Nvidia Reveals Next-Generation "Pascal" GPU for 2016

[RupertJr]

Looks similar to current embedded solutions available from AMD, a socketed embed solution gpu + ram.http://www.amd.com/us/products/embedded/graphics-processors/Pages/embedded-display-graphics.aspxhttp://www.amd.com/us/Documents/AMD-Radeon-E6760-Discrete-GPU-product-brief.pdfThe only difference and most innovative thing in my opinion is the ram. The rest seems like bluff.

 

[thor220]

It would be allot easier to convince mobo manufacturers to use your new proprietary NLink interface if say it was beneficial to them. In Nvidia's case, they expect the mobo manufacturers to hop for them on some niche product that will cost more to add onto motherboards. You can't have your cake and eat it too Nvidia. If you want mobos to use the new interface, you have to open it up to AMD and Intel. It's just not economical for them to accept it as a standard.

from what i understand this NVLink is for HPC and server stuff only although there are possibility for nvidia to incorporate this tech into their SLI technology. maybe they can make something similar to NF200 chipset.

NVLink was created chiefly for use in supercomputing clusters and other enterprise-class deployments where many GPUs may be installed into a single server. Interestingly, as part of today's announcements, IBM revealed that it will incorporate NVLink into future CPUs. We don't have any details yet about which CPUs or what proportion of the Power CPU lineup will use NVLink, though.

http://techreport.com/news/26226/nvidia-pascal-to-use-stacked-memory-proprietary-nvlink-interconnect

EDIT: from anandtech article it seems NVLink will ditch PCI/PCIE slot altogether and will have it's own connector. so i don't think we are going to see NVLink on regular dekstop mobo.

But the rabbit hole goes deeper. To pull off the kind of transfer rates NVIDIA wants to accomplish, the traditional PCI/PCIe style edge connector is no good; if nothing else the lengths that can be supported by such a fast bus are too short. So NVLink will be ditching the slot in favor of what NVIDIA is labeling a mezzanine connector, the type of connector typically used to sandwich multiple PCBs together (think GTX 295).

With all of that said, while NVIDIA has grand plans for NVLink, it’s also clear that PCIe isn’t going to be completely replaced anytime soon on a large scale. NVIDIA will still support PCIe – in fact the blocks can talk PCIe or NVLink – and even in NVLink setups there are certain command and control communiques that must be sent through PCIe rather than NVLink. The best case scenario for NVLink right now is that it takes hold in servers, while workstations and consumers would continue to use PCIe as they do today.

http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

I don't get why a company like Nvidia would want to create a new interface. They state that it will improve performance but they have no recent experience making interfaces (they stopped making chipsets awhile back) and they want to pigeon-hole it by forcing it to be nvidia only. Assuming that PCIe doesn't offer the same kind of performance at the time, what's going to encourage motherboard manufacturer's to create new designs, chipsets, and interfaces all to accommodate Nvidia? I would really like to know if anyone can guess.

 

[Mousemonkey]

It would be allot easier to convince mobo manufacturers to use your new proprietary NLink interface if say it was beneficial to them. In Nvidia's case, they expect the mobo manufacturers to hop for them on some niche product that will cost more to add onto motherboards. You can't have your cake and eat it too Nvidia. If you want mobos to use the new interface, you have to open it up to AMD and Intel. It's just not economical for them to accept it as a standard.

from what i understand this NVLink is for HPC and server stuff only although there are possibility for nvidia to incorporate this tech into their SLI technology. maybe they can make something similar to NF200 chipset.

NVLink was created chiefly for use in supercomputing clusters and other enterprise-class deployments where many GPUs may be installed into a single server. Interestingly, as part of today's announcements, IBM revealed that it will incorporate NVLink into future CPUs. We don't have any details yet about which CPUs or what proportion of the Power CPU lineup will use NVLink, though.

http://techreport.com/news/26226/nvidia-pascal-to-use-stacked-memory-proprietary-nvlink-interconnect

EDIT: from anandtech article it seems NVLink will ditch PCI/PCIE slot altogether and will have it's own connector. so i don't think we are going to see NVLink on regular dekstop mobo.

