Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
You have to realize how silly the method you are using really is. Let's say the full Ada die had the same transistor count as Turing did, and then they gave you the 4090 using the full, 100% die. According to your method, that would be FANTASTIC. Even though that hypothetical 4090 would be much much much slower than the one you are currently getting.
Well said. Puts it all into perspective. Clearly I’m favouring the latter (70% of rich persons wealth lol)
That’s a good way to base purchase decisions.
The 3080 die cost around $65 to make assuming a 90% yield, 4080 die built on a AD102 would cost $197 with a 90% yield so an increase of $132 but then the card is $500 more so plenty of profit in there for Nvidia.
Wasn't my method, you brought up the transistor count somehow being related to how big the die is. If they used "the full Ada die" it would be at the maximum reticle size of (iirc) 300mm², there's no relation between transistor count and the size of the die, you could make a 300mm² die with one transistor on it if you wanted.
Nvidia is not selling you dies. They are selling you gpus, and you are buying them for the performance metrics. Again, let's flip the script. Let's say nvidia made a massive Uber expensive die that costs them 900$ to produce. If it underperformed massively, would you be paying 1.5k to buy just because it cost a lot? Of course not. So why do you care about the die cost now?
And that's the only method that makes sense. Comparing dies is complete nonsense
Exactly, so why are you comparing % of dies between models as if It means anything?
Because the costs of the die dictate if it's good value or not, like you say if it was a 300mm² die costing $900, had 1k transistors on it, and performed like a GeForce 256 then nobody would buy it because it would be bad value. Like i said transistor count is meaningless in terms of size, cost, performance and pretty much everything. The size of the die is only relevant in terms of cost.
I'm not.
I’d argue that most of the ADA gpus do under perform in terms of their costs, for the new higher prices they are asking we should be getting +100% performance at each tier.
yeah but they can make even more money by not doing that cause **** you thats why
So that's what we should be talking about. Performance and price. Who gives a duck about dies transistors and cores. The 4090 is a great card and I don't care what % of the full die it is. It could be 1%, don't matter to me.
True, however die size dictates price or at least value. If it cost $9000 to make a 300mm² die then few, if any, people are going to buy them.
A 4090 offers +70% performance for +7% more cost, a 4080 on the other hand offers +48% more performance at +72% more cost, can you not see just how poor that is?
Sure never said the 4080 is great value.
AMD are actually slightly better but still really poor, at least with their new stuff.
Yeah with the price drops sure, but on launch holy cow they were bad. Which is my point, as offensive as the 4070tis price was, it basically offered better performance per dollar in both rt and raster than the 7900xt, while consuming much less power and offering more features. I would honestly feel like an idiot complaining about the 4070ti price back then, yet people did.
The 4070ti is bad regardless, just because AMD was worse at that time didn’t suddenly make the 4070ti any better.
Of course. That doesn't stop me from feeling it's stupid to complain about the lesser of 2 evils.
