Overclock.net banner

1 - 20 of 42 Posts

·
Premium Member
Joined
·
6,181 Posts
Discussion Starter #1
Source (TechPowerUp)
Source (Samsung)

Quote:
Samsung Electronics Co., Ltd., announced today that it has begun mass producing the industry's first 4-gigabyte (GB) DRAM package based on the second-generation High Bandwidth Memory (HBM2) interface, for use in high performance computing (HPC), advanced graphics and network systems, as well as enterprise servers. Samsung's new HBM solution will offer unprecedented DRAM performance - more than seven times faster than the current DRAM performance limit, allowing faster responsiveness for high-end computing tasks including parallel computing, graphics rendering and machine learning.

"By mass producing next-generation HBM2 DRAM, we can contribute much more to the rapid adoption of next-generation HPC systems by global IT companies," said Sewon Chun, senior vice president, Memory Marketing, Samsung Electronics. "Also, in using our 3D memory technology here, we can more proactively cope with the multifaceted needs of global IT, while at the same time strengthening the foundation for future growth of the DRAM market."

The newly introduced 4GB HBM2 DRAM, which uses Samsung's most efficient 20-nanometer process technology and advanced HBM chip design, satisfies the need for high performance, energy efficiency, reliability and small dimensions making it well suited for next-generation HPC systems and graphics cards.

Following Samsung's introduction of a 128GB 3D TSV DDR4 registered dual inline memory module (RDIMM) last October, the new HBM2 DRAM marks the latest milestone in TSV (Through Silicon Via) DRAM technology.

The 4GB HBM2 package is created by stacking a buffer die at the bottom and four 8-gigabit (Gb) core dies on top. These are then vertically interconnected by TSV holes and microbumps. A single 8Gb HBM2 die contains over 5,000 TSV holes, which is more than 36 times that of a 8Gb TSV DDR4 die, offering a dramatic improvement in data transmission performance compared to typical wire-bonding based packages.
 

·
Registered
Joined
·
536 Posts
Will nVidia get any of this?
 

·
Registered
Joined
·
536 Posts
Isn't Nvidia suing Samsung for some patents. Will that be a problem?
 

·
Top kek
Joined
·
3,595 Posts
Quote:
Originally Posted by Myst-san View Post

Isn't Nvidia suing Samsung for some patents. Will that be a problem?
Actually, nVidia got denied for the patent rights, so now Samsung is suing back.. But money are money, Samsung will still sell it to them.
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by iBruce View Post

EVGA GTX1080 Kingpin please.
thumb.gif


OR is Pascal skipping 10 and moving to 11 nomenclature?

IF EVGA only uses Samsung chips on their Kingpin cards, from whom do they source memory for all the other cards? Micron?

http://s296.photobucket.com/user/iBruceEVGA/media/IMG_4730_zps52b3zsum.jpg.html
Hynix. Elpida is another one but rare on newer Nvidia cards IIRC. It's a lot more common on AMD cards.

We'll see how Pascal is compared to Maxwell in terms of voltage scaling. I'm hoping it's the same as Maxwell so I don't have to pay anything extra for a custom card and wait even longer for a waterblock.
 

·
Banned
Joined
·
2,915 Posts
Quote:
Originally Posted by xxdarkreap3rxx View Post

Hynix. Elpida is another one but rare on newer Nvidia cards IIRC. It's a lot more common on AMD cards.

We'll see how Pascal is compared to Maxwell in terms of voltage scaling. I'm hoping it's the same as Maxwell so I don't have to pay anything extra for a custom card and wait even longer for a waterblock.
I waited 5months for the 980 Kingpin to release, then EK and EVGA took their time getting the blocks released, the Titan X launched and still no Kingpin blocks, so I just waited 5 more months for the $799 price to come down, picked up my 980 Kingpin for $499 during the EVGA 16th anniversary event.

Was hoping to do the same this year during the 17th anniversary, save some money.
thumb.gif


After (4) Asus cards, love the EVGA so much, now I'll never go back.

