rumor has it nVidia will begin mass production of Pascal GPUs before 2nd half of this year

[zMeul]

source: https://translate.google.com/translate?hl=en&sl=auto&tl=en&u=http%3A%2F%2Fwww.hwbattle.com%2Fbbs%2Fboard.php%3Fbo_table%3Dnews%26wr_id%3D15876

this rumor comes straight after Samsung officially announced they have begun mass production of 4GB HBM2 stacks: http://linustechtips.com/main/topic/530231-samsung-begins-mass-production-of-4gb-hbm2-stacks/

the interesting part about this: according to a quote from Digitimes Korea, Hynix hasn't yet started mass producing HBM2 and are expected to do so in August

SK Hynix is planning to mass production as early as August the HBM 2 generations. Detailed specifications are tentative, but reportedly Samsung has seen a performance similar to the plan to 8GB HBM2 production in the first half, especially from the second half of the year Pascal architecture GPU, NVIDIA is into the first half enough to rush in full production ends the trial production, reliability test such as plans.

I know, google translate .. don't blame me for it  

if this news is to be believed and AMD sourced HBM2 only from SK-Hynix, their video cards will be late to the party - I would not want to be in AMD's shoes if the rumors are true

AMD's exclusivity deal with Hynix turns and bites them in the arse .. would be hilarious 

[Enderman]

8GB nvidia pascal cards

hell yes

[daniellearmouth]

8GB nvidia pascal cards

hell yes

 

Sign me the hell up!

[AresKrieger]

I mean they do have the financial resources to push things out faster, and unlike intel have no incentive to kept amd partially alive, so there may be some truth to this.

[Arc_Jester]

Even if they start production it doesn't mean they will be releasing them.

[zMeul]

Even if the if they start production it doesn't mean they will be releasing them.

no, it does not

but the same deal goes for AMD too

[dragosudeki]

I understand that AMD and Nvidia are both going to use HBM2, but that literally has nothing to do with the article unless there's news that Samsung is the sole supplier to one side and Hynix is the sole supplier to the other. If anything, this is a battle between Samsung's HBM and Hynix's HBM. If Samsung takes the market first, Hynix can be left in the dust.

[-BirdiE-]

8GB nvidia pascal cards

hell yes

 

First ones are going to be 16GB HBM2 (New Titan). Will be glorious!

 

Will definitely make our framerates high, and our temperatures low.

[-BirdiE-]

I understand that AMD and Nvidia are both going to use HBM2, but that literally has nothing to do with the article unless there's news that Samsung is the sole supplier to one side and Hynix is the sole supplier to the other. If anything, this is a battle between Samsung's HBM and Hynix's HBM. If Samsung takes the market first, Hynix can be left in the dust.

 

I think that what they're saying is that because nVidia is using both, and AMD is using only Hynix (which is behind by a couple months), then nVidia will be able to start producing their cards a while before AMD will, which will likely have an impact on release date.

[zMeul]

I understand that AMD and Nvidia are both going to use HBM2, but that literally has nothing to do with the article unless there's news that Samsung is the sole supplier to one side and Hynix is the sole supplier to the other. If anything, this is a battle between Samsung's HBM and Hynix's HBM. If Samsung takes the market first, Hynix can be left in the dust.

the news broke about the same time Fury X launched that AMD will have priority of HBM2 production from Hynix

at that time Samsung was nowhere near HBM, at least not publicly 

[dragosudeki]

I think that what they're saying is that because nVidia is using both, and AMD is using only Hynix (which is behind by a couple months), then nVidia will be able to start producing their cards a while before AMD will, which will likely have an impact on release date.

 

the news broke about the same time Fury X launched that AMD will have priority of HBM2 production from Hynix

at that time Samsung was nowhere near HBM, at least not publicly 

Where does that confirm the fact that AMD cannot use HBM from Samsung? Samsung is already manufacturing AMD's 14nm chips, so its not like AMD cannot do business with Samsung.

 

EDIT: If you take into consideration that AMD can use different memory manufacturers (since I haven't seen any report of AMD not being able to do so) and AMD's priority right to Hynix's production, then AMD would theoretically have more access to HBM2 than Nvidia (assuming Samsung's HBM2 is shared evenly between them)...

[Trik'Stari]

First ones are going to be 15.5GB HBM2 (New Titan). Will be glorious!

 

Will definitely make our framerates high, and our temperatures low.

FTFY

 

(Couldn't resist)

[VagabondWraith]

Uh oh AMD.

RIP.

[Guest Generallee]

FTFY

 

(Couldn't resist)

 

pew pew pew pew pew

[zMeul]

Where does that confirm the fact that AMD cannot use HBM from Samsung? Samsung is already manufacturing AMD's 14nm chips, so its not like AMD cannot do business with Samsung.

it doesn't

but at the same time, AMD only confirmed to be sourcing HBM2 from Hynix; will AMD also be sourcing HBM2 from Samsung? perhaps

but there's also the possibility that nVidia has priority over HBM2 production from Samsung

 

 

if indeed AMD sources HBM2 from Samsung, there should be news in the next few days/weeks

[Geralt]

No more 15.5GB this time?

