Gemini (FuryX2) looms near

[Mondozei]

It's that that we're necessarily long away from 14/16 nm GPUs, but that both AMD/NV will continue the path that NV in particular have trailblazed since the slowing of Moore's law.

In other words, we should probably expect a more normal generational performance increase from the GP104 and it's AMD equivalent at GDDR5X, maybe 40% or so, and in such light, it makes sense to delay the release since whatever you can buy next year may not necessarily smoke the Gemini.

Honestly I never understood why some people thought we'd get the full fat archs from both companies right away. Have these people not been paying attention to these last few years? We'll see Kepler Redux. GP100 will come out right away because it has to compete with KNL in the HPC space, but other than that, both companies will likely space their releases out.

And this hasn't even touched on HBM2 yield issues.

Realizations like this are depressing. Much like discovering that, mathematically, the GTX 980 is the 560 of its architecture but costs twice as much.

In other news: Moore's law has not died and I am in denial.

I'm starting to laugh at this meme by now - and the people perpetuating it.

 

[maddie]

A Gemini release around the same time as hype for the Occulus Rift imminent release increases makes a lot of sense. A dual high end GPU will be the best option for a good VR experience.

Maybe AMD is hoping to capitalize on the VR angle to make significant sales this late.

 

[railven]

A Gemini release around the same time as hype for the Occulus Rift imminent release increases makes a lot of sense. A dual high end GPU will be the best option for a good VR experience.

Maybe AMD is hoping to capitalize on the VR angle to make significant sales this late.

If that's AMD's intention, I'm sure they'll miss it by a month.

 

[kawi6rr]

When AMD gets beaten by nVidia, just go for a dual GPU card that only guys with the right mainboard can go along with. OH and don't forget to buy a 1200 W PSU too!

AMD is pathetic at this point. No wonder the GTX 680 GTX to 980 TI went from 500€ to 700€. The last good AMD was the 7970 GHZ edition. The Fury X gets completely destroyed by the 980 TI.

A village lost it's idiot! I wonder who's alt account this is? One person comes to mind.

So AMD can't take the top card crown but they sure do have some great options below that.

Infraction issued for callout and personal attack.
-- stahlhart

 

[Vesku]

With the top end GPUs it's hardly ever "too late", these aren't high volume products. I don't expect FinFet GPU equivalents of 390/980 until May or later unless both AMD and Nvidia pull a GTX 680 and use smaller than expected chips for the price category. 980Ti/Fury equivalents by end of 2016, hopefully.

 

[railven]

With the top end GPUs it's hardly ever "too late", these aren't high volume products. I don't expect FinFet GPU equivalents of 390/980 until May or later unless both AMD and Nvidia pull a GTX 680 and use smaller than expected chips for the price category. 980Ti/Fury equivalents by end of 2016, hopefully.

If they get products at 20-25% faster than current 980 Ti/Fury X, I pretty much expect them to put them out as new "flag ship" cards. With their bigger robust brethren down the road as either "Ultras" or "flagships" for a second series.

Frankly, I'm surprised NV kept Maxwell 2 in one family (GTX 900). Were not for Fury, I wouldn't have been surprised if they used GM200 for GTX 1080.

 

[maddie]

If that's AMD's intention, I'm sure they'll miss it by a month.

I'm not sure what you mean. Can you explain with more detail.

 

[as9hW]

AIOs becoming much more affordable[/URL], many gamers would probably rather save $80-100 on the cooling solution and buy their own water blocks instead.

At least for AMD + Fiji, I think the interposer is a bit fragile to officially sanction end user modification. One scratch and no more very expensive GPU.

 

[Timmah!]

It's that that we're necessarily long away from 14/16 nm GPUs, but that both AMD/NV will continue the path that NV in particular have trailblazed since the slowing of Moore's law.

In other words, we should probably expect a more normal generational performance increase from the GP104 and it's AMD equivalent at GDDR5X, maybe 40% or so, and in such light, it makes sense to delay the release since whatever you can buy next year may not necessarily smoke the Gemini.

Honestly I never understood why some people thought we'd get the full fat archs from both companies right away. Have these people not been paying attention to these last few years? We'll see Kepler Redux. GP100 will come out right away because it has to compete with KNL in the HPC space, but other than that, both companies will likely space their releases out.

