• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

3D XPoint vs HBM - The 3D Memory War

S

S0L0

n00b
Joined
Oct 11, 2012
Messages
44
In short, 3D XPoint is essentially a DIMM slot SSD that performs at DDR4 speed. So fast that there is no need for actual RAM in a system and it doesn't wear out like flash ram.

some features:
same read write speeds
non-volatile (can store data when powered off)
cheaper than DRAM, more expensive than flash

The first generation of what Intel is branding "Optane" will be available via SATA, SATA Express, PCIe due in 2016. While the second generation will likely be launched with Cannonlake in 2017 for the Xeon.


HBM is not located where GDDR5 would be, it is actually located on the package where you would have a GPU/CPU but separated by an interposer. HBM can also be stacked vertically to allow for a theoretical 4 times the memory per chip. HBM is stacked by using tiny holes in the silicon called TSVs.

some features;
one gigabyte of HBM is smaller than one gigabyte of GDDR5
HBM is more power efficient allowing more power-hungry GPU/CPUs
HBM has a large bandwidth

The first generation of HBM is already available however its based on 28nm manufacturing process so it is not until the 2nd generation in 2016 we will see significant performance gains when the process switches to 14nm. Furthermore, it is likely AMD will use HBM with their APU line instead of DDR4.
 
you are so off base it isn't even funny. XPOINT IS NOT A REPLACEMENT FOR RAM!!!!! (Seriously I swear people don't even read the press release and make stuff up. This was explicitly stated.)

Ideally, we would see HBM X(2,3...) as RAM replacing DDR 4 with XPoint within the RAM as the OS and major programs drive with regular NAND as fast mass storage going through PCH or PCI.

why i will get my grubby little hands on a DIMM and PCIe of XPoint the first chance i get.
 
Last edited:
SomeGuy133 said:
you are so off base it isn't even funny. XPOINT IS NOT A REPLACEMENT FOR RAM!!!!! (Seriously I swear people don't even read the press release and make stuff up. This was explicitly stated.)

Ideally, we would see HBM X(2,3...) as RAM replacing DDR 4 with XPoint within the RAM as the OS and major programs drive with regular NAND as fast mass storage going through PCH or PCI.

why i will get my grubby little hands on a DIMM and PCIe of XPoint the first chance i get.
Click to expand...

Intel is planning on implementing 3DXPoint differently from Micron who will likely have variants of it for PCIe 4.0 (due in 2017). However when PCIe 5.0 comes around it will likely have a completely different architecture from 4.0 (not going to be backwards compatible) and its likely we will see a convergence in memory and storage along with it. IMO by the time PCIe 5.0 comes around with Intel's 10nm tock computer architecture is going to be much different than it is in the current generation.

Also, 3DXPoint is also designed for mobile performance similar to HBM.


http://www.pcworld.com/article/29735...h-a-catch.html

http://go.redirectingat.com/?id=83512X1535351&site=anandtech.com&xs=1&isjs=1&url=http%3A%2F%2Fwww.kitguru.net%2Fcomponents%2Fgraphic-cards%2Fanton-shilov%2Fpci-express-4-0-with-16gts-data-rates-and-new-connector-to-be-finalized-by-2017%2F&xguid=&xuuid=197bcd3324ec9ddc06b02b0e55dbd1f2&xsessid=&xcreo=0&xed=0&sref=http%3A%2F%2Fforums.anandtech.com%2Fshowthread.php%3Fp%3D37688455%26posted%3D1%23post37688455&pref=http%3A%2F%2Fforums.anandtech.com%2Fnewreply.php%3Fdo%3Dnewreply%26p%3D37687930&xtz=-180
 
S0L0 said:
Intel is planning on implementing 3DXPoint differently from Micron who will likely have variants of it for PCIe 4.0 (due in 2017). However when PCIe 5.0 comes around it will likely have a completely different architecture from 4.0 (not going to be backwards compatible) and its likely we will see a convergence in memory and storage along with it. IMO by the time PCIe 5.0 comes around with Intel's 10nm tock computer architecture is going to be much different than it is in the current generation.

Also, 3DXPoint is also designed for mobile performance similar to HBM.


http://www.pcworld.com/article/29735...h-a-catch.html

http://go.redirectingat.com/?id=83512X1535351&site=anandtech.com&xs=1&isjs=1&url=http%3A%2F%2Fwww.kitguru.net%2Fcomponents%2Fgraphic-cards%2Fanton-shilov%2Fpci-express-4-0-with-16gts-data-rates-and-new-connector-to-be-finalized-by-2017%2F&xguid=&xuuid=197bcd3324ec9ddc06b02b0e55dbd1f2&xsessid=&xcreo=0&xed=0&sref=http%3A%2F%2Fforums.anandtech.com%2Fshowthread.php%3Fp%3D37688455%26posted%3D1%23post37688455&pref=http%3A%2F%2Fforums.anandtech.com%2Fnewreply.php%3Fdo%3Dnewreply%26p%3D37687930&xtz=-180
Click to expand...

la ten cy....xpoint will not be used as a system memory period plus it has order of magnitudes lower endurance. Everything you just said does not change that plus one source is broken and PCIe has nothing to do with this being a memory. HBM and XPoint are complimentary not competing. Your whole post is utterly pointless and wrong and is continuing this misinformation as if XPoint is some sort of system memory. IT IS NOT.
 
You must log in or register to reply here.
Back
Top