Intel says integrating RAM into Lunar Lake SoC was a mistake, might abandon desktop GPUs again
Intel says integrating RAM into Lunar Lake SoC was a mistake, might abandon desktop GPUs again
During the company's third-quarter 2024 earnings call, Intel confirmed that its future laptop chips will return to the traditional use of RAM sticks, reversing Lunar Lake's radical...
Panther Lake and Nova Lake laptops will return to traditional RAM sticks
Reverting to RAM sticks is good, but not shutting down GPU line. GPU market needs more competiter, not less.
Intel can't afford to keep making GPUs because it doesn't have the reliable CPU side to soak up the losses. The GPU market has established players and Intel, besides being a big name, didn't bring much to the table to build a place for itself in the market. Outside of good Linux support (I've heard, but not personally used) the Intel GPUs don't stand out for price or performance.
Intel is struggling with its very existence and doesn't have the money or time to explore new markets when their primary product is cratering their own revenue. Intel has a very deep problem with how it is run and will most likely be unable to survive as-is for much longer.
As a Linux user of an Intel Arc card. I can safely say that the support is outstanding. In terms of price to performance, I think it’s pretty good too. I mainly enjoy having 16GB of VRAM and not spending $450-$500+ to get that amount like Nvidia. I know AMD also has cards around the same price that have that amount of VRAM too though
Basically there is only money at the top of the gpu range. Everything else is a budget card with razor thin margins.
AI specific chips will take off over time but even then the ship is starting to sail . These are mostly data center projects.
besides being a big name, didn't bring much.
Absolutely wrong. A lot of old and dated information in your post.
They have something no one else has: manufacturing, and very low price and great performance after recent driver updates. They just lack the driver stability which has been making leaps and bounds.
I do not think anyone else can enter the market, let alone with an edge.
coming up next: Intel fires 25% of their staff, CEO gets a quarterly bonus in the millions
Gelsinger said the market will have less demand for dedicated graphics cards in the future.
No wonder Intel is in such rough shape! Gelsinger is an idiot.
Does he think that the demand for AI-accelerating hardware is just going to go away? That the requirement of fast, dedicated memory attached to a parallel processing/matrix multiplying unit (aka a discreet GPU) is just going to disappear in the next five years‽
The board needs to fire his ass ASAP and replace him with someone who has a grip on reality. Or at least someone who has a some imagination of how the future could be.
Does he think that the demand for AI-accelerating hardware is just going to go away? That the requirement of fast, dedicated memory attached to a parallel processing/matrix multiplying unit (aka a discreet GPU) is just going to disappear in the next five years‽
Maybe the idea is to put it on the CPU/NPU instead? Hence them going so hard on AI processors in the CPU, even though basically nothing uses it.
Unless he thinks he's going to serve all that from the die in the next 5 years.
You think Intel is going to have 500-850mm^2 dies?
That's what they need to compete in the GPU space.
I don't think Lunar lake wasn't a "mistake" so much as it was a reaction. Intel couldn't make a competitive laptop chip to go up against Apple and Qualcomm. (There is a very weird love triangle between the three of them /s.) Intel had to go to TSMC to get a chip to market that satisfied this AI Copilot+ PC market boom(or bust). Intel doesn't have the ability to make a competitive chip in that space (yet) so they had to produce lunar lake as a one off.
Intel is very used to just giving people chips and forcing them to conform their software to the available hardware. We're finally in the era where the software defines what the cpu needs to be able to do. This is probably why Intel struggles. Their old market dominant strategy doesn't work in the CPU market anymore and they've found themselves on the back foot. Meanwhile new devices where the hardware and software are deeply integrated in design keep coming out while Intel is still swinging for the "here's our chip, figure it out for us" crowd.
In contrast to their desktop offerings, looking at Intel's server offerings shows that Intel gets it. They want to give you the right chips for the right job with the right accelerators.
