Nvidia GeForce RTX 40x0 rumors and speculation

Yep. A full $300 increase from the 3070 Ti. Nvidia seems to be overplaying their hand here. Maybe there are still lots of 3080s sitting on shelves.

I don't know about other places but here there's no 3080 or 3090 to be found for quite some time. There are slightly more 4090 now but they are still mostly limited to one per customer for now. 4080 16GB are quite plenty as always.
 
Yeah, GA102 seems pretty much cleaned up. There are likely still some in the channel, but the main concern would be GA104 now.
Assuming the RTX4070Ti performs around the RTX3090Ti, pricing it at $699 would put too much downward pressure on the remaining GA104.
Even a $799 RTX4070Ti would likely move the ~$600-700 RTX3070Ti down to <$500, since the 4070Ti offers ~3-4 tiers more performance.

It would make more sense, IMO, to sandwich the AMD cards.
Price RTX4070TI at $849 and RTX4080 at $1049-$1099.
I personally think the $999 is a good price point for the 7900XTX but the $899 7900XT baffles me.
I was really expecting $799 or $849 but maybe that weird pricing is intentional.

Nvidias harvesting/binning is likely going to cause some crazy naming schemes, I wonder if we will see both Supers and Tis again.
 
Yes, which means that it would be nothing more than a rename which resulted in a delay.
Hope everyone who weren't happy with a "3080 12GB" naming will be happy now ¯\_(ツ)_/¯

If this happens then hopefully it'll get torn apart by the media and no-one will buy it like the 4080.

I was absolutely committed to getting an NV card this gen but if they really go ahead with this then I might just end up going AMD out of principle.
 
I don't know about other places but here there's no 3080 or 3090 to be found for quite some time. There are slightly more 4090 now but they are still mostly limited to one per customer for now. 4080 16GB are quite plenty as always.

Interesting, there's still a LOT of 3080's in the US.

Regards,
SB
 

Which would basically make it a 3070 with higher clocks, bigger caches and a 192-bit / 12GB memory.
 

Which would basically make it a 3070 with higher clocks, bigger caches and a 192-bit / 12GB memory.

Considering there's a good chance it'll be priced higher than a 3080 while also being slower than it, or at least only in the same ball park, this does not seem like a good value proposition at all.
 
Considering there's a good chance it'll be priced higher than a 3080 while also being slower than it, or at least only in the same ball park, this does not seem like a good value proposition at all.
It's a salvage die based on AD104 which is what? 250mm? We'll see but I think it can be priced rather aggressively if there will be some pressure from N3x N6 dies.
 
This depends on perspective and of course how things eventually shake out.

Personally speaking I feel pushing the price of entry for >8GB, and especially that important 12GB demarcation line, makes this better than the old rumoured configuration with more SMs but only 10GB. Ultimately the SM count will still be the biggest factor in determining relative performance, and therefore relative pricing and increases the likelihood a lower price.
 
nVidia has easily won the notebook round. Efficiency is 30%+ better on Lovelace under full load. So much to "piss poor" efficiency...
I will withhold judgment until smaller chips, which would be, you know, actually put into notebooks.
 
People that were claiming Lovelace would have poor efficiency were simply saying, "I dont know how efficiency works" the entire time.
nah, it's more like "we see how power hungry Ampere's high end is getting, and Nvidia has thrown efficiency out the window before to maintain performance supremacy, so I don't trust Nvidia to focus on efficiency based on the information we have"
we were proven wrong. kudos to nvidia, they defied our expectations for ADA's efficiency.
 
nah, it's more like "we see how power hungry Ampere's high end is getting, and Nvidia has thrown efficiency out the window before to maintain performance supremacy, so I don't trust Nvidia to focus on efficiency based on the information we have"
we were proven wrong. kudos to nvidia, they defied our expectations for ADA's efficiency.
They were using a Samsung 10nm family process with Ampere. We all knew that's exactly why Ampere lacked some efficiency at the high end(not that it was ever super bad). Making the move to TSMC 5nm was always going to result in an absolutely massive improvement here. It's a similar level of node improvement that we had from 28nm to 16nm with Maxwell to Pascal.

There should have been no reason to expect that Lovelace would be some unreasonable power hog.
 

So roughly half way between a 3060 and 3060Ti in SM count and only 8GB VRAM.

Looks like another huge disappointment from Nvidia if true.
 
Back
Top