Exclusive: Intel urges PC makers to use cutting-edge CPUs amid shortage
Posted by SirActionhaHAA@reddit | hardware | View on Reddit | 102 comments
Posted by SirActionhaHAA@reddit | hardware | View on Reddit | 102 comments
Droid_pro@reddit
"Shovel company encourages consumers to buy more shovels"
Uptons_BJs@reddit
It’s really “shovel company encourages customers to adopt new shovel because they’re discontinuing old shovel, customers unhappy because they prefer old shovel”
b84ui@reddit
The customers as in the OEMs are unhappy, but I can pretty much assure you most consumers would prefer even a lower end laptop with Wildcat Lake over an ancient Alder/Raptor Lake chip that just throttled itself to worse than WCL performance as soon as it gets unplugged, while probably not even getting half of the WCL's battery life.
I know because I have a clunker of a 12900H laptop in front of me. It's fine plugged in all the time, but severely compromised as soon as it's on battery power.
certainlystormy@reddit
even arrow lake is pretty bad in power management at the higher end. my 275hx with the dgpu disabled has like an hour of doing anything that isnt just web browsing before it dies on battery lol
soggybiscuit93@reddit
ARL is pretty good as a laptop chip.
It's the HX line, which is not really a laptop chip that's the problem with battery life
certainlystormy@reddit
ahhh. just the "we give up, it's a gaming laptop" line of parts
soggybiscuit93@reddit
It's actually literally a desktop chip. It's the exact same silicon as a 285K, just detuned 300Mhz and they changed the socket to BGA.
The "H" line from ARL was actually designed to be a laptop chip. Some people are willing to give up the battery life advantages of a purpose-built laptop chip to get the extra performance.
certainlystormy@reddit
oh, that's interesting. i knew it was good but didn't know it was like that lol
lukfi89@reddit
Let's see how that strategy works out for them. If their OEM partners have to design new machines to accomodate the 18A chips, they might as well turn to AMD.
DontchaOpe@reddit
You think AMD has historically or in the current moment been able to supply laptop OEMs with products?
Dangerman1337@reddit
I think we'll see maybe Medusa Point being announced at Computex soon?
DontchaOpe@reddit
Doesn’t really matter if there is no supply. Which AMD has not had for OEMs for the last decade.
nanonan@reddit
I've been able to buy a wide variety of AMD laptops in the last decade. What makes you think there are supply issues at all?
SirActionhaHAA@reddit (OP)
*incorrect buzzer sound*
Luggage-Lock@reddit
This is just factually incorrect.
SirActionhaHAA@reddit (OP)
incorrect buzzer sound
nanonan@reddit
Sure. Name a point in time where you couldn't buy an AMD laptop.
SirActionhaHAA@reddit (OP)
Amd mobile x86 unit share: 28.3%
Amd mobile x86 revenue share: 28.9%
Their mobile share has increased every quarter against the reddit myth of no supply, closing in on 1/3 of the market now.
soggybiscuit93@reddit
AMD, like everyone else, is putting their efforts (even more so than usual) into datacenter first. When they're still supply constrained in DC, there's no chance they want to divert wafers to capture the budget laptop segment
lukfi89@reddit
Historically, no, but in the last few years, AMD has gained more market share in consumer segments. It is now around 30 %. Of course I don't know how much inventory they have on hand, etc.
Sevastous-of-Caria@reddit
With old chips and its bins yes. High end goes to servers
DontchaOpe@reddit
*incorrect buzzer sound*
Whirblewind@reddit
So you asked a question you didn't actually know the answer to, then.
Wonderful-Lack3846@reddit
AMD is busy with AI. They give zero fucks about you or other consumers.
Intel does give a little shit because they want to gain reputation back and are capable of getting good margins.
steve09089@reddit
Turn to AMD for zero capacity and even pricier chips?
Because AMD priorities data center too
elkond@reddit
turn down to what, imaginary volume?
Arachnapony@reddit
Does AMD have the capacity?
TwanToni@reddit
intel 7 was a fantastic node, such a shame....
grahaman27@reddit
the 14900k on intel 7 is a marvel of achievement. Beating AMD TSMC 3 in everything but gaming (because of 3d cache)
Geddagod@reddit
I swear I specifically made this same comment under your post in the Intel stock subreddit several months ago lol, but:
AMD does not use "TSMC 3" in client. It's N4P. And even then, claiming that the 14900k beats the 9950x in everything but gaming isn't even true.