He's not wrong that GPUs in the desktop space are going away because SoCs are inevitably going to be the future. This isn't because the market has demanded it or some sort of conspiracy, but literally we can't get faster without chips getting smaller and closer together.
Even though I'm burnt on Nvidia and the last two CPUs and GPUs I've bought have been all AMD, I'm excited to see what Nvidia and mediatek do next as this SOC future has some really interesting upsides to it. Projects like ashai Linux proton project and apple GPTK2 have shown me the SoC future is actually right around the corner.
Turns out, the end of the x86 era is a good thing?
contrast to their desktop offerings
That's because server offerings are real money, which is why Intel isn't fucking those up.
AMD is in the same boat: they make pennies on client and gaming (including gpu), but dumptrucks of cash from selling Epycs.
IMO, the Zen 5(%) and Arrow Lake bad-for-gaming results are because uarch development from Intel and AMD are entirely focused on the customers that pay them: datacenter and enterprise.
Both of those CPU families clearly show that efficiency and a focus on extremely threaded workloads were the priorities, and what do you know, that's enterprise workloads!
end of the x86 era
I think it's less the era of x86 is ended and more the era of the x86 duopoly putting consumer/gaming workloads first has ended because, well, there's just no money there relative to other things they could invest their time and design resources in.
I also expect this to happen with GPUs: AMD has already given up, and Intel is absolutely going to do that as soon as they possibly can without it being a catastrophic self-inflicted wound (since they want an iGPU to use). nVidia has also clearly stopped giving a shit about gaming - gamers get a GPU a year or two after enterprise has cards based on the same chip, and now they charge $2000* for them - and they're often crippled in firmware/software so that they won't compete with the enterprise cards as well as legally not being allowed to use the drivers in a situation like that.
ARM is probably the consumer future, but we'll see who and with what: I desperately hope that nVidia and MediaTek end up competitive so we don't end up in a Qualcomm oops-your-cpu-is-two-years-old-no-more-support-for-you hellscape, but well, nVidia has made ARM SOCs for like, decades, and at no point would I call any of the ones they've ever shipped high performance desktop replacements.
- Yes, I know there's a down-stack option that shows up later, but that's also kinda the point: the ones you can afford will show up for you... eventually. Very much designed to push purchasers into the top end.
He’s not wrong that GPUs in the desktop space are going away because SoCs are inevitably going to be the future. This isn’t because the market has demanded it or some sort of conspiracy, but literally we can’t get faster without chips getting smaller and closer together.
While I agree with you on a technical level, I read it as Pat Gelsinger intends to stop development of discrete graphics cards after Battlemage, which is disappointing but not surprising. Intel's GPUs while incredibly impressive simply have an uphill battle for desktop users and particularly gamers to ensure every game a user wishes to run can generally run without compatibility problems.
Ideally Intel would keep their GPU department going because they have a fighting chance at holding a significant market share now that they're past the hardest hurdles, but they're in a hard spot financially so I can't be surprised if they're forced to divest from discrete GPUs entirely
I would like to see further development but I always had a sneaking suspicion that its life was limited due to the fact that ARC does not come from Intel's fabs either. Like lunar lake, Arc is also made at TSMC.
Gelsinger said the market will have less demand for dedicated graphics cards in the future.
In other news, Intel is replaced by Nvidia in the Dow Jones, a company that exclusively produces dedicated graphics cards: https://lemmy.world/post/21576540
Nvidia does more than just GPUs.
Nvidia makes both SoCs like the Tegra series and server CPUs (Grace; ARM based to be used with their ML/AI cards with much higher bandwidths than regular CPUs).
Nvidia also just announced that they are working on a consumer desktop CPU.
Well I had the same thought a few years ago when APUs started getting better. But then I'm not the CEO of a huge tech company, so nobody lost their job because I was wrong.
Can you be certain that no company monitors your brain and uses that as their CEO somehow?
