AMD's Biggest Mic Drop

[kapone32]

AMD Instinct MI300X is the Cat's meow for all enthusiasts. AMD is live streaming it right now.

AMD Radeon Instinct MI300X Specs

AMD Aqua Vanjaram, 2100 MHz, 19456 Cores, 1216 TMUs, 0 ROPs, 196608 MB HBM3, 2525 MHz, 8192 bit

www.techpowerup.com

 

[Space Lynx]

link to live stream? i checked youtube and dont see it

edit nm

 

[Fouquin]

MI300X isn't for enthusiasts. It's a pure compute chip, it costs more than a new car, and it is not available to retail. Unless you're an AI enthusiast this isn't for you.

 

[Space Lynx]

MI300X isn't for enthusiasts. It's a pure compute chip, it costs more than a new car, and it is not available to retail. Unless you're an AI enthusiast this isn't for you.

AI is what will keep AMD alive as a company, so this is important for gamers too. otherwise AMD bankrupt.

 

[Fouquin]

link to live stream? i checked youtube and dont see it

Check again. It's the only thing happening on AMD's YouTube page.

AI is what will keep AMD alive as a company, so this is important for gamers too. otherwise AMD bankrupt.

Delivering on schedules will keep AMD alive as a company. AI is a bubble, what they need is a bedrock product lineup that actually launches when they say it does so when the bubble pops they aren't left holding the bag. Even NVIDIA, a self proclaimed AI company, knows and operates this way.

 

[Space Lynx]

Check again. It's the only thing happening on AMD's YouTube page.

i already edited my post with the link

 

[kapone32]

MI300X isn't for enthusiasts. It's a pure compute chip, it costs more than a new car, and it is not available to retail. Unless you're an AI enthusiast this isn't for you.

I get it but you cannot deny that the specs are absolutely spectacular.

 

[Space Lynx]

I get it but you cannot deny that the specs are absolutely spectacular.

lisa su said its most powerful AI computer ever made. and its a drop in device for customers.

this is awesome. this is huge for AMD stock price. and future competitive behavior

 

[kapone32]

lisa su said its most powerful AI computer ever made. and its a drop in device for customers.

this is awesome. this is huge for AMD stock price. and future competitive behavior

1.92 TB of HBM. 8 of them make one AI PC. I am blown away with what these are. No wonder we are getting all of these new CPUs. They will make money for days. I am buying some AMD stock right now. When they announce their next quarter it should be insane.

 

[Space Lynx]

its faster than H100.... well done AMD.

 

[thesmokingman]

Jensun's sphincter is puckered right now.

 

[kapone32]

I am supposed to be watching the Premier league right now but this is on the Main TV. I am blown away with the details that are coming out of this stream. The Level 1 visit to AMD by Wendell was very eye opening for me and gave me the feeling of how the Company I work for used to be.

 

[Fouquin]

1.92 TB of HBM.

No, 192 GB. That's 1.5 TB per 4U node.

 

[Space Lynx]

Jensun's sphincter is puckered right now.

yes and no, Jensen has the money to buy all the production time of TSMC, similar to Apple, there won't be much production time left for AMD, a little sure, but not enough to cause real fear. Nvidia will always outbid and buy the majority of wafer time.

ryzen 7040 apu just came out... wtf 8040 is already out?

 

[kapone32]

I am loving the potential of AI from this stream.

The MI300A is a true display of where APU technology is going.

 

[Fouquin]

ryzen 7040 apu just came out... wtf 8040 is already out?

Phoenix launched in January, the reason it feels like it "just came out" is because AMD couldn't seem to get any chips into any customer's hands for 6+ months. This generation is a refresh anyway so it's just extending the clock on getting current silicon out there.

 

[Space Lynx]

Phoenix launched in January, the reason it feels like it "just came out" is because AMD couldn't seem to get any chips into any customer's hands for 6+ months. This generation is a refresh anyway so it's just extending the clock on getting current silicon out there.

yeah AMD and its vaporware... love AMD but yeah they constantly fail in this regard.

 

[tabascosauz]

MI300X isn't for enthusiasts. It's a pure compute chip, it costs more than a new car, and it is not available to retail. Unless you're an AI enthusiast this isn't for you.

I get it but you cannot deny that the specs are absolutely spectacular.

Semianalysis email newsletter has a brief overview of the AMD press release. The hardware specs have everyone drooling but the performance isn't really at the level that is expected of raw hardware of that caliber (yet).

Conclusion was that AMD has solid footing that will be a winning recipe on their hands........if they deliver on the software optimization. Seems like they've got some good motivation there, though, considering the expected customer base here.

Also, H100 launched what, 13 months ago? Better late than never, I guess, but MI300X *needed* to be impressive to even be worth talking about.

yeah AMD and its vaporware... love AMD but yeah they constantly fail in this regard.

