• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

digitimesAMD expected to unveil Polaris GPU in June

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I'm not sure why AMD supporters feel the need to constantly insult people they disagree with. If you go to the Pascal thread the latest info is May.

It is just that whenever I read one of your posts, my IQ drops by 3.

You just insult any logic saying that if some product is being revealed in may, it is 6 month ahead of another product being revealed in june. You have 1 brain and 2 hands, think double next time before brain farting whenever you post something.
 
So if both target mid-2016 release, then GDDR5X is out of the question, right? They'll reserve that for a 2017 refresh I suppose.
 
You just insult any logic saying that if some product is being revealed in may, it is 6 month ahead of another product being revealed in june. You have 1 brain and 2 hands, think double next time before brain farting whenever you post something.
He's been happily trolling this forum, getting replies even from experienced members who should know better. Maybe... stop replying to him? There's also the ignore function, can help tremendously with that.
 
Definitely wouldn't bet on GDDR5X from either camp, frankly. GDDR5 for all but highest end; highest end gets HBM2.

I think we will see a GP104 with 4096 cc, 256 bit GDDR5X, 8 GB, 128 ROPs. The Polaris competitor would be 4096 sp, 2048 bit HBM2, 8GB, 128 ROPs. But I think both these chips will launch later towards the end of 2016. First we will see smaller chips from both camps.
 
Definitely wouldn't bet on GDDR5X from either camp, frankly. GDDR5 for all but highest end; highest end gets HBM2.

I wouldn't be surprised if they refresh GP104 sometime in 2017 with GDDR5X and slightly higher clocks while maintaining prices. By then they will stop optimizing for Maxwell so the gap will be much larger than it will look at GP104's initial launch.

HBM2 seems destined to be a GP100/GP102 exclusive.
 
I think we will see a GP104 with 4096 cc, 256 bit GDDR5X, 8 GB, 128 ROPs. The Polaris competitor would be 4096 sp, 2048 bit HBM2, 8GB, 128 ROPs. But I think both these chips will launch later towards the end of 2016. First we will see smaller chips from both camps.

Could you explain why you think that NVIDIA would use an inferior/less efficient memory technology relative to what AMD is using for a comparable part?
 
Could you explain why you think that NVIDIA would use an inferior/less efficient memory technology relative to what AMD is using for a comparable part?

Because I think AMD's first Polaris flagship will also double up as a FirePro flagship with fp64. For Nvidia I think they are likely to follow the Kepler model with the GP104 based first flagship being a cost optimized and easier to produce GDDR5X part without FP64 focus. Then Nvidia brings out its big die Tesla/Geforce king of the hill GPU with FP64 and HBM2. Anyway thats just my opinion. I am also curious to see Nvidia and AMD's memory choices.
 
Because I think AMD's first Polaris flagship will also double up as a FirePro flagship with fp64. For Nvidia I think they are likely to follow the Kepler model with the GP104 based first flagship being a cost optimized and easier to produce GDDR5X part without FP64 focus. Then Nvidia brings out its big die Tesla/Geforce king of the hill GPU with FP64 and HBM2. Anyway thats just my opinion. I am also curious to see Nvidia and AMD's memory choices.

Thank you for your input. We shall see how it all plays out.
 
I think we will see a GP104 with 4096 cc, 256 bit GDDR5X, 8 GB, 128 ROPs. The Polaris competitor would be 4096 sp, 2048 bit HBM2, 8GB, 128 ROPs. But I think both these chips will launch later towards the end of 2016. First we will see smaller chips from both camps.

Wow, you got this way wrong. Would you like to make any more predictions?
 
What part was wrong ???

The "NVIDIA will use crappy, power hungry GDDR5/5X while AMD's comparable part will use fancy new HBM2."

The not-so-subtle implication was that AMD will have better technology & the leadership part while NVIDIA would put out an under-performer by way of using inferior technology.
 
The "NVIDIA will use crappy, power hungry GDDR5/5X while AMD's comparable part will use fancy new HBM2."

