6 ms·

Always perplexed me why nobody else kept up with Nvidia for the last decade in GPUs. Gaming has been a lucrative market. Why is there only Nvidia?

packetlost 3y ago
There's not? Both AMD and Intel have GPUs that are competitive in certain market segments. AMD in particular is pretty much even in terms of rasterization performance on the high end. Intel is just new (again) to the game of discrete graphics. Nvidia, on the other hand, invested tons and tons of money to make CUDA good over the last like 15 years, while AMD has not really put any serious effort into AI/ML workloads. It's mostly a software support problem, not a hardware problem to my understanding.
- Lord-Jobo 3y ago
  AMD is even on performance and significantly more affordable for consumers. but people wont stop buying nvidia's ripoff products like the 4060. really frustrating
  - bryanlarsen 3y ago
    Even more frustrating is on the Linux side. We've had a couple of threads about how bad Wayland is, when most of the problems are due to Nvidia binary drivers. It's been 11 years since Linus gave Nvidia the finger. Nobody who is thinking of possibly running Linux should ever buy an Nvidia product. Yet Nvidia is still very commonly spec'd into Linux builds even today.
    - shmerl 3y ago
      On Linux people are clearly leaning to AMD for a while already, especially when gaming is concerned: https://www.gamingonlinux.com/index.php?module=statistics&view=trends#GPUVendor-top https://www.gamingonlinux.com/index.php?module=statistics&vi... In the Windows land it's indeed mostly Nvidia.
      - robotnikman 3y ago
        Ironically, ARM drivers for Nvidia are available for Linux, but not for Windows
    - Filligree 3y ago
      AMD cards are _very bad_ at running CUDA. If you're running linux on a desktop, there's a decent chance you're the type who wants to play with ML as well...
      - latchkey 3y ago
        AMD has a translation layer called HIP as part of the ROCm toolset. This doesn't get us 100% of the way there, but it is an alternative that is gaining traction as they work to improve it.
    - smoldesu 3y ago
      KDE's Wayland session is fine on Nvidia, at least with my 3070. There used to be more visual artifacts, but everything I'm using works, from CUDA to Mesa to DLSS. The largest remaining bug I'm aware of has to do with how blurred windows are presented. It's pretty functionally solid, at least with newer drivers.
      - bryanlarsen 3y ago
        I am still getting a lot of visual artifacts on a 970.
    - anon291 3y ago
      I've used NVIDIA on linux for decades and the support has always been great in my experience
  - jfim 3y ago
    For ML workloads, AMD wasn't competitive due to lack of software support for PyTorch. Nvidia isn't making most of their profit from selling individual GPUs to gamers, they're making money from data center GPUs that are bought by the thousands at five figure prices each for ML training.
  - opportune 3y ago
    I made the same argument to myself when I bought an AMD GPU for gaming 4 years ago. What doesn’t show up in a side by side comparison is that AMD’s drivers (or Windows’ treatment of them) fucking suck. Until I bought an NVIDIA replacement recently I was very frequently having to reinstall the drivers and also intermittently encountering driver crashes in games and browsers. That’s not to mention having to fiddle with the drivers/games more than I was accustomed to, which IIUC is more a consequence of having less market share and thus not being a priority for game developers to work out of the box. I can excuse that, but not crashes or reinstalling every couple months.
    - tokai 3y ago
      On linux its the opposite. Its was such a relive to get an AMD card and finally have a good stable driver.
    - kulahan 3y ago
      Heh, I remember this being the reputation they had back in the early 2000s. Guess they never got past it.
    - gabeaweg 3y ago
      Completely agree. I bought an AMD GPU like 6 years ago. The software support fucking sucked. The drivers were bad, games weren't optimised for AMD. AMD looks great on paper in benchmarks. As a product, overall, it is bad. That's why people keep buying nvidia. I will gladly pay double for a product that works.
    - fenesiistvan 3y ago
      I am using a Radeon RX 6900 XT GPU since two years and never ever had any stability problem or crash running my PC ~10 hours/day, playing a some (various, demanding) games almost every day.
  - dotnet00 3y ago
    AMD might've gotten more traction if it wasn't schizophrenic with regards to compute support on its hardware. GCN was such a great architecture for heavy compute, yet it was mostly wasted by shit software. Crypto really only managed to use it because there was more money to be made using AMD, justifying the wrangling needed to get compute to work on it. Now they continue to waste their potential with ROCm's design choices which make its hardware support range very limited.
    - latchkey 3y ago
      In my eyes, as someone who ran 150k+ AMD gpus for crypto... it really boiled down to being able to heavily tune various aspects of the GPUs. Each GPU is a snowflake and being able to tweak the knobs individually, really made a huge difference on performance. Super low level performance enhancements in code helped a lot too. I worked with some of the popular mining apps to improve things there. AMD is already backporting ROCm to "lesser" hardware [0]. I wouldn't expect them to spend too many resources on going back further than that though. I'd rather see them improve their newer stuff. The larger issue is that it is impossible to rent time on the super high end GPUs that they use to build the super computers with... nobody offers that and it is something I'm working on fixing. [0] https://www.tomshardware.com/news/amd-enables-rocm-and-pytorch-on-radeon-rx-7900-xtx https://www.tomshardware.com/news/amd-enables-rocm-and-pytor...
