The stuff Hacker News actually recommends.

GPUs and CPUs

AMD Radeon vs NVIDIA GeForce RTX

Hacker News owners lean toward the NVIDIA GeForce RTX.

AMD RadeonNVIDIA GeForce RTX
Owners with an opinion232617
Positive verdicts59% of 19076% of 416
Positive/negative since 202462/45214/71
Happy after 3+ years710
Longest happy owner14 yrs6 yrs
Praised forperformance and reliabilityperformance and value
Criticized forsoftwarenoise
Newest comment

Updated

What owners wrote about both

AMD Radeon: Doesn’t like itNVIDIA GeForce RTX: Likes it
Because M3 Pro is not a real thing as far as actual agentic workflows go. I run dual R9700 boxes those idle at 150W Because of a Ryzen AM5, and I run Spark boxes, which are decent But still idle at 45-50 watts a pop. My favorite, to be honest, is a 5090 with a Qwen 27B because that one is good for quick hitters and flies, But again, the box itself idles at 140 watts. So it's really the idle power that I don't like And it's too much of an inconvenience to power boxes down and power them back on, so they just end up running and sucking electricity. I am working on getting some sort of a smarter proxy setup where I would give boxes time to wake up and go to the cloud while they're waking up. So the whole point of having these things running in the background is that you do want them to be almost always available. So power consumption is definitely an issue.
AMD Radeon: Likes itNVIDIA GeForce RTX: Doesn’t like it
Personally I think the price is way too high right now. It’s a power hungry gaming GPU. The efficient single card equivalent would be a 4500 Blackwell which launched at about $3500. Or you could get a 9700 32GB or an Arc B70 for well under $2k, today. You only buy a 5090 if you want absolute speed. 32GB is still not that much. I would rather get a Spark and have the RAM to experiment with larger LLMs, even if it was slow.
AMD Radeon: Likes itNVIDIA GeForce RTX: Doesn’t like it
I had a laptop with a mobile GeForce RTX 4060 - it was a pain to run Linux on, mainly because of Nvidia. I got rid of it and built myself a PC (not just because of Nvidia, gaming laptops are just a disappointing experience). And went all in with Team Red - Ryzen and Radeon. It's a joy to run Linux on, I work on it and have played several games - 0 problems. Nvidia may be the leader in performance, but if it means you're stuck with Windows - I'd say it's not worth it.
AMD Radeon: Doesn’t like itNVIDIA GeForce RTX: Likes it
Just an FYI, this is writeup from August 2023 and a lot has changed (for the better!) for RDNA3 AI/ML support. That being said, I did some very recent inference testing on an W7900 (using the same testing methodology used by Embedded LLM's recent post to compare to vLLM's recently added Radeon GGUF support [1]) and MLC continues to perform quite well. On Llama 3.1 8B, MLC's q4f16_1 (4.21MB weights) performed +35% faster than llama.cpp w/ Q4_K_M w/ their ROCm/HIP backend (4.30MB weights, 2% size difference). That makes MLC still the generally fastest standalone inference engine for RDNA3 by a country mile. However, you have much less flexibility with quants and by and large have to compile your own for every model, so llama.cpp is probably still more flexible for general use. Also llama.cpp's (recently added to llama-server) speculative decoding can also give some pretty sizable performance gains. Using a 70B Q4_K_M + 1B Q8_0 draft model improves output token throughput by 59% on the same ShareGPT testing. I've also been running tests with Qwen2.5-Coder and using a 0.5-3B draft model for speculative decoding gives even bigger gains on average (depends highly on acceptance rate). Note, I think for local use, vLLM GGUF is still not suitable at all. When testing w/ a 70B Q4_K_M model (only 40GB), loading, engine warmup, and graph compilation took on avg 40 minutes. llama.cpp takes 7-8s to load the same model. At this point for RDNA3, basically everything I need works/runs for my…
AMD Radeon: Doesn’t like itNVIDIA GeForce RTX: Likes it
Same, but R9 380X back in college. Bang per buck it seemed better than NVidia. But despite my distaste for NVidia back then, after that experience, I resolved to never get an AMD GPU again unless things drastically change. Drivers would constantly break something, some games would run poorly and I would have to fiddle with driver versions to address this (often involving annoying version rollbacks), it would randomly start overheating for no reason, sometimes it would just crash the system, etc. Upgraded to 480 later, and the exact same story repeated. You know what happened once I got an NVidia GPU years later (1080Ti)? I never had to think about the GPU ever again. I install a driver update occasionally, it “just werks”, and I would go on my merry way. Upgraded to a 3080 about half a year after it got released, and had the same story - it just works, and I never have to think or worry about it ever. I would love to support AMD GPUs, but I cannot, in good conscience, buy their GPU again and waste time troubleshooting it and stress over every driver update. On the flip side, AMD CPUs are amazing. Been with them since the original Ryzen release, upgraded once (from 1700 to 3900x), no issues whatsoever and am very happy with those. My next build will be an AMD Ryzen CPU as well, but it will have an NVidia GPU.
AMD Radeon: Doesn’t like itNVIDIA GeForce RTX: Likes it