Ok-Parfait-9856@reddit
Where does the 14900k lose? I thought they were comparable for productivity tasks
nanonan@reddit
Lifespan for one.
Geddagod@reddit
Tons of specific benchmarks. Check out this TPU review, for example. Just from the rendering section, V-ray, keyshot, and Corona. The 14900k is behind the 9950x in these nT workloads.
I never said the 14900k is outright worse though, I think they are about comparable too. More specifically though....
3D center's meta review averaging data from many different reviewers has the 9950x as having a marginal lead, and comparable nT perf to the 285K, but making the claim that Intel wins in everything but gaming with RPL is just not accurate.
Lastly, the 9950x's uniform cores and support of AVX-512 are more advantages for productivity tasks that care about those features. Frankly, it can easily be argued AMD is outright better, especially when looking at Phoronix's data, but again, let's just say they are comparable, since I don't really want to get into the weeds of this.
raptorlightning@reddit
Takes a power plant to run and melts itself out of the box, but go off.
TwanToni@reddit
same can be said with the burn rates on the recent x3d chips
grahaman27@reddit
indeed it does, but a wins a win
Geddagod@reddit
It's not a win though
elkond@reddit
even in gaming it can pull off better 0.1% lows because u can have any amount of cache but if stuff aint cached, u go against infinity fabric dogshit io
Tower21@reddit
I too enjoyed Intel 10 +++&1/2+ nm
TwanToni@reddit
cool story? Not sure what your point is. Intel is basically caught up with TSMC with 18A
Tower21@reddit
Intel 7 is rebranded Intel 10nm, and all those +.
How you gonna call Intel 7 fantastic, when you don't even know that?
smh
SomeMobile@reddit
Except intel 10 is comparable to tsmc's 7 ?
nanonan@reddit
Sure. Still their rebranded 10nm node.
Geddagod@reddit
Depends on the variant. Early 10nm? Not likely. Later versions such as 10 ESF or 10SF? Prob.
TwanToni@reddit
Again, what Is your point?????????????????????????????
Tower21@reddit
I provided context, you really couldn't rub your last two brain cells together to figure that out?
Let me know if you need help with other low hanging ideas in future, I help people with special needs all the time.
ryanvsrobots@reddit
No need for that.
TwanToni@reddit
Intel 7 with alder lake was fantastic. Your context doesn't help your case. It's just whining for the sake of whining.
Geddagod@reddit
A node behind. Technically a subnode rn since ig no N2 products are out, but in a few months...
Really there's not much of a change in competitiveness (N-1) if you imagine Intel 3 is a N5/4 class node, and TSMC was producing N3 stuff. Or when Intel was on 10SF while TSMC was on N5.
Though I suppose there is an advantage of node shrinks also gaining less and less PPA benefits over their predecessors. Despite that , it seems like customers are still willing to pay top dollar for the newest stuff, based on demand for N2.
TwanToni@reddit
18A-P? Again, not sure what the point was when intel hasn't been stuck on a node since 14nm that the previous person was whining about
Geddagod@reddit
Hopefully N3P level.
I mean they were stuck on the Intel 10nm node family too. Intel 7nm (now called Intel 4/3) was delayed for a while too.
He also isn't wrong about Intel 7 being 10nm with a tons of plusses. Intel 7 is just renamed 10ESF, afterall.
The exact amount of plusses would look something like this:
10nm CNL
10nm+ ICL
10nm++ (SF) TGL
10nm+++ (ESF) ADL aka Intel 7
10nm++++ (UESF) RPL aka Intel 7 ultra
There's nothing wrong with Intel getting their moneys worth out of the node though (if they just didn't have to delay Intel 4). It's pretty impressive how they managed to squeeze so much out of the node.
cp5184@reddit
didn't intel just rename intel 10 to 'intel 7'?
Geddagod@reddit
Yes, Intel 10 enhanced super fin specifically.
Intel 10nm+ server parts had no better perf/watt than Intel's older 14nm plus plus etc etc parts.
But the density gains were real.
Intel 10nm vs Intel 14nm was a node shrink though, even if it took them several generations to fully realize the perf/watt benefits (or maybe SNC sucked so bad it ate up an entire node shrinks worth of perf/watt lol).
Later 10nm variants were comparable to N7, on paper and in products.
