AMD is doing well with CPUs these days. GPUs on the other hand… they just aren’t making the same progress there.
Their software is behind in the consumer market and WAY behind in enterprise market.
Nvidia hitting production bottlenecks should help AMD move a few GPUs in Enterprise though, so long as they don’t throw away their opportunity here.
Don’t think there’s the same level of focus there. Even when it comes to AI, AMD is pushing hard with ROCm. Recently progress on that has rapidly accelerated. They’re getting deals for their upcoming data center products. In comparison, how much revenue can be expected to be made with gaming? Especially since to make a gaming GPU they sacrifice being able to make a more profitable CPU or HPC/AI Accelerator. Furthermore when you look at the size difference between AMD and Nvidia (Nvidia’s profits are greater than AMD’s entire revenue) you see why the current situation has occurred. There’s no way AMD can keep pace with Nvidia given their smaller, divided resources. Not an excuse, just the reality.
Pushing hard with ROCm?
There are millions of Devs who develop for CUDA. Nvidia I believe has north of a thousand (can’t remember if it’s like 1 or 2 thousand) people working on Cuda. CUDA is 17 years old. There is SO MUCH work already done in CUDA, Nvidia is legit SO far ahead and I think people really underestimate this.
If AMD hired like 2000 engineers to work on ROCm they would still take maybe 5 years to get to where Nvidia is now, and still be 5 years behind Nvidia. Let’s not even get into the magnitudes more CUDA GPUs floating around out there compared to ROCm GPUs, because CUDA GPUs started being made earlier at higher volumes and even really old stuff is still usable for learning/home lab. As far as I know, they’re hiring a lot less, they just open sourced it and are hoping they can convince enough other companies to write stuff for ROCm.
I don’t mean to diminish AMD’s efforts here, Nvidia is certainly scared of ROCm, ROCm I expect to make strides in the consumer market in particular as hobbyists try and get their cheaper AMD chips to work with Diffusion models and whatever. When it comes to more enterprise facing stuff though CUDA is very very far ahead and the lead is WIDENING and the only real threat to that status quo is that there literally are not enough NVIDIA GPUs to go around.
CUDA’s moat is being undone by things like OpenAI Triton. Soon most ML code will be coded in interfaces that allow for any supported hardware vendor to run the code. AMD doesn’t have to replicate all of Nvidia’s work, especially when the industry has multiple giants all working on undoing Nvidia’s software moat.
Nvidia’s dominance won’t last forever, they have the advantage but one day all this AI hardware/software will be commoditized.
The crazy thing is that’s not even really true. RDNA2 competed extremely well with Ampere. In Raster it was generally slightly better than its direct competition, and excluding the insanity of Covid and supply shortages, it was always cheaper as well.
I paid $360 for my 6700xt in Dec 2022. Now 6700xt’s are routinely like $300 on sale. The equivalent from Nvidia is… what? The 4060ti is slightly faster but only has 8gb of VRAM, and launched at $400 fucking dollars. The 4060ti 16gb is the same performance with more VRAM and launched at $500 fucking dollars.
At this point what’s propping up Nvidia is unfathomable levels of mind share. The product stack doesn’t actually deserve to sell anywhere near as well as it is currently.
It’s certainly not unheard of in the tech world for this to happen. Look at Apple releasing a fucking $1600 MBP with 8gb of RAM, and then trying to gaslight their customers into believing it’s better than it is, like they aren’t being ripped off. Shit’s absurd, but the average consumer is a pre-programmed midwit that relies almost entirely on brand loyalty to make their purchasing decisions.
And while I’m not super impressed with RDNA3 at launch, it’s significantly better as the prices slowly creep down. I kind of figure the 7800xt will slowly creep down towards $450 being a normal price, and the 7700xt will probably creep under $400.
The 7600, when priced below 250, especially in the low 200’s, like $220-230, is fantastic value as well.
Even at the high end, the 7900xt at $700-750 is a solid buy, and the 7900xt at $900 or less is very good as well.
Lovelace is fantastic from a technical perspective, but Nvidia outdid their own previous records for greed when it comes to the product stack and the pricing.
Radeon really needs their Ryzen moment
Nvidia is all hands on deck going pedal to the metal right now trying to stop AMD from gaining any market share right now. They’re worried about becoming a victim of their own success here and the AI boom allowing AMD to gain a foothold in AI with their open source strategy. They’re also worried about Intel and Google for similar reasons.
Nvidia is quite the formidable foe especially compared to Intel, and they have a massive headstart because they have a LOT of advantages beyond having merely the best hardware on the market, but I’m still a bit bullish on AMD’s chances here.