  - vinculuss 3y ago
    The thing I hear a lot from people who skew towards AMD is that raytracing and DLSS and all the software solutions that nvidia has are gimmicks, or somehow actually not a big deal. They’re stuck on the concept of rasterized graphics being the ground truth in gpu value. Then they’re shocked when the wider consumer market doesn’t hold their perspective. AMD hasn’t had the big picture vision that nvidia has had. For example, with Vega AMD had a gpu architecture with exceptional async compute performance a generation before nvidia was competitive. AMD then abandoned Vega, switched to a leaner and less compute friendly rDNA architecture while nvidia went in the opposite direction, pushed compute heavy features like raytracing and dlss and AMD hasn’t had an exceptional response since then.
    - Sohcahtoa82 3y ago
      Raytracing being a gimmick is highly subjective. Creating reflections using rasterization requires shortcuts and sacrifices. And once you see them, you can't unsee them, and then you find them visually distracting. There are basically 3 ways to create reflections without ray-tracing: 1. Double the geometry and render it in the reflection. This is rarely the approach taken because it's incredibly resource intensive. Oddly, Duke Nukem 3D back in what, 1996? did this, but that had pretty simple graphics. Genuinely not feasible for any sort of reflect surface that is curved. 2. Screen-space reflection: Just take the rendered image, invert it, and overlay it onto the reflective surface. This is probably the most common way to do it. It's fast and looks decent. The problem is that it requires the reflected image to be visible on the screen. When it's not, there's no reflection. It creates a bizarre effect when, for example, you're looking over a body of water at something in the distance, and you see the reflection in the water, but as you move the camera down, the reflection gets cut off and eventually disappears entirely. 3. Using a static image as an environment mapped texture. Apex Legends does this for shiny surfaces. It creates a great illusion of a reflection, but taking even a second to actually look at it, you'll notice that the reflection doesn't actually show what it should unless you're standing in the exact right place. I will concede that gamers typically don't notice the issues created using methods 2 and 3, especially in a game like Apex Legends where you're moving around too much to notice. But me? I dunno. I notice and get distracted. EDIT: DLSS, on the other hand, is basically a crutch. It's not surprise that DLSS became a big thing at the same time RTX did. RTX requires more juice to run at full resolution than the GPU has, so they introduced DLSS to render at a low resolution and upscale it.
  - paulmd 3y ago
    4060 is one of the better-value options in the stack, actually. The RX 7600 is still on 6nm and is no faster, much less efficient, and skips many of the microarchitectural improvements to rdna3 to try and keep cost down. $250-270 is fine for this type of product even today. Similarly, the 4070 is also fairly attractively priced even compared to 6800XT/6950XT or 7800XT. More efficient, better feature set, same performance tier, similar perf/$. But people still get upset about intangibles like memory bus or die size (despite this already being measured in the performance!), or just don’t want to admit that wafer costs and R&D costs are rising in the post-Moores law era. Operating margins were actually down relative to the last few gens (until AI hit), and were broadly similar to 2012 levels. This is just what it costs now, and smaller die take the worst of the cost increases because memory controllers and pcie don’t shrink anywhere near as much as logic (the gpu cores). There is a “minimum cost floor” that has been rising substantially due to this.
- roughly 3y ago
  CUDA’s definitely the “moat”, in the Warren Buffet sense - it’s not impossible for someone like AMD or Intel to make something CUDA compatible, but it might cost enough to wreck the economics of trying to compete in the space.
  - LeanderK 3y ago
    I don't think you will need something coda-compatible. Apple has relatively solid pytorch support without copying CUDA, but PyTorch+ROCm is still not working on windows (and I think there's no ROCm support for the consumer GPUs from AMD). I think Apple proves that it is possible, and possible relatively quickly.
    - kimixa 3y ago
      Pytorch supports directML on windows - is that not sufficient (or performant) for your usecase? It feels like AMD are less willing to step on Microsoft's toes, possibly as they just don't have the market share clout. And Rocm supports the higher RDNA2 and the highest-end RDNA3 consumer GPUS - [0], though only the top RNDA3 cards in linux [1] (officially - other cards often do "work" but YMMV) One thing Nvidia are much better at is supporting the whole stack - while I doubt many people are actually running much ML on 8gb mid-low end cards, it seems a hook for sales if people "could" do that, and a starting point for personal experimentation. I know multiple people who buy low end Nvidia GPUs because they "might" experiment with CUDA some time in the future. [0] https://rocm.docs.amd.com/en/latest/release/windows_support.html https://rocm.docs.amd.com/en/latest/release/windows_support.... [1] https://rocm.docs.amd.com/en/latest/release/gpu_os_support.html https://rocm.docs.amd.com/en/latest/release/gpu_os_support.h...
      - LeanderK 3y ago
        > One thing Nvidia are much better at is supporting the whole stack - while I doubt many people are actually running much ML on 8gb mid-low end cards, it seems a hook for sales if people "could" do that, and a starting point for personal experimentation. I know multiple people who buy low end Nvidia GPUs because they "might" experiment with CUDA some time in the future. but this is very important. Many develop locally before sending it to a cluster, especially if you only have limited credits available.
