Samsung Bets on GDDR6 for 2018 Rollout

[btarunr]

Even as its fellow-Korean DRAM maker SK Hynix is pushing for HBM3 to bring 2 TB/s memory bandwidths to graphics cards, Samsung is betting on relatively inexpensive standards that succeed existing ones. The company hopes to have GDDR6, the memory standard that succeeds GDDR5X, to arrive by 2018.

GDDR6 will serve up bandwidths of up to 16 Gbps, up from the 10 Gbps currently offered by GDDR5X. This should enable memory bandwidths of 512 GB/s over a 256-bit wide memory interface, and 768 GB/s over 384-bit. The biggest innovation with GDDR6 that sets it apart from GDDR5X is LP4X, a method with which the memory controller can more responsively keep voltages proportionate to clocks, and reduce power-draw by up to 20% over the previous standard.



View at TechPowerUp Main Site

 

[ZeppMan217]

Why doesn't Nvidia use HBM?

 

[btarunr]

Why doesn't Nvidia use HBM?

Because it can make GPUs that are faster than AMD's without needing HBM.

 

[mroofie]

Because it can make GPUs that are faster than AMD's without needing HBM.

 

[arbiter]

Why doesn't Nvidia use HBM?

HBM isn't and wasn't ready in time to be used. It was supposed to be full production end of last year but that was delayed til q3 and even yet i don't think its in full mass production yet. Before you ask about HBM1 well that is worthless cause 4gb limit made it that way. Throw in issue the interposer add's in another issue of complex issues and creating high chance of bad chips. GDDR5x is just a better option since yields are not as good on the new proc node gpu's are using atm.

 

[FYFI13]

Because it can make GPUs that are faster than AMD's without needing HBM.

This one made my day

 

[ZoneDymo]

Because it can make GPUs that are faster than AMD's without needing HBM.

Yep and sadly thats all the are after, not actually pushing the envelope, actually propelling humanity forward with cutting edge tech, just tiny mouse steps, just enough to beat the competition for easy maximum profit.

 

[Caring1]

Why doesn't Nvidia use HBM?

Because they are cheapskates and don't care about the consumer, only profits!

 

[RejZoR]

It seems with HBM3, there is very little point in framebuffer compression, it's only useful for crippled low end cards. Also, if GDDR6 becomes a reality, low and mid end models will not even get HBM memory for the next 10 years, because they'll slowly migrate through GDDR5X and GDDR6 with those and high ends will get HBM treatment...

 

[Assimilator]

Because it can make GPUs that are faster than AMD's without needing HBM.

We're gonna need a bigger burn ward for all the AMD users.

 

[Solidstate89]

Why doesn't Nvidia use HBM?

They do; in their Tesla P100 card because it still isn't available in high enough volume for a consumer card.

Because they are cheapskates and don't care about the consumer, only profits!

What a naive view to have about a for-profit company. Of course they care about profits. So does AMD. Newsflash: companies are all about profits. Any belief that they "care" about the consumer is just horseshit. They care enough about the consumer to try and keep them as repeat customers. These companies aren't charities - of course they care about profit. It's their whole reason for being.

 

[Chaitanya]

They do; in their Tesla P100 card because it still isn't available in high enough volume for a consumer card.


What a naive view to have about a for-profit company. Of course they care about profits. So does AMD. Newsflash: companies are all about profits. Any belief that they "care" about the consumer is just horseshit. They care enough about the consumer to try and keep them as repeat customers. These companies aren't charities - of course they care about profit. It's their whole reason for being.

Tesla is not a consumer card, maybe to a certain level Quadros are consumer cards but Tesla is in different category altogether.

 

[RejZoR]

Caring about customers and profit are mutual goals. One cannot exist without the other. Many people often buy technically inferior products only because companies have superior support or customer service.

 

[Nokiron]

Tesla is not a consumer card, maybe to a certain level Quadros are consumer cards but Tesla is in different category altogether.

That's precisely what he meant. HBM2 isn't available in high enough volumes.

 

[zo0lykas]

becouse open google, nvidia told they go for ddr5x

Why doesn't Nvidia use HBM?

 

[TheGuruStud]

Screw off, Sammy. HBM is clearly the future. I bet they're just mad they turned down AMD for HBM lolz (I'm assuming that AMD approached them b/c why wouldn't you).

 

[TheDeeGee]

Yep and sadly thats all the are after, not actually pushing the envelope, actually propelling humanity forward with cutting edge tech, just tiny mouse steps, just enough to beat the competition for easy maximum profit.

So Pascal is a tiny improvement compared to Maxwell?

Ooookey...

 

[TheGuruStud]

So Pascal is a tiny improvement compared to Maxwell?

Ooookey...

It is. The only real advantage is clock speed, but I doubt little of that is from chip design (was already in Maxwell).

Nvidia didn't do anything except get high clocks and sell you a smaller chip for the same price. Typical, but no massive improvement.

 

[Frick]

So Pascal is a tiny improvement compared to Maxwell?

Ooookey...

Pascal clocks massively, that's where the difference is. I think that wss his idea.

 

[Solidstate89]

Tesla is not a consumer card, maybe to a certain level Quadros are consumer cards but Tesla is in different category altogether.

I didn't say Tesla was a consumer card. You could at least correctly interpret what I wrote before saying something totally nonsensical that has nothing to do with what I said.

 

[ZoneDymo]

So Pascal is a tiny improvement compared to Maxwell?

Ooookey...

its no more then expected, and only a fraction a tiny tiny fraction of where Nvidia could put us right now, but nope, why sell improvement A B and C right away for 800 dollars when you can sell them along the way for 800 dollars each?
maximum profit, minimum progress, screw you real time raytracing etc!!

 

[RejZoR]

It is. The only real advantage is clock speed, but I doubt little of that is from chip design (was already in Maxwell).

Nvidia didn't do anything except get high clocks and sell you a smaller chip for the same price. Typical, but no massive improvement.

Pascal is also suppose to be using tile based rendering, something not found on Maxwell 1/2.

 

[64K]

It is. The only real advantage is clock speed, but I doubt little of that is from chip design (was already in Maxwell).

Nvidia didn't do anything except get high clocks and sell you a smaller chip for the same price. Typical, but no massive improvement.

36% to 42% faster in overall gaming than Maxwell according to reference card reviews here. More shaders. Faster Clocks. Brought to market way faster than AMD can deliver on something comparable to a 1070, 1080 or a Pascal Titan X.

 

[BiggieShady]

It is. The only real advantage is clock speed, but I doubt little of that is from chip design (was already in Maxwell).

Nvidia didn't do anything except get high clocks and sell you a smaller chip for the same price. Typical, but no massive improvement.

You are wrong, by rearranging the cuda cores into more modules (less cores per module), they improved amount of cache per core. Pascal also has twice beefier dispatchers. Clocks are major part of the performance boost, though.

 

[TheoneandonlyMrK]

It is. The only real advantage is clock speed, but I doubt little of that is from chip design (was already in Maxwell).

Nvidia didn't do anything except get high clocks and sell you a smaller chip for the same price. Typical, but no massive improvement.

No they also improved their lossy colour compression so as to further mitigate the bandwidth issues they are facing ,while Hbm production ramps to useable levels sometime in 2017-18.
Facts not fanboy bullshit. <Not aimed at you guru more bta and his comment shit.