But the rabbit hole goes deeper. To pull off the kind of transfer rates NVIDIA wants to accomplish, the traditional PCI/PCIe style edge connector is no good; if nothing else the lengths that can be supported by such a fast bus are too short. So NVLink will be ditching the slot in favor of what NVIDIA is labeling a mezzanine connector, the type of connector typically used to sandwich multiple PCBs together (think GTX 295).

With all of that said, while NVIDIA has grand plans for NVLink, it’s also clear that PCIe isn’t going to be completely replaced anytime soon on a large scale. NVIDIA will still support PCIe – in fact the blocks can talk PCIe or NVLink – and even in NVLink setups there are certain command and control communiques that must be sent through PCIe rather than NVLink. The best case scenario for NVLink right now is that it takes hold in servers, while workstations and consumers would continue to use PCIe as they do today.

http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

I don't get why a company like Nvidia would want to create a new interface. They state that it will improve performance but they have no recent experience making interfaces (they stopped making chipsets awhile back) and they want to pigeon-hole it by forcing it to be nvidia only. Assuming that PCIe doesn't offer the same kind of performance at the time, what's going to encourage motherboard manufacturer's to create new designs, chipsets, and interfaces all to accommodate Nvidia? I would really like to know if anyone can guess.

Well it seems to be aimed at the server and workstation market and they already have a good foothold in that sector but I would be surprised if they launched this for the desktop sector.

 

[Dexter Kairaiuak]

Now that FB owns Oculus VR I would like nvidia to get into VR.

 

[Mousemonkey]

Now that FB owns Oculus VR I would like nvidia to get into VR.

Really?

It looks like she has a box of tissues strapped to her face! :lol:

Sorry but that's not for me so I'm going to hold out for the Holo-deck.

 

[renz496]

It would be allot easier to convince mobo manufacturers to use your new proprietary NLink interface if say it was beneficial to them. In Nvidia's case, they expect the mobo manufacturers to hop for them on some niche product that will cost more to add onto motherboards. You can't have your cake and eat it too Nvidia. If you want mobos to use the new interface, you have to open it up to AMD and Intel. It's just not economical for them to accept it as a standard.

from what i understand this NVLink is for HPC and server stuff only although there are possibility for nvidia to incorporate this tech into their SLI technology. maybe they can make something similar to NF200 chipset.

NVLink was created chiefly for use in supercomputing clusters and other enterprise-class deployments where many GPUs may be installed into a single server. Interestingly, as part of today's announcements, IBM revealed that it will incorporate NVLink into future CPUs. We don't have any details yet about which CPUs or what proportion of the Power CPU lineup will use NVLink, though.

http://techreport.com/news/26226/nvidia-pascal-to-use-stacked-memory-proprietary-nvlink-interconnect

EDIT: from anandtech article it seems NVLink will ditch PCI/PCIE slot altogether and will have it's own connector. so i don't think we are going to see NVLink on regular dekstop mobo.

But the rabbit hole goes deeper. To pull off the kind of transfer rates NVIDIA wants to accomplish, the traditional PCI/PCIe style edge connector is no good; if nothing else the lengths that can be supported by such a fast bus are too short. So NVLink will be ditching the slot in favor of what NVIDIA is labeling a mezzanine connector, the type of connector typically used to sandwich multiple PCBs together (think GTX 295).

With all of that said, while NVIDIA has grand plans for NVLink, it’s also clear that PCIe isn’t going to be completely replaced anytime soon on a large scale. NVIDIA will still support PCIe – in fact the blocks can talk PCIe or NVLink – and even in NVLink setups there are certain command and control communiques that must be sent through PCIe rather than NVLink. The best case scenario for NVLink right now is that it takes hold in servers, while workstations and consumers would continue to use PCIe as they do today.

http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

I don't get why a company like Nvidia would want to create a new interface. They state that it will improve performance but they have no recent experience making interfaces (they stopped making chipsets awhile back) and they want to pigeon-hole it by forcing it to be nvidia only. Assuming that PCIe doesn't offer the same kind of performance at the time, what's going to encourage motherboard manufacturer's to create new designs, chipsets, and interfaces all to accommodate Nvidia? I would really like to know if anyone can guess.

i've read a few comment from other tech site that this NVLink thing was supposed to solve the problem around PCI-E limitation that happen in supercomputer or server that take advantage GPGPU. for gaming aspect current PCI-E implementation did not cause the same problem that happen in supercomputer/server. so there is no need for nvidia to convince motherboard maker to adopt NVLink because the tech wasn't meant to be used with our consumer motherboard. for now IBM already announce their support for NVLink to be used with their Power CPU

 

[Beezy]

nothing new. we know there will be maxwell (in name only) since fermi.