And this hasn't even touched on HBM2 yield issues.



In other news: Moore's law has not died and I am in denial.

I'm starting to laugh at this meme by now - and the people perpetuating it.

If they are going to have GP100 Teslas right away coming spring, but at the same time the top end product of the GTX series will be GP104 part, they are not going to get my money. For 800-1000 EUROs better be a proper flagship.

 

[crisium]

In other news: Moore's law has not died and I am in denial.

I'm starting to laugh at this meme by now - and the people perpetuating it.

You argue against both the mathematical configuration and the die size comparisons as multiple users already outlined to you. But if you want to base your position on laughter rather than facts, please continue. It's not like I claim there is no reason for the roughly doubling of GPU prices per tier from 5 years ago, I'm just lamenting the fact that they have doubled.

 

[Techhog]

It's that that we're necessarily long away from 14/16 nm GPUs, but that both AMD/NV will continue the path that NV in particular have trailblazed since the slowing of Moore's law.

In other words, we should probably expect a more normal generational performance increase from the GP104 and it's AMD equivalent at GDDR5X, maybe 40% or so, and in such light, it makes sense to delay the release since whatever you can buy next year may not necessarily smoke the Gemini.

Honestly I never understood why some people thought we'd get the full fat archs from both companies right away. Have these people not been paying attention to these last few years? We'll see Kepler Redux. GP100 will come out right away because it has to compete with KNL in the HPC space, but other than that, both companies will likely space their releases out.

And this hasn't even touched on HBM2 yield issues.



In other news: Moore's law has not died and I am in denial.

I'm starting to laugh at this meme by now - and the people perpetuating it.

Well, you got the denial part right. Unless you have an explanation for why the successor to GF100 and GF110 is somehow GK104, and how GK200 is suddenly a new tier that popped out of thin air. Also, what does that have to do with Moore's Law?

 

[Glo.]

There is a rumor that Apple may announce new iPhone 6C/updated 5S in january with availability later in February. Why I post it here? Because it may indicate there will be january event from Apple, where they can announce new Mac Pro, and latest indications in build of OS X Beta show it may be the case.

Device ID for Fury is already in the OS X. So Apple may need A LOT of Fury chips from AMD, and that may result in delay of Gemini chips.

 

[Timmah!]

There is a rumor that Apple may announce new iPhone 6C/updated 5S in january with availability later in February. Why I post it here? Because it may indicate there will be january event from Apple, where they can announce new Mac Pro, and latest indications in build of OS X Beta show it may be the case.

Device ID for Fury is already in the OS X. So Apple may need A LOT of Fury chips from AMD, and that may result in delay of Gemini chips.

As someone who owns iPhone 5S and would not want bigger phone, i hope that is true.

 

[RussianSensation]

At least for AMD + Fiji, I think the interposer is a bit fragile to officially sanction end user modification. One scratch and no more very expensive GPU.

I could see how the interposer may be fragile, but it sits 1 vertical level below the HBM + GPU die. Therefore, I do not see how it's possible to damage it with a water-block install, considering the vertical force on the interposer will first be absorbed/dissipated at the HBM+GPU level.

Maybe they are just not confident that most users could install their own cooling safely and then they'd have a hard time administering and creating warranty rules that are fair to AMD and the consumer simultaneously.

With the top end GPUs it's hardly ever "too late", these aren't high volume products. I don't expect FinFet GPU equivalents of 390/980 until May or later unless both AMD and Nvidia pull a GTX 680 and use smaller than expected chips for the price category. 980Ti/Fury equivalents by end of 2016, hopefully.

http://forums.anandtech.com/showthread.php?t=2456544

There is a rumor that Apple may announce new iPhone 6C/updated 5S in january with availability later in February. Why I post it here? Because it may indicate there will be january event from Apple, where they can announce new Mac Pro, and latest indications in build of OS X Beta show it may be the case.

Device ID for Fury is already in the OS X. So Apple may need A LOT of Fury chips from AMD, and that may result in delay of Gemini chips.