IamGeoMan@reddit
I'm surprised they have enough capacity to divert 18A to consumer CPUs. I thought they were inundated with customers for 18A hence the insane SP. Guess not 🤡
grahaman27@reddit
They don't yet, but yields are improving quickly. They just can't afford to maintain legacy nodes.
Soon Intel 7 will be retired and they will shutdown the fabs, including Israel. I can see that coming soon
DontchaOpe@reddit
Intel 7 isn’t going to be retired until after this decade at the earliest, then those fabs will either be converted to an EUV node or a customer DUV node (16, 12, 7)
grahaman27@reddit
I'm just reading the room. What do you think this announcement from Intel means?
nanonan@reddit
It means they are scared that Apple is about to kick its arse in the laptop game and AMD sales will pick up when they cut off old supply.
DontchaOpe@reddit
Long term, higher margin agreements were signed shifting wafers to data center segment. 18A is currently only producing consumer chips (ignoring the small volume of Clearwater forest), so they want OEMs to buy and sell those chips.
OEMs obviously want these lower priced chips that they can bundle into budget builds.
TechTechTerrible@reddit
It means they are using all the capacity from pre 18a nodes on data center products and so the only thing left to buy is products built on 18a.
Capital-Froyo-4359@reddit
Intel isn't producing a single 18A product for an external client.
Johnny_Oro@reddit
AFAIK 18A has good yield, just not good parametric yield. Basically means high yield, but low chance to meet strict performance standards. For low end CPUs like Wildcat and lower clock speed lower core count Panther Lake, it's fine.
DontchaOpe@reddit
All the conversation in the media is on 18AP which Intel hasn’t even released a product on yet. Design commitments take 2\~ years to come to market on a good node.
What are you implying lol
grahaman27@reddit
18A , for CPU
LegitimateMost2602@reddit
kind of a weird situation where they're pushing oems toward newer platforms during a supply crunch instead of just clearing older stock. practically speaking this means new builds might be forced onto different socket requirements, so if you're planning a near-term upgrade it's worth double checking mobo compatibility before assuming your current platform gets refreshed parts.
soggybiscuit93@reddit
At this point it seems that the older stock has been getting cleared out
OfficialHavik@reddit
That’s the thing…. I don’t think there IS any older stock and manufacturing more Raptor Lake when you could be making say Emerald Rapids on that same tooling doesn’t make much sense….
Capital-Froyo-4359@reddit
Shutting down Fab 42 was a big mistake. I swear Intel has gotta be the poster child for terrible management. Its like every decision they make turns out to be wrong. You'd think just by random chance they'd do better.
DontchaOpe@reddit
When was that announced lol. News to me
Capital-Froyo-4359@reddit
These sort of things aren't ever announced.
DontchaOpe@reddit
Yes they do. it’s 20B of capex. They either get sold or used.
Capital-Froyo-4359@reddit
I've never been able to find a public document outlining what fab is currently producing what process etc. Theres general announcements made, but trying to figure out any sort of specifics is very hard without digging.
DontchaOpe@reddit
But there are plenty of public documents outlining fab sales and transfers… there are multiple from Intel within the last 10 years. Companies don’t let 20B dollar assets sit unutilized.
Capital-Froyo-4359@reddit
Intel does, lol. They have 2 fab shells in Isreal fully built but not operating.
tiradium@reddit
Pretty sure they meant shutting it down and resuming. Fab is operational but Intel took it offline for a couple of years and that continues to hurt them
Minced-Juice@reddit
Fab 42, assuming it made Intel 7 wafers during the time in question, was also the fab responsible for the defective wafers that partly caused the Raptor Lake and Raptor Lake refresh instability issues.
If it was taken offline for a thorough investigation into how the contamination happened, then it is a good thing IMO.
ryanvsrobots@reddit
It wasn't.
Capital-Froyo-4359@reddit
No, its not. It was shit down to re-tool for 18A/14A.
animi0155@reddit
Fab 42 was never shut down since it started production. There have been changes in output in response to changes in demand and some retooling for 18A, but none of those stopped all 10nm/Intel 7 production.
Capital-Froyo-4359@reddit
Source? And even if its like 50% offline the point still stands that they should have left it at full capacity and built out Fab 62 instead.
animi0155@reddit
With what money? It doesn't make financial sense to run at full capacity when you don't have demand or forecasted demand. They would be spending money to produce and store stuff that they did not think they would have committed customers for, for long enough, to justify maintaining the capacity.
