What happens when NVLink launches?

[PlayStation 2]

yeah i know

 

i never said nvlink was going to take over in 2 years, thats ridiculous

 

but its likely that we will be using that (or something even better) in 10 or 20 years when PCIe does go obsolete

Yeah, I see where that comes from. PCI Express will be relevant for what I'll guess is another decade but yeah, we'll have something like NVLink around that time

[Enderman]

Yeah, I see where that comes from. PCI Express will be relevant for what I'll guess is another decade but yeah, we'll have something like NVLink around that time

that's my point

 

we cant just say "oh I dont want this to come to PC because I would have to change my motherboard and pay extra and blablabla"

things naturally go obsolete, and get replaced by better things

and by the time that happens, nvlink could be as cheap as PCIe to implement on consumer motherboards

[DXMember]

 

Ok so we should all stick to PCIe for the next 200 years of PC computing because that way people can use backwards compatible cards, right?
Let's just completely stop innovating and improving technology for the next few centuries.

 

Who told you that you need to pay $300 royalty fees and $200 more for a CPU?
Intel CPUs already cost over $200
Also you obviously don't know how gsync works, because the point of gsync is to NOT hit the maximum refresh rate of your monitor
If you're running games at 4k ultra, even 3 GPUs wont keep you over 60fps or 144fps 100% of the time
Frame drops and stuttering is what gsync reduces

 

lower latency, higher CPU-GPU communication, possibly even using higher speed ram for vram
its all about having options and alternatives in the future

 

I don't think you realize how many times motherboards have changed expansion slots...
remember PCI?
yeah, that old slot that is no longer on motherboards? yup, its becoming obsolete.
Of course if you want backwards compatibility there are still motherboards with BOTH TYPES OF SLOTS
like seriously, motherboards are already CPU specific
if you lived back when AMD and intel used the same socket you would have probably made the same argument
look how far we've come now
it seems like you're just afraid of change, like most people here

If the performance and latency is your concern then HSA is what you're looking for, not nvlink

 

 

that's my point

 

we cant just say "oh I dont want this to come to PC because I would have to change my motherboard and pay extra and blablabla"
things naturally go obsolete, and get replaced by better things
and by the time that happens, nvlink could be as cheap as PCIe to implement on consumer motherboards

like variable refresh rate is not cheap to implement, look at the numerous models that can be firmware updated to support dp 1.2a
doesn't stop nvidia charging you 200+$ premium on g-sync

[Enderman]

like variable refresh rate is not cheap to implement, look at the numerous models that can be firmware updated to support dp 1.2a

doesn't stop nvidia charging you 200+$ premium on g-sync

ok...?

gsync is still better than freesync, and the price premium is determined by the monitor manufacturer, not nvidia

the reason gsync monitors are more expensive is because the manufacturers like asus want a higher profit margin due to the supply-demand economics

 

you will see how 5 years from now gsync is dirt cheap and there are plenty of better technologies

[WhiteSkyMage]

yeah i know

i never said nvlink was going to take over in 2 years, thats ridiculous

but its likely that we will be using that (or something even better) in 10 or 20 years when PCIe does go obsolete

That is why I was saying - NVLink should only be implemented, as of now, to Tesla cards. At some point, when Nvidia start to see bottlenecks from the PCIe when testing their next gen GeForce cards, then they would go "Ok since PCIe X gen has not enough bandwidth to keep your next gen GPUs feeded with data, you should consider looking for NVLink support when buying a new motherboard to take advantage of the full potential of the next generation GPUs....and blablabla...we will support those CPUs ...." And they will partner with Intel for next gen CPU support for NVLink (who else when AMD are their competitors)...

I highly doubt Nvidia's next gen Pascal GPUs to require higher bandwidth than the PCIe x16 offers. So then why do we, customers, need NVLink right now? It is just going to be a waste...What I am afraid is that if they go even more aggressive and decide to have NVLink exclusive boards (No more PCIe). I know I sound really ridiculus here but who knows how they think...Of course AMD will try to counter this with something that's their's but can also be used by Nvidia. In other words, in the end of the day, you would really want to actually have 2 PCs, each for different brands of GPUs to get their full advantage...

