Rambus Develops HBM2E Controller & PHY: 3.2 Gbps, 1024-Bit Bus

[erek]

I like rambus, just interesting

“The Rambus HBM2E controller core (originally developed by Northwest Logic) is DFI 3.1 compatible (with appropriate extensions) and supports AXI, OCP or proprietary interfaces to connect to integrator logic. Meanwhile, the controller also supports Look-Ahead command processing (a standard way to trim latencies) as well as channel densities of up to 24 Gb.

Licensees of Rambus’s HBM2E solution will get everything they need to integrate it into their designs, including source code of the controller(in a bid to synthesize it for a particular process technology) as well as fully-characterized hard macros (GDSII) of the interface. Alternatively, engineers from Rambus can help integrate the HBM2E IP support for a fee”

https://www.anandtech.com/show/15599/rambus-develops-hbm2e-controller-phy-32-gbps-1024bit-bus

 

[The Mad Atheist]

Anything actually use RB today beside a few stuff.
N64 used it, but haven't seen it elsewhere.
Didn't they try to be the new memory standard back in the day IIRC?

 

[Thunderdolt]

"Develops?" I was under the impression that the company was just bottom feeding patent trolls at this point.

 

[Ebernanut]

Anything actually use RB today beside a few stuff.
N64 used it, but haven't seen it elsewhere.
Didn't they try to be the new memory standard back in the day IIRC?

Intel used it in their systems for a short time(RDRAM) but it was too expensive so they phased it out for DDR SDRAM.

Basically RAMBUS was involved in creating the standard and talked the consortium into using several features that they then later claimed they held patents on and therefore owned the tech.

TLDR: Fuck RAMBUS

 

[Nenu]

Rambus fragmented the Intel platform with no positive outcome.
I took the common sense approach and avoided spending stupid money for little performance benefit but some friends ended up with mobos and ram they couldnt sell come upgrade time unless they accepted less than 1/2 what they wanted.
This made people give up on the PC as a gaming platform because it was seen as scammy.
My friends turned to consoles and they were high end PC gamers, granted with little knowledge, but plenty of cash.
It was sad to lose them.

 

[Jim Kim]

Rambus could develop the cure for cancer and i'd take a pass.

 

[IdiotInCharge]

Rambus fragmented the Intel platform with no positive outcome.

Intel did a lot of things here; they'd introduced Netburst in the Pentium IV with RDRAM, which made very little sense; they introduced Netburst over refining the P6 in the Pentium III, which also made very little sense; and instead of moving x86 on IA32 to 64bit, they decided to make IA64 VLIW instead, and develop an entirely new processor architecture and paradigm in the process.

We know now that Intel might have had no problem keeping ahead of AMD if they'd stuck with the P6, which was resurrected as the Core series, and that RDRAMs latency was really never going to be that useful on the desktop with branching code. And that we're still not ready for processors like the Itanium, with the way that programming and compiling is done.


I don't really blame RAMBUS for all that, they just provided one of the solutions from their patent-troll portfolio. Intel didn't have to use it, but they did, along with a raft of pie-in-the-sky ideas.

 

[GiGaBiTe]

Intel used it in their systems for a short time(RDRAM) but it was too expensive so they phased it out for DDR SDRAM.

Not before they tried to jimmy rig a fix. Consumers and the industry were revolting at the high cost of RDRAM, and Intel didn't have time to spin up a new chipset with SDR or DDR support, so they shat out the MTH (Memory Translation Hub). The MTH translated RDRAM to SDRAM and allowed existing chipsets to be used and minimal changes to motherboards to adapt SDRAM. In practice the MTH was a dumpster fire, it was extremely buggy and slower than if a chipset had native SDR support. It was so bad that Intel eventually issued a recall for it.

Basically RAMBUS was involved in creating the standard and talked the consortium into using several features that they then later claimed they held patents on and therefore owned the tech.

Rambus joined the JEDEC in the early 90s, but left the group in 1995 when they refused to follow JEDEC practices of licensing their technology in "reasonable and non-discriminatory ways". Five years later they started their patent lawsuit war against the memory industry claiming they owned SDR and DDR, which they eventually lost, but had other wins against various manufacturers.

The memory industry had the last laugh though, through their illegal price fixing tactics, they pushed Rambus out of the market; And with Rambus continuing to bite the hand that fed them, nobody was going to manufacture their products. Samsung was an outlier though, I guess they had a bit of stockholm syndrome because they kept making RDRAM long after everyone else, and made several of the successors of RDRAM, like XDR and XDR2 (the former being used in the PS3).

Now they seem to just be a patent troll, filing a new lawsuit every few years to keep extorting memory manufacturers and keep themselves going.

 

[noko]

Rambus is a partner with GloFlo who also makes HBM2e. Since AMD uses GloFlo for their IO chips and still some of their processors, the memory controller which is on the stack of HBM2e I would think would be a Rambus design. So we may hear more about this in the near future. That is if AMD uses GloFlo HBM2e.

https://www.rambus.com/interface-ip/ddrn-phys/hbm/

 

[EniGmA1987]

I thought Samsung made HBM2E which they dubbed "Flashbolt"? How can RAMBUS come in and say they made the controller and physical connections for it now and are going to license it?

edit: https://www.sammobile.com/news/sams...irst-third-generation-hbm2e-flashbolt-memory/
And according to the article, it sounds like a JEDEC standard?

One year after releasing the HBM2 ‘Aquabolt’ memory, the South Korean chip giant had initially announced the Flashbolt HBM memory last year during Nvidia’s 2019 GTC (GPU Technology Conference). This memory technology will likely be used in upcoming high-end GPUs from AMD and Nvidia. Late last month, JEDEC (Joint Electron Tube Engineering Council) had published an updated revision of the HBM2 standard.

https://semiengineering.com/hbm2e-the-e-stands-for-evolutionary/

Samsung introduced the first memory products in March that conform to JEDEC’s HBM2E specification ...

 

[Lakados]

I thought Samsung made HBM2E which they dubbed "Flashbolt"? How can RAMBUS come in and say they made the controller and physical connections for it now and are going to license it?

edit: https://www.sammobile.com/news/sams...irst-third-generation-hbm2e-flashbolt-memory/
And according to the article, it sounds like a JEDEC standard?


https://semiengineering.com/hbm2e-the-e-stands-for-evolutionary/

The wonderful things about "Standards" is everybody has their own.... So Rambus simply took the spec and built their own controllers to that spec, so anybody who wants to use HBM2e can either go through the process of developing their own controllers to use the memory or they can license Rambus' instead. Samsung was big enough that they built their own and they probably license that one as well. Each will perform slightly different as they put their own spin on the requirements and specifications and yay competition is born. I mean I don't trust Rambus as far as I can throw them but I do trust them to work to the benefit of their bottom line and their shareholders which at least keeps them predictable.

 

[erek]

https://www.guru3d.com/news-story/n...192-cores-and-boost-speed-up-to-2200-mhz.html