People who complain about Nvidia pricing and threaten to go to AMD never actually do it, they just want AMD to undercut Nvidia and force Nvidia to lower prices so they can buy Nvidia again.
Some theories say they control inventory a lot better.
By making it seem like a commodity/not as readily available as AMD GPU's they end up selling more. It's obvious that they have a not of RTX 3000 stock now, but instead of dump they'll make RTX 4000 more expensive. With RTX 2000 and 3000 it's obvious they never wanted to flood the market.
They also sell directly to a lot of OEM's, as well as miners (before the crash).
Better hashrate and better resell value.
Seeing these used to pay themselves, in the long run better made it more profitable per unit.
They are, but FSR2 is closing the gap on DLSS.
It's not that they needed to use tensor cores for DLSS as you had DLSS 2.0 running on regular shader units on Control before they revamped it to require them. It's just that it was free to do DLSS there seeing it wasn't being used for anything else in a gaming scenario.
Nvidia using it for that though, most likely limits it's use for AI or other game related task if developers wanted to. So AMD doesn't really need to have them if they have more performance in the first place. For games that is.
Tensor Cores of course are valuable for professional workloads and AMD doesn't have a solution for that.
With RT, AMD's solution is seemingly similar, adding ray accelerators to the CU's. It can pay off later on as well against dedicated units, if the GPU's have overhead that is. Again, it's "free" on Nvidia because they'd be sitting unused otherwise.
They should, yes.
Undercutting them in price shouldn't be hard though.
In contrast, Nvidia has almost never dropped the ball so even if Radeon innovates, GeForce is still ahead.
In other words, maybe AMD just got lucky against Intel and it's unlikely to happen against Nvidia.
I can kinda agree with that since I've also been using both for two decades now. But I think people tend to conflate stability issues with performance issues and I don't think nvidia has been any better in terms of stability. I think both are actually quite fine in that regard (overall) but nvidia's drivers have caused me more than a few BSODs when their cards are still new. Things like sleep mode causing the KM driver to crash.Yeah I've actually jumped between NVIDIA and AMD quite a lot over the years and I can say I've never had any issues with NVIDIA GPUs but have semi-regularly had problems with AMD (ATI at the time) cards including driver issues and heat issues.
Its not luck if you actually innovate and the competition does not push forward. And unlike Intel Nvidia seems to be moving forward every generation.
This is a bit off on a tangent from the main topic but I haven't seen it mentioned elsewhere on Gaf:Honestly i don't like having 30% of the silicon dedicated for AI/RT, Nvidia also increased the L2 cache
by massive amount which will eat even more silicon, if they made a GPU without all that specialized
silicon we could end up with much faster GPUs that can handle 4k with ease also nvidia is using this
tech to make their previous GPUs obsolete very fast
Branding and marketing plays a role but believe me, no marketing gonna help if ur card crashing like a mofo coz of unstable drivers like was so often case with amd products(and yup even if 5% of their cards crash like that, it can make huge wave, customer lost once not to price or performance but actual usuability/stability of a card is customer u lose for a very long time if no forever.NVidia branding and marketing is a huge advantage. Just like it was against Intel on the CPU front, AMD only has a chance if they completely outclass NVidia.
That sounds like a sound strategy if you need those features, but while nvidia has the creativity to go ahead and implement these "next gen" features in their hardware, they're always proprietary leaving the rest of the industry to figure out how to do it in an open environment.People just trust Nvidia more when it comes to software support and support for new features.