NVIDIA Pascal officially called PK100 and PK104 w/ HBM2

[MrRavens]

PK... why not GP I wonder...

 

GF = Geforce Fermi

GK = Geforce Kepler

GM = Geforce Maxwell

PK = Pascal (Kepler?)...

 

PK because PK Fire

 

The new GPUs are going to be straight fiya!

[SurvivorNVL]

Hnng 14nm Greenland CPU with 12GB HBM, water cooled, 2GHZ, with a 34" ultra wide curved 3440x1440 OLED with freesync with an interval of 9-240hz. -snip image-

I'd prefer a 29" ultra wide curved 3440x1440 OLED with freesync  The pixel density of 3440x1440 at 29" would be immaculate, and then 9-120hz panel would be perfect for freesync.  *_*  It'd be the best $3000 ever on a monitor.

[Opcode]

PK... why not GP I wonder...

 

GF = Geforce Fermi

GK = Geforce Kepler

GM = Geforce Maxwell

PK = Pascal (Kepler?)...

Code name Pyro King. Shoots five foot flame for effective ignition of the home.

[Notional]

I'd prefer a 29" ultra wide curved 3440x1440 OLED with freesync  The pixel density of 3440x1440 at 29" would be immaculate, and then 9-120hz panel would be perfect for freesync.  *_*  It'd be the best $3000 ever on a monitor.

 

On the 34", the pixel density is about the same as a 1440 27" 16:9 right? I think that is ok, but really, for gaming, I just want the larger screen, taking up more of my field of view.

 

つ ◕_◕ ༽つ LG, you work with AMD on Adaptive Sync, and you make the only OLED TV's. LG GIB ultra wide curved OLED monitors with AS つ ◕_◕ ༽つ  #LGGIBUWOLED

[patrickjp93]

And this surprises people because?

32GB of HBM 2 is pretty damn surprising, because even in Scientific Computing with NVLINK it makes 0 sense. It will take longer to load the card than to crunch the numbers. You'll never keep such a thing fed.

[SurvivorNVL]

On the 34", the pixel density is about the same as a 1440 27" 16:9 right? I think that is ok, but really, for gaming, I just want the larger screen, taking up more of my field of view.

 

つ ◕_◕ ༽つ LG, you work with AMD on Adaptive Sync, and you make the only OLED TV's. LG GIB ultra wide curved OLED monitors with AS つ ◕_◕ ༽つ  #LGGIBUWOLED

I can still see the pixels at 1440 27" 16:9 >_>  I need more density.  *_*  Density~  But--hopefully LG and Samsung working with AMD  for Adaptive-sync and all that--will actually yield quantum dot or oled/amoled displays with variable refresh technology and -preferably- below $1000.

[Opcode]

32GB of HBM 2 is pretty damn surprising, because even in Scientific Computing with NVLINK it makes 0 sense. It will take longer to load the card than to crunch the numbers. You'll never keep such a thing fed.

It's not to me as HBM has become a standard for AMD. It really gave Nvidia no choice but to follow suit or to be the other company stuck on old tech. The question in the back of my mind is will they be using the standard that AMD invented (Hynix manufactures) or will they invest into their own HBM technology.

[shert73]

You know, I wish AMD had pulled an Nvidia with hbm, and used that as leverage against Nvidia keeping all their tech proprietary and closed.

[Dabombinable]

*throws 970 out the window*

[Dabombinable]

It's not to me as HBM has become a standard for AMD. It really gave Nvidia no choice but to follow suit or to be the other company stuck on old tech. The question in the back of my mind is will they be using the standard that AMD invented (Hynix manufactures) or will they invest into their own HBM technology.

Nvidia does have the money to invest in their own HBM memory, the question is would it be worth the cost to them?

[patrickjp93]

It's not to me as HBM has become a standard for AMD. It really gave Nvidia no choice but to follow suit or to be the other company stuck on old tech. The question in the back of my mind is will they be using the standard that AMD invented (Hynix manufactures) or will they invest into their own HBM technology.

Hynix created it. AMD just jumped in to ensure their GPUs would be ready for it. And don't forget Nvidia is part of the Hybrid Memory Cube Consortium. Chances are it will be HBM 2.0 made by Hynix or Samsung, but Micron may deliver a surprise as well given the supplies it's delivering to Intel for Knight's Landing. And really, 32GB?! It's wholly unreasonable no matter how strong they could make their GPUs at this point in time.

[ivan134]

And AMD is still lagging behind... why can't they keep up with Intel and Nvidia's process shrinks?

 

 

I'm looking forward to HBM, the GPU cores and architectures of late are starting to get stagnant and it would be nice to see something actually driving the market forward instead of AMD failing to keep up and Nvidia sitting on their hands doing sweet fuck all.

