R9 390X Fiji XT - Extreme Memory Bandwidth & SPs

[Opcode]

There is exactly one punctuation flaw in the opening sentence. The rest is perfect. I put work into my phrasing. It's a more archaic style, but it's completely valid.

You can't have an exclusive tech without a patent, and since Hynix does have parents on HBM, it stands to reason Nvidia can't just outsource it. You're wrong four times for every one I am, as this thread demonstrates.

Being the young teenager that you are the "I'm always right" attitude goes without saying. I suggest you actually look over my posts and understand that in every single one of them you've been proven wrong.

[patrickjp93]

Being the young teenager that you are the "I'm always right" attitude goes without saying. I suggest you actually look over my posts and understand that in every single one of them you've been proven wrong.

You haven't proven me wrong except in 2 posts toward the beginning. If only I could sue for public defamation...it'd be a cakewalk, but enough derailing the thread. Find someone else to troll.

[Opcode]

You haven't proven me wrong except in 2 towards the beginning. If only I could sue for public defamation...it'd be a cakewalk, but enough derailing the thread. Find someone else to troll.

You're clearly the one promoting trolling here (I'm pretty sure the entire community will stand behind that). Being the young teenager that you are the "I'm always right" attitude goes without saying. I suggest you actually look over my posts and understand that in every single one of them you've been simply proven wrong. I don't mind replying to you on a consistent basis as it gives me more time to teach you things that you clearly don't know. Tho if you're only going to manipulate my posts and throw them back at me as if they're wrong (in which case they're not) and you're somehow magically right. There's not much room for you to be taught something let alone enough room for you to actually learn something, keep that in mind.

[patrickjp93]

You're clearly the one promoting trolling here (I'm pretty sure the entire community will stand behind that). Being the young teenager that you are the "I'm always right" attitude goes without saying. I suggest you actually look over my posts and understand that in every single one of them you've been simply proven wrong. I don't mind replying to you on a consistent basis as it gives me more time to teach you things that you clearly don't know. Tho if you're only going to manipulate my posts and throw them back at me as if they're wrong (in which case they're not) and you're somehow magically right. There's not much room for you to be taught something let alone enough room for you to actually learn something, keep that in mind.

Oh, I can be taught, but you are no teacher. You've proven nothing, and no academic institution or court of law would say otherwise. You're caught in your own web of speculation and you won't admit it. Fine by me, but quit derailing the thread and spreading your opinions as facts to others. At least admit you're speculating.

I've proven HBM can't be outsourced, and you said Hynix's HBM would be AMD-exclusive, which isn't remotely possible since Nvidia has it on their roadmap for Pascal and Hynix doesn't license manufacturing of its technologies to anyone.

[Opcode]

Oh, I can be taught, but you are no teacher. You've proven nothing, and no academic institution or court of law would say otherwise. You're caught in your own web of speculation and you won't admit it. Fine by me, but quit derailing the thread and spreading your opinions as facts to others. At least admit you're speculating.

I've proven HBM can't be outsourced, and you said Hynix's HBM would be AMD-exclusive, which isn't remotely possible since Nvidia has it on their roadmap for Pascal and Hynix doesn't license manufacturing of its technologies to anyone.

If learning from from bad sources is your thing I can tell how that's working out for you. Don't take it the wrong way but the thread derailed the moment you started posting in it.

 

HBM can be outsourced if the company feels fit to market it to said buyer. I said Hynix's branded HBM will probably be exclusive to AMD as in I doubt Nvidia is going to pull up on using HBM that was developed in cooperation with AMD. Nvidia will get HBM, tho it will more than likely come from another manufacture as Hynix won't be the only manufacture of HBM by that time.

 

Next question?

[FatalityDiablo]

-

lel^^

[OlekKing]

Intriguing

[Sonefiler]

If learning from from bad sources is your thing I can tell how that's working out for you. Don't take it the wrong way but the thread derailed the moment you started posting in it.

 

HBM can be outsourced if the company feels fit to market it to said buyer. I said Hynix's branded HBM will probably be exclusive to AMD as in I doubt Nvidia is going to pull up on using HBM that was developed in cooperation with AMD. Nvidia will get HBM, tho it will more than likely come from another manufacture as Hynix won't be the only manufacture of HBM by that time.

 

Next question?

Dude, patents exist. You're wrong. I always see you on these types of threads starting flamewars. Stop.

[Opcode]

Dude, patents exist. You're wrong. I always see you on these types of threads starting flamewars. Stop.

Me starting flame wars? Feel free to show me proof, otherwise don't post garbage just like your little buddy intentionally to stir things up.

 

Apparently you're unfamiliar with JEDEC's patent policy. Let me know the last time you heard of someone trying to patent DDR, DDR2, DDR3, or even DDR4 technology.

 

[bogus]

Read it again. Stacked memory was developed for those sensors because if you want a 60 Megapixel photo, you need to move all that data extremely fast over very high bandwidth at low power. The bulk of the work had already been done.

AMD only came in and asked it be brought to GPus as well. The difficult engineering work (through-silicon Vias being used in memory and computational chips) had already been done. The only thing new was getting a set of interconnects for a GPU PCB and rearranging a reference card for multi-channel RAM, which isn't much work.

