News Intel LGA1851 Socket For Future Arrow Lake CPUs Detailed

[Admin]

Igor's Lab provides a ton of information on Intel's forthcoming LGA1851 socket for Arrow Lake processors.

Intel LGA1851 Socket For Future Arrow Lake CPUs Detailed : Read more

 

[bit_user]

From what I can tell, neither this nor the Igor's Lab article says what's happening to DMI (i.e. the chipset link). Currently, it's PCIe 4.0 x8. What's happening with its speed and width?

 

[usertests]

The maximum dynamic pressure for the LGA1815 socket has increased from 489.5N to 923N, an 89% increase. In layman's terms, it'll require more pressure from CPU coolers. You may be able to conserve your CPU cooler but will need a new mounting kit.

What? Why?

 

[who_farted]

I am surprised we still have to deal with the fragility of pins. Why can't we just have a pad to pad connection?

 

[Dantte]

I am surprised we still have to deal with the fragility of pins. Why can't we just have a pad to pad connection?

Pads dont have tolerances and must be on a perfectly level surface to make contact.

Your question is like saying: "why cant we have heat sinks that dont require TIM?" You know why we have TIM, because the surfaces are not perfectly level and we need a (Thermal Interface) Material between them to fill in the gaps. The "pins" on a processor have tolerances and fill in those gaps between the socket and processor, some go a little deeper, some a little shallow, but they still make contact, pads cant do this.

 

[Friesiansam]

Whizzo! Another new socket, another new motherboard etc...

On the other hand you could go AM5 and be able to upgrade the CPU in a few years, to the latest thing, without needing a new motherboard.

 

[InvalidError]

What? Why?

I'm guessing the doubling of mounting pressure is because the pin density increase is achieved using stubbier pins.

Pads dont have tolerances and must be on a perfectly level surface to make contact.

It doesn't have to be perfectly flat. Only flat enough that mounting pressure, compressibility and flex can take care of whatever gaps remains. Plate extra thickness on the pads and it should be fairly doable.

 

[gggplaya]

Alder Lake launched in November 2021 which was 20 months ago and Raptor lake launched 8 months ago. So 2 gens on a socket? Assuming Arrow lake comes out in q4 2024 or Meteor lake comes out sooner with LGA1851 or as Mobile only.

I'll stick with AMD, I like what they did with AM4. It wasn't perfect, it was hard to foresee future feature updates added to new chipsets. But it was much better than this, where Intel changes the socket completely after only 2 gens. At least AMD attempted to make older x370 motherboard future compatible with fragmented bios updates.

 

[Amdlova]

Will come with the alluminium thing to not bending the socked...
Will need the bender of Futurama

 

[InvalidError]

On the other hand you could go AM5 and be able to upgrade the CPU in a few years, to the latest thing, without needing a new motherboard.

That is only relevant to maybe 2-3% of PC owners. Between laptops, SBCs and similar highly integrated solutions people use as x86 PCs, 80+% of PC owners have CPUs soldered to the motherboard.

By the time I want to upgrade my i5-11400, I'll likely want new-everything just like every other time I have upgraded.

Alder Lake launched in November 2021 which was 20 months ago and Raptor lake launched 8 months ago. So 2 gens on a socket? Assuming Arrow lake comes out in q4 2024 or Meteor lake comes out sooner with LGA1851 or as Mobile only.

Intel has been doing two CPU generations per socket for the last ~25 years apart for one or two cases where there was only one, nothing new there.

 

[cyrusfox]

The only questions I have that are still unclear,
•Mounting, will it be cross compatible with LGA1700?
•PCIE cpu lanes and DMI, Any different from 700 series?
•DLVR?

 

[TerryLaze]

What? Why?

Because of this, which is right before it.

Intel's Alder Lake processors suffered from warping and bending in the socket, which has made LGA1700 frames very popular. While the information talks about the Board Transient Bend Strain on the LGA1851 socket, it doesn't provide any guidelines. It just suggests that vendors contact a local quality engineer.

Unless there is a physicist out there that knows better, for me maximum dynamic pressure means the maximum that the mobo can take without bending.
So they made the socket be able to take more abuse and bend less.
I can only find explanations about aircraft...but it seems to be about the maximum difference of the dynamic pressing down force and the static mobo laying there.

The max q, or maximum dynamic pressure, condition is the point when an aerospace vehicle's atmospheric flight reaches the maximum difference between the fluid dynamics total pressure and the ambient static pressure.

 

[who_farted]

Pads dont have tolerances and must be on a perfectly level surface to make contact.

Your question is like saying: "why cant we have heat sinks that dont require TIM?" You know why we have TIM, because the surfaces are not perfectly level and we need a (Thermal Interface) Material between them to fill in the gaps. The "pins" on a processor have tolerances and fill in those gaps between the socket and processor, some go a little deeper, some a little shallow, but they still make contact, pads cant do this.

No, that's where mounting pressure comes in to this equation, if they can make it precise enough for 1700 microscopic pins to line up and touch I have faith they could get it flat enough to make them touch.

 

[InvalidError]

No, that's where mounting pressure comes in to this equation, if they can make it precise enough for 1700 microscopic pins to line up and touch I have faith they could get it flat enough to make them touch.

