NVIDIA Maxwell Speculation Thread

[spworley]

The stuff about the interconnect is false.

It sounds odd to me too. "Interconnect" seems to have multiple definitions depending on what level of abstraction you're looking at, so I'd expect every complex chip routing data around would have interconnect of some form. (I'll be happy to learn more if there's some special definition for GPU interconnect.)

Jen-hsun complained about the difficulty of fabbing Fermi's interconnect.

 

[LordEC911]

Where?

Swore I saw a slide with that... Looking for it now can't find it.

Maybe it was in with the pre-release fake info.

 

[DSC]

http://blogs.nvidia.com/blog/2014/03/05/cuda-6-available/

https://developer.nvidia.com/cuda-pre-production

The following are known issues with the CUDA 6.0 Release Candidate that will be resolved in the production release:
‣ The minBlocksPerMultiprocessor parameter for the launch_bounds() qualifier only accepts values up to 16 when used in compiling for sm_50, even
though values up to 32 are possible on that architecture.
‣ There is a performance issue with the new SIMD video intrinsics __v*2() and __v*4() when used in compiling for the sm_50 architecture.
‣ The sm_50 architecture supports 48 KB of shared memory per block; however, the check for this limit is not functioning properly in the compiler. This can allow
programs that use more than 48 KB of shared memory per block to compile successfully, although they will fail to run because the driver component does check
the limit properly.
‣ The MT19937 random number generator in the cuRAND library generates non-deterministic results for curandGenerateUniformDouble().
‣ The NPP library function nppiAlphaComp_8u_AC4R() generates incorrect results when used with the NPPI_OP_ALPHA_ATOP_PREMUL option.
‣ The NPP library functions FilterSobelHorizSecondBorder() and FilterSobelVertSecondBorder() may generate incorrect results.

CUDA 6.0 RC now available for public download.

 

[UniversalTruth]

Swore I saw a slide with that... Looking for it now can't find it.

Maybe it was in with the pre-release fake info.

I remember the slide saying "Gen 1 Maxwell" which is on page 37 of this thread!
However, I think someone just made speculations about Gen 2 Maxwell and its timeframe...

If anything, those who post those slides can say more! For instance, DSC!

 

[xDxD]

DirectX 12, i hope 2nd generation Maxwell will support it videocardz.com/49835/microsoft-announce-directx-12-march-20th

 

[Wynix]

DirectX 12, i hope 2nd generation Maxwell will support it videocardz.com/49835/microsoft-announce-directx-12-march-20th

It may only require a software update to DX11.2 capable cards.

 

[xDxD]

It may only require a software update to DX11.2 capable cards.

Mmm...so in this case Kepler vgas would be not dx12 compliant, right?

 

[jlippo]

It may only require a software update to DX11.2 capable cards.

Really hope that DX12 will bring changes that are beyond software update.
Changes ROPs, TEX units, tesselator, perhaps new pipeline stages.

 

[DSC]

https://devtalk.nvidia.com/default/...hats-new-about-maxwell-/post/4145602/#4145602

Maxwell NVENC benchmarked, performance improvement is more than 1.5x - 2x claimed by Nvidia.

 

[homerdog]

Mmm...so in this case Kepler vgas would be not dx12 compliant, right?

Probably will be compliant save for the few 11.1 features that are not supported in Kepler or Maxwell. Remember, those architectures already support the new 11.2 feature set and the "optional" features in 11.1. 11.2 debuted on a Kepler card after all.

 

[DSC]

http://www.engadget.com/2014/03/09/msi-gaming-laptop-gs60-ghost-pro-nvidia/

 

[UniversalTruth]

http://www.engadget.com/2014/03/09/msi-gaming-laptop-gs60-ghost-pro-nvidia/

MSI have no clue that 15 inches is too small, especially for gaming.

and perhaps they are even more clueless why it has so big bezels, instead of bigger useful screen occupying the same dimensions!

 

[Wynix]

http://www.engadget.com/2014/03/09/msi-gaming-laptop-gs60-ghost-pro-nvidia/

Gaming laptop with 3k res? lol good luck trying to run games at native res.

 

[LiXiangyang]

However that rig will be perfect for a developer, 3K res and latest CPU/GPU archs.

 

[Mize]

MSI have no clue that 15 inches is too small, especially for gaming.

totally OT, but try gaming with a 15+ laptop in economy class...14" all the way with an external monitor back at the office.

 

[Blazkowicz]

15" 4:3 used to be enough, something you can live with, but 15" 16:9 is quite smaller, more like the area of 14" 4:3.
14" 16:9 is slightly smaller than 13.3" 4:3.

Times are rude
I would like "3K" to mean 3200x1800, too.

 

[UniversalTruth]

totally OT, but try gaming with a 15+ laptop in economy class...14" all the way with an external monitor back at the office.

I cannot and usually won't even try because of potential head-ache. This is the last place for gaming, isn't it?

 

[pjbliverpool]

Gaming laptop with 3k res? lol good luck trying to run games at native res.

But it would allow perfect 1:4 pixel scaling (so no scaling blur) at a resolution of 1440x810 which for this very small screen size and the GPU's available power would be near perfect.

 

[Mize]

I cannot and usually won't even try because of potential head-ache. This is the last place for gaming, isn't it?

Not if you're on a 14 hour flight. Gaming is the perfect way to pass those hours quickly. There have been years where flights like this were my main opportunity to game.

 

[DavidGraham]

But it would allow perfect 1:4 pixel scaling (so no scaling blur) at a resolution of 1440x810 which for this very small screen size and the GPU's available power would be near perfect.

I always thought that concept is quite fascinating, So let's say I have a 1440p panel, would gaming at 720 be the same thing as playing at native res? no loss of IQ whatsoever? not even a little bit?