AMD Could Be Prepping a Flagship Next-Gen GPU Featuring 96 Compute Units with Memory on a 512-Bit Bus

Realistically they're unlikely to allocate resources for flagship consumer parts unless they somehow get extremely lucky with yields, or get extremely unlucky with commercial demand.
 
DAPUNISHER said:
Su said their focus is enterprise. Won't be surprised if dates keep getting pushed back, products canceled, paper launches, etc. A real AMD flagship gaming card sounds great, but unlikely while there is still the infinite money glitch.
Click to expand...

AMD recognizes their focus is on the enterprise, yet no one says a peep. nvidia is RUMORED to abandon pc gaming and everyone gets their torches and pitchforks :rolleyes: :rolleyes:
 
Riccochet said:
Realistically AMD has a golden opportunity here to catch up if Nvidia drops the ball on consumer products.
Click to expand...
Catch up to what exactly? Steal a slice of the smallest Pie in IT?

AMD wanted to win mindshare when they were entering the CPU market for servers so they did lots of stuff targeting consumers that were in charge of those purchases.

Now with AI farms the gloves are off... every cpu and every GPU available via enterprise channels is being purchased with insane prices for support and same day replacement. So not only do vendors need ALL of the GPU's, they need them all PLUS 25% more of whatever they sell so they can fulfill their insanely priced service contracts.

AMD could completely abandon the consumer space, sell ONLY enterprise parts or to enterprise SI's and grow their profit margins even more.

so could all of them. But hopefully nobody is willing to bet on the AI market being the golden goose that lays an unlimited number of golden eggs per hour forever.
 
Honestly if you wanted to get into an industry that will see astronomical demand in areas where it wasn't before... Get into small city size power generation. Develop something that can be deployed and reliably ran to power these data centers. License it to them, so they pay you to license the tech for however many years and you don't have to do any of the actual work.

Just sit back and let them build it maintain it and support it and collect your check. Just never do an exclusive license on it. And you'll be golden.
 
Grimlakin said:
Catch up to what exactly? Steal a slice of the smallest Pie in IT?

AMD wanted to win mindshare when they were entering the CPU market for servers so they did lots of stuff targeting consumers that were in charge of those purchases.

Now with AI farms the gloves are off... every cpu and every GPU available via enterprise channels is being purchased with insane prices for support and same day replacement. So not only do vendors need ALL of the GPU's, they need them all PLUS 25% more of whatever they sell so they can fulfill their insanely priced service contracts.

AMD could completely abandon the consumer space, sell ONLY enterprise parts or to enterprise SI's and grow their profit margins even more.

so could all of them. But hopefully nobody is willing to bet on the AI market being the golden goose that lays an unlimited number of golden eggs per hour forever.
Click to expand...

That actually sounds accurate. On the consumer GPU side AMD has little to gain, they can't compete at a flagship level which make them look like a 2nd tier. So even if they gain market share, the revenue won't be as great. They also have little market share on the AI market, but the potential growth and revenue is orders of magnitude higher than the consumer market. They are doing pretty well on the CPU side and that can help them by providing a complete solution.

Thing is like nvidia, AMD forfeiting the consumer market would mean losing many other markets its in (as the hardware is basically the same). Although contrary to nvidia, AMD doesn't dominate any of them except the console market.

I don't expect any new next gen products from intel/AMD/nvidia till well past 2027 if only becuase of the RAM supply.
 
Grimlakin said:
Honestly if you wanted to get into an industry that will see astronomical demand in areas where it wasn't before... Get into small city size power generation. Develop something that can be deployed and reliably ran to power these data centers. License it to them, so they pay you to license the tech for however many years and you don't have to do any of the actual work.