Funny I feel the exact opposite about motherboards, only Asus mobos.
thumb.gif
 

·
Registered
Joined
·
2,181 Posts
Quote:
Originally Posted by DNMock View Post

AMD's 14nm GPU's are being made by Samsung, I think TSMC is doing the 16nm Nvidia chips. Since HBM2 is on the chip itself, I highly doubt any of the samsung produced hbm2 will be going to Nvidia.
HBM1 on Fury/X cards is made by SK Hynix.

Now connect the dots.
 

·
Registered
Joined
·
3,472 Posts
Quote:
Originally Posted by xxdarkreap3rxx View Post

Hynix. Elpida is another one but rare on newer Nvidia cards IIRC. It's a lot more common on AMD cards.

We'll see how Pascal is compared to Maxwell in terms of voltage scaling. I'm hoping it's the same as Maxwell so I don't have to pay anything extra for a custom card and wait even longer for a waterblock.
off topic, actually can we check what RAM chip brand our GPU is using without taking out the cooler? or even check via software?
tongue.gif
 

·
Registered
Joined
·
2,517 Posts
Quote:
Originally Posted by DNMock View Post

AMD's 14nm GPU's are being made by Samsung, I think TSMC is doing the 16nm Nvidia chips. Since HBM2 is on the chip itself, I highly doubt any of the samsung produced hbm2 will be going to Nvidia.
GF/Samsung are making the low powered chips for AMD. Rumor is TSMC is still making the big boys chips for NVidia and AMD. HBM2 will be on the big chips from TSMC for Nvidia and AMD from Samsung or Hynix, whoever makes the right deal.
 

·
Premium Member
Joined
·
6,735 Posts
Quote:
Originally Posted by DNMock View Post

AMD's 14nm GPU's are being made by Samsung, I think TSMC is doing the 16nm Nvidia chips. Since HBM2 is on the chip itself, I highly doubt any of the samsung produced hbm2 will be going to Nvidia.
HBM is on package, not in/on the chip. It just gets soldered to an interposer. You can have DRAM from one company and a GPU die from another.
 

·
Registered
Joined
·
945 Posts
Quote:
Originally Posted by DNMock View Post

AMD's 14nm GPU's are being made by Samsung, I think TSMC is doing the 16nm Nvidia chips. Since HBM2 is on the chip itself, I highly doubt any of the samsung produced hbm2 will be going to Nvidia.
Not sure if serious.
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by iBruce View Post

I waited 5months for the 980 Kingpin to release, then EK and EVGA took their time getting the blocks released, the Titan X launched and still no Kingpin blocks, so I just waited 5 more months for the $799 price to come down, picked up my 980 Kingpin for $499 during the EVGA 16th anniversary event.

Was hoping to do the same this year during the 17th anniversary, save some money.
thumb.gif


After (4) Asus cards, love the EVGA so much, now I'll never go back.

Funny I feel the exact opposite about motherboards, only Asus mobos.
thumb.gif
I got duped with the 980 Kingpin thinking it was like the 780 Ti one. Little did I know that Maxwell is different in terms of voltage scaling so having a really beefy PCB for gaming under air or water is next to useless (won't OC any more, limited by the silicon). Luckily I only bought 2 and both were used but they were a pain to sell when the 980 Ti came out because nobody wanted to pay more than the price of reference 980 cards so I had to sell them for, IIRC, $475 each + S&H (with EK block). That's why I'm hoping Pascal is the same, so I can buy another reference card and get a block for it early on instead of having to wait for a card and extra time for a block.
Quote:
Originally Posted by Clocknut View Post

off topic, actually can we check what RAM chip brand our GPU is using without taking out the cooler? or even check via software?
tongue.gif
GPU-Z, NVIDIA Inspector
 

·
Registered
Joined
·
710 Posts
Quote:
Originally Posted by EniGma1987 View Post

HBM is on package, not in/on the chip. It just gets soldered to an interposer. You can have DRAM from one company and a GPU die from another.
I could be totally wrong, but i thought interposers are a special design for being able to test your chips.
 
1 - 20 of 42 Posts
Top