JEDEC Updates Groundbreaking High Bandwidth Memory (HBM) Standard

[csbin]

http://www.techpowerup.com/219175/j...aking-high-bandwidth-memory-hbm-standard.html

JEDEC Solid State Technology Association, the global leader in the development of standards for the microelectronics industry, today announced the publication of an update to JESD235 High Bandwidth Memory (HBM) DRAM standard. HBM DRAM is used in Graphics, High Performance Computing, Server, Networking and Client applications where peak bandwidth, bandwidth per watt, and capacity per area are valued metrics to a solution's success in the market. The standard was developed and updated with support from leading GPU and CPU developers to extend the system bandwidth growth curve beyond levels supported by traditional discrete packaged memory. JESD235A is available for free download from the JEDEC website.

JESD235A leverages Wide I/O and TSV technologies to support up to 8 GB per device at speeds up to 256 GB/s. This bandwidth is delivered across a 1024-bit wide device interface that is divided into 8 independent channels on each DRAM stack. The standard supports 2-high, 4-high and 8-high TSV stacks of DRAM at full bandwidth to allow systems flexibility on capacity requirements from 1 GB - 8 GB per stack.

Additional improvements in the recent update include a new pseudo channel architecture to improve effective bandwidth, and clarifications and enhancements to the test features. JESD235A also defines a new feature to alert controllers when DRAM temperatures have exceeded a level considered acceptable for reliable operation so that the controller can take appropriate steps to return the system to normal operation.

"GPUs and CPUs continue to drive demand for more memory bandwidth and capacity, amid increasing display resolutions and the growth in computing datasets. HBM provides a compelling solution to reduce the IO power and memory footprint for our most demanding applications," said Barry Wagner, JEDEC HBM Task Group Chairman.

 

[ShintaiDK]

In other words, HBM2 specification is first finished now.

And what's with the image? It serves no purpose.

 

[Stuka87]

The image shows who has their hands in the development of the technology. AMD, Intel, all the memory fabs, and a few other systems builders.

Its good to hear it has been finalized, and that this hopefully means we will see products with it this year.

 

[ShintaiDK]

The Board doesn't cover all , and its just for any JEDEC.

JEDEC members that contribute:
https://www.jedec.org/about-jedec/member-list

 

[Krteq]

Yes, but some of them were not involved in this standard developement.

 

[ShintaiDK]

Yes, but some of them were not involved in this standard developement.

Neither was all the board members. They just approved it.

 

[Azix]

In other words, HBM2 specification is first finished now.

And what's with the image? It serves no purpose.

it shows that nvidia is not involved and will likely fail miserably when they first try to use the tech :sneaky:

I thought it was specs till I read your comment and enlarged it. ah well. Hopefully chips are going to be in mass production. Would really benefit the industry. I'm thinking CPUs even using it instead of the System RAM when possible.

 

[ShintaiDK]

Its more likely you see HMC for CPUs, its much better fitted for that task. And it will be what Xeons choose. HMC also offers higher densities and higher speed (320GB/sec).

HBM is cheaper and offers what GPUs need.

 

[96Firebird]

JESD235A also defines a new feature to alert controllers when DRAM temperatures have exceeded a level considered acceptable for reliable operation so that the controller can take appropriate steps to return the system to normal operation.

How was this feature not included from the beginning? Seems like something that would have been thought of right off the bat...

 

[monstercameron]

How was this feature not included from the beginning? Seems like something that would have been thought of right off the bat...

Hence the locked down hbm on amd fury cards. Now we are connecting the dots.

 

[Stuka87]

How was this feature not included from the beginning? Seems like something that would have been thought of right off the bat...

Its not a feature that is in standard RAM, which is why if it gets hot you just start getting memory corruption. This feature is new to memory outside of the server space to my knowledge.

 

[guskline]

it shows that nvidia is not involved and will likely fail miserably when they first try to use the tech :sneaky:

I thought it was specs till I read your comment and enlarged it. ah well. Hopefully chips are going to be in mass production. Would really benefit the industry. I'm thinking CPUs even using it instead of the System RAM when possible.

Azix, if you look at the membership list for JEDEC, Nvidia IS a member and most likely will use HBM2.

 

[Stuka87]

Azix, if you look at the membership list for JEDEC, Nvidia IS a member and most likely will use HBM2.

nVidia is a member, but they were not involved in the creation of HBM. The two primary companies were AMD and Hynix.

 

[3DVagabond]

Azix, if you look at the membership list for JEDEC, Nvidia IS a member and most likely will use HBM2.

I'm pretty sure there was much sarcasm in his post.

 

[Azix]

Azix, if you look at the membership list for JEDEC, Nvidia IS a member and most likely will use HBM2.

Yeah they will use it of course. I was just joking that they would fail miserably. Though I kinda hope they do so we can have some balance in the force again.

They weren't involved in making it afaik. Tons of companies are a member of the JEDEC but they would all be in different subgroups working on different technologies.

 

[ShintaiDK]

Yeah they will use it of course. I was just joking that they would fail miserably. Though I kinda hope they do so we can have some balance in the force again.

They weren't involved in making it afaik. Tons of companies are a member of the JEDEC but they would all be in different subgroups working on different technologies.

So you need NVidia to fail? Wouldn't it be better for AMD not to fail? Then consumers wouldn't have to be punished in your case

 

[Azix]

So you need NVidia to fail? Wouldn't it be better for AMD not to fail? Then consumers wouldn't have to be punished in your case

Sure AMD needs to do well for the sake of PC gaming. But I was not thinking about AMD or the greater good. It would be a more amusing outcome. A lot of people aren't even considering anything but pascal. It would simply be hilarious watching them still buy pascal if it was hot, slower and crap on memory.

Very interested in seeing how this turns out. Not only was AMD involved in the creation of HBM, but they already have experience in the manufacture of GPUs with HBM. Whatever issues might have plagued the fury cards likely won't do the same with polaris. On the other hand, nvidia has no experience and was not involved in creating it, at the same level at least. Fun times ahead.

Beyond the fun I am more interested in what HBM will do for mobile and APUs. All that memory in a couple stacks of DRAM coupled with a tiny chip is crazy. Especially if it could be passively cooled or use a small fan

 

[guskline]

nVidia is a member, but they were not involved in the creation of HBM. The two primary companies were AMD and Hynix.

I was aware of that. However, the assumption that Nvidia does not have any expertise in HBM is, most likely, incorrect.

I doubt Hynix signed any exclusivity agreement in favor of AMD and against Nvidia, in light of Nvidia's market share vs AMD in the GPU arena.

BTW as an owner of both Nvidia and AMD higher end GPUs (see sig below) I want them BOTH to succeed.

Moreover, I agree with the sentiment above that since the new JEDEC standard has been finalized that we might see speed up of adoption of the HBM concept.

 

[Azix]

I was aware of that. However, the assumption that Nvidia does not have any expertise in HBM is, most likely, incorrect.

I doubt Hynix signed any exclusivity agreement in favor of AMD and against Nvidia, in light of Nvidia's market share vs AMD in the GPU arena.

BTW as an owner of both Nvidia and AMD higher end GPUs (see sig below) I want them BOTH to succeed.

Moreover, I agree with the sentiment above that since the new JEDEC standard has been finalized that we might see speed up of adoption of the HBM concept.

I dont think its a question of exclusivity. You used the word "expertise" yourself. But I expect nvidia was in the loop somehow and for a while. They had to be, otherwise they couldn't make pascal to use it even at the high end. If they weren't and are rushing from 2015 to now to sort out that aspect... unlikely but that wouldn't be good.