schmidtbag:

Why are the 4080s so different? The 16GB seems like it'd be a 4080 Ti.

Caesar:

If the number is called in the USA, the recorded message "Tell us, how fast would you like to go?" can be heard https://pics.computerbase.de/1/0/5/0/3/0-7ffe720534b675df/3-1080.3706703a.jpg

Horus-Anhur:

I fear ProjectBeyond will result in more price increases for GPUs. Again. https://media2.giphy.com/media/3o6wrebnKWmvx4ZBio/giphy.gif?cid=ecf05e47etz0kugp7u6jkoxqtamtr2685tfopcnd2msdig24&rid=giphy.gif&ct=g

cucaulay malkin:

9700 cuda vs 16300 isn't a 4080ti

schmidtbag:

Huh? Are you looking at the right specs? It's 9700 for the 16GB and 7600 for the 12GB. That's a 12% difference, which is significant. Not only that but the memory size, memory bandwidth, and power consumption are all pretty dramatically different.

schmidtbag:

Huh? Are you looking at the right specs? It's 9700 for the 16GB and 7600 for the 12GB. That's a 12% difference, which is significant. Not only that but the memory size, memory bandwidth, and power consumption are all pretty dramatically different.

cucaulay malkin:

are you ? 4090 has 16300 cuda. 9700 cuda card isn't a 4080ti if full 4090 has 16300. more likely to have at least 14000.

Denial:

4080 TI (9700 CUDA) 4090 (16300 CUDA)

cucaulay malkin:

makes zero sense, btw you know the 9700 cuda card is a different die ? no cut down ad102 card ? surely there will be one

Denial:

Why does that make zero sense? Wouldn't it being a different die make it make more sense?

cucaulay malkin:

zero sense

Denial:

one hundred cents

Denial:

That being said I'm pretty sure Nvidia has had previous offerings with same name, different CUDA configurations - 1060 3GB vs 6GB comes to mind. I just think it's bad practice, especially when you have a Ti moniker that everyone is already aware of.

cucaulay malkin:

zero sense

cucaulay malkin:

zero. it would mean that all not fully functional ad102 won't be sold as 4080ti and also that 4090 will have 70% more shaders than 4080, for 3090 vs 3080 it's like 20%.

schmidtbag:

Indeed they did have different CUDA configurations. It was about a 10% difference, or only about 100 cores. The difference in CUDA cores between the 4080 12GB and 16GB is enough to make a whole RTX 2050. A 2050 is no slouch. But unlike the 1060s, there's clearly more happening with the different 4080 models, when you consider the 30% increase in power draw, despite the 12% increase in CUDA cores. How mature. As I just mentioned, it's not just the amount of VRAM and CUDA cores. The 16GB is bound to be more than 12% faster. No matter what, there is going to be some hefty gap between the 4080 (of either size) and 4090. Nvidia knows that at this price point and scale, the 12GB will not be of any interest if it's so close in performance. That being said, I still can't help but wonder who would buy the 12GB model over the 16GB. They're both most likely going to have quadruple-digit prices and consume a lot of power, so, it's not like the 12GB is going to be enticing to anyone on a budget.