Historical success/performance of DOW is biased by these strategic decisions to replace losers before their bankruptcy but where decline is obvious.
And here I was thinking Arc and storage were the only semi-competitive wings of intel... They just needed a couple of years for adoption to increase
I've commented many times that Arc isn't competitive, at least not yet.
Although they were decent performers, they used twice the die size for similar performance compared to Nvidia and AMD, so Intel has probably sold them at very little profit.
Still I expected them to try harder this time, because the technologies to develop a good GPU, are strategically important in other areas too.
But maybe that's the reason Intel recently admitted they couldn't compete with Nvidia on high end AI?Yeah true, plus I bought my a770 at pretty much half price during the whole driver issues and so eventually got a 3070 performing card for like $250, which is an insane deal for me but no way intel made anything on it after all the rnd and production costs
The main reason Intel can't compete is the fact CUDA is both proprietary and the industry standard, if you want to use a library you have to translate it yourself which is kind of inconvenient and no datacentre is going to go for that
Still I expected them to try harder this time, because the technologies to develop a good GPU, are strategically important in other areas too
I think I read somewhere that they're having problems getting AIB partners for Battlemage. That would be a significant impediment for continuing in the consumer desktop market unless Battlemage can perform better (business-wise) than Alchemist.
They probably will continue investing in GPU even if they give up on Arc, it might just be for the specialized stuff.
I have no confidence in Intel's long-term prospects.
Intel has been a mistake since 1978. But evil doesn't generally die.
Intel's long term prospects rely on China invading Taiwan.
Intel is a CIA champion. Vector for backdoor spying and kill switches. Why not embed plastic explosives on every motherboard, since US/Trump praised the Israel strategy?
Taiwan declaring independence and offering to host US nuclear missile bases... incoming.
Blaming loss on SoC? Lmfao. SoC is better. Just stop offering a lower tier and make all SoC 32gb+
… looking at you too, Apple.
I'm wondering, the integrated RAM like Intel did for Lunar Lake, could the same performance be achieved with the latest CAMM modules? The only real way to go integrated to get the most out of it is doing it with HBM, anything else seems like a bad trade-off.
So either you go HBM with real bandwidth and latency gains or CAMM with decent performance and upgradeable RAM sticks. But the on-chip ram like Intel did is neither providing the HBM performance nor the CAMM modularity.
The transfer speed isn't the big issue, it's the density and reliability. Packing more heat generating stuff onto the SoC package just makes it more difficult to dissipate. The transfer of data to where it needs to be is still the same, so the trade-off is pretty null in that sense except reduction of overall power consumption.
Look up intel i7 8709g
I see the idea of Intel dropping arc as good news for AMD. Intel was going to chip at AMD’s marketshare well before Nvidia’s. It would be better to have more competition though.
AMD would never close their GPU department because they sell their apu to Xbox, playstation, steam deck
Intel could have conceivably competed with them there too. If they were still a competent business and not in a crisis of mismanagement. Its amazing how much better AMD is managed compared to Intel.
I wasn’t saying AMD would shut down, but that Intel would take market share from them before truely affecting Nvidia’s market share. ie AMD and Intel would be fighting over the same 25ish% of the pc market.
With RAM access being the one big bottleneck of a modern PC, can anyone in the know tell me about those SoCs? How much RAM did they have, and was it faster than external DIMMs?
You shouldn't be comparing with DIMMs, those are a dead end at this point. CAMMs are replacing DIMMs and what future systems will use.
Intel likely designed Lunar lake before the LPCAMM2 standard was finalized and why it went on package. Now that LPCAMMs are a thing, it makes more sense to use those as they provide the same speed benefits while still allowing user replaceable RAM.
For laptops, it seems like a winner to me. Do you need to expand your laptop's GPU memory? It was justified by lower power, and likely better transfer rates.
They have to try revive their idea like they did in intel core i7 8709g first though