True, Nvidia has always shown us two or three generations ahead on a map of the future. But I don't ever remember them giving this much technical detail before in a product two generations away from the current one.

planned obsolescence at its best. Weve really hit a wall it appears. Im not sure ANY upgrades will be "needed" or truly warranted for a while now (how else do they already know so much about the next next gen tech??? its because its ready, maybe not cheap enough yet, but it is a business afterall). the only push well see for new tech in the next couple years is developers making games run buggy for the old, and polished for the new tech, slamming us with a BS excuse, and effectively pushing us all against a wall to buy the new generation.

 

[thor220]

It would be allot easier to convince mobo manufacturers to use your new proprietary NLink interface if say it was beneficial to them. In Nvidia's case, they expect the mobo manufacturers to hop for them on some niche product that will cost more to add onto motherboards. You can't have your cake and eat it too Nvidia. If you want mobos to use the new interface, you have to open it up to AMD and Intel. It's just not economical for them to accept it as a standard.

from what i understand this NVLink is for HPC and server stuff only although there are possibility for nvidia to incorporate this tech into their SLI technology. maybe they can make something similar to NF200 chipset.

NVLink was created chiefly for use in supercomputing clusters and other enterprise-class deployments where many GPUs may be installed into a single server. Interestingly, as part of today's announcements, IBM revealed that it will incorporate NVLink into future CPUs. We don't have any details yet about which CPUs or what proportion of the Power CPU lineup will use NVLink, though.

http://techreport.com/news/26226/nvidia-pascal-to-use-stacked-memory-proprietary-nvlink-interconnect

EDIT: from anandtech article it seems NVLink will ditch PCI/PCIE slot altogether and will have it's own connector. so i don't think we are going to see NVLink on regular dekstop mobo.

But the rabbit hole goes deeper. To pull off the kind of transfer rates NVIDIA wants to accomplish, the traditional PCI/PCIe style edge connector is no good; if nothing else the lengths that can be supported by such a fast bus are too short. So NVLink will be ditching the slot in favor of what NVIDIA is labeling a mezzanine connector, the type of connector typically used to sandwich multiple PCBs together (think GTX 295).

With all of that said, while NVIDIA has grand plans for NVLink, it’s also clear that PCIe isn’t going to be completely replaced anytime soon on a large scale. NVIDIA will still support PCIe – in fact the blocks can talk PCIe or NVLink – and even in NVLink setups there are certain command and control communiques that must be sent through PCIe rather than NVLink. The best case scenario for NVLink right now is that it takes hold in servers, while workstations and consumers would continue to use PCIe as they do today.

http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

I don't get why a company like Nvidia would want to create a new interface. They state that it will improve performance but they have no recent experience making interfaces (they stopped making chipsets awhile back) and they want to pigeon-hole it by forcing it to be nvidia only. Assuming that PCIe doesn't offer the same kind of performance at the time, what's going to encourage motherboard manufacturer's to create new designs, chipsets, and interfaces all to accommodate Nvidia? I would really like to know if anyone can guess.

i've read a few comment from other tech site that this NVLink thing was supposed to solve the problem around PCI-E limitation that happen in supercomputer or server that take advantage GPGPU. for gaming aspect current PCI-E implementation did not cause the same problem that happen in supercomputer/server. so there is no need for nvidia to convince motherboard maker to adopt NVLink because the tech wasn't meant to be used with our consumer motherboard. for now IBM already announce their support for NVLink to be used with their Power CPU

That makes sense. It's just odd that they are announcing features like this without stating that they are for enterprise/supercomputers. Does this mean that Nvidia is focusing on GPGPU and exascale computing for pascal?

