[TT] Pascal rumored to use GDDR5X..

[PPB]

GDDR5 got higher densities. How do you think pro cards can get 24/32GB on 384/512bit? Or Titan X with 12GB?

Pro cards also cost an arm and a leg, they can sink the cost of higher density RAM ICs. Seems cost structure dissapeared completely from the argument. I remember it was a thing when people raved about power efficiency and how it helps reduce BoM of GPUs.

So in one corner we will have the usual +20CM reference PCBs with 4GB of memory (8GB cards will cost the +100 USD offset price we say in 770 2GB vs 4GB at launch), in the other one you could have as tiny as 15CM PCBs and up to 16GBs. Any measure you take to shorten your PCB and card size with GDDR5/X adds to cost, and negates any cost advantage versus going with HBM/2 (we may as well see HBM cards for the mid range next year if you dont need more than 4GB), then add the power difference, the BoM difference higher power RAM makes (those little chips and gray boxes in the PCB arent for show, you know).

Also remember GDDR5 isnt particulary cheap, especially when you consider Nvidia already has to buy +1750mhz binned chips because of their design decisions.

 

[thesmokingman]

You got 3 Fiji cards. You conveniently pick the lowest. A binned and down clocked version. Then compare it to 980TI instead of the Fury/Fury X. It was too obvious wasn't it.

The Nano is the TDP focused product from their stack. It would make the obvious choice.

A 120-150W Pascal card with GDDR5X could easily look like this (GTX970).

The Nano is already 25w-50w away from that goal, lmao!~

 

[AtenRa]

You got 3 Fiji cards. You conveniently pick the lowest. A binned and down clocked version. Then compare it to 980TI instead of the Fury/Fury X. It was too obvious wasn't it.

yes it is, a 175W TDP Fiji is only 10-15% slower than the 250W TDP Maxwell GTX980Ti. What is wrong with the comparison ?? I was talking about Laptops if you dont remember and those need low TDPs.

 

[ShintaiDK]

Also remember GDDR5 isnt particulary cheap, especially when you consider Nvidia already has to buy +1750mhz binned chips because of their design decisions.

HBM1/HBM2 and Interposer seems to be the high cost and limited availability. Until that changes GDDR5(X) will be the cheaper.

7Ghz GDDR5 have been a standard product for ages (since 2010). Its cheap.

 

[ShintaiDK]

yes it is, a 175W TDP Fiji is only 10-15% slower than the 250W TDP Maxwell GTX980Ti. What is wrong with the comparison ?? I was talking about Laptops if you dont remember and those need low TDPs.

How much power would slower GM200 use with comparable speeds to the Nano?

 

[vissarix]

Really ?? Nano at only 175W TDP is only 10-15% slower than 250W TDP GTX980Ti in latest games.

Take Nano,
Shrink it to 16nm FF
add DX-12 games
Include Laptops in the equation

I dont think Pascal + GDDR5X will have fun competing against it.

in which planet nano is 10% slower vs the gtx 980ti?

http://www.bug.hr/_cache/7734f4865bf8a0a6d60af9ff66ee050a.31bcb9f0f250e4c57b131733a5632296.jpg

also the gtx980 consumes less power is 1 year older and doesnt have hbm memory and beats nano at 1080p, at 1440p is almost on par...at 4k nano wins hands down for a silky smooth slideshow gameplay..

so yeah the future looks bad for amd, as always...

 

[ShintaiDK]

The Nano is already 25w-50w away from that goal, lmao!~

No performance was mentioned. You could just as well say its power consumption was twice of a GTX750TI and then go crazy in laughter.

The context was that a 120-150W GDDR5(X) card wont have to be 24cm.

 

[LTC8K6]

Exactly see, as if DDR power consumption was ever at a head? It will now be as it never existed!

That's how you run a railroad!

 

[AtenRa]

in which planet nano is 10% slower vs the gtx 980ti?

Im talking about latest games like SW Battlefront and DX-12.