- tverbeure 3y ago
  It’s important to have a vision of what you want to do as a company. For as long as I can remember, Jensen has been telling us during all hands meetings that Nvidia is a visual computing solutions company, not a semiconductor company. (The “visual” part has only been dropped in the past few years.) Nvidia sells hardware, but they also sell complete solutions, and everything in between.
  - J_Shelby_J 3y ago
    Indeed. I don't really care what GPU I use. I do care that Geforce experience works flawlessly. I mean, it doesn't, and my grandkids will probably still use DDU, but when shadowplay works it's indispensable as a gamer (or even for recording video calls.) I'm sure AMD has an alternative or I could use OBS, but I'd rather just pay a premium and rarely have to think about it. And also CUDA.
- latchkey 3y ago
  AMD is coming out with the MI300x on Dec 6th, which has 192GB and runs on the same hardware platform as H100s (OAM-UBB). They've doubled down on ROCm/Pytorch and other projects. Don't expect them to stay behind for much longer.
  - packetlost 3y ago
    One can hope!
    - latchkey 3y ago
      Doing everything I can to help them, I don't see it as a need for only hope. =)
  - esafak 3y ago
    I see MS (and AWS?) support it. https://techcommunity.microsoft.com/t5/azure-high-performance-computing/azure-announces-new-ai-optimized-vm-series-featuring-amd-s/ba-p/3980770 https://techcommunity.microsoft.com/t5/azure-high-performanc...
    - latchkey 3y ago
      Yup! Part of a partnership. Exciting news.
djyaz1200 3y ago
Agree, can someone(s) more knowledgeable about this field share their thoughts on how defensible Nvidia's position is in the short and long term? Like are companies like this a credible threat? https://www.cerebras.net/ https://www.cerebras.net/ I hear a big part of Nvidia's lock-in is that they provide software that runs these AI projects?
- faeriechangling 3y ago
  They also own a high end networking company and will sell you an entire rack or datacenter integrated for you. Also yeah, the decade or so they’re had thousands of in-house engineers and people around the world working on CUDA is HUGE. Also the crypto boom encouraged them to buy up tons of fab capacity. Nvidia had foresight on multiple levels here. Nvidia’s problem is that every company is a credible threat because in the ongoing gold rush they can’t make enough shovels to meet demand and damn there is gold in them thar hills. People are going to do literally whatever it takes to make their competition’s products work. If you are selling a shovel, even a shitty shovel, you are a competitive threat to Nvidia right now.
jdiff 3y ago
AMD and Intel are competitive in gaming. Where Nvidia excels is other types of acceleration. AMD and Intel can't hold a candle to Nvidia when it comes the raytracing renderers used in every 3D and CGI workflow, for example.
AlexandrB 3y ago
There's also AMD - but it seems like a hard industry to get into. Intel keeps trying, most recently with Intel Arc[1], but they've never been able to (or perhaps wanted to) crack the high end. [1] https://www.tomshardware.com/news/intel-arc-graphics-cards-are-the-best-deals-this-black-friday https://www.tomshardware.com/news/intel-arc-graphics-cards-a...
- Night_Thastus 3y ago
  Saying they've "never been able to" is a kind of weird perspective. They're really only just starting on their dedicated GPUs. The Alchemist architecture is their very first attempt, and they've been putting a lot of effort into repairing and refining the drivers. The hardware is actually pretty decent when it's firing on all cylinders. But Nvidia and AMD have a huge time advantage on Intel. It's going to be a few generations and billions of dollars before Intel can catch up, assuming they keep with it. Intel has been in the integrated GPU market for awhile but with little pressure to innovate and very different design goals.
  - AlexandrB 3y ago
    Maybe I'm thinking of the Intel i740? I remember there were some Intel products early in the discrete 3D GPU era[1]. [1] https://www.computer.org/publications/tech-news/chasing-pixels/intels-gpu-history https://www.computer.org/publications/tech-news/chasing-pixe...
    - Night_Thastus 3y ago
      If you are, that's 25 years ago now. Intel's ARC only released in 2022 with really nothing before that. Nvidia has been making dedicated GPUs that entire time, while Intel basically only started. Intel has a LOT of catching up to do.
  - Foobar8568 3y ago
    How many SDK for parallel computing both Intel and AMD released over the last 15years while Nvidia just focused on iterating on CUDA? Intel and AMD just keep burning developers over and over.
automatic6131 3y ago
AMD's Radeon group has, by any definition, kept up. In gaming, in the tier of cards that people actually buy, they trade blows with NVidia. But making a GPU is hard. Making a good GPU is even harder. This becomes a chicken and egg problem. If NVidia's hardware has quirks, and you're a game developer, you optimize your game for Nvidia's hardware. If someone else tries and builds a GPU, they discover that many, popular, games run like sh*t. Because of tens of years of strange workarounds to ship a game, drivers rewrite commands for specific games, game devs abuse the directx spec or do it wrong. However, Nvidia set out 10-15 years ago to corner the professional general purpose GPU compute market with a software layer called CUDA. Software was written and optmised for CUDA, and not any other generic graphics library for reasons I don't know (I'm not a graphics developer, just a gamer). So now Nvidia enjoys a moat in gpGPU (as it was called).
- rmckayfleming 3y ago
