Amd "gemini" dual GPU card specs (Leak/rumour/wccf)

[TechGod]

Yeah never gonna happen

Ashes of Singularity devs might. 

[patrickjp93]

Yeah, 1 980 ti > 1 fury x. But the fury x2 will be liquid cooled, and has better scaling so will be > dual gpu maxwell

Unless dual Maxwell is also liquid cooled. The Titan Z had way better scaling than 2 Titan Blacks or 2 780TIs.

[TechGod]

Unless dual Maxwell is also liquid cooled. The Titan Z had way better scaling than 2 Titan Blacks or 2 780TIs.

Look at the post up above. The 980Ti isn't scaling very well in comparison to the Fury X.

[Curufinwe_wins]

Yeah, 1 980 ti > 1 fury x. But the fury x2 will be liquid cooled, and has better scaling so will be > dual gpu maxwell

You assume that the dual gpu for nvidia won't also be liquid cooled.

If the titan z told nvidia anything it's that liquid cooling is basically a must for dual gpu these days.

Also dual Fiji at lowered clock rates is really disappointing, not that Fiji overclocks well anyways, but still.

[Curufinwe_wins]

Look at the post up above. The 980Ti isn't scaling very well in comparison to the Fury X.

Yes cause that 10% difference is still so small that assuming both use liquid cooling, the top end of gm200 will be a good 20% better than dual Fiji.

[TechGod]

Yes cause that 10% difference is still so small that assuming both use liquid cooling, the top end of gm200 will be a good 20% better than dual Fiji.

You are soooooo sure of that....

The dual Fiji will be better. 

[patrickjp93]

Look at the post up above. The 980Ti isn't scaling very well in comparison to the Fury X.

Emphasis on yet. SLI improves over time. This isn't news.

[patrickjp93]

You are soooooo sure of that....

The dual Fiji will be better.

If the dual Fiji runs at Nano's clocks, it will be about 10% worse assuming 80% SLI scaling.

[TechGod]

Emphasis on yet. SLI improves over time. This isn't news.

But if the tables were turned, then you would have been praising nvidia and saying AMD will never be able to improve Crossfire. 

 

You seriously think that AMD is just gonna sit there and not improve cross fire scaling for dual Fiji and let nvidia take the crown? You're a bigger fanboy than I thought. 

[Curufinwe_wins]

You are soooooo sure of that....

I am. Call it right now. Save the quote and call me out if wrong.

Top end of dual gm200 will be at least 20% better than top end of dual Fiji if both use water cooling. (Hopefully jayztwocents can get a review sample of each, even if he can't keep it).

And I feel very confident about water cooling on a dual gm200.

[TechGod]

If the dual Fiji runs at Nano's clocks, it will be about 10% worse assuming 80% SLI scaling.

That's a big if.

[TechGod]

I am. Call it right now. Save the quote and call me out if wrong.

Top end of dual gm200 will be at least 20% better than top end of dual Fiji if both use water cooling. (Hopefully jayztwocents can get a review sample of each, even if he can't keep it).

And I feel very confident about water cooling on a dual gm200.

That's a ridiculous expectation. Gonna make it my sig so I can easily rub AMD's victory in your face. 

[patrickjp93]

But if the tables were turned, then you would have been praising nvidia and saying AMD will never be able to improve Crossfire.

You seriously think that AMD is just gonna sit there and not improve cross fire scaling for dual Fiji and let nvidia take the crown? You're a bigger fanboy than I thought.

I'd never say that. Technology is always improving until it slams into barriers based in the laws of physics. AMD's XFire scaling is around 95%, which is about as good as you'll do when you add the communication overhead. Squeezing blood from a rock isn't really possible.

[patrickjp93]

That's a big if.

Nope, as that's the current purported Gemini clock rate, same as Nano both in core and memory. At 95% scaling there's no way it'll beat a dual GM 200 under water stock to stock. Overclocking will narrow the gap if not close it, but this is a foregone conclusion of the rumored speeds are correct.

