SK hynix Debuts Prototype of First GDDR6-AiM Accelerator Card 'AiMX' for Generative AI

[GFreeman]

SK hynix unveiled and demonstrated a prototype of AiMX (Accelerator-in-Memory based Accelerator), a generative AI accelerator card based on GDDR6-AiM, at the AI Hardware & Edge AI Summit 2023 held September 12-14 at the Santa Clara Marriott, California.Hosted annually by the UK marketing firm Kisaco Research, the AI Hardware & Edge AI Summit brings together global IT companies and high-profile startups to share their developments in artificial intelligence and machine learning. This is SK hynix's third time participating in the summit.

At the event, the company showcased the prototype of AiMX, an accelerator card that combines multiple GDDR6-AiMs to further enhance performance, along with the GDDR6-AIM itself under the slogan of "Boost Your AI: Discover the Power of PIM (Processing-In-Memory) with SK hynix's AiM (Accelerator in Memory)." As a low-power, high-speed memory solution capable of handling large amounts of data, AiMX is set to play a key role in the advancement of data-intensive generative AI systems. The performance of generative AI improves as it is trained on more data, highlighting the need for high-performance products which can be applied to an array of generative AI systems.



SK hynix also demonstrated Meta's generative AI Open Pretrained Transformer (OPT) 13B model on a server system equipped with the AiMX prototype. The AiMX system featuring GDDR6-AiM chips reduces data processing time by more than 10 times compared to systems with GPUs, while consuming one-fifth the power. The company's demonstration piqued the interest of global companies providing AI services by showing that it can deliver higher performance than the most recent accelerators.

In addition, the company held a session outlining the benefits of AiMX. In a presentation titled "Cost-Effective Generative AI Inference Acceleration using AiM," Eui-cheol Lim, vice president of the Solution Development division, compared the performance of GPUs and AiMX and discussed the future of next-generation intelligent semiconductor memories.

"SK hynix's AiMX is a solution that delivers higher performance while consuming less power, and costing less than conventional GPUs," Lim explained. "We will continue to develop memory technologies that will lead the way in the era of artificial intelligence."

View at TechPowerUp Main Site | Source

 

[kondamin]

An asic for an early model, I guess it could make sense if you were to use it to generate assets for a Game or something but still.
creating silicon for that now is anything but cost effective.

 

[TumbleGeorge]

"5"nm lithography process? Were is not bleeding edge and premium charged.

 

[notaburner]

An asic for an early model, I guess it could make sense if you were to use it to generate assets for a Game or something but still.
creating silicon for that now is anything but cost effective.

Could be useful for services that provide generative AI for users as they could have the necessary scale. Not sure how adobe's implementation for photoshop works, but if it's all handled in the cloud than something like this could make sense. For a game studio sort of situation, unless they went all in on AI assets, it probably doesn't make sense to go the dedicated hardware route at this point.

 

[kondamin]

Could be useful for services that provide generative AI for users as they could have the necessary scale. Not sure how adobe's implementation for photoshop works, but if it's all handled in the cloud than something like this could make sense. For a game studio sort of situation, unless they went all in on AI assets, it probably doesn't make sense to go the dedicated hardware route at this point.

That’s just the thing, a game makes sense.
in a game (online)with a load of users generating dungeons with a certain type of monsters works for years.

adobe users will want better ai models not an old one, dumping server farms of hardware just because you have a new and better model

 

[notaburner]

That’s just the thing, a game makes sense.
in a game (online)with a load of users generating dungeons with a certain type of monsters works for years.

adobe users will want better ai models not an old one, dumping server farms of hardware just because you have a new and better model

My point was more about if a game would require enough resources to really require them purchasing their own dedicated hardware to support it. Game companies usually rent server space to run their online services and don't go out and build their own server infrastructure. Would think that a company that provides "AI services" would be interested in this sort of hardware and then sell that capacity to game developers or anyone else who would be interested. Probably cheaper for everyone involved to run it at scale and then sell the capacity to a large group of game developers.

 

[kondamin]

My point was more about if a game would require enough resources to really require them purchasing their own dedicated hardware to support it. Game companies usually rent server space to run their online services and don't go out and build their own server infrastructure. Would think that a company that provides "AI services" would be interested in this sort of hardware and then sell that capacity to game developers or anyone else who would be interested. Probably cheaper for everyone involved to run it at scale and then sell the capacity to a large group of game developers.

Ah, yes.
I can't see an infrastructure provider offering something that task specific.
You would need an entire industry agreeing on running on 1 model for x amount of time so the investment is earned back and profits are made.