The not-so-subtle implication was that AMD will have better technology & the leadership part while NVIDIA would put out an under-performer by way of using inferior technology.

Which do you think is wrong? AMD having HBM2 or nVidia using GDDR5X?
 
The "NVIDIA will use crappy, power hungry GDDR5/5X while AMD's comparable part will use fancy new HBM2."

The not-so-subtle implication was that AMD will have better technology & the leadership part while NVIDIA would put out an under-performer by way of using inferior technology.

Really, the only thing he might have been wrong on is the name. The GP104 competitor might be out end of 2016, with 8GB HBM2 and a large number of shaders. They might just call it Vega instead of Polaris.
 
I highly doubt either camp will be using GDDR5X on Polaris/Pascal chips. GDDR5X is going to be much more expensive, and not available in the quantities that either want. AMD has pretty much stated they will not use GDDR5X because of cost. Polaris is GDDR5, and Vega is HBM2.

I would expect big Pascal (out next year sometime) to use HBM2. While the first round of Pascal coming late this year is GDDR5 (not X).
 
Which do you think is wrong? AMD having HBM2 or nVidia using GDDR5X?

My guess is that both are wrong. HBM parts will be out later this year or early 2017. I can't see either camp letting the 4GB HBM1 limit spoil their party and GDDR5X would delay the launch past May.

Mid-range consumer parts from both camps and 1 Tesla from Nvidia. Vega 10 and 11 are what everyone is really waiting for, but we'll be waiting a while 😎
 
My guess is that both are wrong. HBM parts will be out later this year or early 2017. I can't see either camp letting the 4GB HBM1 limit spoil their party and GDDR5X would delay the launch past May.

Mid-range consumer parts from both camps and 1 Tesla from Nvidia. Vega 10 and 11 are what everyone is really waiting for, but we'll be waiting a while 😎

If I understand we are talking the Vega parts in Q4/Q1, not Polaris. Those should have HBM2 and GDDR5X (Silly name. Why not GQDR?) by then.
 
Really, the only thing he might have been wrong on is the name. The GP104 competitor might be out end of 2016, with 8GB HBM2 and a large number of shaders. They might just call it Vega instead of Polaris.

Yea i was going to say that the name is not Polaris but Vega and so far that is the only thing we know.

So if he was wrong or not is not established yet. Oh and Polaris could come with HBM1. Nobody said its GDDR 5 only.
 
If I understand we are talking the Vega parts in Q4/Q1, not Polaris. Those should have HBM2 and GDDR5X (Silly name. Why not GQDR?) by then.

Do you think they would do 2 different memory techs on the same GPU? I'm assuming that 10 would be the big boy and 11 a cut down variant. That makes the most sense financially.
 
Do you think they would do 2 different memory techs on the same GPU? I'm assuming that 10 would be the big boy and 11 a cut down variant. That makes the most sense financially.

I would think they are 2 different GPU's like Polaris is. But no, I don't think they will use 2 different memory types. I just mean that both should be available by then.
 
Maybe NVIDIA bought out the first round of HBM2

I don't think that's the case at all. Vega is rumored to be fabbed at TSMC as opposed to Polaris which is at GloFo. AMD's probably been focusing their efforts on getting Polaris out and Vega is just simply behind.
 
Polaris and Vega are both bright Stars in the northern skies. It's probably the same architecture but Vega was made to use HBM2 while Polaris needs a MC on die for GDDR5.

GDDR5X is not ready for mass production. It's scheduled for June+ from official public release. Anything that wants to use it has to wait a quarter for validation/QA. That's Q3 for production of cards with GDDR5X in the factories in China. When will it be ready for retail? Maybe later in Q3.

HBM2 Samsung is ramping up already now. The problem with this is the interposer and TSV process takes time and QA. So presumably cards could also be ready for a June+ release.

If anything, HBM2 is more ready than GDDR5X.
 
Back
Top