 

[Headfoot]

wow this got pretty ridiculous pretty fast. GDDR5X could be a lifesaver for mid-range cards. Id be really happy if we actually got a decent MSRP $250-300 midrange card like the days of yore. I don't really care what memory tech it uses as long as it performs decently and isn't a low end chip badged as a midrange chip with the midrange chip badged as a flagship chip. Max performance is interesting but I'm personally most interested in the best high-midrange perf/$ card

 

[ShintaiDK]

Im talking about latest games like SW Battlefront and DX-12.

DX12 like Fable?
http://www.pcgameshardware.de/DirectX-12-Software-255525/Specials/Spiele-Benchmark-1172196/

Anyway, since you dont have any GM200 downclocked numbers we can stop here.

 

[ShintaiDK]

GDDR5X could be a lifesaver for mid-range cards. Id be really happy if we actually got a decent MSRP $250-300 midrange card like the days of yore. I don't really care what memory tech it uses as long as it performs decently and isn't a low end chip badged as a midrange chip with the midrange chip badged as a flagship chip. Max performance is interesting but I'm personally most interested in the best high-midrange perf/$ card

Exactly. Mainstream and low end will benefit greatly from this since everything else would more or less be cost prohibitive. And perhaps high end depending on memory type availability.

 

[railven]

Was going to post when I first saw this thread, but man did it already go down the toilet.

If this is NV's answer to HBM1/2 perceived issues, I wonder how it will do performance wise. If it's going to cover small Pascal, I might just skip it if big Pascal is gonna come with HBM.

Welps, now I'm excited about 2016.

 

[96Firebird]

I'm with Headfoot... They could reuse DDR3 for all I care, if the card performs how I want it to and there isn't a complete bottleneck due to the memory, call it whatever you want.

Anyone who cares because it isn't the newest tech out there (which it looks like it is, set for release in 2016) I must laugh at...

 

[AtenRa]

DX12 like Fable?
http://www.pcgameshardware.de/DirectX-12-Software-255525/Specials/Spiele-Benchmark-1172196/

Anyway, since you dont have any GM200 downclocked numbers we can stop here.

Highlights of AMD Catalyst™ 15.9.1 Beta Windows Driver
Performance Optimizations

• • Star Wars™: Battlefront Beta - Performance and quality optimizations
  • Fable Legends : Includes the latest DirectX
    ® 12 optimizations for the Fable Legends: Benchmark​

http://www.computerbase.de/2015-10/...eta/#diagramm-star-wars-battlefront-3840-2160

SW Battlefront 1080p

SW Battlefront 1440p

SW Battlefront 3840x2160

 

[tential]

Was going to post when I first saw this thread, but man did it already go down the toilet.

If this is NV's answer to HBM1/2 perceived issues, I wonder how it will do performance wise. If it's going to cover small Pascal, I might just skip it if big Pascal is gonna come with HBM.

Welps, now I'm excited about 2016.

I mean, how I think it will be is nvidia will skip hbm all together this gen, or at least for mid/low end. They'll be first to market, get the sales from amd. Then amd will launch their lineup right after big pascal launches. Amd will have good competitive, even better cards maybe. It'll be too late, people will have upgraded to mid range pascal earlier in the year.
Amd prices will drop (r9 390 anyone?) then amd will make much smaller margins off a smaller number of remaining consumers.

Considering hbm didn't allow for any real disruptive movement in the market, I'm not worried about the lack of inclusion of it in mid range cards

 

[ShintaiDK]

I mean, how I think it will be is nvidia will skip hbm all together this gen, or at least for mid/low end. They'll be first to market, get the sales from amd. Then amd will launch their lineup right after big pascal launches. Amd will have good competitive, even better cards maybe. It'll be too late, people will have upgraded to mid range pascal earlier in the year.
Amd prices will drop (r9 390 anyone?) then amd will make much smaller margins off a smaller number of remaining consumers.

Considering hbm didn't allow for any real disruptive movement in the market, I'm not worried about the lack of inclusion of it in mid range cards

Its the native quad core all over. Technology before performance.

The question still begs if HBM is even viable outside the 650$ cards due to cost. And even at that price HBM is questionable for the time being.

