AMD has 2 server platforms, trash-lowend-premium-highend level mobile chips, two core IP variants, hybrid chips, 3D variants, etc. They aim at making multiple different accelerator lines. They want to create an ultra-premium mobile line. Not even mentioning the Xilinx IPs.People are stuck in the past. AMD is a very different company compared to ten and even 5 years ago.
Yeah, they should just give up and settle for an unprofitable business with low market share. They would be better off exiting dGPUs that use that 'business' model. They screwed up on RDNA3 and apparently 4. They have a plan to Make RDNA5 a winner - hope they are able to pull it off, grab some decent market share and profits. Then we will have more choices between competitive products in each market segment.Radeon has to admit they are #2 by lowering the price and profiling themselves as such.
Yet, RTG products are still abysmal with no magic change on the horizon.
People don't care about most things you've listed.Missing a true DLSS alt, various subpar implementations of FSR, unusable VR months after the release on RDNA3, long standing multi-monitor/video power consumption problems, often broken OpenCL driver releases, recent broken HW browser video acceleration, no OptiX acceleration, broken DaVinci resolve, ROCm is no alternative to the CUDA stack, long standing OpenGL Windows issues, etc
Neither of the three.Not enough talent, overly ambitious, bad management?
Am trying to understand what MI300 is.RTG products are still abysmal with no magic change on the horizon.
Uh, those are the three biggest reasons products fail to hit TTM targets (and underfunding, which you rejected). Thus the question of why still stands out rather starkly.Neither of the three.
Some things just inherently take a lot of time and they have a schedule to adhere to.those are the three biggest reasons products fail to hit TTM targets (and underfunding, which you rejected).
AMD R&D is up 30m QoQ so ughhhh.Cut funding to chase AI hype?
AMD R&D is up 30m QoQ so ughhhh.
No.
Also never ever happening.I mean cut dGPU funding.
Also never ever happening.
Why would they?
Which is why you build the biggest stick and charge whatever for it.Because people only buy AMD dGPUs when the margins suck
They're doing that as well.Better to chase AI Hype.
Which is why you build the biggest stick and charge whatever for it.
What they have now is very scalable, the question is always building big enough while praying nothing ever goes wrong, unlike RDNA3.Sounds like they need the scalable arch to be able to do that.
What they have now is very scalable, the question is always building big enough while praying nothing ever goes wrong, unlike RDNA3.
That was all RDNA4, yes.By scalable I mean with chiplets.
Those graphs are cool and shiny. However, the reality is Radeon brand compared to GeForce is like comparing SsangYong to Chrysler. Both offer usable products, although in the market PoV only one of them is feature complete, offers a wide range of support, is the "way to go", ...There s a lot of viral marketing around AMD GPUs, not to point the finger on some users but reading here and there it seems that most NV users act so at the behalf of their brand, and at the end thoses purposely fabricated urban legends are taken at face value.
Better to lOok at real numbers and you ll understand why AMD is vilified, i.E, because that s the only mean to keep them from gaining marketshare as their GPUs are very competitive perfs and perf/price wise
The rant concerns Radeon dGPUs but whatever.Am trying to understand what MI300 is.
Can you help me?
Looks too leadership to be abysmal with no magic change on the horizon.
This is ATi/AMD's classic!Which is why you build the biggest stick and charge whatever for it.
Those graphs are cool and shiny. However, the reality is Radeon brand compared to GeForce is like comparing SsangYong to Chrysler. Both offer usable products, although in the market PoV only one of them is feature complete, offers a wide range of support, is the "way to go", ...
Graphs like that do not include the feature set, brand name, or its legacy. There is no vilification needed.
The rant concerns Radeon dGPUs but whatever.
It is a Radeon dGPU.The rant concerns Radeon dGPUs but whatever.
You don't know anything then.My crystal ball tells me the MI300 series is a line of products with availability in H1 2024
You can do the math yourself, there's an endnote for MI300A quoting 2500TF sparse FP8 achieved (guess what % of theorethical it is).Yet, the theoretical performance in the "AI precision" is behind 2022 H100, right?
Oh it works well enough.Also it will have to rely on the ROCm stack which is ... not supported the way you would expect.
Wow how wonderful that you've skipped R580 or Cypress or Hawaii or N21.All those big bad chips hyped as "nVidia killers"... the memories: R520, R600, Cayman, Tahiti, Fiji, Vega 10, Vega 20, or Navi 31
No it has to be the biggest stick.Betting on the biggest stick is nice but it has to be the real flagship product with polished features, SW support, etc.
Indeed, that's always been the case. ATi/AMD is the a bit cheaper alternative to nV. But is this enough? The market doesn't think so.Graphs perfs do not need any comment, they speak loud and right, reality is that AMD s GPUs are of good performance and at more afordable prices than their NV counterparts.
This is ATi/AMD's classic!
All those big bad chips hyped as "nVidia killers"... the memories: R520, R600, Cayman, Tahiti, Fiji, Vega 10, Vega 20, or Navi 31. NVidia always just released a faster card - often with no sweat, no fancy tech, no wide bus, nor Ti variants.
Betting on the biggest stick is nice but it has to be the real flagship product with polished features, SW support, etc. People somehow don't want to shell $1-2k.
Indeed, that's always been the case. ATi/AMD is the a bit cheaper alternative to nV. But is this enough? The market doesn't think so.
The market is heavily influenced by the constant viral marketing against AMD, Nvidia doesnt even need to pay too much people to do the job, their users are constantly bashing AMD while ignoring anything problematic in their brand GPU.
Indeed you act the same, you dont even look at the perfs and instead are telling me that cluless and influenced people choice is the prove that Radeon cards are no good, that s akin to say that 10 people saying that 1 + 1 = 11 are right against the one guy who say that the result is actually 2.
That was all RDNA4, yes.
The halo part was an egregiously expensive tiled silicon spam wunderwaffen.
Tiled GPU isn't about cheaper, it's about winning more.