Thursday, December 28th 2023
NVIDIA RTX 4080 SUPER Sticks with AD103 Silicon, 16GB of 256-bit Memory
Recent placeholder listings of unreleased MSI RTX 40-series SUPER graphics cards seem to confirm that the RTX 4070 Ti SUPER is getting 16 GB of memory, likely across a 256-bit memory interface, as NVIDIA is tapping into the larger "AD103" silicon to create it. The company had maxed out the "AD104" silicon with the current RTX 4070 Ti. What's also interesting is that they point to the RTX 4080 SUPER having the same 16 GB of 256-bit memory as the RTX 4080. NVIDIA carved the current RTX 4080 out of the "AD103" by enabling 76 out of 80 SM (38 out of 40 TPCs). So it will be interesting to see if NVIDIA manages to achieve the performance goals of the RTX 4080 SUPER by simply giving it 512 more CUDA cores (from 9,728 to 10,240). The three other levers NVIDIA has at its disposal are GPU clocks, power limits, and memory speeds. The RTX 4080 uses 22.4 Gbps memory speed, which it can increase to 23 Gbps.
The current RTX 4080 has a TGP of 320 W, compared to the 450 W of the AD102-based RTX 4090, and RTX 4080 cards tend to include an NVIDIA-designed adapter that converts three 8-pin PCIe connectors to a 12VHPWR with signal pins denoting 450 W continuous power capability. In comparison, RTX 4090 cards include a 600 W capable adapter with four 8-pin inputs. Even with the 450 W capable adapter, NVIDIA has plenty of room to raise the TGP of the RTX 4080 SUPER up from the 320 W of the older RTX 4080, to increase GPU clocks besides maxing out the "AD103" silicon. NVIDIA is expected to announce the RTX 4070 Ti SUPER and RTX 4080 SUPER on January 8, with the RTX 4080 SUPER scheduled to go on sale toward the end of January.
Source:
VideoCardz
The current RTX 4080 has a TGP of 320 W, compared to the 450 W of the AD102-based RTX 4090, and RTX 4080 cards tend to include an NVIDIA-designed adapter that converts three 8-pin PCIe connectors to a 12VHPWR with signal pins denoting 450 W continuous power capability. In comparison, RTX 4090 cards include a 600 W capable adapter with four 8-pin inputs. Even with the 450 W capable adapter, NVIDIA has plenty of room to raise the TGP of the RTX 4080 SUPER up from the 320 W of the older RTX 4080, to increase GPU clocks besides maxing out the "AD103" silicon. NVIDIA is expected to announce the RTX 4070 Ti SUPER and RTX 4080 SUPER on January 8, with the RTX 4080 SUPER scheduled to go on sale toward the end of January.
60 Comments on NVIDIA RTX 4080 SUPER Sticks with AD103 Silicon, 16GB of 256-bit Memory
4080 Super is a dud tho. The only saving grace is if it price matches 7900 XTX at 1000 but still using AD103 for 4080 is missed opportunity to use a cut down AD102 with 20GB.
So NVIDIA ends up selling a 4K card at 1000+ with 16GB still.
Based on rtx 4080 super spec, there is still room for a rtx 4080 ti as well. But so far no news about such a card. Not what i have seen so far at least.
With the rumors of rtx 4090 ti/super cancelled and the fact that rtx 4090 price has gone up again and if a rtx 4090 to actually comes out to a price like 2000 use+. I must say still even a year after. I am still glad i just went for rtx 4090 at release. As the super cards is not going to be close to as fast as 4090 and the fact that apparently 4090 ti is cancelled and rtx 5000 is still at least 1 year away from release.
Waiting for example for rtx 4080 ti. As i usually do wait for the 80 series card ti version to come before upgrade. looks to have been in vain this time if i had waited... So far.
It not to brag. But for once i am happy with the choice i made about gpu compared to other choices before. Either do to my own ignorance or unforseen out comes like the mining crase with rtx 3000 cards lagge to be in be Stock do meterial shortage and high demand for gpus do to mining and covid lockdowns.
Will it lose 0.2W/frame or not? hohohoho can't wait.
Anyway, I applaud your positivism in what is otherwise a completely pointless update.
