521–530 of 722 posts
Why don't they just release a basic GPU with 128GB RAM and eat NVidia's local generative AI lunch? The networking effect of all devs porting their LLMs etc. to that card would instantly put them as a major CUDA threat. But beancounters running the company would never get such an idea...
Earlier quoted context omitted.
Intel GPUs don't sell well to gamers. They've been on the market for years now. >market of maybe 1000 people The market of people interested in local ai inference is in the millions. If it's cheap enough the data center market is at least 10 million.
Yes, Intel cards have sucked. But they are trying again!
/s
12GB max is a non-starter for ML work now. Why not come out with a reasonably priced 24gb card even if it isn't the fastest and target it at the ML dev world? Am I missing something here?
I still don't understand why graphics cards haven't evolved to include sodimm slots so that the vram can be upgraded by the end user. At this point memory requirements vary so much from gamer to scientist so it would make more sense to offer compute packages with user-supplied memory. tl;dr GPU's need to transition from being add-in cards to being a sibling motherboard. A sisterboard? Not a daughter board.
It's one of the reasons why ARM Macbooks get great performance/watt, memory being even "closer" than mainboard soldered RAM so getting more of those benefits, though naturally less flexibility.
Earlier quoted context omitted.
It is possible to have multiple memory ranks to reduce the bus width requirements for a given amount of memory. Nvidia has demonstrated that this is doable with GDDR6X on the RTX 3090. The RTX 3090 has a 384-bit bus with 24 memory ICs, despite only needing 12 to reach 384-bit. That means it has every two chips sharing one 32-bit interface, which is a dual rank configuration. If you look at the history of computer mem…
Having four chips per channel is exactly why this is implausible. DDR5 can barely operate with four ranks per channel, at severely reduced speeds. Pulling that off with GDDR6 or GDDR7 is not something we can presume to be possible without specific evidence. The highest-density configurations possible for LPDDR5x are dual-rank and byte mode (one chip per 8 bits of the memory bus, so two chips ganged together to popula…
That said, I am not an electrical engineer (although I work alongside one and have had a minor role in picking low end components for custom PCBs), I think if Intel were to make a GPU with 128GB VRAM using GDDR7 in the next year or two, the engineer who does the trace routing to make it possible should make a go fund me page for people to send beer money.
Earlier quoted context omitted.
> Intel's discrete GPU is hanging by a thread, so they're not hoping on that bandwagon Why would that matter? You buy one GPU, in a few years you buy another GPU. It's not a life decision.
>Why would that matter? The game devs are going to spend all their time & effort targetting amd/nvidia. Custom code paths etc. It's not a one size fits all world. OpenCL etc abstraction are good at covering up differences, but not that good. So if you're the player with <10% market share you're going to have an uphill battle to just be on par.
From my experience they target NVidia and Consoles. AMD might get a look at the code just before release if they notice any big problems.
I'd be surprised if many Gamedevs even pick up the phone for an Intel GPU developer.
Earlier quoted context omitted.
By subsidizing it more they'll lose less money?
Increasing VRAM would differentiate intel GPUs and allow driving higher ASPs, into the green.
You'd need hundreds of thousands of units to really make much of a difference.
Earlier quoted context omitted.
The 600 mini is on the m4 with the anemic 10 core GPU. This is for grandama or your 10 year old. The good one which is still slower than m4 max is 2200. If you want the max you need at least a macbook pro starting at 3200 and if you want the better one with 128G RAM it starts at about 5k
No matter how few GPU cores the M4 has, it is still an extremely potent product on the whole. Better than most of the pc's out there.
I'm really curious to see if these still rely heavily on resizable BAR. Putting these in old computers in linux without reBAR support makes the driver crash with literally any load rendering the cards completely unusable. It's a real shame, the single slot a380 is a great performance for price light gaming and general use card for small machines.
What is the newest platform that lacks resizable BAR? It was standardized in 2006. Is 4060-level graphics performance useful in whatever old computer has that problem?
Earlier quoted context omitted.
It is possible to have multiple memory ranks to reduce the bus width requirements for a given amount of memory. Nvidia has demonstrated that this is doable with GDDR6X on the RTX 3090. The RTX 3090 has a 384-bit bus with 24 memory ICs, despite only needing 12 to reach 384-bit. That means it has every two chips sharing one 32-bit interface, which is a dual rank configuration. If you look at the history of computer mem…
Having four chips per channel is exactly why this is implausible. DDR5 can barely operate with four ranks per channel, at severely reduced speeds. Pulling that off with GDDR6 or GDDR7 is not something we can presume to be possible without specific evidence. The highest-density configurations possible for LPDDR5x are dual-rank and byte mode (one chip per 8 bits of the memory bus, so two chips ganged together to popula…
That is objectively false. See, for instance, V-color’s threadripper RAM[0]. If 96GB quad-rank modules @ 6000Mhz in octo-channel counts as “barely operating” maybe we have different definitions of operation requirements.
As a side note, their quad-channel 3-rank RAM [1] hits 8000MHz, out of the box. Admittedly only 24GB modules, but still.
[0] https://v-color.net/products/ddr5-ocrdimm-amd-wrx90-workstat... [1] https://v-color.net/products/ddr5-oc-rdimm-amd-trx50-worksta...