• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 577 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I think the sensible conclusion is that you're comparing a product that hasn't launched yet to a product that has launched l, and maybe realise that's why you can get Snapdragon now and you can't get strix point.
Sure, but the Snapdragon X Elite was only just released in store on 18 June. I.e. the release dates are not that far apart. So it should be possible to roughly compare number of announced models.
 
Ah. Trying to predict x86's demise from the number of announced models. Sure, that will work just great.
I’m not. I simply asked what possible reasons there could be for number of announced Snapdragon X Elite models being higher. And it was not I who initially made the observation about the difference in number of announced models either. So chill.
 
Sure, but the Snapdragon X Elite was only just released in store on 18 June. I.e. the release dates are not that far apart. So it should be possible to roughly compare number of announced models.
I dunno chief. I see zero zen 6 laptops on preorder and I see zero of Nvidia arm CPU laptops on order. Probably a bad look for their share growth forecasts.

Seriously though, the launch dates are close but one isn't out yet. It's obvious you're gonna see fewer models on the unlaunched model. I had to write that sentence.
 
I dunno chief. I see zero zen 6 laptops on preorder and I see zero of Nvidia arm CPU laptops on order. Probably a bad look for their share growth forecasts.

Seriously though, the launch dates are close but one isn't out yet. It's obvious you're gonna see fewer models on the unlaunched model. I had to write that sentence.
Yes, that will of course explain a part of it. But there could be other explanations too. Just as an example, it could e.g. be because Qualcomm is pricing the Snapdragon X Elite really low to get it into more laptop models, since Windows on Arm is a new unproven kid on the block to a lot of users, so they have to dump the price to grab market share.
 
Yes, that will of course explain a part of it. But there could be other explanations too. Just as an example, it could e.g. be because Qualcomm is pricing the Snapdragon X Elite really low to get it into more laptop models, since Windows on Arm is a new unproven kid on the block to a lot of users, so they have to dump the price to grab market share.
Could also be they want to ride on the Copilot+ hypetrain.
 
Yes, that will of course explain a part of it. But there could be other explanations too. Just as an example, it could e.g. be because Qualcomm is pricing the Snapdragon X Elite really low to get it into more laptop models, since Windows on Arm is a new unproven kid on the block to a lot of users, so they have to dump the price to grab market share.
..... or Microsofts relentless drive to make WoA a thing and twisting the arm of every OEM to make models or else?
 

Strix Halo with 128GB of RAM would be able to slot nicely into the Mobile ML Dataset R&D Mobile Station niche. I'm starting to see it the appeal beyond Gaming Laptops.
 
Or cluster decoding is not the best solution for a single thread. However, this is the best solution for hyper-threading (SMT), similar to the Gracemont-Skymont concept.

Epyc is a priority for AMD, so Zen 5 is also a project with more Epyc in mind.
They way I understand it, you can get two clusters to work on 1 thread, but there has to be a branch happening in that cycle for the second decoder to get employed (and there is the belief that there's a branch in x86 code every 6 instructions on average).

The NOP benchmark probably didn't have branches in the part that measured the decoding width and that's why it only got 4-wide result in single thread?
 
I’m not. I simply asked what possible reasons there could be for number of announced Snapdragon X Elite models being higher. And it was not I who initially made the observation about the difference in number of announced models either. So chill.
AMD always gets cold shoulder, apparently OEMs will jump with bags of money on everyone else.
Those three lappies are all Asus tho? So probably just that brand is on preorder yet. Not sure now how many Snapdragon X ones Asus has on offer.
 

Strix Halo with 128GB of RAM would be able to slot nicely into the Mobile ML Dataset R&D Mobile Station niche. I'm starting to see it the appeal beyond Gaming Laptops.
Not having on-package memory?? This will lead to huge board sizes.

This was never meant for gaming, it will target >$2000.
 
Not having on-package memory?? This will lead to huge board sizes.