Intel wouldn't run Intel 7 if they could get away with it. It's an expensive process for what it is, and the products on it cannot be sold at high margins to remain remotely competitive. Intel 4/3 are better but the demand and margins also weren't quite there, and it's not made in Arizona (source: basically every 10-K and quarterly earnings report from the past few years).
For Fab 62, it's cheaper to tool up existing, depreciated fab space that isn't utilized or expect to free up in the near-ish future. TSMC does not build out a fab until it knows it has customers for it, and Intel has to do the same. They're not going to take the depreciation and operating cost hit to occupy a fab they don't have the demand to start filling.
Capital-Froyo-4359@reddit
Intel has massive demand for their DCAI chips which use 7nm. The inability to meet demand is costing them billions.
RedTuesdayMusic@reddit
Strange of intel to recommend AMD, no?
Uptons_BJs@reddit
Considering that the only 18A parts are on mobile, I'm assuming everyone in this article is talking about mobile chips.
Raptor Lake is a 4 year old product, 5 generations old for most mobile parts. The fact that Raptor Lake demand is still so high even with prices are high is really a sign that Intel's Raptor Lake replacements in the 100 and 200 series has not really been able to replace Raptor Lake in many use cases. If you're gaming or need single thread performance, the best Intel chip to go with on mobile is still Raptor Lake.
This is really a problem with Intel's chip design that Raptor Lake's replacements haven't been able to dislodge its position in the market until Panther Lake (and even that is semi-forced).
soggybiscuit93@reddit
It's really just that RPL-U is (was) very cheap, and good enough. For the type of consumer buying RPL-U, everything after it is better for them. But cost more.
So now that WCL is out, RPL-U will be more expensive than WCL and become the new budget-volume product. Selling large volumes of cheap RPL-U is not a good use of Intel's fabs atm
Seanspeed@reddit
Exactly this. Arrow Lake may not be the big performance leap over Raptor Lake that many hoped, but it *is* way more efficient, which is very important for laptop segment. So it very much has value in the market.
And for like 98% of users on laptop, some examples of Raptor Lake having faster ST performance are not making any deciding factor.
This is just a cost thing.
Ok-Parfait-9856@reddit
WCL gets the same single core performance as RPL
ResponsibleJudge3172@reddit
The situation in the laptop sphere is opposite the desktop. Mobile Arrowlake is better unless you go for HX where its more tied. More efficient in all scenarios though
Appropriate_Name4520@reddit
yeah Intel maybe try to make good affordable hardware yourself instead of encouraging others?
sussy_ball@reddit
Just read about Intel Project Firefly, which aims at building cheaper budget laptops
WASynless@reddit
I would have upgradedy PC by now if it wasn't for the elephant in the room. I do not know how Intel can fix it or mitigate it, but it sure looks like they are not trying
DontchaOpe@reddit
RAM? Intel doesn’t manufacture ram
sussy_ball@reddit
Currently* Intel has partnered with Citi Bank to again enter into the RAM market producing what they call Z-Angled Memory, but its only coming to Datacenters when *if* it enters production in 2028
spicesucker@reddit
Hence
bb999@reddit
Buy slower lower capacity RAM and live with it until RAM prices drop? (if they ever)
Minced-Juice@reddit
Buy Lunar Lake?
They essentially don't pass on any cost for the on-package memory to the OEMs,
SirActionhaHAA@reddit (OP)
Archived here
https://archive.is/eHwFi
tldr Intel is forcing pc manufacturers to upgrade to 18a cpus because it moved almost all intel7 capacity to dc products. It's tellin them to take it or leave it and it's gonna drive up prices of consumer products soon as demand > supply
Ok-Parfait-9856@reddit
What’s dc products? Sucks that Intel 7 products will become more rare soon
jrp70@reddit
Data centers.
INITMalcanis@reddit
I mean we might as well use the expensive CPUs when the RAM alone costs like £400-600
Any-Pop-4795@reddit
yeah no thanks get fucked
soggybiscuit93@reddit
Now that WCL is out, that's where Intel wants OEMs to go for their cheap, commodity volume CPUs. There's high demand for Intel 7 in datacenter. Intel has a (for the most part), better cheap CPU in the form of WCL (vs RPL-U), and selling low cost RPL-U is just a waste of their fab capacity at this point.