GDDR5X looks to be the cheap and easy solution for now. Stacked DRAM is the future, but the future may be some time away.

 

[AtenRa]

I mean, how I think it will be is nvidia will skip hbm all together this gen, or at least for mid/low end. They'll be first to market, get the sales from amd. Then amd will launch their lineup right after big pascal launches.

So you believe AMD will not launch a new 14/16nm GPU in 2016 ?? because i can hardly see big Pascal in 2016 for the consumer market.

 

[railven]

I mean, how I think it will be is nvidia will skip hbm all together this gen, or at least for mid/low end. They'll be first to market, get the sales from amd. Then amd will launch their lineup right after big pascal launches. Amd will have good competitive, even better cards maybe. It'll be too late, people will have upgraded to mid range pascal earlier in the year.
Amd prices will drop (r9 390 anyone?) then amd will make much smaller margins off a smaller number of remaining consumers.

Considering hbm didn't allow for any real disruptive movement in the market, I'm not worried about the lack of inclusion of it in mid range cards

Yah, basically what they've been doing since 2012. It's worked wonders for their greedy corporate stockholders, why rock the money ship when your primary competitor is AMD.

Its the native quad core all over. Technology before performance.

The question still begs if HBM is even viable outside the 650$ cards due to cost. And even at that price HBM is questionable for the time being.

GDDR5X looks to be the cheap and easy solution for now. Stacked DRAM is the future, but the future may be some time away.

I wouldn't be surprised if AMD puts HBM1 into a <$500 card. AMD has a habit of giving consumers the whole cow for the cost of a glass of milk. Then everyone scratches their heads wondering why AMD is doing so bad.

 

[Mondozei]

Rumor makes sense. HBM is a future proof tech, but the current-day benefits are limited (in terms of raw performance). Fury proved that. The massive amounts of bandwidth didn't get utilised that well.

 

[Headfoot]

To be fair, at the same die size on 14/16nm there is going to be a whole lot more GPU hardware to feed so the increased bandwidth of HBM/GDDR5X is going to be a lot more relevant. Besides, Fiji seems to have ROP issues and other weird bottlenecks. I dont think its related to the HBM IMO. High bandwidth is not a panacea.

Mid range cards will probably still not require the bandwidth provided by HBM. But things like Knights Corner (HMC), and the high end 14/16nm stuff probably will.

my personal opinion is that Fiji was the 4770-style pipecleaner product for HBM.

 

[3DVagabond]

Not too worried about power given fiji vs maxwell. Seems nvidia is ok on that front.

Space wise, this is perfect for nano. It reigns supreme in fps per inch (lol) and its successor has the potential to do the same.

However, without more mitx style cards like nano being mainstream, we won't get cases specifically designed for these builds.

Sucks, but nvidia is the gpu industry apple.

I don't think Pascal is going to be a stripped down design sacrificing functionality for power consumption like Kepler and Maxwell. We'll see, though.

 

[3DVagabond]

HBM2 for the 4K gaming solutions, GDDR5X for the 1080p gamer. Makes perfect sense.

Sure, nVidia will be able to create a niche for it to target, rather than pushing it as an all around. My biggest concern is does this mean there are issues with HBM?

 

[tviceman]

AtenRA has been, by far, the most vocal AMD troll/shill on this site for quite some time. Everyone is best served not feeding the troll.

GDDR5X being used is a sign that HBM2 won't be ready by the time Pascal chips are. Also, no one here has bothered to speculate on what the power consumption of GDDR5X would be of it ran at a slower 8-9ghz in the lower end desktop and mobile variants.

Nvidia competed, and beat, AMD at 28nm in performance and efficiency without HBM. I'm sure they will do just fine with Pascal 1.0.

Infraction issued for member callout.
-- stahlhart

 

[lilltesaito]

Sure, nVidia will be able to create a niche for it to target, rather than pushing it as an all around. My biggest concern is does this mean there are issues with HBM?

Maybe not issues with HBM but maybe issues with Nvidia getting HBM working. I am sure AMD had a lot of the delays for Fury because of getting HBM working correctly.