Those extra shaders aren't going to add very much without the wider bus to feed them, and the 4080 is already short on bandwidth and VRAM capacity for its price point and target resolution of 4K with RT.
For most gaming scenarios memory bandwidth are a constraint long before capacity.
With the right memory speed these Super cards could offer a nice performance uplift.
Speaking of "duds", the RTX 4060 Ti 16 GB which so many "demanded", turned out to be totally pointless for gaming. Unfortunately, people are far too fixated on memory capacity rather than real world performance. Perhaps finally he can afford a decent shirt? :P
Historically, mid-generation refreshes have turned out to be the greatest value offerings, in terms of performance per dollar, even compared to the following generation. In addition you often get proven cooler designs and board designs, and often more well "balanced" performers with better matched memory speeds. So generally speaking, we should expect this to be a good time buying cards, even though new generations are always more over-hyped.
I'm more concerned about Nvidia diverting their attention to the "AI" nonsense though.
4070 is much faster than 3060 or 2060 !!!
u should say "Used to be x80 performance"
4070 is faster than 3080 and cost less.. also 2080Ti cost lot more and 4070 is still lot faster.
So stop trolling thx.
The point is x60 GPUs used to be as fast as previous x80 class GPUs, now the 4070 fits that profile. The performance uplift from the 4070 is anemic compared to previous generations. Not trolling, just use your head thx.
Vya Domus, you're too fixated on naming, as the product segmentation has changed a lot over the years.
If you go all the way back to like the GeForce 400 series, the 60-card was a tremendous value, the 80-card cost 150% more and only offered 50% more performance. The 70-card was closer to the 80-card in performance, but midway in pricing.
This distribution has changed over the years, and in the past two generations we've basically had "two more tiers" above that, at least if we think in terms of relative performance vs. a baseline. RTX 4090 offers about 200% more performance than RTX 4060, so proportionally the current 60-card is actually a lower mid-range (possibly upper low-end), and as I've argued in previous discussions, should have been called 4040 to make the naming fit the segmentation (so skew the lineup and rename the Ti-models).
There is also the cost aspect, and if we consider the horrendous inflation many countries have experienced in the past three years (~20% in some European countries and ~40% in the US, real inflation), I'm actually glad hardware hasn't risen more in pricing.
In terms of what it offers, RTX 4070 is arguably the best positioned card in the current lineup, and if I were building today, it would be one of my primary contender for a decent gaming rig. :)
The 3060 is somewhere in between because they put more VRAM in it, so they had to (so to speak) skimp on the rest.
Net of the fact that a clean comparison with the price cannot be made much, given the unfortunate context in which it rose...
A fully enabled AD103 is likely a much better choice, although, as you may guess, you probably won't get 10% of performance over the vanilla 4080 with this configuration. The 10 GB 3080 released Sep 1. 2020 and went on sale on the 17th, more than a full month before AMD announced the 6800 XT and almost 3 before they released the 6900 XT. You might mean the 12 GB 3080, but that's basically a 3080 Ti that didn't make the cut for some quality assurance reason. Likely the GA102's bin. The best ones with 82 and 84 functional SMs went on the 3090 and 3090 Ti, with the 3080 Ti getting one that had 80 units, 3080 12GB 70 and original 3080 68. There's a pro card (A4500) that gets only 56 SMs, being the worst GA102 bin.
At current pricing, not really. The only things to bother this GPU are:
• RX 7800 XT, yet y'know, it's already more expensive and less available, and also is an AMD GPU so it doesn't have DLSS, CUDA, power efficiency, and ray tracing performance to brag about.
• RX 6900 XT. Same crap but also even less power efficient.
• RTX 3080. Technically faster than 4070, yet it consumes almost double the power and has two less GB of VRAM so not ideal by any stretch of imagination.
• Aftermarket 3080 Ti and 6950 XT. Both are faster but they are also both used and power hogs.
And that's it. Compared to other price segments, 4070 gets you the best bang per buck as well. So it's not 4070 that's bad, it's the market that's suboptimal.
(but having a $500 GPU that handles any game at 1080p/DLSS1440p and will handle any game at 1080p/DLSS1440p for another five years is really not bad)