Saturday, May 22nd 2021

AMD Socket AM5 an LGA of 1,718 Pins with DDR5 and PCIe Gen 4

A reliable source with AMD and NVIDIA leaks, ExecutableFix has shared some interesting bits of early information on AMD's next-generation Socket AM5. Apparently this will be AMD's first mainstream-desktop socket that does away with pins on the processor package, shifting them to the motherboard, in a Land Grid Array (LGA) format. This won't be AMD's first client LGA, though, as it was the Quad FX platform from 2006, which used a pair of Socket F LGAs. Socket AM5 will have a pin-count of 1,718 pins, 18 more than Intel's upcoming Socket LGA1700, on which its 12th Gen Core "Alder Lake-S" is expected to be based.

AMD will give the I/O of its client desktop platform a major update, with the introduction of DDR5 memory. Socket AM5 processors are expected to feature a dual-channel DDR5 memory interface. With Intel "Alder Lake-S" implementing DDR5, too, you now know why every major memory manufacturer is unveiling their first DDR5 U-DIMM product development. Interestingly, the PCI-Express interface on Socket AM5 will remain PCI-Express 4.0, even though PCI-Express 5.0 is being rumored for "Alder Lake-S." The switch to PCI-Express 5.0 may not be significant from a graphics cards perspective immediately, but paves the way for next-gen M.2 NVMe SSDs with double the transfer-rates of current drives that use PCI-Express 4.0. AMD is developing the new 600-series chipset to do with its next-generation Socket AM5 processors.
Source: ExecutableFix (Twitter)
Add your own comment

138 Comments on AMD Socket AM5 an LGA of 1,718 Pins with DDR5 and PCIe Gen 4

#26
Wirko
When AMD eventually decides to start supporting PCIe 5, what will they do? Change sockets again and introduce AM5.5? No way. I'm very sure they've designed AM5 for PCIe 5.

This latest leak just indicates that the first generation of AM5 CPUs, chipsets and boards will support version 4 only.
Posted on Reply
#27
jesdals
The switch to PCI-Express 5.0 may not be significant from a graphics cards perspective immediately, but paves the way for next-gen M.2 NVMe SSDs with double the transfer-rates of current drives that use PCI-Express 4.0.
Haveing gen 4 drives my self - I do find it doubtfull that 14.000MB/sek transfer rate desktop parts is high demand any time soon - 4tb drives to a affordable price though
Posted on Reply
#28
Jill Valentine
Xex360Hopefully they'll make the CPUs more secure, it's annoying to have the CPU stuck to the cooler.
Protip: Twist the heatsink/block before removing it. I learned that when I removed the stock cooler first time on my A64 15 years ago.
Posted on Reply
#29
silentbogo
Chloe PriceProtip: Twist the heatsink/block before removing it. I learned that when I removed the stock cooler first time on my A64 15 years ago.
And after 15 years we still have no secure bracket for PGA CPUs.
If only there was a solution.... :wtf:
Posted on Reply
#30
_Flare
I never bend any socket or CPU pin. How can anybody be so untalented?
Maybe there are people out there poking nose and eyes eveytime washing their face,
or cutting themselves everytime they use a knife.
Maybe some people do not have the ability to get in the state of being calm and just don´t break anything.
Posted on Reply
#31
qubit
Overclocked quantum bit
Shame it's not 1717 pins. Totally nerd out to that lol.
Posted on Reply
#32
Chaitanya
Chloe PriceProtip: Twist the heatsink/block before removing it. I learned that when I removed the stock cooler first time on my A64 15 years ago.
And that will damage the socket on board after couple of times to the point that wont be secure and detected. I learned that with Asus A8N-Sli Deluxe being sent to RMA for using twisting method to remove CPU cooler. Better to use some heat cycle and IPA to loosen paste.
Posted on Reply
#33
Jill Valentine
silentbogoAnd after 15 years we still have no secure bracket for PGA CPUs.
If only there was a solution.... :wtf:
Ah yeah, I saw that some time ago. Kinda weird how something like that didn't exist in P4/A64 days already, or at least I've never seen a CPU holder like that before.
ChaitanyaAnd that will damage the socket on board after couple of times to the point that wont be secure and detected. I learned that with Asus A8N-Sli Deluxe being sent to RMA for using twisting method to remove CPU cooler. Better to use some heat cycle and IPA to loosen paste.
Weird as I've always heard that twisting is okay as long you use pure sense with it, not twist it with a gorilla's rage.
Posted on Reply
#34
bencrutz
WirkoWhen AMD eventually decides to start supporting PCIe 5, what will they do? Change sockets again and introduce AM5.5? No way. I'm very sure they've designed AM5 for PCIe 5.