 

[somebodyspecial]

I don't understand the confusion...We already had Volta on the roadmaps, pascal is just volta's real name and it's been expected in 2016 for 2yrs already. This is no surprise or even news except how they're doing it. We already new Volta had 3d stacked memory etc ages ago, now they are outing it a LITTLE more. This is no different than the way they've done fermi, kepler, maxwell etc. Now that maxwell is upon us (28nm first, 20nm in Oct/Nov most likley for the real ones), you SHOW rather than talk about your next chips. Intel/AMD do the same thing. I'd expect an AMD announcement shortly unless they are just out of the game now due to funding and 290/290x horrible retail card launch. Clearly they didn't have enough to get a decent cooling solution launched with the hawaii cards (you don't launch a chip that can't be cooled adequately on purpose), so the question is, is that situation worse now with no Pascal/Volta answer or will we see something shortly from AMD to respond to all of NV's announcements?

 

[Mousemonkey]

I don't understand the confusion...We already had Volta on the roadmaps, pascal is just volta's real name and it's been expected in 2016 for 2yrs already. This is no surprise or even news except how they're doing it. We already new Volta had 3d stacked memory etc ages ago, now they are outing it a LITTLE more. This is no different than the way they've done fermi, kepler, maxwell etc. Now that maxwell is upon us (28nm first, 20nm in Oct/Nov most likley for the real ones), you SHOW rather than talk about your next chips. Intel/AMD do the same thing. I'd expect an AMD announcement shortly unless they are just out of the game now due to funding and 290/290x horrible retail card launch. Clearly they didn't have enough to get a decent cooling solution launched with the hawaii cards (you don't launch a chip that can't be cooled adequately on purpose), so the question is, is that situation worse now with no Pascal/Volta answer or will we see something shortly from AMD to respond to all of NV's announcements?

Here's a thought, do you think that Pascal is possibly just the workstation/server version of Volta?

 

[Christopher Shaffer]

Whats AMD's solution to this?

A card that is 5% weaker and half the price...

Actually, it's called HSA and hUMA: http://www.amd.com/us/products/technologies/hsa/Pages/hsa.aspxI'd like to see Nvidia incorporate this in a way that makes it viable on an x64 Intel based system.

 

[somebodyspecial]

Understand I say that because everything from AMD lately is beta. Mantle, Freesync, Gaming Evolved (the raptor software is far from NV quality still, most review sites commented on it's beta state), bad retail cooling, phase3 drivers, enduro problems, etc. I hope their next cooling solution isn't a beta too as mining won't save them next time and not even sure yet that it did this time since they gained no share and NV revenues on gpus were up in a down PC cycle.

 

[somebodyspecial]

Whats AMD's solution to this?

A card that is 5% weaker and half the price...

That isn't true and never has been. How's that working out on your $400/$550 290/290x? Have they been sold yet at this price? Not that I've seen. Cheapest 290 at newegg still $460 NOT 400. Maybe they'll be at MSRP by xmas? 😉 They are not 5% weaker either...ROFL. Not to mention nearly everything they have released in the last 2yrs is apparently beta or we wouldn't have phase3 drivers still coming, freesync vaporware, mantle in beta basically, raptor is beta IMHO, cooling that can't get the job done on retail cards etc etc...You get what you pay for huh? I say that as a very happy radeon 5850 owner... 😉

I'm sure AMD will launch cheap, but not sure one will ever be sold at what it launches at 😉 The only REAL price, is the one you have to pay when you buy it, all review sites should list THAT price when they review it, not MSRP which apparently at times is fake. I would have owned a LOT more AMD cards if what you said was EVER true...LOL.

 

[rwinches]

Stop with the mega quotes already!

 

[Mousemonkey]

Stop with the mega quotes already!

What's wrong with the quoting system?

 

[tomfreak]

may be should replace the SLI connector with NVlink, since Nvidia GPU can only SLI similar cards, means there is no worry about backward compatible

 

[alextheblue]

Have a 680 GTX, not interested in any video card until DirectX 12 comes out. Useless to spend money on top-end now when DirectX 12 is around the corner.

um dx 12 is coming to older generation cards that currently have dx 11

Although the current crop of cards may not support all of the DX12 features. However, it doesn't really matter much at this point, as DX12 isn't going to hit for a while - at least on the PC side. It's possible they'll push out an XBO-specific variant by the end of the year.

 

[Guest]

What happened to "Volta" Nvidia? Also considering Maxwell is getting downsized in terms of features I'm now actually happy that I got a 780Ti instead of waiting for the 800 series to come out.