In current state Nvidia has no competition, and you buy AMD only because you hate Nvidia, not because it's better. I had tons of AMD cards my first being ATI 9000 but if you want to do more than game the hassle is not worth it. My last AMD was Radeon VII, once I got a bit serious into Blender there's just no comparison, even with enterprise drivers the crashes, random issues and slowness comparatively is just not worth it. What took me 3 minutes to render on Radeon VII takes 10s on 3090 Ti, StableDiffusion renders take 5-6s without any hassle playing with rocm, gaming is also no comparison with RTX (I don't even use DLSS). Fun fact I sold my RVII after 3 years and added $500 for a new 3090 Ti. Nvidia sux for their business practices but technically they dominate because they invested in software really early on and established themselves with Cuda, OptiX, RTX, DLSS. Older AMD cards are nice for hackintosh if you're into that though, Apple dropped Nvidia hard. Also the Linux driver blob thing if you want to be a purist, but IIRC it's supposedly changing (sorry don't have a source right now).
AMD Radeon: Doesn’t like itNVIDIA GeForce RTX: Likes it
> That just lives AMD as competition, and only if RDNA 3 comes out a winner. The RX 6650 XT is in the same price bracket as the RTX 3050. I'm not sure they would be considered competition even if RX 77whatever were in the same price bracket as the RTX 3050. Nvidia just seems to push just the right buttons on the software features which AMD just cannot seem to get. Don't get me wrong. I would love AMD to be competition. They should frankly team up with Intel. AMD has the hardware and Intel can provide the software. Don't see that happening considering AMD went alone on their FP8 proposal and everyone else joined up on a separate paper.
AMD Radeon: Likes itNVIDIA GeForce RTX: Likes it
Modern GPUs like RTX 5080 are much faster for the applications that are limited by computational capabilities, mainly because they have more execution units, whose clock frequencies have also increased. I suppose that most games are limited by computation, so they are indeed much faster on modern GPUs. However, there are applications that are limited by memory throughput, not by computation, including AI inference and many scientific/technical computing applications. For such applications, old GPUs with higher memory throughput are still faster. This is why I am still using an old Radeon VII and a couple of other ancient AMD GPUs with high memory throughput. Last year I have bought an Intel GPU, which is still slower than my old GPUs, but it at least had very good performance per dollar, competitive with that of the old GPUs, because it was very cheap, while the current AMD and especially NVIDIA GPUs have poor performance per dollar.
AMD Radeon: Likes itNVIDIA GeForce RTX: Likes it
If unfamiliar: it is a big deal that AAPL & NVDA again have an official relationship. For well over the previous decade Apple has not allowed newer nVidia GPUs (by not allowing drivers). A seven year old GPU (e.g. VEGA64, RTX1080Ti) can still process more tokens/second than most Apple Silicon (particularly the lower-ends). As discussed elsewhere, Apple MAX/Ultra processors are best-suited for huge models (but are not as fast as e.g. RTX5090).
AMD Radeon: Likes itNVIDIA GeForce RTX: Likes it
I used an AMD Radeon 7770 for years, then I upgraded to a GTX 1060, the 7770 became my younger sister's GPU for her PC. It has only recently gave out, after some 14 years of service. Likewise for the 1060, its still going strong. I upgraded to a 3060 round about the time my younger brother decided he wanted a PC so he is now using it without any issues. About 10 years use out of it, plenty more to go. GPUs are pretty damn resilient if you aren't pissing around with them.
AMD Radeon: NeutralNVIDIA GeForce RTX: Neutral
I have an AMD 290 or whatever GPU to run NixOS (will be upgrading to Intel soon) and a 3060 that I pass into a Windows VM for gaming, I feel very rational. Unused PCIe slots in a laptop is hard, haven't tried but I imagine Thunderbolt could work for this purpose.
AMD Radeon: Likes itNVIDIA GeForce RTX: Likes it
Contrary to what other posters are saying, I do think AMD is actually trying to compete; for example, they finally gave up on OpenCL and are trying to play catch-up to Nvidia via ROCm, effectively allowing you to run CUDA on AMD's server GPU offerings. And on the gaming side their top-of-the-line offering, the 7900 XTX, is ~$600 cheaper (aka 60%!) at MSRP than Nvidia's top-of-the-line RTX 4090 — and the prices you'd pay from e.g. resellers on eBay are even more skewed towards Nvidia. That being said, they're just way, way behind. Nvidia can charge what they do simply because AMD isn't great at making GPUs. In gaming AMD never caught up to Nvidia's raytracing perf, and AMD's attempt to clone DLSS is significantly less performant. Meanwhile in AI/ML the industry consolidated around CUDA because OpenCL was never that great, and — Nvidia owning the standard, and having built around it for over a decade, gives them a really big advantage. Plus, every GPU sold in the last decade+ by Nvidia supports CUDA, whereas only the new server GPUs from AMD support ROCm. Nvidia has executed incredibly well for decades: they pioneered shader programming, CUDA, raytracing, and DLSS. Arguably they even invented the GPU itself with the GeForce 256 in 1999 (as opposed to earlier "accelerator" cards). I struggle to think of anything AMD has done in the GPU space that matches even one of those accomplishments... Their only slight win is having OSS drivers for Linux, but honestly, that doesn't sell a…