This latest leak just indicates that the first generation of AM5 CPUs, chipsets and boards will support version 4 only.
IIRC, rumored genoa will support pcie 5, that's gonna be zen 4 epyc
Posted on Reply
#35
evernessince
_FlareI never bend any socket or CPU pin. How can anybody be so untalented?
Maybe there are people out there poking nose and eyes eveytime washing their face,
or cutting themselves everytime they use a knife.
Maybe some people do not have the ability to get in the state of being calm and just don´t break anything.
You say that but damaged pins have been the number 1 issue for Intel boards for the longest time.

You can blame the user all you want but that isn't going to improve the reality of the situation. The fact of the matter is as a company if users are experiencing an issue you should seek to fix it.

Mind you there are some people who simply don't have steady hands, have a medical issues (like diabetes, lazy eye, or any form of eye giggle), or are just old. It's rather insensitive to simply discount those people as untalented.
Posted on Reply
#36
Minus Infinity
WirkoWhen AMD eventually decides to start supporting PCIe 5, what will they do? Change sockets again and introduce AM5.5? No way. I'm very sure they've designed AM5 for PCIe 5.

This latest leak just indicates that the first generation of AM5 CPUs, chipsets and boards will support version 4 only.
You mean how socket AM4 managed to go unchanged when moving from PCI-E 3.0 to 4.0. Gees already worrying about non-issues when it is Intel that drops a new socket every 2 years.
Posted on Reply
#37
repman244
evernessinceYou say that but damaged pins have been the number 1 issue for Intel boards for the longest time.

You can blame the user all you want but that isn't going to improve the reality of the situation. The fact of the matter is as a company if users are experiencing an issue you should seek to fix it.

Mind you there are some people who simply don't have steady hands, have a medical issues (like diabetes, lazy eye, or any form of eye giggle), or are just old. It's rather insensitive to simply discount those people as untalented.
If you are somehow unable to not damage pins on CPU/MOBO leave it to a shop or someone qualified to do it for you then (that goes for untalented people as well).

I really don't get all the moaning about bent pads/pins on LGA or PGA...I've removed/installed CPUs countless times without damaging any of it.
Posted on Reply
#38
Chaitanya
Chloe PriceAh yeah, I saw that some time ago. Kinda weird how something like that didn't exist in P4/A64 days already, or at least I've never seen a CPU holder like that before.


Weird as I've always heard that twisting is okay as long you use pure sense with it, not twist it with a gorilla's rage.
You dont have to use gorilla rage but do it enough times(on old A8N-SLI Deluxe it was about 7-8 times) to loosen the socket. Since then I use IPA after a heat cycle to remove heatsinks.
Posted on Reply
#39
lexluthermiester
erockerI find damaging socket pins easier. But 1718 pins on a PGA I'd probably end up messing up too.
I find it sixes either way.
Posted on Reply
#40
umano
Pci 5 at the moment it seems pointless in the consumer market, even prosumer. There is no gpu able to saturate even the pci 3.0 16x and current nvmes offer plenty of sequential speed, random iops increase is what we need.
Even 8k uncompressed raw video cannot saturate pci 4 nvme speeds. Considering the future cost of ddr5, storage and gpu I'd rather save some money on the motherboard.
Posted on Reply
#41
Nanochip
umanoPci 5 at the moment it seems pointless in the consumer market, even prosumer. There is no gpu able to saturate even the pci 3.0 16x and current nvmes offer plenty of sequential speed, random iops increase is what we need.
Even 8k uncompressed raw video cannot saturate pci 4 nvme speeds. Considering the future cost of ddr5, storage and gpu I'd rather save some money on the motherboard.
People said the same thing about pcie4 when intel didn’t support it until 11th gen. It’s good that Intel is bringing support and hopefully amd does so sooner rather than later.
Posted on Reply
#42
trparky
ChaitanyaSince then I use IPA after a heat cycle to remove heatsinks.
What's "IPA" stand for?
Posted on Reply
#43
Chaitanya
trparkyWhat's "IPA" stand for?
Isopropyl Alcohol.
Posted on Reply
#44
Darmok N Jalad
voltageThey cant even come up with their own product naming scheme. :roll:
In 2001, AMD launched the Athlon XP Palomino. To counter Pentium 4, a CPU with high clock speeds but lower IPC, the XP was given model numbers based on performance that lined up with the P4's frequency. At the time, clockspeed was everything, and there was a lot of controversy around the differences in IPC. I even remember a few bad analogies being made from people trying to explain that there was more to CPU performance than clockspeeds. Something about a kid having to move his legs faster to keep up with an adult. Naming conventions pretty much stuck around with AMD ever since the Athlon XP.

