- Mar 3, 2017
- 1,659
- 6,101
- 136
Well yeah but that's a cost structure problem, Intel is themselves to blame.
Like QC will be shipping your purvas and AMD will be shipping krakens just fine.
4 months for ramping just a new PHX revision. OkHawk point was launched in early January and it took barely 4 months to land at BestBuy, what make you think that for Strix it would be 6 months + ?..
The issue really is MTL/ARL-U being not cost-competitive for lower end segments due to higher die/assembly costs and LNL is explicitly a premium tablet part which they can't price down that much.Maybe the low cost of Raptor is not the full story, there may be more to the story.
That's fast for a pretty pricey part to reach $600 crapbooks (also took Intel as much to ramp Alder revision aka RPL).4 months for ramping just a new PHX revision. Ok
Because OEMs prefered to get rid of Phoenix chips and laptops inventories while stockpiling at their place the forcibly more sellable Hawk point ones..?4 months for ramping just a new PHX revision. Ok
Yes they can, they dumped a ton of HWK into the channel, units are up gigantic double digits YoY!
Thank you for saying the obvious.Looking at past and current quarters:
- Intel client declined much more QoQ than AMD. AMD nearly flat in Q1
- Intel Q2 forecast for client is flat. AMD client forecast is up (likely single digits).
I think this corroborates very decent ramp of Hawk Point, and if Lisa Su says Strix design wins > Hawk, that would indicate accelerating momentum.
Another tidbit: AMD is forecasting higher units and higher ASPs going into H2. Which would indicate AMD shifting volume from legacy (Zen2, Zen3) to current gens (Zen4, Zen5), moving up the ladder, possibly pushing Intel down the ladder.
I don't know where you live (well i actually know, but i think you know what i mean) , but if i'm not completely wrong Cezanne was the beginning of the Ryzen Mobile "Meme Launches".You're skipping like CZN ramp which was on shelves before the actual launch date or HWK ramp which is everywhere now
That's RMB.but if i'm not completely wrong Cezanne was the beginning of the Ryzen Mobile "Meme Launches".
CZN was a fast launch, you could get -H parts in China before even the actual on shelves launch date.It launched at CES2021, then Review Outlets received few Devices to Review it end of February, first Devices from Asus were April, but like 95% of Customers had to wait for June to get one.
Raptor is just Alder and it took about 4 months to ramp.Also Hawk is basically Phoenix (is it even a new stepping?). So they needed like 3-4Months after "Launch" to sell the same chip with a new name.
That's a channel issue.Overall Phoenix Was just a big mess.
Looking at past and current quarters:
- Intel client declined much more QoQ than AMD. AMD nearly flat in Q1
- Intel Q2 forecast for client is flat. AMD client forecast is up (likely single digits).
I think this corroborates very decent ramp of Hawk Point, and if Lisa Su says Strix design wins > Hawk, that would indicate accelerating momentum.
Another tidbit: AMD is forecasting higher units and higher ASPs going into H2. Which would indicate AMD shifting volume from legacy (Zen2, Zen3) to current gens (Zen4, Zen5), moving up the ladder, possibly pushing Intel down the ladder.
China is something different imo, you probably would even get some GNR parts there already, if you search long enough. I'm from Europe, so people in China being able to buy something in January, when I can't buy it until June after it launched at the beginning of January doesn't really count as fast launch for me.CZN was a fast launch, you could get -H parts in China before even the actual on shelves launch date.
Naa, they got deprived of RMB/PHX goodness pretty hard, only with HWK did China channel rebound for AMD.China is something different imo
Naa pretty sure I could fetch a few CZN-H devices by mid-April in Germany.when I can't buy it until June after it launched at the beginning of January doesn't really count as fast launch for me.
"AI" is found 7 times in this quote."but PCs will also let users interact with AI every day. We’re witnessing a significant growth of local AI applications for both personal and business purposes. With this increase of local AI comes a higher demand for incorporating AI engines in client and end devices. We were the first company to bring AI to PCs over a year ago by integrating a dedicated neural processing unit (NPU) on an x86 processor. These dedicated AI engines handle local AI applications efficiently, offering better performance and battery life in popular devices like ultrathin laptops. We have exciting announcements at Computex this June."
That's RMB.
