** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[Mauller]

lol... Maybe because HBM2 modules are somewhat larger than HBM :s

Not necessarily, higher capacity GDDR modules didn't get bigger, just denser. The stacks are also a fixed height whether 2 or 8 hi, they just use different height packing on the top of the stack to equalise their height overall.

like how it shows here http://hardwarebg.com/38195-samsung-започва-масово-производство-на-най-н/

 

[Rroff]

Right now though AFAIK they are only producing ~90mm2 HBM2 packages compared to 37 or 40mm2 or something for HBM1. That isn't to say that won't change at some point.

 

[Mauller]

Right now though AFAIK they are only producing ~90mm2 HBM2 packages compared to 37 or 40mm2 or something for HBM1. That isn't to say that won't change at some point.

Ah, i had not seen info about that. Found this on Anand.

HBM2 memory stacks are not only faster and more capacious than HBM1 KGSDs, but they are also larger. SK Hynix’s HBM1 package has dimensions of 5.48 mm × 7.29 mm (39.94 mm2). The company’s HBM2 chip will have dimensions of 7.75 mm × 11.87 mm (91.99 mm2). Besides, HBM2 stacks will also be higher (0.695 mm/0.72 mm/0.745 mm vs. 0.49 mm) than HBM1 KGSDs, which may require developers of ASICs (e.g., GPUs) to install a heat-spreader on their SiPs to compensate for any differences in height between the memory stacks and GPU die, to protect the DRAM, and to guarantee sufficient cooling for high bandwidth memory.

 

[Rroff]

Looks like the whole package for the first round of proper high end GPUs could have a fairly large slab of space taken up under the heatsink with the space required for various considerations - layout, size of HBM2, heat, etc.

 

[Mauller]

Looks like the whole package for the first round of proper high end GPUs could have a fairly large slab of space taken up under the heatsink with the space required for various considerations - layout, size of HBM2, heat, etc.

at 1200mm^2 though, they could fit 4 90mm^2 HBM2 stacks and a 350mm^2 core with room to spare. Unless they are trying to go with having 6 - 8 HBM 2 stacks.

Edit - would make the rumours around them having trouble with interposers even more legitimate if 6 - 8 stacks were the case. that is a lot of traces to route for a first time use of interposers.

 

[eddyr]

at 1200mm^2 though, they could fit 4 90mm^2 HBM2 stacks and a 350mm^2 core with room to spare. Unless they are trying to go with having 6 - 8 HBM 2 stacks.

Edit - would make the rumours around them having trouble with interposers even more legitimate if 6 - 8 stacks were the case. that is a lot of traces to route for a first time use of interposers.

Dawn of the staggered multi-substrate interposer stack with TSV's.

 

[Orangey]

They are not trying to fit a 350mm2 die on there, they want a ~500mm2 die on there.

Unless certain rumours turn out to be true in 2017.

 

[drunkenmaster]

Sampling interposers.... now, but people think GP100 is coming in the next couple of months. Sampling interposers now would mean they haven't even made a Gp100 + interposer + hbm package yet, which means they are a very very long way from production level capability.

 

[Mauller]

They are not trying to fit a 350mm2 die on there, they want a ~500mm2 die on there.

Unless certain rumours turn out to be true in 2017.

500mm^2 is way too big this early on in the node. I doubt they would be getting reasonable yields anytime soon with a die that large.

 

[RichardH754]

So whats the odds on them bringing out the 1080Ti as well as the other 2?

 

[NexusK]

So whats the odds on them bringing out the 1080Ti as well as the other 2?

3.141 : 1

 

[RichardH754]

lol

 

[CyberCrow]

Hopefully the 1070 will be a bit faster than a 980Ti or at least on a par.

April fools has been an gone!

The 980 is roughly neck an neck with the 780ti (both beating the other in various situations)
so how is an 1070 going be faster than a 980Ti.

Doing so will would kill off the market for older hardware.

 

[FoxEye]

April fools has been an gone!

The 980 is roughly neck an neck with the 780ti (both beating the other in various situations)
so how is an 1070 going be faster than a 980Ti.

Doing so will would kill off the market for older hardware.

Careful. Certain people here start foaming at the mouth if you say things like that. Because apparently the 1070 is going to be faster than a 980ti, and cost £250. JSH is apparently taking over from Santa Claus this year.

 

[easyrider]

Careful. Certain people here start foaming at the mouth if you say things like that. Because apparently the 1070 is going to be faster than a 980ti, and cost £250. JSH is apparently taking over from Santa Claus this year.

JSH ?

 

[CyberCrow]

Careful. Certain people here start foaming at the mouth if you say things like that. Because apparently the 1070 is going to be faster than a 980ti, and cost £250. JSH is apparently taking over from Santa Claus this year.

Lol love it

 

[NMBlack]

JSH ?

Jen-Hsun Huang....nvidia boss....?

 

[Rroff]

April fools has been an gone!

The 980 is roughly neck an neck with the 780ti (both beating the other in various situations)
so how is an 1070 going be faster than a 980Ti.

Doing so will would kill off the market for older hardware.

Careful. Certain people here start foaming at the mouth if you say things like that. Because apparently the 1070 is going to be faster than a 980ti, and cost £250. JSH is apparently taking over from Santa Claus this year.

For a number of reasons Pascal is unlikely to be cheap.

But history is against a lot of the claims in regards to "1070" performance:

GTX670 was convincingly faster than the GTX580, GTX470 stomped all over the GTX285 the GTX260 was often almost twice as fast as the 9800GTX(+) - nVidia has no problems killing off the value of older hardware, often the next generation x70 position card was priced keenly against the previous generation high end card or even significantly cheaper while matching or beating its performance and the 980ti certainly won't be immune to that.

 

[easyrider]

For a number of reasons Pascal is unlikely to be cheap.

But history is against a lot of the claims in regards to "1070" performance:

GTX670 was convincingly faster than the GTX580, GTX470 stomped all over the GTX285 the GTX260 was often almost twice as fast as the 9800GTX(+) - nVidia has no problems killing off the value of older hardware, often the next generation x70 position card was priced keenly against the previous generation high end card or even significantly cheaper while matching or beating its performance and the 980ti certainly won't be immune to that.

I think the 980 not ti

 

[Simon Belmont]

Nvidia’s CEO Jen-Hsun Huang’s keynote on pascal starts at 9am PT (California pacific standard time) That's 5pm UK. Tuesday April 5th

News better be good?