Just sit back and let them build it maintain it and support it and collect your check. Just never do an exclusive license on it. And you'll be golden.
Click to expand...
There are already companies in that market space. NuScale and Westinghouse lead the pack. SMR's are waiting to be deployed. Government red tape holds everything back. Realistically we won't see the first SMR go online until 2033 at the earliest. Google and Meta are pushing HARD to fast track SMR's. We're talking 100's of billions of dollars in investment sitting there waiting.

My wife is in the industry. ;)
 
Grimlakin said:
Get into small city size power generation.
Click to expand...
That is exactly what I do at work.

There aren't enough electrical generators right now to support the data center demand. Gas turbine and large reciprocating engine availability is out until 2030 (or beyond). Nuclear isn't rapid deployment at any scale. Renewables require so much lithium to back them up and companies are afraid of the fire risk, even though they like the appearance of being "carbon neutral". It's gone totally plaid - people are buying old used assets at rediculous prices and refurbing them and hoping for the best. I figure that, in reality, out of the gigawatts of "potential" data centers going in; what looks like 20 or 30 projects on the surface will end up being 3 or 4 at best, and a lot of this is purely speculative.
 

TinyCorp pressures AMD to build a 96GB GPU that doesn’t exist yet​

TinyCorp Is Challenging AMD to Build a 96GB RDNA 5 GPU, But the Memory Crisis Makes It Nearly Impossible
March 9, 2026

The company wants to raise $20 million at a $200 million valuation, buy an $11.5 million building in Oregon with 5 megawatts of power capacity, pre-order 3,000 units of an AMD RDNA 5 GPU with 96GB of VRAM at around $2,500 each,
Hotz tagged AMD CEO Lisa Su directly and wrote:
“RDNA 5 should be GDDR7 with a 512-bit bus. Get the 3GB modules and double side it like the Blackwell. We’ll build the board if they don’t.”


A 96GB GPU that only NVIDIA makes, for $8,000 to $10,000​

Right now, the only GPU on the market with 96GB of VRAM is NVIDIA’s RTX PRO 6000 Ada Blackwell, which retails between $8,000 and $10,000 per unit. That is a workstation card aimed at professionals and large enterprises, not something a small AI lab or independent developer can easily budget for. For context, TinyCorp is asking for that same memory capacity at less than a third of the price.

(and we all know AMD will sell for nv - $500 dollars. tiny corp is asking for 1/3rd of that price)


TinyCorp and AMD have a long and bumpy history​

TinyCorp’s frustration with AMD isn’t new, and neither is their commitment to working with Team Red anyway. The original TinyBox Red ships with six Radeon RX 7900 XTX cards, built on the belief that AMD’s price-to-performance ratio makes them the smarter pick for accessible AI computing compared to NVIDIA’s more expensive hardware.

But the relationship has had real friction. AMD’s ROCm software stack caused enough problems that TinyCorp eventually decided to build their own solution from scratch. They developed a completely independent software stack for AMD GPUs, their own driver, runtime, libraries, and emulator, all in roughly 12,000 lines of code, bypassing ROCm entirely. That is a serious engineering investment, and it underlines just how committed they are to making AMD hardware work, even when AMD doesn’t make that easy.


https://geekrealmhub.com/tinycorp-amd-96gb-rdna5-gpu/
 
Brian_B said:
That is exactly what I do at work.

Renewables require so much lithium to back them up and companies are afraid of the fire risk, even though they like the appearance of being "carbon neutral".
Click to expand...

Sodium Ion will be a solution soon. Looks like, for now, some datacentres are going with ferrous oxide batteries (really!)

Brian_B said:
It's gone totally plaid - people are buying old used assets at rediculous prices and refurbing them and hoping for the best. I figure that, in reality, out of the gigawatts of "potential" data centers going in; what looks like 20 or 30 projects on the surface will end up being 3 or 4 at best, and a lot of this is purely speculative.
Click to expand...

There's a lot of money on fusion coming online by the early 2030s. We'll see how THAT goes.
 
You must log in or register to reply here.
Become a Patron!
Back
Top