  I do wonder how much of an impact the PS5 and Xbox Series have on this though since they both have RDNA2 GPUs. Consoles tend to face higher optimization pressure, especially as the generation wears on. It might be why AMD has kept so competitive in gaming as of late.
wing-_-nuts 3y ago
If you mean AI/ML/Cuda? I fully agree with you. AMD is suffering the consequences of their half hearted efforts with ROCm right now. If you mean gaming? AMD has had better mid range gpus since the rx 480 was released in '16. Nvidia has been resting on their laurels and AMD has recently seemed to be fine playing the 'price is right' game where they price their gpus ~ 10% less than nvidia and don't even attempt to take market share.
- latchkey 3y ago
  AMD is committed to actively improving ROCm and directly supporting projects like Pytorch. They were caught with their pants down, but working hard to catch back up.
chongli 3y ago
Nvidia got out in front with CUDA. They got a huge first-mover and vendor lock-in advantage with that one. AMD dropped the ball and is playing catch-up.
pavlov 3y ago
Gaming isn't the lucrative part of the business — it's selling data center GPUs for AI/ML. That's where Nvidia is raking in the cash. It's in the article: "...of the company's $18.12 billion in revenue, $14.51 billion was generated by its data center division." They spent over a decade building up CUDA into the industry standard for GPU compute programming. That's why it's so hard for anyone to compete in this domain. AMD is doing better in gaming, but that's practically a sideshow business for Nvidia now.
- perryizgr8 3y ago
  > Gaming isn't the lucrative part of the business I know, but my point is there were heavy incentives to compete with Nvidia. But nobody did, at least not with the same intent. And now they're (Nvidia) raking it in due to the AI/ML wave. If others had comparable GPU capacity and know how, the market wouldn't be so skewed in favour of Nvidia.
  - swatcoder 3y ago
    Hardware development has much higher inertia than software development, and so you can't just chase the wind and see market-disrupting results in a year or two. Nvidia's success is rooted in design and business choices made a long time ago. They were bets at that time, as ambitious hardware initiatives need to be, and other vendors like AMD made different bets. Nvidia won their bet more spectacularly than others, and now gets to benefit from their lead as others slowly work to reorient their ships in its direction.
shmerl 3y ago
AMD kept up with them. Others - not really. But it would be good if AMD would try to get ahead too, not just keep up.
doctorpangloss 3y ago
Absolutely, positively nothing stops AMD from shipping affordable GPUs with more VRAM. It might not be a popular device - most people are buying $100-200 GPUs. But I would buy a 64GB 7900XTX, not a 24GB one. The simple answer is, like many markets with few competitors, they like prices high. It's like a point of no return if they undercut their crappy workstation graphics.
- faeriechangling 3y ago
  Well there’s the looming threat of the fact that Nvidia is making margins that are way bigger so they can easily match any AMD undercutting and then undercut AMD if they want to. The products that are red hot right now also use HBM not VRAM.
  - michaelt 3y ago
    Right now, there's a big gap in nvidia's product offering: nvidia is happy to let their gaming/affordable cards languish, to stop them competing with the server products at 4x the price. And the ML community isn't very motivated to port their code from nvidia to amd because the nvidia tax isn't too bad. If AMD could take one of their existing 24GB consumer cards and quadruple the RAM, though? Open source ML efforts would have support within days. And nvidia couldn't respond to undercut them without also undercutting their cash cow server GPUs.
    - bonton89 3y ago
      AMD is kind of ironically hamstring by x86. They make a lot more profit off their CPUs than they do their GPUs and they only have so many TSMC wafers for both products. If they try to buy their way into more GPU market share they're losing margin on the GPUs they do sell and denying themselves more profitable CPU sales. And Nvidia has historically been pretty good about aggressively back filling any holes in their product stack AMD figures out how to exploit with a late product introduction. AMD seems like they've mostly given up on gaining in the discrete market. Intel is making some better moves lately but the impression I get is they're losing money on every sale and need to buy their wafers from TSMC like AMD. I'm not sure how long they'll stomach that situation or if they'll eventually get their GPUs onto their own process.
englishspot 3y ago
AMD typically loses out in certain features. Video encoding/decoding is worse on AMD, if I'm not mistaken. They've also had a reputation for bad drivers, at least in the past. Not sure how true that is now.
mrbonner 3y ago
AMD has real good hardware to compete at least in the gaming sector. But, the hardware is only as good as the software/driver supporting it. In the ML and Cryptor sector, Nvidia has built a moat around their hardware called CUDA, a pretty much standard if you want your GPU specific code to be “portable”. I’m afraid OpenCL won’t matter anymore unless Nvidia supports it. Anyone with more industrial domain knowledge wants to chime in?
- esafak 3y ago
  If it runs pyTorch and Tensorflow they'll be in the industrial running.
seanalltogether 3y ago
I think Nvidia has built a firmware and library support moat around their hardware that is difficult for others to keep up with. They invest time to work with studios to make sure they have game ready drivers available the same day as new AAA games are launched.
HDThoreaun 3y ago
80% of Nvidia's revenue this quarter is from datacenter. The gaming market is not what makes their business so lucrative, right now AI is.