This was never meant for gaming, it will target >$2000.
One of Halo bullet points was that OEMs can shed off the Discrete GPUs platform costs with a single solution. There should be one or two gaming designs from ASUS to be presented at CES next year.
 
They way I understand it, you can get two clusters to work on 1 thread, but there has to be a branch happening in that cycle for the second decoder to get employed
That's how Tremont clustered decode works. However, since Gracemont (2021) the routing can be actively load balanced.
On of Halo bullet points was that OEMs can shed off the Discrete GPUs platform costs with a single solution. There should be one or two gaming designs from ASUS to be presented at CES next year.
Shedding off dGPU cost but having no "green RTX sticker" on a $2k G4M1NG laptop? Nah
 
Shedding off dGPU cost but having no "green RTX sticker" on a $2k G4M1NG laptop? Nah
Exactly, I know some will say but green is not needed any more but for most gamers it’s important.

You would have live and breathe AMD’s bubble to believe that RTX branding isn’t important to the majority of PC gamers. Just like how Ryzen X3D brand is the de facto for gaming CPUs now.

The problem with Halo is that for professional workloads you need Nvidia GPU and gamers prefer RTX, so the only benefit Halo has is the unified memory for LLMs which would make it a M3 Max/Pro competitor.

—-
I know you obviously can game on this but I’m largely referring to market expectations.
 
Meh, Strix is still kinda big for a mobile APU with 225mm². They wouldn't have done a 300mm² one even with competition.
AMD will make larger APUs.

Just not like "standard" Strix Point. Maybe even not like Strix Halo.

If there would be a market for larger, monolithic APUs, like M3 Pro - M3 Max in consumer space - AMD would do it in a heartbeat, assuming costs are feasible.
 
You would have live and breathe AMD’s bubble to believe that RTX branding isn’t important to the majority of PC gamers. Just like how Ryzen X3D brand is the de facto for gaming CPUs now.
Branding is nothing, all that matters is performance and cost. Nobody buys X3D chips because they have a cooler model number, they buy em because they're faster. Strix Halo definitely isn't going to dethrone Nvidia. On paper it should be a bit faster than a 4070m, 36 SM vs 40 CU, but still way under flagship GPUs. So the people buying $2000 laptops with monster dGPUs are not the target market.

With Strix Halo I'm thinking models will be more the $999-$1499 segment. Similar price range to 4060 and 4070 laptops, however it should have considerably better battery life and thermals. And potentially cheaper for OEMs to build out since its all on one package.

Now i'm not trying to say that a longer battery life is all they need to make people change their mind from buying a laptop with the coveted green sticker, but IMO it still has a place in the current market. OEMs are the real customers after all and on paper it should be a very potent APU in the mid range, price it right and I bet they'll buy some.
 
Qualcomm/Microsoft are pushing the Copilot+ PC very hard. Just look at home page of Bestbuy.com site. Also in store as well they have prime real estate dedicated to the PC. Bestbuy is offering it for 24 months no interest for its credit card customers.

I am not sure AMD would be able to get something like that going. They have never done something like this. its like Wintel campaign in 90s.
 
Branding is nothing, all that matters is performance and cost. Nobody buys X3D chips because they have a cooler model number, they buy em because they're faster.
That’s the bubble we live in. the majority buy on branding. Most gamers look for the Ryzen and RTX sticker.
 

Qualcomm/Microsoft are pushing the Copilot+ PC very hard. Just look at home page of Bestbuy.com site. Also in store as well they have prime real estate dedicated to the PC. Bestbuy is offering it for 24 months no interest for its credit card customers.

I am not sure AMD would be able to get something like that going. They have never done something like this. its like Wintel campaign in 90s.
I hope it fails tbh. If in future the translation layer works well I could get behind it. Right now it feels like using some newer features (copilot) to bamboozle customers into accepting no legacy compatibility. I doubt that's the intention, but that's what's happening and I hope it crashes and burns.
 
Back
Top