 

 

I have a feeling when the 300 series launches from AMD, they will be winning. But as soon as Pascal comes out we will see AMD fade out and maybe even go brankrupt and disappear forever. Nvidia is so much better setup right now, their architecture is ahead and as soon as they get their hands on the latest process and memory tech their cards are going to make a big leap in performance that we havent seen in years, and will not see again for a while.

Lol, just lol.

[NeverNotExhausted]

Lol, just lol.

 

And what might your lol be referring to?

[Opcode]

You know, I wish AMD had pulled an Nvidia with hbm, and used that as leverage against Nvidia keeping all their tech proprietary and closed.

Unfortunately they couldn't as HBM is technically a JDEC open standard even though AMD invented a lot of the technology in the HBM that Hynix manufactures.

 

Nvidia does have the money to invest in their own HBM memory, the question is would it be worth the cost to them?

They do although will they invest into it to refrain from using AMD innovation? We all know how smug Nvidia is especially when it comes to their competitor.

 

Hynix created it. AMD just jumped in to ensure their GPUs would be ready for it. And don't forget Nvidia is part of the Hybrid Memory Cube Consortium. Chances are it will be HBM 2.0 made by Hynix or Samsung, but Micron may deliver a surprise as well given the supplies it's delivering to Intel for Knight's Landing. And really, 32GB?! It's wholly unreasonable no matter how strong they could make their GPUs at this point in time.

Hynix manufactured it, AMD invented a lot of industry standards for the HBM that Hynix manufactures (it's not a straight JDEC standard). You'll learn more about this later this year at HotChips. This is why I question Nvidia's use of it.

[patrickjp93]

Unfortunately they couldn't as HBM is technically a JDEC open standard even though AMD invented a lot of the technology in the HBM that Hynix manufactures.

 

 

They do although will they invest it to refrain from using AMD innovation? We all know how smug Nvidia is especially when it comes to their competitor.

 

Hynix manufactured it, AMD invented a lot of industry standards for the HBM that Hynix manufactures (it's not a straight JDEC standard). You'll learn more about this later this year at HotChips. This is why I question Nvidia's use of it.

No proof exists that AMD did any developmental work, and AMD's patents have nothing relating to the tech in HBM nor its manufacturing process. It's more than safe to assume AMD's partnership with Hynix is purely for AMD getting there first and having well-tuned hardware.

 

Oh please Nvidia uses the best there is regardless of who makes it as long as the tech is proven.

 

Again, having extensive knowledge of AMD's patents would reveal you're full of crap, and frankly Semiwiki disagrees with you as well.

[Opcode]

No proof exists that AMD did any developmental work, and AMD's patents have nothing relating to the tech in HBM nor its manufacturing process. It's more than safe to assume AMD's partnership with Hynix is purely for AMD getting there first and having well-tuned hardware.

 

Oh please Nvidia uses the best there is regardless of who makes it as long as the tech is proven.

 

Again, having extensive knowledge of AMD's patents would reveal you're full of crap, and frankly Semiwiki disagrees with you as well.

You can't patent an open standard. This is another reason why Nvidia (or any other companies) can easily get their hands on it. AMD is not the company known to hold back technologies either. If they make something that benefits everyone they usually open it up from the moment they make it.

[Guest Kloaked]

It's not to me as HBM has become a standard for AMD. It really gave Nvidia no choice but to follow suit or to be the other company stuck on old tech. The question in the back of my mind is will they be using the standard that AMD invented (Hynix manufactures) or will they invest into their own HBM technology.

 

So what video cards currently out from AMD has HBM? You have a weird definition of standard.

 

Nvidia was doing this anyways, mate. They showed their roadmap with stacked memory a looong time ago, so you attempting to imply AMD is pushing Nvidia to do this is flat out absurd.

[Poofu]

I can't wait for this but i really want AMD to give some strong competition 

[patrickjp93]

You can't patent an open standard. This is another reason why Nvidia (or any other companies) can easily get their hands on it.

Not true. Adaptive Sync and other technologies in the VESA standards are patented, but they are released to VESA. The same thing is true of JEDEC; and Hynix, Micron, Elpida, and others have plenty of patents related to SDRAM and DDR3 specifically. However, their technologies must be licensed under the JEDEC alliance charter.

 

Seriously how much are you willing to BS just to call AMD an innovator? I can find nothing proving beyond speculation that AMD was instrumental in the technical aspects of HBM development. If you can I'm willing to read through it.

[Opcode]

So what video cards currently out from AMD has HBM? You have a weird definition of standard.

 

Nvidia was doing this anyways, mate. They showed their roadmap with stacked memory a looong time ago, so you attempting to imply AMD is pushing Nvidia to do this is flat out absurd.