packetlost 3y ago

There's not? Both AMD and Intel have GPUs that are competitive in certain market segments. AMD in particular is pretty much even in terms of rasterization performance on the high end. Intel is just new (again) to the game of discrete graphics. Nvidia, on the other hand, invested tons and tons of money to make CUDA good over the last like 15 years, while AMD has not really put any serious effort into AI/ML workloads. It's mostly a software support problem, not a hardware problem to my understanding.

Lord-Jobo 3y ago
AMD is even on performance and significantly more affordable for consumers. but people wont stop buying nvidia's ripoff products like the 4060. really frustrating
- bryanlarsen 3y ago
  Even more frustrating is on the Linux side. We've had a couple of threads about how bad Wayland is, when most of the problems are due to Nvidia binary drivers. It's been 11 years since Linus gave Nvidia the finger. Nobody who is thinking of possibly running Linux should ever buy an Nvidia product. Yet Nvidia is still very commonly spec'd into Linux builds even today.
  - shmerl 3y ago
    On Linux people are clearly leaning to AMD for a while already, especially when gaming is concerned: https://www.gamingonlinux.com/index.php?module=statistics&view=trends#GPUVendor-top https://www.gamingonlinux.com/index.php?module=statistics&vi... In the Windows land it's indeed mostly Nvidia.
    - robotnikman 3y ago
      Ironically, ARM drivers for Nvidia are available for Linux, but not for Windows
  - Filligree 3y ago
    AMD cards are _very bad_ at running CUDA. If you're running linux on a desktop, there's a decent chance you're the type who wants to play with ML as well...
    - latchkey 3y ago
      AMD has a translation layer called HIP as part of the ROCm toolset. This doesn't get us 100% of the way there, but it is an alternative that is gaining traction as they work to improve it.
  - smoldesu 3y ago
    KDE's Wayland session is fine on Nvidia, at least with my 3070. There used to be more visual artifacts, but everything I'm using works, from CUDA to Mesa to DLSS. The largest remaining bug I'm aware of has to do with how blurred windows are presented. It's pretty functionally solid, at least with newer drivers.
    - bryanlarsen 3y ago
      I am still getting a lot of visual artifacts on a 970.
  - anon291 3y ago
    I've used NVIDIA on linux for decades and the support has always been great in my experience
- jfim 3y ago
  For ML workloads, AMD wasn't competitive due to lack of software support for PyTorch. Nvidia isn't making most of their profit from selling individual GPUs to gamers, they're making money from data center GPUs that are bought by the thousands at five figure prices each for ML training.
- opportune 3y ago
  I made the same argument to myself when I bought an AMD GPU for gaming 4 years ago. What doesn’t show up in a side by side comparison is that AMD’s drivers (or Windows’ treatment of them) fucking suck. Until I bought an NVIDIA replacement recently I was very frequently having to reinstall the drivers and also intermittently encountering driver crashes in games and browsers. That’s not to mention having to fiddle with the drivers/games more than I was accustomed to, which IIUC is more a consequence of having less market share and thus not being a priority for game developers to work out of the box. I can excuse that, but not crashes or reinstalling every couple months.
  - tokai 3y ago
    On linux its the opposite. Its was such a relive to get an AMD card and finally have a good stable driver.
  - kulahan 3y ago
    Heh, I remember this being the reputation they had back in the early 2000s. Guess they never got past it.
  - gabeaweg 3y ago
    Completely agree. I bought an AMD GPU like 6 years ago. The software support fucking sucked. The drivers were bad, games weren't optimised for AMD. AMD looks great on paper in benchmarks. As a product, overall, it is bad. That's why people keep buying nvidia. I will gladly pay double for a product that works.
  - fenesiistvan 3y ago
    I am using a Radeon RX 6900 XT GPU since two years and never ever had any stability problem or crash running my PC ~10 hours/day, playing a some (various, demanding) games almost every day.
- dotnet00 3y ago
  AMD might've gotten more traction if it wasn't schizophrenic with regards to compute support on its hardware. GCN was such a great architecture for heavy compute, yet it was mostly wasted by shit software. Crypto really only managed to use it because there was more money to be made using AMD, justifying the wrangling needed to get compute to work on it. Now they continue to waste their potential with ROCm's design choices which make its hardware support range very limited.
  - latchkey 3y ago
    In my eyes, as someone who ran 150k+ AMD gpus for crypto... it really boiled down to being able to heavily tune various aspects of the GPUs. Each GPU is a snowflake and being able to tweak the knobs individually, really made a huge difference on performance. Super low level performance enhancements in code helped a lot too. I worked with some of the popular mining apps to improve things there. AMD is already backporting ROCm to "lesser" hardware [0]. I wouldn't expect them to spend too many resources on going back further than that though. I'd rather see them improve their newer stuff. The larger issue is that it is impossible to rent time on the super high end GPUs that they use to build the super computers with... nobody offers that and it is something I'm working on fixing. [0] https://www.tomshardware.com/news/amd-enables-rocm-and-pytorch-on-radeon-rx-7900-xtx https://www.tomshardware.com/news/amd-enables-rocm-and-pytor...
- vinculuss 3y ago
  The thing I hear a lot from people who skew towards AMD is that raytracing and DLSS and all the software solutions that nvidia has are gimmicks, or somehow actually not a big deal. They’re stuck on the concept of rasterized graphics being the ground truth in gpu value. Then they’re shocked when the wider consumer market doesn’t hold their perspective. AMD hasn’t had the big picture vision that nvidia has had. For example, with Vega AMD had a gpu architecture with exceptional async compute performance a generation before nvidia was competitive. AMD then abandoned Vega, switched to a leaner and less compute friendly rDNA architecture while nvidia went in the opposite direction, pushed compute heavy features like raytracing and dlss and AMD hasn’t had an exceptional response since then.
  - Sohcahtoa82 3y ago