Intel didn't introduce consumer CPU model numbers until around 2006 with the Pentium 4. Netburst had pretty much failed by that time, so in order to prepare us for the lower-clocked but far superior Core architecture, Pentium 4 needed to adopt lesser model numbers than Core, which was clocked lower but performed way better. I guess in a weird sense, our model numbers today are still based on Netburst.
Posted on Reply
#45
zlobby
newtekie1LGA pins are LGA pins. At least they are less likely to get damaged than pins on the CPU. I'm so glad AMD is finally moving to LGA for their mainstream desktop processors.
I for one never had an issue with pins being on the CPU.
ChaitanyaIsopropyl Alcohol.
Or Indian Pale Ale?
Posted on Reply
#46
Jism
DeathtoGnomesIts a shame they are gonna use PCI4, if intel does go PCI5, AMD will be under the gun to push it too.
For consumers we dont need PCI-E 5.0 really. A fully loaded Epyc in DC provides more then enough lanes, bandwidth for whatever you want to throw at it. Difficult to not max it out and still claim you dont have enough bandwidth.

PCI-E 4.0 is'nt even fully taxed; let alone PCI-E 3.0.
Darmok N JaladIn 2001, AMD launched the Athlon XP Palomino. To counter Pentium 4, a CPU with high clock speeds but lower IPC, the XP was given model numbers based on performance that lined up with the P4's frequency. At the time, clockspeed was everything, and there was a lot of controversy around the differences in IPC. I even remember a few bad analogies being made from people trying to explain that there was more to CPU performance than clockspeeds. Something about a kid having to move his legs faster to keep up with an adult. Naming conventions pretty much stuck around with AMD ever since the Athlon XP.

Intel didn't introduce consumer CPU model numbers until around 2006 with the Pentium 4. Netburst had pretty much failed by that time, so in order to prepare us for the lower-clocked but far superior Core architecture, Pentium 4 needed to adopt lesser model numbers than Core, which was clocked lower but performed way better. I guess in a weird sense, our model numbers today are still based on Netburst.
The PR rating was based on AMD's own CPU's, not Intel's counterpart. If you'd buy a 1800+ or so it was guaranteed to be equal like a 1800Mhz chip.

Netburst was horrible. The A64 beat it to it while being clocked lower.
Posted on Reply
#47
zlobby
umanoPci 5 at the moment it seems pointless in the consumer market, even prosumer. There is no gpu able to saturate even the pci 3.0 16x and current nvmes offer plenty of sequential speed, random iops increase is what we need.
Even 8k uncompressed raw video cannot saturate pci 4 nvme speeds. Considering the future cost of ddr5, storage and gpu I'd rather save some money on the motherboard.
Yeah, so let's stay with any tech for 10 years just because there is no demand for better one.

Oh, and it just so happens that there is demand! For example nvme RAID cards. 4 SSD could easily saturate all the bandwidth.
Or you could split one PCIe into few slower ones.
Posted on Reply
#48
thegnome
cyberlonermsrp IS FAKE now........ all price will UP
5600x under msrp, 5800x under msrp, etc, etc.
Posted on Reply
#49
z1n0x
It seems to me, one of the main reasons to not opt for PCIe 5.0 is platform cost.
This might aswell end up been a good move by AMD, despite the complaints by some folks.
Posted on Reply
#50
TheLostSwede
NanochipOne cooler mount for AM5 or LGA1700?
We can but dream...
srsbsnsPCIE 5.0 is for DirectStorage
Seemingly not this time around. Could be fake though.
And if I'm reading that correctly, Intel will have support for three generations of PCIe on a single platform.

silentbogoTell that to foxconn. Also, it's helluvalot more pins, so if LGA1150 is easy to mess up, then this gold-plated bristle on a whooping 1718 pins is going to be even thinner and even less durable.
But the physical socket is also bigger...
_FlareI never bend any socket or CPU pin. How can anybody be so untalented?
Maybe there are people out there poking nose and eyes eveytime washing their face,
or cutting themselves everytime they use a knife.
Maybe some people do not have the ability to get in the state of being calm and just don´t break anything.
I guess you've never dropped anything in your entire life either?
umanoPci 5 at the moment it seems pointless in the consumer market, even prosumer. There is no gpu able to saturate even the pci 3.0 16x and current nvmes offer plenty of sequential speed, random iops increase is what we need.
Even 8k uncompressed raw video cannot saturate pci 4 nvme speeds. Considering the future cost of ddr5, storage and gpu I'd rather save some money on the motherboard.
And PCIe 5.0 will have even more reduced PCB trace length, which will require ever more expensive redrivers to maintain signal integrity...
Posted on Reply
Add your own comment