I've read it again in both sources it says AMD and Hynix developed HBM. So I don't get it.

[patrickjp93]

I've read it again in both sources it says AMD and Hynix developed HBM. So I don't get it.

The work for stacked memory IN GENERAL was done before AMD was even in the picture. The only thing new about HBM was the interface with a graphics processing unit. That is why AMD should not really be given any credit. The bulk of the work and research for stacked memory was already done, and making a system to connect to it is a joke by comparison since it's still DRAM.

[Opcode]

The work for stacked memory IN GENERAL was done before AMD was even in the picture. The only thing new about HBM was the interface with a graphics processing unit. That is why AMD should not really be given any credit. The bulk of the work and research for stacked memory was already done, and making a system to connect to it is a joke by comparison since it's still DRAM.

You honestly think AMD developed a JEDEC standard?

 

:lol:  :lol:  :lol:

[patrickjp93]

If learning from from bad sources is your thing I can tell how that's working out for you. Don't take it the wrong way but the thread derailed the moment you started posting in it.

 

HBM can be outsourced if the company feels fit to market it to said buyer. I said Hynix's branded HBM will probably be exclusive to AMD as in I doubt Nvidia is going to pull up on using HBM that was developed in cooperation with AMD. Nvidia will get HBM, tho it will more than likely come from another manufacture as Hynix won't be the only manufacture of HBM by that time.

 

Next question?

You can't outsource technology Hynix has full rights to and has given no license to anyone else to produce. If you mean competing products, that would be where the hybrid memory cube comes in, but you can't currently outsource HBM.

[patrickjp93]

You honestly think AMD developed a JEDEC standard?

 

:lol:  :lol:  :lol:

They developed Mantle and half of the VESA standards for Graphics output since DP came onto the scene, so it's possible for them to given AMD has competent people, but I was referring to Hynix developing a Jedec standard. Anyone can develop a Jedec standard, though it does have to be accepted by the consortium, and of course we have companies who are known to develop many of them.

[Opcode]

You can't outsource technology Hynix has full rights to and has given no license to anyone else to produce. If you mean competing products, that would be where the hybrid memory cube comes in, but you can't currently outsource HBM.

You need to catch up with the reply's in this thread, the post I made last night already corrects this post (it's the one with the link in it up there ^). 

 

They developed Mantle and half of the VESA standards for Graphics output since DP came onto the scene, so yes. 

Please, share with us which memory technology (focusing on HBM) that you think AMD developed independently.

 

Nice quick edit, finally reading the material that I am providing. 

[brockelley]

Am I the only one waiting for the generation after this one? The one we know will be able to push 4k decently? The one that will most likely have better support for Oculus/ be coming out around the same time as the Rift? The one that will come out around the same time HDMI 2.0 will be the standard, and we'll be able to get good 4k 60 inch TVs for under $1000 that will refresh at 120hz with less than 20ms of input lag?

 

 

Guys, I'd much rather a discussion about that then whatever it is this thread turned into.. anyone wanna talk about the actual topic and what effect it will have on your future investments in PC tech?

[patrickjp93]

You need to catch up with the reply's in this thread, the post I made last night already corrects this post (it's the one with the link in it up there ^). 

 

Please, share with us which memory technology (focusing on HBM) that you think AMD developed independently.

 

Nice quick edit, finally reading the material that I am providing. 

Memory tech? None. Though I do realize I partly messed up and assumed you meant "Hynix developed a Jedec standard?" 

 

AMD didn't develop one, though they could. Hynix has.

[bogus]

The work for stacked memory IN GENERAL was done before AMD was even in the picture. The only thing new about HBM was the interface with a graphics processing unit. That is why AMD should not really be given any credit. The bulk of the work and research for stacked memory was already done, and making a system to connect to it is a joke by comparison since it's still DRAM.

But we are talking about HBM - wich is for GPUs and APUs. This is what it's being said: AMD and Hynix developed HBM. Not only by the media, but by AMD and Hynix.

How can you say AMD shouldn't receive any credit if you don't know what was their contribution?  They (AMD/ATI) developed GDDR so you think they have zero contribution, either know-how, IP, what ever?

In fact, why would Hynix claim a development partnership if they don't need AMD?

Jesus, you really think AMD are a bunch of dumb guys who have no idea what they are doing and just glue themselfs to others work.

[patrickjp93]

But we are talking about HBM - wich is for GPUs and APUs. This is what it's being said: AMD and Hynix developed HBM. Not only by the media, but by AMD and Hynix.

How can you say AMD shouldn't receive any credit if you don't know what was their contribution?  They developed GDDR so you think they have zero contribution, either know-how, IP, what ever?

In fact, why would Hynix claim a development partnership if they don't need AMD?

Jesus, you really think AMD are a bunch of dumb guys who have no idea what they are doing and just glue themselfs to others work.

GDDR 1 and 3 were based on DDR 1 and 3. 5 isn't even that much of a departure. Again, most of the work was already done. Rearranging the interface on a GPU PCB to be appropriate for it was a simple task. 