    Raytracing being a gimmick is highly subjective. Creating reflections using rasterization requires shortcuts and sacrifices. And once you see them, you can't unsee them, and then you find them visually distracting. There are basically 3 ways to create reflections without ray-tracing: 1. Double the geometry and render it in the reflection. This is rarely the approach taken because it's incredibly resource intensive. Oddly, Duke Nukem 3D back in what, 1996? did this, but that had pretty simple graphics. Genuinely not feasible for any sort of reflect surface that is curved. 2. Screen-space reflection: Just take the rendered image, invert it, and overlay it onto the reflective surface. This is probably the most common way to do it. It's fast and looks decent. The problem is that it requires the reflected image to be visible on the screen. When it's not, there's no reflection. It creates a bizarre effect when, for example, you're looking over a body of water at something in the distance, and you see the reflection in the water, but as you move the camera down, the reflection gets cut off and eventually disappears entirely. 3. Using a static image as an environment mapped texture. Apex Legends does this for shiny surfaces. It creates a great illusion of a reflection, but taking even a second to actually look at it, you'll notice that the reflection doesn't actually show what it should unless you're standing in the exact right place. I will concede that gamers typically don't notice the issues created using methods 2 and 3, especially in a game like Apex Legends where you're moving around too much to notice. But me? I dunno. I notice and get distracted. EDIT: DLSS, on the other hand, is basically a crutch. It's not surprise that DLSS became a big thing at the same time RTX did. RTX requires more juice to run at full resolution than the GPU has, so they introduced DLSS to render at a low resolution and upscale it.
- paulmd 3y ago
  4060 is one of the better-value options in the stack, actually. The RX 7600 is still on 6nm and is no faster, much less efficient, and skips many of the microarchitectural improvements to rdna3 to try and keep cost down. $250-270 is fine for this type of product even today. Similarly, the 4070 is also fairly attractively priced even compared to 6800XT/6950XT or 7800XT. More efficient, better feature set, same performance tier, similar perf/$. But people still get upset about intangibles like memory bus or die size (despite this already being measured in the performance!), or just don’t want to admit that wafer costs and R&D costs are rising in the post-Moores law era. Operating margins were actually down relative to the last few gens (until AI hit), and were broadly similar to 2012 levels. This is just what it costs now, and smaller die take the worst of the cost increases because memory controllers and pcie don’t shrink anywhere near as much as logic (the gpu cores). There is a “minimum cost floor” that has been rising substantially due to this.
roughly 3y ago
CUDA’s definitely the “moat”, in the Warren Buffet sense - it’s not impossible for someone like AMD or Intel to make something CUDA compatible, but it might cost enough to wreck the economics of trying to compete in the space.
- LeanderK 3y ago
  I don't think you will need something coda-compatible. Apple has relatively solid pytorch support without copying CUDA, but PyTorch+ROCm is still not working on windows (and I think there's no ROCm support for the consumer GPUs from AMD). I think Apple proves that it is possible, and possible relatively quickly.
  - kimixa 3y ago
    Pytorch supports directML on windows - is that not sufficient (or performant) for your usecase? It feels like AMD are less willing to step on Microsoft's toes, possibly as they just don't have the market share clout. And Rocm supports the higher RDNA2 and the highest-end RDNA3 consumer GPUS - [0], though only the top RNDA3 cards in linux [1] (officially - other cards often do "work" but YMMV) One thing Nvidia are much better at is supporting the whole stack - while I doubt many people are actually running much ML on 8gb mid-low end cards, it seems a hook for sales if people "could" do that, and a starting point for personal experimentation. I know multiple people who buy low end Nvidia GPUs because they "might" experiment with CUDA some time in the future. [0] https://rocm.docs.amd.com/en/latest/release/windows_support.html https://rocm.docs.amd.com/en/latest/release/windows_support.... [1] https://rocm.docs.amd.com/en/latest/release/gpu_os_support.html https://rocm.docs.amd.com/en/latest/release/gpu_os_support.h...
    - LeanderK 3y ago
      > One thing Nvidia are much better at is supporting the whole stack - while I doubt many people are actually running much ML on 8gb mid-low end cards, it seems a hook for sales if people "could" do that, and a starting point for personal experimentation. I know multiple people who buy low end Nvidia GPUs because they "might" experiment with CUDA some time in the future. but this is very important. Many develop locally before sending it to a cluster, especially if you only have limited credits available.
tverbeure 3y ago
It’s important to have a vision of what you want to do as a company. For as long as I can remember, Jensen has been telling us during all hands meetings that Nvidia is a visual computing solutions company, not a semiconductor company. (The “visual” part has only been dropped in the past few years.) Nvidia sells hardware, but they also sell complete solutions, and everything in between.
- J_Shelby_J 3y ago
  Indeed. I don't really care what GPU I use. I do care that Geforce experience works flawlessly. I mean, it doesn't, and my grandkids will probably still use DDU, but when shadowplay works it's indispensable as a gamer (or even for recording video calls.) I'm sure AMD has an alternative or I could use OBS, but I'd rather just pay a premium and rarely have to think about it. And also CUDA.
latchkey 3y ago
AMD is coming out with the MI300x on Dec 6th, which has 192GB and runs on the same hardware platform as H100s (OAM-UBB). They've doubled down on ROCm/Pytorch and other projects. Don't expect them to stay behind for much longer.
- packetlost 3y ago
  One can hope!
  - latchkey 3y ago
    Doing everything I can to help them, I don't see it as a need for only hope. =)
- esafak 3y ago
  I see MS (and AWS?) support it. https://techcommunity.microsoft.com/t5/azure-high-performance-computing/azure-announces-new-ai-optimized-vm-series-featuring-amd-s/ba-p/3980770 https://techcommunity.microsoft.com/t5/azure-high-performanc...
  - latchkey 3y ago
    Yup! Part of a partnership. Exciting news.