Fiji has HBM which was confirmed at FAD and should be out not long from now. In a better way to put it AMD invented industry standards that made HBM possible that Hynix manufactures for them. People get the impression that Hynix created the HBM that they are manufacturing for AMD entirely on their own. Although really AMD came to them (with joint efforts from the Joint Electron Device Engineering Council) looking for someone to collaborate with on their efforts and to help manufacture it. Like said HotChips 2015 will be this August where you'll learn more.

 

Not true. Adaptive Sync and other technologies in the VESA standards are patented, but they are released to VESA. The same thing is true of JEDEC, and Hynix, Micron, Elpida, and others have plenty of patents related to SDRAM.

Show me DRAM on the market (lets try 240-pin DDR3) that is patented by the manufacture (G.Skill, Corsair, etc).

[patrickjp93]

Fiji has HBM which was confirmed at FAD and should be out not long from now. In a better way to put it AMD invented industry standards that made the HBM possible that Hynix manufactures for them. People get the impression that Hynix created the HBM that they are manufacturing for AMD entirely on their own. Although really AMD came to them looking for someone to collaborate on their efforts and to help manufacture it. Like said HotChips 2015 will be this August where you'll learn more.

 

Show me DRAM on the market (lets try 240-pin DDR3) that is patented by the manufacture (G.Skill, Corsair, etc).

Google much?

http://www.law360.com/articles/148200/hynix-others-hit-with-dram-chip-patent-suit

http://www.conversantip.com/patent-category/memory-portfolio/

http://www.theregister.co.uk/2012/09/24/judge_reduces_rambus_royalties/

 

Clearly companies can have patents on DRAM chip technologies given money is being won in lawsuits to that effect.

[Opcode]

Google much?

http://www.law360.com/articles/148200/hynix-others-hit-with-dram-chip-patent-suit

Not what I asked for.

[patrickjp93]

Not what I asked for.

That's exactly what you asked for. Look at the referenced patents. DDR3 SDRAM patents up and down the line, and there's two more links for you including a library of 4000 memory patents related to DDR SDRAM technology. Opcode, give it up. For once, you have outright been proven to be full of garbage. If Hynix and Rambus can patent this stuff, AMD could have patented plenty of the tech related to HBM technology development, and yet there are none in the U.S., British, Chinese, Japanese, nor South Korean patent offices.

[Opcode]

That's exactly what you asked for. Look at the referenced patents. DDR3 SDRAM patents up and down the line, and there's two more links for you including a library of 4000 memory patents related to DDR SDRAM technology. Opcode, give it up. For once, you have outright been proven to be full of garbage. If Hynix and Rambus can patent this stuff, AMD could have patented plenty of the tech related to HBM technology development, and yet there are none in the U.S., British, Chinese, Japanese, nor South Korean patent offices.

Face it your many months of rambling in regards of irrelevant information hasn't changed. You did not answer my question at all even in the slightest relevance. There's a difference between technology standards that contribute to an existing standard and just a standard in general. You wont see HBM patented because it's an open JEDEC standard. I would refresh your memory of what open source hardware is and how it won't stand up at the patent office. Like said in my previous post unless you can prove DRAM is patented by anyone (not just underlying changes in DRAM) then your argument is moot. What technologies that AMD came up with for HBM they wouldn't patent from the start if it benefits the greater good of the entire industry. Look at all the places HBM can be used beyond just GPUs. I don't think AMD wants to pull an Apple any time soon.

[patrickjp93]

Face it your many months of rambling in regards of irrelevant information hasn't changed. You did not answer my question at all even in the slightest relevance. There's a difference between technology standards that contribute to an existing standard and just a standard in general. You wont see HBM patented because it's an open JEDEC standard. I would refresh your memory of what open source hardware is and how it won't stand up at the patent office. Like said in my previous post unless you can prove DRAM is patented by anyone (not just underlying changes in DRAM) then your argument is moot. What technologies that AMD came up with for HBM they wouldn't patent from the start if it benefits the greater good of the entire industry. Look at all the places HBM can be used beyond just GPUs. I don't think AMD wants to pull an Apple any time soon.

HAHAHAH! YES I DID! RamBus and Micron have all the fundamental patents of DRAM used in DDR2 and DDR3 technology which are referenced in the lawsuit. Now, JEDEC can impose a standard which all system SDRAM manufacturers are supposed to adhere to and force these companies under the umbrella to license their tech, but that has no bearing on whether or not the tech can be patented, and when it comes to process tech and additional metrics JEDEC has 0 control, a fact you would know if you actually read the JEDEC charter. Actually read through those sources and look at the patents. They are all the basis for the SDRAM standards which eventually became the cornerstones of various DDR generation standards. The same patents exist for HBM, but none filed by AMD, and over 200 filed by Hynix.

 

Furthermore, SERIOUSLY?! AMD is a business. If AMD isn't looking out for #1 (itself), it's not going to last, especially with a mountain of money due at the end of 2018. AMD has to pull an Apple or it'll be bled dry by Intel's debt collection.