Well yeah but slow DDR5 ramp is all on Intel DCAI.RMB was a big setback for AMD. Big opportunity missed.
I meant the product in general. Phoenix isn't/wasn't bad, but still a bit underwhelming. Only real uplift was pure CPUs tasks. Good, but not what the majority of people need in their devices. iGPU wasn't a good uplift both in efficiency and performance, battery life was more of a slight regression from Rembrandt.That's a channel issue.
Which I mentioned in my comment. Few ASUS devices in April.Naa pretty sure I could fetch a few CZN-H devices by mid-April in Germany.
iGP is the overall RDNA3 issue and BL stuff was LPDDR-related (sounds illogical but yes.png), fixed now.iGPU wasn't a good uplift both in efficiency and performance, battery life was more of a slight regression from Rembrandt.
Well yeah but slow DDR5 ramp is all on Intel DCAI.
Not AMD's fault SPR didn't make it on time and wasn't a mid'21 launch.
Yeah, Hawk Point seems to be what Phoenix should've been from the beginning. AMD should definitely leave the 1 Year cadence for completely new mobile APUs, it's just too fast to do proper products. Phoenix just wasn't ready. I really wonder how Strix Point will turn out in terms of matureness.iGP is the overall RDNA3 issue and BL stuff was LPDDR-related (sounds illogical but yes.png), fixed now.
Raptor/ADL U mobile were sold very cheap and Intel flooded the channel (and nuked their margins) with them. IIRC an i3 ADL/RDL U costed more or less the same as AMD Mendoccino.- Intel said that their costs for pre-EUV nodes is very high and expects lowering the costs as EUV nodes come on line. If Intel says the cost of Intel 7 is very high, then how can cost Raptor be very cheap?
I belive it was still under the old arrangements where the foundry was basically giving them away to CCG. Now that Foundry is separated, CCG needs to pay the fair share for the RPL.- Last quarter, Intel reported higher than expected profits, mainly by some pulling Raptor dies from inventory. Next quarter, intel forecasts lower than expected profits, since the storage is empty, and more new Raptor dies have to be made
PHX 2 is 137mm². RPL-U/ADL-U is around 180 - 200mm² iirc? (Less P cores but GT3 Iris Xe GPU compared to DT GT1 GPU).- I don't know the small Raptor and small Phoenix die sizes, but big Raptor is ~280mm2, big Phoenix is 178mm2
For this year, yes. Next year, Barcelo-R, Rembrandt-R and HWK are RIP (Except for some very budget designs). Kraken will come with force to replace them. It's an AI PC Ready SoC and will introduce AI PC for the mainstream designs.In late 2024 Phoenix and Hawk Point will be relegated as mainstream cheap products while Strix will take the higher segments, so far Intel can only compete, barely, with the formers, at this (Strix...) point OEMs who want to gain or simply keep marketshares at same level will have no other choice than to jump in the Hawk/Strix bandwaggon.
Strix Point Next Gen AI PC is what Papermaster is mainly hinting at. Computex will have Z5 introduction, Strix and Granite introduction and design wins and (maybe?) RDNA 4."AI" is found 7 times in this quote.
Just curious, if that means anything... )
a) yes yuo dobut I don't really need the CPU perfomance
155 or so.RPL-U/ADL-U is around 180 - 200mm² iirc?
Sorry, I don't see why. If we assume end of this year there will be my dream laptop, one with Strix, one with LNL. Given battery life and iGPU aren't much (>10%) worse on LNL than on Strix and Strix is meaningfully more expensive (200€ at a general price range of ~1500€), I would definitely take the LNL one. I don't see how 30% ST and 100% MT would help me, even if ZEN5 is such a beast. It should do some offic work and some light gaming from time to time, not more, not less.a) yes yuo do
b) 1t perf delta!
Your literal web browser.Which programs for casuals benefit ZEN5s ST perf in a meaningful way?
LNL itself is the expensive part.and Strix is meaningfully more expensive (200€ at a general price range of ~1500€)
Raptor is just Alder and it took about 4 months to ramp.
Very technically so.Well, it's new silicon.
No?Took the server variant Golden Cove cores from Sapphire Rapids, called the Raptor Cove,
More cache, yes. But again, not on mobile?and improved the e-core implementation a bit.