They may be small but they still have about 0.4mm of compliance to accommodate some pretty significant non-uniformity across the socket. If you wanted to use bumps instead of springy contact points without requiring absurd mounting pressure, you'd need to aim for flatness around 10 microns.

 

[2Be_or_Not2Be]

I thought it was BGA instead of PGA? The graphic says the "ball" count is 1851, so no pins there.

Also, the mounting pressure sounds like they are going to require mobos to be much stronger at the CPU socket, if not the whole board. The higher allowed mounting pressure hopefully indicates a stronger socket/board, which would reduce or even eliminate problems with flex/bending.

 

[InvalidError]

Also, the mounting pressure sounds like they are going to require mobos to be much stronger at the CPU socket, if not the whole board.

Just have a full-contact steel backplate bear all of the CPU+HSF mounting load so the board itself only has to bear the compression and weight across the entire socket area. They already have a retention mechanism backplate, they just need to make it a little thicker and have full flat coverage with a thin insulation sheet to prevent shorting anything that may be on the PCB's surface layer that isn't ground.

Most brackets today only link the four holes together and aftermarket hardware cannot touch the board beyond screw holes due to back-side SMD clearance concerns.

 

[bit_user]

Is there a single person here who's contemplating an upgrade to Raptor Refresh?

I'll admit I was thinking about it. When I heard the rumor of it, I thought maybe I'd buy an i5-12600 (because it's one of the cheaper models with ECC support) and upgrade to Raptor Refresh i9 sometime probably next year. However, that initial purchase got delayed for various reasons, and we're now close enough that I might just wait to buy my W680 board until Raptor Refresh launches. The benefit would be ensuring the board will support the new CPU models, which seems a pretty safe assumption, but also maybe they'll respin it to support faster memory (currently, they've qualified only DDR5-4800 for it).

By the time I want to upgrade my i5-11400, I'll likely want new-everything just like every other time I have upgraded.

I'm normally in the same boat, which makes this situation an exception for me. I usually have the luxury of being able to afford one of the faster CPUs, which means there's typically no need to upgrade until the socket, memory standard, and just about everything else has advanced a couple generations.

 

[bit_user]

•PCIE cpu lanes and DMI, Any different from 700 series?

From my reading, it seems they're going to have the following PCIe lanes directly from the CPU:

• PCIe 5.0 x16
• PCIe 5.0 x4
• PCIe 4.0 x4

But, we don't yet know about DMI. If they keep DMI @ x8, then their total lane count will slightly exceed that of AM5. However, unless they upgrade their DMI to PCIe 5.0 x8, their aggregate PCIe bandwidth will match AM5's.

 

[InvalidError]

I'm normally in the same boat, which makes this situation an exception for me. I usually have the luxury of being able to afford one of the faster CPUs, which means there's typically no need to upgrade until the socket, memory standard, and just about everything else has advanced a couple generations.

Being able to afford something and actually needing it are two different things. In my case, I upgraded mainly for preemptive maintenance reasons and refreshing IOs, not really for performance. As I expected, my i5-11400 didn't feel particularly faster in most of my day-to-day stuff than my i5-3470. Anything fancier would have been a complete waste of money for me. (And anything less would also be a waste of money due to the much higher likelihood I may end up wanting to upgrade prematurely.)

 

[bit_user]

Being able to afford something and actually needing it are two different things. In my case, I upgraded mainly for preemptive maintenance reasons and refreshing IOs, not really for performance.

I buy towards the upper end of the range and don't replace until something comes along that will be noticeably faster. I have a quad-core Sandybridge today, making a 6-core i5-12600 more than 2x as fast. Heck, between IPC improvements and clock speed, just the single-thread performance should be about 2x.

When I use an Alder Lake PC for work, it does feel noticeably faster than my home PC.

 

[dtemple]

The wording and interpretation in this article is a bit suspicious. I don't think new coolers are "required" to have nearly double the pressure:

For any given size of CPU retention mechanism from Intel, there has only ever been one official cooler mount. LGA775, LGA115x/1200, and LGA1700 each have their own cooler mount which has very specific spacing and Z-height tolerance. LGA17xx/18xx use the same CPU retention mechanism and identical Z-height; as such it's almost a guarantee that existing coolers for LGA1700 will be physically compatible with the mounting for LGA18xx. The static pressure *maximum* has increased substantially; that does not mean coolers *require* to have more pressure, it means they're *allowed* to have more pressure. The pressure in question is not the pressure of the CPU onto the pins in the socket, it's the pressure of the cooler mount onto the IHS.

Unsurprising it's DDR5-only considering the pricing has plummeted. I sure hope AMD gets their memory controller working at higher speeds stably before then.

 

[gapdragon]

Whizzo! Another new socket, another new motherboard etc...

On the other hand you could go AM5 and be able to upgrade the CPU in a few years, to the latest thing, without needing a new motherboard.

Intel has guaranteed 3 generation support for this socket.

 

[TJ Hooker]

Intel has guaranteed 3 generation support for this socket.

Source?

 

[InvalidError]

Intel has guaranteed 3 generation support for this socket.

Three years isn't necessarily three generations. The next platform is supposed to launch "early" in 2024. If that sticks around until the next-gen launch in 2025 followed by the next launch after that at some point in 2026, the platform will have technically spanned three production calendar years even if the 2026 launch comes with a whole new socket.