 

Hynix was given a lot of money for the HBM project. No need to toss AMD aside. It's just a matter of knowing who wore the pants and who bought them.

 

No, AMD is not full of stupid people. AMD is full of people who have no experience in chip stacking and for the last 6 years proved they could barely design competitive CPUs while relying on the IP and expertise of the ATI team to keep the whole operation afloat. With Keller and some other hard hitters from the past being back, AMD can redeem itself, but don't delude yourself. AMD was out of its element with regards to HBM. It's bleeding edge memory technology, not logic circuitry. AMD hasn't touched the stuff in a decade. Their cache techniques are the same as they were in 2005 as well.

[Opcode]

GDDR 1 and 3 were based on DDR 1 and 3. 5 isn't even that much of a departure. Again, most of the work was already done. Rearranging the interface on a GPU PCB to be appropriate for it was a simple task. 

 

Hynix was given a lot of money for the HBM project. No need to toss AMD aside. It's just a matter of knowing who wore the pants and who bought them.

 

No, AMD is not full of stupid people. AMD is full of people who have no experience in chip stacking and for the last 6 years proved they could barely design competitive CPUs while relying on the IP and expertise of the ATI team to keep the whole operation afloat. With Keller and some other hard hitters from the past being back, AMD can redeem itself, but don't delude yourself. AMD was out of its element with regards to HBM. It's bleeding edge memory technology, not logic circuitry. AMD hasn't touched the stuff in a decade. Their cache techniques are the same as they were in 2005 as well.

What would be the point of AMD spending millions (that they don't have) into making their own HBM manufacturing facility. When they can turn to a company that has specialized in manufacturing memory for decades. Trade off is AMD get's a price cut on their HBM. In the long run everybody wins including you the consumer.

[patrickjp93]

What would be the point of AMD spending millions (that they don't have) into making their own HBM manufacturing facility. When they can turn to a company that has specialized in manufacturing memory for decades. Trade off is AMD get's a price cut on their HBM. In the long run everybody wins including you the consumer.

This is exactly why AMD went to Hynix. The only part of the story you seem to be missing is that most of the work was already done. It only took a repackaging of existing technologies and a rearranged connection for a GPU PCB (or for the Carrizo APU). That's why AMD didn't have to sink quite as much money into Hynix. As per AMD getting a good deal, we'll see, and we'll see how much those costs get passed onto consumers. I withhold judgment on that until the 300 series launches.

[blAckspAde]

Yay for freakishly huge memory bandwidth.

[dalekphalm]

GDDR 1 and 3 were based on DDR 1 and 3. 5 isn't even that much of a departure. Again, most of the work was already done. Rearranging the interface on a GPU PCB to be appropriate for it was a simple task. 

 

Hynix was given a lot of money for the HBM project. No need to toss AMD aside. It's just a matter of knowing who wore the pants and who bought them.

 

No, AMD is not full of stupid people. AMD is full of people who have no experience in chip stacking and for the last 6 years proved they could barely design competitive CPUs while relying on the IP and expertise of the ATI team to keep the whole operation afloat. With Keller and some other hard hitters from the past being back, AMD can redeem itself, but don't delude yourself. AMD was out of its element with regards to HBM. It's bleeding edge memory technology, not logic circuitry. AMD hasn't touched the stuff in a decade. Their cache techniques are the same as they were in 2005 as well.

This statement is wrong. I can't find much information on the original "GDDR1" implementation, but GDDR3 is most definitely NOT based on DDR3.

 

http://en.wikipedia.org/wiki/DDR2_SDRAM#Relation_to_GDDR_memory

ATI has since designed the GDDR technology further into GDDR3, which is based on DDR2-SDRAM ...

 

http://en.wikipedia.org/wiki/GDDR3

GDDR3 is not related to the JEDEC DDR3 specification.

 

So, at this point, you're going to need to provide sources for your claims, since you've already gotten other information wrong.

[patrickjp93]

This statement is wrong. I can't find much information on the original "GDDR1" implementation, but GDDR3 is most definitely NOT based on DDR3.

http://en.wikipedia.org/wiki/DDR2_SDRAM#Relation_to_GDDR_memory

http://en.wikipedia.org/wiki/GDDR3

So, at this point, you're going to need to provide sources for your claims, since you've already gotten other information wrong.

Apologies. That happens to be a typo. On my damn phone the digital keys are just a hair too small for my thumbs.

[Opcode]

This is exactly why AMD went to Hynix. The only part of the story you seem to be missing is that most of the work was already done. It only took a repackaging of existing technologies and a rearranged connection for a GPU PCB (or for the Carrizo APU). That's why AMD didn't have to sink quite as much money into Hynix. As per AMD getting a good deal, we'll see, and we'll see how much those costs get passed onto consumers. I withhold judgment on that until the 300 series launches.

All of the work was already done as once memory becomes a JEDEC standard manufactures essentially have a guideline of how to produce said technology. AMD didn't invest any money into R&D with Hynix but more over covered production costs in order for them to mass produce it. As for being a cost issue we both know HBM more than likely won't drive up AMD pricing. AMD would of never invested into using such a technology if it drove up production costs of their products.