[TechGod]

I'd never say that. Technology is always improving until it slams into barriers based in the laws of physics. AMD's XFire scaling is around 95%, which is about as good as you'll do when you add the communication overhead. Squeezing blood from a rock isn't really possible.

I don't know. Some of your posts about AMD are all about how they're inferior etc. I almost don't want to take you seriously. 

[Curufinwe_wins]

That's a ridiculous expectation. Gonna make it my sig so I can easily rub AMD's victory in your face.

Using the same graphs you loved to use against Patrick:

At the top end, single 980ti is easily 15-25% better than fury x top end on average.

Assuming 179% scaling on a 980ti you get 223% (1.25*179) of a fury x top end.

Consider now that the fury x is 187% scaling in cf.

That's a 19% difference right there (223/187), and I predict even relative to the dual gm200, dual Fiji will have even more top end issues as it has had thus far overclocking.

Hence 20%. Plus nvidia will put more resources into developing and optimizing the card than amd will (especially on launch day). As they have for quite a long time.

[patrickjp93]

I don't know. Some of your posts about AMD are all about how they're inferior etc. I almost don't want to take you seriously.

Just because AMD's products are garbage now (frankly the 200 series was that bad under DX 11 until the most recent driver updates), doesn't mean AMD will always make garbage products. That said, I doubt they have the ability to dig themselves out of their financial hole.

[TechGod]

Just because AMD's products are garbage now (frankly the 200 series was that bad under DX 11 until the most recent driver updates), doesn't mean AMD will always make garbage products. That said, I doubt they have the ability to dig themselves out of their financial hole.

Arctic Islands and Zen can get them out of their dire situation. I firmly believe that. OEMs might jump on board if AMD can offer them a competitive package at a competitive price. 

They don't need to wreck the i7, they just need to offer chips with equivalent performance with maybe slightly lower pricing. 

 

EDIT: I literally fail to see how the Fury X, Fury, Fury Nano and 390 cards are garbage.  

[patrickjp93]

Arctic Islands and Zen can get them out of their dire situation. I firmly believe that. OEMs might jump on board if AMD can offer them a competitive package at a competitive price.

They don't need to wreck the i7, they just need to offer chips with equivalent performance with maybe slightly lower pricing.

EDIT: I literally fail to see how the Fury X, Fury, Fury Nano and 390 cards are garbage.

That relies upon AMD selling enough Zen parts at high enough margins to offset their R&D expenses and additionally give them breathing room for research on Zen+ and its successor, not to mention the Raven Ridge APUs and their successor.

You're also assuming Arctic Islands will be wildly popular when all indicators say otherwise. I didn't say the 300 series was garbage. I said the 200 series was for most of its lifetime, which is true.

And you expect Intel won't price them into the floor if AMD becomes a problem? Intel's already got the business realm locked up for workstations and desktops. AMD will have to break in on tablets, netbooks, and laptops in higher price ranges than entry. That would require prestige to get the sales, something AMD is not going to build enough of to get Zen out the door in the quantities needed before Kaby Lake arrives. As long as Intel has near exclusive rights with Dell (the biggest buyer of CPUs in the world), I don't see AMD making the comeback it needs to. Data centers are looking more to Scale-Up architecture, something Intel's adapting to but AMD isn't. AMD's Zen APUs won't be ready until 2017 either, and no one's buying until they're proven. I don't see AMD making it through the end of the decade. That's not to say the products won't be any good. That's a separate conversation.

[TechGod]

That relies upon AMD selling enough Zen parts at high enough margins to offset their R&D expenses and additionally give them breathing room for research on Zen+ and its successor, not to mention the Raven Ridge APUs and their successor.

You're also assuming Arctic Islands will be wildly popular when all indicators say otherwise. I didn't say the 300 series was garbage. I said the 200 series was for most of its lifetime, which is true.