If NV doesn't get the Apple design win with GM200, given its 6-12GB of VRAM and superior perf/watt, that's as clear of a message as can be that Apple does not care about the CUDA eco-system enough since pushing/maintaining CUDA requires Apple to keep paying NV premiums that they could otherwise pocket as profits.

 

[Mondozei]

But how did the last Mac Pro really sell? Does anyone have any hard numbers? Are we really talking about earth-shattering numbers for the Fury here? (If we assume the new Mac Pro will sell around the same ballpark in the initial phase, say half-year or so, of selling).

 

[Silverforce11]

If NV doesn't get the Apple design win with GM200, given its 6-12GB of VRAM and superior perf/watt, that's as clear of a message as can be that Apple does not care about the CUDA eco-system enough since pushing/maintaining CUDA requires Apple to keep paying NV premiums that they could otherwise pocket as profits.

1. Nano is very power efficient. Even the normal Fury Air, Asus model, is very efficient. Per a few sites, matching the 980.

2. Apple is behind OpenCL, open source, where they can build upon it.

3. Metal is Mantle-derived just like Vulkan. It was trialed for iOS, deemed a success by Apple due to massive perf and perf/w gains and later plans to include into OSX itself.

What GPU uarch excel with OpenCL and Mantle?

 

[Pariah]

1. Nano is very power efficient. Even the normal Fury Air, Asus model, is very efficient. Per a few sites, matching the 980.

2. Apple is behind OpenCL, open source, where they can build upon it.

3. Metal is Mantle-derived just like Vulkan. It was trialed for iOS, deemed a success by Apple due to massive perf and perf/w gains and later plans to include into OSX itself.

What GPU uarch excel with OpenCL and Mantle?

Why would Apple care about Metal/Mantle performance for a new Mac Pro? That would be like Lenovo optimizing their workstations for DirectX performance. Graphics performance certainly matters for various industries, but not gaming optimized performance.

Not sure where you are getting that Metal is Mantle-derived. They aren't related at all. Metal is targetted for mobile platforms and was designed by Apple to replace OpenGL ES.

 

[Silverforce11]

Not sure where you are getting that Metal is Mantle-derived. They aren't related at all. Metal is targetted for mobile platforms and was designed by Apple to replace OpenGL ES.

It's well known Metal is Mantle derived. Maybe its time for you to catch up.

You are free to access the SIGGRAPH 2015 talks on the future APIs. Easy to read powerpoint slides on the topic.

 

[ShintaiDK]

It's well known Metal is Mantle derived. Maybe its time for you to catch up.

You are free to access the SIGGRAPH 2015 talks on the future APIs. Easy to read powerpoint slides on the topic.

Maybe easy to read, but obviously hard to understand. :thumbsdown:

I know you want to make AMD the center of the universe, but that isn't so.

 

[ShintaiDK]

I wonder if we could end up see AMD being 6-9 months behind NVidia this time, if NVidia comes out in the spring with GP100. The same time that AMD will be able to deliver the Fury X2.

 

[Glo.]

If NV doesn't get the Apple design win with GM200, given its 6-12GB of VRAM and superior perf/watt, that's as clear of a message as can be that Apple does not care about the CUDA eco-system enough since pushing/maintaining CUDA requires Apple to keep paying NV premiums that they could otherwise pocket as profits.

In sheer compute horsepower, Nvidia is nowhere near AMD Fiji.

1. Nano is very power efficient. Even the normal Fury Air, Asus model, is very efficient. Per a few sites, matching the 980.

2. Apple is behind OpenCL, open source, where they can build upon it.

3. Metal is Mantle-derived just like Vulkan. It was trialed for iOS, deemed a success by Apple due to massive perf and perf/w gains and later plans to include into OSX itself.

What GPU uarch excel with OpenCL and Mantle?

Why would Apple care about Metal/Mantle performance for a new Mac Pro? That would be like Lenovo optimizing their workstations for DirectX performance. Graphics performance certainly matters for various industries, but not gaming optimized performance.

Not sure where you are getting that Metal is Mantle-derived. They aren't related at all. Metal is targetted for mobile platforms and was designed by Apple to replace OpenGL ES.