djyaz1200 3y ago

Agree, can someone(s) more knowledgeable about this field share their thoughts on how defensible Nvidia's position is in the short and long term? Like are companies like this a credible threat? https://www.cerebras.net/ https://www.cerebras.net/ I hear a big part of Nvidia's lock-in is that they provide software that runs these AI projects?

faeriechangling 3y ago
They also own a high end networking company and will sell you an entire rack or datacenter integrated for you. Also yeah, the decade or so they’re had thousands of in-house engineers and people around the world working on CUDA is HUGE. Also the crypto boom encouraged them to buy up tons of fab capacity. Nvidia had foresight on multiple levels here. Nvidia’s problem is that every company is a credible threat because in the ongoing gold rush they can’t make enough shovels to meet demand and damn there is gold in them thar hills. People are going to do literally whatever it takes to make their competition’s products work. If you are selling a shovel, even a shitty shovel, you are a competitive threat to Nvidia right now.

jdiff 3y ago

AMD and Intel are competitive in gaming. Where Nvidia excels is other types of acceleration. AMD and Intel can't hold a candle to Nvidia when it comes the raytracing renderers used in every 3D and CGI workflow, for example.

AlexandrB 3y ago

There's also AMD - but it seems like a hard industry to get into. Intel keeps trying, most recently with Intel Arc[1], but they've never been able to (or perhaps wanted to) crack the high end. [1] https://www.tomshardware.com/news/intel-arc-graphics-cards-are-the-best-deals-this-black-friday https://www.tomshardware.com/news/intel-arc-graphics-cards-a...

Night_Thastus 3y ago
Saying they've "never been able to" is a kind of weird perspective. They're really only just starting on their dedicated GPUs. The Alchemist architecture is their very first attempt, and they've been putting a lot of effort into repairing and refining the drivers. The hardware is actually pretty decent when it's firing on all cylinders. But Nvidia and AMD have a huge time advantage on Intel. It's going to be a few generations and billions of dollars before Intel can catch up, assuming they keep with it. Intel has been in the integrated GPU market for awhile but with little pressure to innovate and very different design goals.
- AlexandrB 3y ago
  Maybe I'm thinking of the Intel i740? I remember there were some Intel products early in the discrete 3D GPU era[1]. [1] https://www.computer.org/publications/tech-news/chasing-pixels/intels-gpu-history https://www.computer.org/publications/tech-news/chasing-pixe...
  - Night_Thastus 3y ago
    If you are, that's 25 years ago now. Intel's ARC only released in 2022 with really nothing before that. Nvidia has been making dedicated GPUs that entire time, while Intel basically only started. Intel has a LOT of catching up to do.
- Foobar8568 3y ago
  How many SDK for parallel computing both Intel and AMD released over the last 15years while Nvidia just focused on iterating on CUDA? Intel and AMD just keep burning developers over and over.

automatic6131 3y ago

AMD's Radeon group has, by any definition, kept up. In gaming, in the tier of cards that people actually buy, they trade blows with NVidia. But making a GPU is hard. Making a good GPU is even harder. This becomes a chicken and egg problem. If NVidia's hardware has quirks, and you're a game developer, you optimize your game for Nvidia's hardware. If someone else tries and builds a GPU, they discover that many, popular, games run like sh*t. Because of tens of years of strange workarounds to ship a game, drivers rewrite commands for specific games, game devs abuse the directx spec or do it wrong. However, Nvidia set out 10-15 years ago to corner the professional general purpose GPU compute market with a software layer called CUDA. Software was written and optmised for CUDA, and not any other generic graphics library for reasons I don't know (I'm not a graphics developer, just a gamer). So now Nvidia enjoys a moat in gpGPU (as it was called).