And you expect Intel won't price them into the floor if AMD becomes a problem? Intel's already got the business realm locked up for workstations and desktops. AMD will have to break in on tablets, netbooks, and laptops in higher price ranges than entry. That would require prestige to get the sales, something AMD is not going to build enough of to get Zen out the door in the quantities needed before Kaby Lake arrives. As long as Intel has near exclusive rights with Dell (the biggest buyer of CPUs in the world), I don't see AMD making the comeback it needs to. Data centers are looking more to Scale-Up architecture, something Intel's adapting to but AMD isn't. AMD's Zen APUs won't be ready until 2017 either, and no one's buying until they're proven. I don't see AMD making it through the end of the decade. That's not to say the products won't be any good. That's a separate conversation.

Apple. That is all.

They've switched to AMD for all their GPUs and it sounds extremely unlikely but I can see the Mac Pro getting the server class Zen processor and maybe the iMac getting a high performance APU. Apple will easily keep AMD afloat if any of this pans out.

[Curufinwe_wins]

Apple. That is all.

They've switched to AMD for all their GPUs and it sounds extremely unlikely but I can see the Mac Pro getting the server class Zen processor and maybe the iMac getting a high performance APU. Apple will easily keep AMD afloat if any of this pans out.

Ehh, unlikely. Apple isn't a large volume pc seller, and they NEVER pay close to full price on something they don't need to, which was literally exactly the reason FirePro cards are in the current lineup. The margins on selling to Apple are slim as fuck. (Just like AMD doesn't make much on console sales, because they sell near cost).

 

Kinda sad, but the way things go. Perhaps if Apple users started giving a shit about hardware, then the inclusion of AMD in Apple products would increase awareness for AMD and help their perception, but that's literally the only way Apple could make a difference for AMD long term.

[TechGod]

Ehh, unlikely. Apple isn't a large volume pc seller, and they NEVER pay close to full price on something they don't need to, which was literally exactly the reason FirePro cards are in the current lineup. The margins on selling to Apple are slim as fuck. (Just like AMD doesn't make much on console sales, because they sell near cost).

 

Kinda sad, but the way things go. Perhaps if Apple users started giving a shit about hardware, then the inclusion of AMD in Apple products would increase awareness for AMD and help their perception, but that's literally the only way Apple could make a difference for AMD long term.

Isn't a large volume PC seller? You kidding me right?

http://www.geekwire.com/2014/apple-now-one-top-5-pc-manufacturers/

 

And that was back in October 2014. Don't underestimate Apple's ability to save struggling companies. 

[Curufinwe_wins]

Isn't a large volume PC seller? You kidding me right?

http://www.geekwire.com/2014/apple-now-one-top-5-pc-manufacturers/

And that was back in October 2014. Don't underestimate Apple's ability to save struggling companies.

Apple is less than 8% of the oem market. And they don't buy at price. Apple doesn't save anyone, because no one but Intel and Nvidia (for pc parts not part of jedec that is) have the leverage to make Apple pay fair market value, and even then I'd be willing to be they paid less.

Saying Apple will save AMD is literally the same statement people made when they said the fact that consoles were all AMD meant AMD was saved. It wasn't true then, and it isn't true now. AMD won't make shit off of sales deals like this since it has literally no leverage.

That said there is always the potential that eventually enough products use AMD that the real money maker (enthusiasts and server space) starts to use AMD in a significant quantity again. But consoles and Apple products are not the best way to make positive awareness for AMD among the /pcmasterrace.

[GloriousPain]

Hope they name it Gemini Fury as a Gemini I'd find that bad ass, and I might sell my Fury X for one...

[patrickjp93]

Apple. That is all.

They've switched to AMD for all their GPUs and it sounds extremely unlikely but I can see the Mac Pro getting the server class Zen processor and maybe the iMac getting a high performance APU. Apple will easily keep AMD afloat if any of this pans out.

Since we all know for a fact AMD didn't win Apple over on efficiency and performance, the margins made from sales to Apple will be razor thin. It won't be nearly enough. Console SOCs make razor thin margins as well.

It's too late to get the next Mac Pro since that's already been contracted for Broadwell Xeons.