Apple is going away from both OpenGL and OpenCL. Metal will be one API for both gaming, and hard work. You don't know that, but lately there has been few signs that Apple will make Metal CUDA-like solution with DirectX in one API. Its hard to explain so far why. MacRumors users already are discussing it in few threads, however future looks like Apple may want to push Metal and Swift - programming language - as hard as they can, rather than OpenCL. They already made Swift Open Source.

One more thing: Nvidia lately has announced on linked they are looking for people with experience in Mac development, to help Nvidia to make something for future Apple projects connected to Metal API. Coincidence? Apple wants to have propertiary solution, that is good for their ecosystem, but is hardware agnostic. The best(read: the highest discount on hardware) hardware in Metal performance - contract win.

And yes, Metal is Mantle based, just like every other API there is right now.

 

[Pariah]

And yes, Metal is Mantle based, just like every other API there is right now.

No, they aren't. Vulkan is. Being designed to serve the same function, a low level graphics API in this case, does not mean they are all derived from one source. AMD didn't invent the low level API.

 

[poofyhairguy]

What would Metal be based on Mantle when it was rolled out first for iOS devices with PowerVR GPUs?

 

[3DVagabond]

What would Metal be based on Mantle when it was rolled out first for iOS devices with PowerVR GPUs?

AMD has always said that Mantle wasn't designed to be specific to any arch. All anyone needed to do is write a driver.

 

[RussianSensation]

I wonder if we could end up see AMD being 6-9 months behind NVidia this time, if NVidia comes out in the spring with GP100. The same time that AMD will be able to deliver the Fury X2.

Why would NV release a card 70-90% faster than 980Ti right away when they can repeat Kepler/Maxwell roll-out? They know NV's loyalists will go buy another mid-range $550-600 GP204, so would NV be interested in launching their flagship in the spring? GP100 may come out as a Tesla/Quadro card and cost $3000+. I could see that. Another alternative could be a heavily cut-down GP100 aka GTX780 that NV can milk for 12 months and then the real flagship comes in 2017.

The 6-9 months timeframe is also something you just made up considering NV would have to launch January 2016 for that to occur based on existing rumors for AI. It's already December 2015 and there are no credible rumours about a Pascal launch in January or even February 2016. It sounds more reasonable that both NV/AMD won't have anything worthwhile to buy on 14nm/16nm until Q2 2016 at the earliest.

"Taiwan Semiconductor Manufacturing Co. has successfully produced the first samples of Nvidia Corp.’s code-named GP100 graphics processing unit. Nvidia has already started to test the chip internally and should be on-track to release the GPU commercially in mid-2016." ~ Kit-Guru

But how did the last Mac Pro really sell? Does anyone have any hard numbers? Are we really talking about earth-shattering numbers for the Fury here? (If we assume the new Mac Pro will sell around the same ballpark in the initial phase, say half-year or so, of selling).

Getting the Mac Pro design isn't about making a lot of $ or market share. It's about mind-share for AMD, and for Apple/AMD it's also about pushing OpenCL and moving away from proprietary closed standards of CUDA.

If Apple chooses AMD for the next Mac Pro, it also reinforces the point that the sales and profits were meeting Apple's expectations. If the current Mac Pro sold poorly as a result of having an AMD GPU per the customer feedback, then surely Apple would do everything possible to switch back to CUDA/NV.

In sheer compute horsepower, Nvidia is nowhere near AMD Fiji.

Even if true, NV's push with CUDA starting with G80 generation in 2008, and NV's dominance in the GPGPU market with supercomputer design wins suggests they strategically locked most of the market into their eco-system. First mover advantage and no competitor from Intel/AMD or anyone ensured that NV built the GPGPU compute eco-system uncontested. As a result, many popular compute apps/programs are still using CUDA. It will take time for the most popular programs to be converted to OpenCL.

If OpenCL could take full advantage of AMD's shaders, then sure, it would dominate NV in performance and perf/watt for compute. In the real world, you still need the software infrastructure to take advantage of AMD's compute capabilities.

The other issue is Fury is 4GB limited but the current Mac has 6GB option. The current Mac Pro is also limited to HDMI1.4a. Surely if Apple is updating it, they will want HDMI 2.0 capability and Fiji doesn't provide that in current form.

I think it makes more sense for Apple to update the Mac Pro in Mid-2016 when they can use Broadwell-E Xeons + next gen 8-12GB GPUs.