[Notional]

AMD already confirmed to release Polaris cards (including the high end versions) in August. That leaves 3 possibilities:

 

1. Source is full of crap
2. AMD also uses Samsung HBM2
3. SK Hynix started production in August 2015

SK Hynix' HBM1 ran at 500mhz. If they clock HBM 1 at 1Ghz, it sill effectively be HBM2. Afaik it is possible to OC HBM1 on FuryX to 1Ghz. Either way, AMD is not the one to worry about. It's obvious that AMD will be first to market with highend next gen (node) GPU's. They have functioning prototypes, that has been showcased.

 

So far we have seen nothing from NVidia. It could be that they are holding their cards close, but it's getting worrisome. With AMD's freesync over hdmi, full HDR support, and high efficiency, it seems that AMD are the ones to beat this time. However it's too soon to say, as we know nothing about Pascal other than 16nm ff+ and HBM2.

 

Either way, with a late Q2 production start, I would assume Pascal products hitting the shelves about the same time as ZEN.

[Prysin]

8GB nvidia pascal cards

hell yes

7,5*

[Enderman]

7,5*

because that extra .5GB is such a dealbreaker, yeah, totally

[Prysin]

because that extra .5GB is such a dealbreaker, yeah, totally

.5GB less

but you still pay for the .5GB

 

so price to performance is worse.

[Enderman]

.5GB less

but you still pay for the .5GB

 

so price to performance is worse.

uh, no...

if it was a 7GB card they were releasing they would still charge the same price

 

so you can equally argue you are getting more for the same price

 

the 980ti was the same price as the 780ti

the 980ti had 6GB of vram compared to the 780ti's 3GB

did they charge double for double the vram? No

they charge the same, for more vram

so what you get for the same $650 is actually more

[Notional]

.5GB less

but you still pay for the .5GB

 

so price to performance is worse.

 

The problem is worse than that on the 970: Games will always see the memory as 4GB full speed and use it accordingly. So NVidia either has to patch performance on the 970 for every single game that can use more than 3.5GB to begin with, or make something that will cheat the game into not use the last 512mb itself, but be controlled by NVidia's drivers instead. With planned obsolescence, we already see the 770 being useless and the 780 series being a joke of it's former self (granted the low amount of vram is to blame a bit too. Remember when people said you would have to run 4k to use more than 3GB?) :lol: ​

[eLucid]

The problem is worse than that on the 970: Games will always see the memory as 4GB full speed and use it accordingly. So NVidia either has to patch performance on the 970 for every single game that can use more than 3.5GB to begin with, or make something that will cheat the game into not use the last 512mb itself, but be controlled by NVidia's drivers instead. With planned obsolescence, we already see the 770 being useless and the 780 series being a joke of it's former self (granted the low amount of vram is to blame a bit too. Remember when people said you would have to run 4k to use more than 3GB?) :lol: ​

 

How many times are you going to bring this up? I've completely lost count now. Nvidia released drivers in june that fixed most of the performance issues with the 700 series. Stop trying to cause a shit storm.

[Prysin]

The problem is worse than that on the 970: Games will always see the memory as 4GB full speed and use it accordingly. So NVidia either has to patch performance on the 970 for every single game that can use more than 3.5GB to begin with, or make something that will cheat the game into not use the last 512mb itself, but be controlled by NVidia's drivers instead. With planned obsolescence, we already see the 770 being useless and the 780 series being a joke of it's former self (granted the low amount of vram is to blame a bit too. Remember when people said you would have to run 4k to use more than 3GB?) :lol: ​

no they wont. Nvidia has blocked the last .5GB with drivers. Not all applications are allowed to see it.

Some applications can see it, but cannot access it.

 

yes, this was confirmed quite some time ago, but it was never properly broadcasted. Yes with todays drivers, the 3.5GB shouldnt affect anything, because you litterally have only 3.5GB, the last .5GB is blocked in most games.

 

Non gaming applications can freely use it from what i know

[dragosudeki]

it doesn't

but at the same time, AMD only confirmed to be sourcing HBM2 from Hynix; will AMD also be sourcing HBM2 from Samsung? perhaps

but there's also the possibility that nVidia has priority over HBM2 production from Samsung

 

 

if indeed AMD sources HBM2 from Samsung, there should be news in the next few days/weeks

"AMD only confirmed to be sourcing HBM2 from Hynix" does not imply "AMD only uses HBM2 from Hynix"

"there's also the possibility that nVidia has priority over HBM2 production from Samsung" is entirely your *cough* biased *cough* speculation...

 

In fact, Nvidia and Samsung have a much more complicated relationship (reason: Nvidia suing Samsung which resulted in Nvidia violating Samsung's patents) than AMD and Samsung (reason: Samsung now manufacturing chips for AMD).

 

Either way, the scope of the Memory market far exceeds just GPUs. The battle between Samsung and Hynik will be about who can quickly access all markets that can use HBM. While Micron is on the side trying to make HMC.