rmckayfleming 3y ago
I do wonder how much of an impact the PS5 and Xbox Series have on this though since they both have RDNA2 GPUs. Consoles tend to face higher optimization pressure, especially as the generation wears on. It might be why AMD has kept so competitive in gaming as of late.

wing-_-nuts 3y ago

If you mean AI/ML/Cuda? I fully agree with you. AMD is suffering the consequences of their half hearted efforts with ROCm right now. If you mean gaming? AMD has had better mid range gpus since the rx 480 was released in '16. Nvidia has been resting on their laurels and AMD has recently seemed to be fine playing the 'price is right' game where they price their gpus ~ 10% less than nvidia and don't even attempt to take market share.

latchkey 3y ago
AMD is committed to actively improving ROCm and directly supporting projects like Pytorch. They were caught with their pants down, but working hard to catch back up.

chongli 3y ago

Nvidia got out in front with CUDA. They got a huge first-mover and vendor lock-in advantage with that one. AMD dropped the ball and is playing catch-up.

pavlov 3y ago

Gaming isn't the lucrative part of the business — it's selling data center GPUs for AI/ML. That's where Nvidia is raking in the cash. It's in the article: "...of the company's $18.12 billion in revenue, $14.51 billion was generated by its data center division." They spent over a decade building up CUDA into the industry standard for GPU compute programming. That's why it's so hard for anyone to compete in this domain. AMD is doing better in gaming, but that's practically a sideshow business for Nvidia now.

perryizgr8 3y ago
> Gaming isn't the lucrative part of the business I know, but my point is there were heavy incentives to compete with Nvidia. But nobody did, at least not with the same intent. And now they're (Nvidia) raking it in due to the AI/ML wave. If others had comparable GPU capacity and know how, the market wouldn't be so skewed in favour of Nvidia.
- swatcoder 3y ago
  Hardware development has much higher inertia than software development, and so you can't just chase the wind and see market-disrupting results in a year or two. Nvidia's success is rooted in design and business choices made a long time ago. They were bets at that time, as ambitious hardware initiatives need to be, and other vendors like AMD made different bets. Nvidia won their bet more spectacularly than others, and now gets to benefit from their lead as others slowly work to reorient their ships in its direction.

shmerl 3y ago

AMD kept up with them. Others - not really. But it would be good if AMD would try to get ahead too, not just keep up.

doctorpangloss 3y ago

Absolutely, positively nothing stops AMD from shipping affordable GPUs with more VRAM. It might not be a popular device - most people are buying $100-200 GPUs. But I would buy a 64GB 7900XTX, not a 24GB one. The simple answer is, like many markets with few competitors, they like prices high. It's like a point of no return if they undercut their crappy workstation graphics.

faeriechangling 3y ago
Well there’s the looming threat of the fact that Nvidia is making margins that are way bigger so they can easily match any AMD undercutting and then undercut AMD if they want to. The products that are red hot right now also use HBM not VRAM.
- michaelt 3y ago
  Right now, there's a big gap in nvidia's product offering: nvidia is happy to let their gaming/affordable cards languish, to stop them competing with the server products at 4x the price. And the ML community isn't very motivated to port their code from nvidia to amd because the nvidia tax isn't too bad. If AMD could take one of their existing 24GB consumer cards and quadruple the RAM, though? Open source ML efforts would have support within days. And nvidia couldn't respond to undercut them without also undercutting their cash cow server GPUs.
  - bonton89 3y ago
    AMD is kind of ironically hamstring by x86. They make a lot more profit off their CPUs than they do their GPUs and they only have so many TSMC wafers for both products. If they try to buy their way into more GPU market share they're losing margin on the GPUs they do sell and denying themselves more profitable CPU sales. And Nvidia has historically been pretty good about aggressively back filling any holes in their product stack AMD figures out how to exploit with a late product introduction. AMD seems like they've mostly given up on gaining in the discrete market. Intel is making some better moves lately but the impression I get is they're losing money on every sale and need to buy their wafers from TSMC like AMD. I'm not sure how long they'll stomach that situation or if they'll eventually get their GPUs onto their own process.

englishspot 3y ago

AMD typically loses out in certain features. Video encoding/decoding is worse on AMD, if I'm not mistaken. They've also had a reputation for bad drivers, at least in the past. Not sure how true that is now.

mrbonner 3y ago

AMD has real good hardware to compete at least in the gaming sector. But, the hardware is only as good as the software/driver supporting it. In the ML and Cryptor sector, Nvidia has built a moat around their hardware called CUDA, a pretty much standard if you want your GPU specific code to be “portable”. I’m afraid OpenCL won’t matter anymore unless Nvidia supports it. Anyone with more industrial domain knowledge wants to chime in?

esafak 3y ago
If it runs pyTorch and Tensorflow they'll be in the industrial running.

seanalltogether 3y ago

I think Nvidia has built a firmware and library support moat around their hardware that is difficult for others to keep up with. They invest time to work with studios to make sure they have game ready drivers available the same day as new AAA games are launched.

HDThoreaun 3y ago

80% of Nvidia's revenue this quarter is from datacenter. The gaming market is not what makes their business so lucrative, right now AI is.