AMD believes AI will be important for all their upcoming products (story correction)

[Alfonso Maruccia]

Story retraction and correction:

The original story we published was titled "The AI desktop PC isn't coming anytime soon, AMD says," but unfortunately the source for that story misrepresented what Justin Galton, AMD's Business Development Director, had said during a recent event.

Other people who were present in the room have confirmed to us that this quote from Justin is not accurate. The official word is that "AMD believes that AI will be important to all our products going forward."

An edited version of the original article follows below:

Generative AI and other machine learning algorithms are apparently the future of everything. Every single technology corporation has adopted the new AI mantra, and chipmakers are scrambling to embed some sort of AI acceleration logic into their products.

AI-based computer chips aren't a thing just yet, and they will not invade the market for quite some time. Justin Galton, worldwide Business Development director at AMD, said that the company isn't seeing a widespread need for AI-dedicated x86 chips except for "the top of the stack."

AMD was first to arrive on the AI market with the Ryzen 7040, a chip featuring a dedicated AI acceleration logic block, and Intel is betting everything on an AI-powered future with the Meteor Lake architecture and the upcoming Core Ultra SoC

The AMD representative stated that many buyers are still investing their money on affordable computers based on Ryzen 5000 and Ryzen 6000 CPUs, therefore the company has decided to extend technical and marketing support for those two processor families into early 2024. AMD is, of course, working on future CPU technology that will exploit the power of ML algorithms, but those products are scheduled to debut in 2024.

Many customers have now adopted a five-year refresh cycle, Galton said, so companies and consumers will likely start to think about getting a new "desktop AI PC" in the upcoming years. More "exciting" computer models and technologies are coming, AMD assured.

According to Galton, AMD now controls between 15 and 20% of the overall commercial PC market. AI or not, the company wants to break the 20% barrier by the end of 2024.

Permalink to story.

https://www.techspot.com/news/100215-ai-desktop-pc-not-coming-anytime-soon-amd.html

 

[Dimitriid]

See, this is the smart approach: there's no guarantees that ML (Yes, I am still purposefully not calling this AI btw) workloads on regular PCs will be anything more than a novelty.

If you look at current use cases, you've got Apple: They've been putting ML cores on their chips for years now and they've got a few tricks they can do with them on certain apps. It's probably not very cost effective for what they get out of the chips but since they can cover everything in typical apple marketing talk it's all good so the loses of needlessly putting ML cores in there before there's a serious need for them is small for apple and their super comfortable margins.

If you look at intel however betting heavy on the PC market being able to utilize ML there's a few problems there: First they'll get the same kind of slow adoption: Microsoft might be able to offload some of the ML loads from the cloud to local but not the big picture stuff. In other words: you won't be running Cortana fully off your intel ML ready SoCs it will still rely heavily on an internet connection. It might at best just better expand the existing models with local info while being able to claim they don't sell as much of your info so Cortana gets smarter according to what you ask and such.

But that's stuff its tiny. What's going to happen is a similar thing to what happened with the Crypto/NFT craze: a lot of people will snatch a lot of these new intel chips and just build farms of cheap, consumer PCs that don't do anything but serve as their on cluster to train ML stuff for money. Intel will of course not be able to keep up with that new demand and fall behind.

So it's better not to open yourself up for such contingencies by purposefully going in second like AMD is planning to. The disadvantage is that intel might dominate the market initially if the ML desktop usecases take off, but I seriously doubt they will.

 

[SalmanIsHere]

Lisa Su has no vision. She is just doing her job as an employee.
Keller saved amd from bankruptcy with Zen. Amd just node shrinked it with tweaks.
Intel was stuck for 5 years. Which let zen to catch up.

She also destroyed the Radeon Pc/Laptop part. Just copy and make cheap knock off of Nvidia's tech and fail.

Nvidia has full booked H100 till end of 2024.

It was already hinted Windows 12 is full AI focusued. Ms already testing Ai tech's on Ms Paint.
Adobe photoshop is enhanced with crazy AI features.
Google and Ms office is also enhanced with AI features.

Looks to me, Lisa will soon leave amd before the ship sinks.

Nvidia and Apple are pre-booking latest nodes.
4090 35% faster in Raster vs 7900xtx and 75% faster in max settings (RT+RASTER) at 4k.
Next gen no radeon gpu over $400.

This is just start of a deep dive.

Nvidia should buy amd's cpu division.

 

[VcoreLeone]

Nvidia should buy amd's cpu division.

lol

 

[blkspade]

Lisa Su has no vision. She is just doing her job as an employee.
Keller saved amd from bankruptcy with Zen. Amd just node shrinked it with tweaks.
Intel was stuck for 5 years. Which let zen to catch up.

She also destroyed the Radeon Pc/Laptop part. Just copy and make cheap knock off of Nvidia's tech and fail.

Nvidia has full booked H100 till end of 2024.

It was already hinted Windows 12 is full AI focusued. Ms already testing Ai tech's on Ms Paint.
Adobe photoshop is enhanced with crazy AI features.
Google and Ms office is also enhanced with AI features.

Looks to me, Lisa will soon leave amd before the ship sinks.

Nvidia and Apple are pre-booking latest nodes.
4090 35% faster in Raster vs 7900xtx and 75% faster in max settings (RT+RASTER) at 4k.
Next gen no radeon gpu over $400.

This is just start of a deep dive.

Nvidia should buy amd's cpu division.

Silicon budget. Remember when the 1080ti was 67% faster than its predecessor, but the 20 series was basically pointless except for the RT cores that really only made sense to use if you got the 2080ti. You also got the tensor cores, but there wasn't a huge use (or desire) for them outside of DLSS. This is that. AI acceleration is already provided by both Nvidia GPUs and now Radeon RX 7000. It's also in Intel's ARC GPU, but no one really wants them and Intel is on the fence about axing the division. Intel needs AI in their CPU to at least be in the conversation. AMD doesn't have a reason to waste die space in their main Desktop line for something that has yet to really even take off in the consumer space. It will be offered in their GPUs and APUs, and they still have to improve the implementation.

If you had been paying attention to what AMD has been doing with previous iterations of Zen CPUs, you'd know that those mobile APU dies eventually get repackaged as desktop G series chips. Those monolithic 7040 APUs will be the next G series desktop chips that have AI, while the main desktop chiplet CPUs can use their die space to be better in other ways. Nearly everyone interested in the regular Ryzen Desktop CPUs is going to pair it with a dGPU that will likely already provide AI capabilities. GPU dies can get so much larger that they'll likely have so many transistors dedicated to AI that it'll always be better than what you can fit in a desktop CPU package.

 

[Julio Franco]

Story retraction and correction:

The original story we published was titled "The AI desktop PC isn't coming anytime soon, AMD says," but unfortunately the source for that story misrepresented what Justin Galton, AMD's Business Development Director, had said during a recent event.

Other people who were present in the room have confirmed to us that this quote from Justin is not accurate. The official word is that "AMD believes that AI will be important to all our products going forward."