work out to 112 Tensor cores, 28 RT cores, 112 TMUs, and 48 ROPs. The GPU's base frequency is set at 1320 MHz, and boost frequency at 1777 MHz—same as the original RTX 3060. Even the typical graphics power is unchanged, at 170 W. The new 8 GB variant doesn't replace the original, but is being positioned a notch below it, possibly to compete against the likes of the Radeon RX 6600 (non-XT), and perhaps even the Arc A750.
Source:
VideoCardz
Besides memory size, bus-width, and bandwidth; NVIDIA hasn't tinkered with the core-configuration with the RTX 3060 8 GB. It still comes with 3,584 CUDA cores across 28 SM, which work out to 112 Tensor cores, 28 RT cores, 112 TMUs, and 48 ROPs. The GPU's base frequency is set at 1320 MHz, and boost frequency at 1777 MHz—same as the original RTX 3060. Even the typical graphics power is unchanged, at 170 W. The new 8 GB variant doesn't replace the original, but is being positioned a notch below it, possibly to compete against the likes of the Radeon RX 6600 (non-XT), and perhaps even the Arc A750.
35 Comments on NVIDIA Partners Quietly Launch GeForce RTX 3060 with 8GB (128-bit) Memory
Edit: Also whats with that LHR label being used even now.
I myself have Nvidia on both my systems as well, but I don't feel like supporting them in the future. Too much bs going on at the green front.
probably better to wait for the 4050 with the same memory capacity, but higher performance!
The silicon still supports LHR. If something were to happen and crypto mining regained popularity, NVIDIA could release updated drivers to stop/reduce hashrates. ETH probably can't come back because of the PoS merge but another crypto token that can be mined on GPUs conceivably could emerge.
(Not sure how NVIDIA would deal with the existence of older non-LHR throttled drivers but that's a separate discussion for another time and place.)
This way NVIDIA can say "We stated it was LHR" and no one could successfully claim that NVIDIA willfully concealed any sort of product inferiority.
Why does this still exist. Production should have stopped ages ago and inventory cleared. We should be getting a 4060 with double the CUDa core by now.
Note that Ampere was manufactured on Samsung's 8nm node. NVIDIA switched Ada Lovelace to TSMC foundries. It's highly unlikely that NVIDIA grabbed wafers destined for 4090 product to make these low-end 3060 cards. Likewise, the 3060 uses lower specced VRAM chips so there is no supply impact on the 4090's GDDR6X needs.
They are probably doing it because there is sharply decreasing market demand for the original 3060 12GB model at its original MSRP. By nerfing the memory controller and putting in less VRAM they can put out a cheaper alternative without completely destroying their retail partners' ability to sell through existing inventory of the better specced 3060 12GB.
NVIDIA is now focused on pushing as many datacenter GPUs out the door before the US ban on exports to China takes effect next year.
The weaker they make the 3060 look, the better the 4060 will look when they try to ask $599 for it next year.
Me, cynical? Nooooooooooooo!
4060 has 2X 3060 Performance
Footnote:
compared with 3060 8GB model not the 12GB
3060 12GB 380,-
2060 12GB 325,-
2060 SUPER 8GB 300,- (some leftovers)
3050 8GB 300,-
2060 6GB 270,-
what the heck will those cut down 3060 cost? between 300 and 325 ?? or 350 :laugh:
MSRP of the 12GB 3060 is still 329,- or not?
Not happy at all with 30/40 series. Although i'm gonna wait a little and see what RDNA3 delivers and how NVIDIA responds.
I don't quite get why, given how 3060Ti is incredibly beefy for a x60 Ti, would Nv disable any number of cores, surely GA106 yields are not this terrible?
And yes to your comment, this "3060 8GB" might perform half way between 3050 and 3060. Given how incredibly unexciting the 3050 is in terms of pushing low-end performance, another 128bit card using the same memory chips must be incredibly cheap to make any sense as replacement for 1650S (77% rel. perf), 1660s/ti (96% & 98%).
What would have made much more sense IMO was a 3060S, full GA107, with G6X chips.
I was intending to comment on how a logical consumer product stack of Ampere for low end, and Ada for high end would look like, but it's too much typing with the amount of marketing fuckery NV has done in the past generations...
From some past leak information, AMD might not be planning any significant upgrade for its x600 in cache size or bus width.
AMD somehow got a 237mm2 128-bit Navi23 to perform close to a 276mm2 192bit GA106.
Since GA107 is nowhere to be seen on desktop and performs worse than Navi23, and as excess RDNA2 and Ampere stock needs to be cleared, we might be stuck with Navi23 and GA106 for a while on low-end.
I'd be perfectly happy if AMD pushes a full Navi23 to something like a 7500 or XT and NV a full GA106 to 4050 non-Ti. (XT tends to mean "not cut down", and Ti "cut down from higher segment", fuck the "super")
Where I live, the 6600 at AU$329 crushes the $399 3050 and $539 3060 in value proposition. That's US$210 which I think is appropriate for the latest gen x500. Same really cannot be said for the NVidia product stack clusterfuckering going on. But used market post eth-crash looks promising already.
Oh and I leak RTX5000 here world first on TPU:
RTX 5090 Ti 24GB
RTX 5090 24GB
RTX 5090 16GB
RTX 5090 12GB
RTX 5090 10GB
RTX 5090 8GB
RTX 5090 6GB
RT 5045 4GB