 

[renz496]

What happened to "Volta" Nvidia? Also considering Maxwell is getting downsized in terms of features I'm now actually happy that I got a 780Ti instead of waiting for the 800 series to come out.

it is more like nvidia put pascal in between maxwell and volta. from what i heard nvidia did not cancel or rename volta to pascal instead they push volta so they can bring in pascal first. as for maxwell the only thing that is missing from last year roadmap was the unified memory stuff and instead will add new hardware for new feature in DX12

 

[renz496]

It would be allot easier to convince mobo manufacturers to use your new proprietary NLink interface if say it was beneficial to them. In Nvidia's case, they expect the mobo manufacturers to hop for them on some niche product that will cost more to add onto motherboards. You can't have your cake and eat it too Nvidia. If you want mobos to use the new interface, you have to open it up to AMD and Intel. It's just not economical for them to accept it as a standard.

from what i understand this NVLink is for HPC and server stuff only although there are possibility for nvidia to incorporate this tech into their SLI technology. maybe they can make something similar to NF200 chipset.

NVLink was created chiefly for use in supercomputing clusters and other enterprise-class deployments where many GPUs may be installed into a single server. Interestingly, as part of today's announcements, IBM revealed that it will incorporate NVLink into future CPUs. We don't have any details yet about which CPUs or what proportion of the Power CPU lineup will use NVLink, though.

http://techreport.com/news/26226/nvidia-pascal-to-use-stacked-memory-proprietary-nvlink-interconnect

EDIT: from anandtech article it seems NVLink will ditch PCI/PCIE slot altogether and will have it's own connector. so i don't think we are going to see NVLink on regular dekstop mobo.

But the rabbit hole goes deeper. To pull off the kind of transfer rates NVIDIA wants to accomplish, the traditional PCI/PCIe style edge connector is no good; if nothing else the lengths that can be supported by such a fast bus are too short. So NVLink will be ditching the slot in favor of what NVIDIA is labeling a mezzanine connector, the type of connector typically used to sandwich multiple PCBs together (think GTX 295).

With all of that said, while NVIDIA has grand plans for NVLink, it’s also clear that PCIe isn’t going to be completely replaced anytime soon on a large scale. NVIDIA will still support PCIe – in fact the blocks can talk PCIe or NVLink – and even in NVLink setups there are certain command and control communiques that must be sent through PCIe rather than NVLink. The best case scenario for NVLink right now is that it takes hold in servers, while workstations and consumers would continue to use PCIe as they do today.

http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

I don't get why a company like Nvidia would want to create a new interface. They state that it will improve performance but they have no recent experience making interfaces (they stopped making chipsets awhile back) and they want to pigeon-hole it by forcing it to be nvidia only. Assuming that PCIe doesn't offer the same kind of performance at the time, what's going to encourage motherboard manufacturer's to create new designs, chipsets, and interfaces all to accommodate Nvidia? I would really like to know if anyone can guess.

i've read a few comment from other tech site that this NVLink thing was supposed to solve the problem around PCI-E limitation that happen in supercomputer or server that take advantage GPGPU. for gaming aspect current PCI-E implementation did not cause the same problem that happen in supercomputer/server. so there is no need for nvidia to convince motherboard maker to adopt NVLink because the tech wasn't meant to be used with our consumer motherboard. for now IBM already announce their support for NVLink to be used with their Power CPU

That makes sense. It's just odd that they are announcing features like this without stating that they are for enterprise/supercomputers. Does this mean that Nvidia is focusing on GPGPU and exascale computing for pascal?

isn't that what nvidia try to do since they came up with CUDA back in 2006/2007? and they have been talking about exascale computing since fermi. and now with GPU has it place in HPC stuff they try to solve the problem that arise from using gpu is such system

 

[sarinaide]

Everyone is has plans about circumventing PCI bandwidth but nobody fully saturates PCI bandwidth. It however makes sense. AMD are interested in this for unleashing the full potency of an APU and Heterogeneous computing, ironically the exact same thing Nvidia want to achieve. I would love to see the day the APU's iGPU is fully extracted.

 

[BasedCereal]

Have a 680 GTX, not interested in any video card until DirectX 12 comes out. Useless to spend money on top-end now when DirectX 12 is around the corner.

Nvidia cards from 4xx and onward are compatible with DirectX 12, so what's your point?