With the Hawk Point announcement, I think it's pretty safe to say that Phoenix fits that description perfectly. They used past tense ("shipped") to describe a product that should still be their mainstay and has barely *actually* reached the market ......hopefully Hawk Point represents an improvement in N4 supply/whatever supply it was that hamstrung Phoenix availability.

It really is just Phoenix that donned a new name tag. From what AMD was saying I was expecting a groundbreaking increase in NPU performance. It looked more like a hardware stepping than anything, NPU arch doesn't change until Strix either.

 

[thesmokingman]

Semianalysis email newsletter has a brief overview of the AMD press release. The hardware specs have everyone drooling but the performance isn't really at the level that is expected of raw hardware of that caliber (yet).

Conclusion was that AMD has solid footing that will be a winning recipe on their hands........if they deliver on the software optimization. Seems like they've got some good motivation there, though, considering the expected customer base here.

Also, H100 launched what, 13 months ago? Better late than never, I guess, but MI300X *needed* to be impressive to even be worth talking about.



With the Hawk Point announcement, I think it's pretty safe to say that Phoenix fits that description perfectly. They used past tense ("shipped") to describe a product that should still be their mainstay and has barely *actually* reached the market ......hopefully Hawk Point represents an improvement in N4 supply/whatever supply it was that hamstrung Phoenix availability.

It really is just Phoenix that donned a new name tag. From what AMD was saying I was expecting a groundbreaking increase in NPU performance. It looked more like a hardware stepping than anything, NPU arch doesn't change until Strix either.

lol, that's one way to look at it, smh. H100 took 4 years to release at 30% uplift and 4x the cost. Bizness as usual? AMD's entry into the space was much anticipated because who the F wants to enrich a supplier at 70%+ margins huh?

AMD Stock Surges With Microsoft, Meta, and OpenAI Set to Use Its Latest AI Chip

AMD shares surged a day after the company launched its latest AI chip, with Microsoft, Meta, and OpenAI all set to use it.

www.investopedia.com

Advanced Micro Devices Inc. (AMD) shares surged over 8% in intraday trading Thursday after the chipmaker launched its latest artificial intelligence (AI) chip, with Microsoft (MSFT), Meta (META), and OpenAI all set to use it.

Let me guess these players don't know wtf they are doing huh with such mediocre hw?

 

[Assimilator]

H100 took 4 years to release at 30% uplift and 4x the cost.

Yeah, that's what happens when a so-called competitor doesn't compete.

Let me guess these players don't know wtf they are doing huh with such mediocre hw?

I'd love to hear you explain how share prices and press releases have any bearing on hardware capabilities and software ecosystem, but I don't want to lose braincells.

 

[TheoneandonlyMrK]

I think this plus the Apu version in servers possibly with a couple of epycs , noice.

 

[tabascosauz]

Let me guess these players don't know wtf they are doing huh with such mediocre hw?

When did I say either of those things? Make your point without putting words in my mouth.

Semianalysis conclusion was that MI300X should be well positioned against both H100 and (maybe) H200. Just inferring some things about MI300X itself, and how AMD is choosing to present its performance in the slides.

There's some software optimizations left, and like I said their customer base this time gives them good reason to keep working on that front.

 

[Assimilator]

Conclusion was that AMD has solid footing that will be a winning recipe on their hands........if they deliver on the software optimization.

AMD's past record in the latter area hasn't been stellar; I've seen no evidence to believe it'll be any better this time around. Hardware costs are insignificant when put up against those of software, after all. I'm expecting the standard cycle of AMD hypes the shit out of its pro-grade GPUs => a bunch of big name customers sign on because AMD is basically paying them to take those GPUs => big-name customers get fed up with AMD's immature drivers and ecosystem and go back to NVIDIA.

But hey, this is an AMD fanboy thread, and the copium is strong.

 

[Belfaborac]

Is it, really? Or is it just that one participant has a peculiar and possibly somewhat unhinged hatred of AMD?

 

[kapone32]

AMD's past record in the latter area hasn't been stellar; I've seen no evidence to believe it'll be any better this time around. Hardware costs are insignificant when put up against those of software, after all. I'm expecting the standard cycle of AMD hypes the shit out of its pro-grade GPUs => a bunch of big name customers sign on because AMD is basically paying them to take those GPUs => big-name customers get fed up with AMD's immature drivers and ecosystem and go back to NVIDIA.

But hey, this is an AMD fanboy thread, and the copium is strong.

I struggle to understand how people continue to use AMD's past to predict it's future. We don't know where these will land but no one can deny the specs. The APUs should also do quite well if priced accordingly. I know you think that AMD's drivers are weak but that is just not true.