[Enderman]

That is why I was saying - NVLink should only be implemented, as of now, to Tesla cards. At some point, when Nvidia start to see bottlenecks from the PCIe when testing their next gen GeForce cards, then they would go "Ok since PCIe X gen has not enough bandwidth to keep your next gen GPUs feeded with data, you should consider looking for NVLink support when buying a new motherboard to take advantage of the full potential of the next generation GPUs....and blablabla...we will support those CPUs ...." And they will partner with Intel for next gen CPU support for NVLink...

I highly doubt Nvidia's next gen Pascal GPUs to require higher bandwidth than the PCIe x16 offers. So then why do we, customers, need NVLink right now? What I am afraid is that if they go even more aggressive and decide to have NVLink exclusive boards (No more PCIe). Of course AMD will try to counter this with something that's their's but can also be used by Nvidia. In other words, in the end of the day, you would really want to actually have 2 PCs, each for different brands of GPUs to get their full advantage...

well to adopt a new technology you need to create support for it first

like putting USB3 ports before USB devices were mainstream

 

for example, no GPUs had HDMI 2.0 when 4k screens first came out, so people had to watch 4k at 30hz

that was because hdmi 2.0 support wasnt deployed before 4k came to market

 

if motherboards with PCIe+nvlink came out AFTER nvlink GPUs were on sale that would be stupid, because nobody would be able to use their GPU until the motherboard for it existed

 

look at sata express

motherboards are already being made with support for it even though there are 0 SSDs that can use it yet

[juretrn]

By the time NVLink becomes relevant for gaming cards, there will be new iterations of PCI-e or equivalent buses. NVLink is (IMO) unlikely to be seen in consumer boards.

[WhiteSkyMage]

By the time NVLink becomes relevant for gaming cards, there will be new iterations of PCI-e or equivalent buses. NVLink is (IMO) unlikely to be seen in consumer boards.

I'm just curious, why is it so heavily advertised when it's not going to be a consumer usable. It would go for GPGPU tasks yes, but then we just subtract this feature from what Pascal would offer since we are not going to use it. I was hoping for them to improve scalability of SLI with this but it is actually aimed for something completely different...

[juretrn]

I'm just curious, why is it so heavily advertised when it's not going to be a consumer usable. It would go for GPGPU tasks yes, but then we just subtract this feature from what Pascal would offer since we are not going to use it. I was hoping for them to improve scalability of SLI with this but it is actually aimed for something completely different...

I'm not really getting an impression NVLink was being advertised really aggressively. It's just a new feature, that deserves some mention. 

The hype is limited to these forums?

[WhiteSkyMage]

I'm not really getting an impression NVLink was being advertised really aggressively. It's just a new feature, that deserves some mention. 

The hype is limited to these forums?

Well I don't know what "aggressive" advertising means to you, but every time I type "Pascal GPU", Nvlink is the first thing they talk about along with HBM. 

 

This is why i was in a panic and decided to ask on this forum cuz for me, that IS aggressive advertising. I think they are trying to get more people to buy those Tesla GPUs...or is all of that targeting to university students and scientists?

[juretrn]

Well I don't know what "aggressive" advertising means to you, but every time I type "Pascal GPU", Nvlink is the first thing they talk about along with HBM. 

 

This is why i was in a panic and decided to ask on this forum cuz for me, that IS aggressive advertising. I think they are trying to get more people to buy those Tesla GPUs...or is all of that targeting to university students and scientists?

I don't think I can give you much more than a guess here, so I'll just say I really don't know. 

[Godlygamer23]

We don't have much information on NVLink, but it wouldn't make sense to abandon PCIe when most(if not all) boards won't support NVLink(if it completely changes the standard for video cards). Really, NVIDIA would lose a lot of money because most people can't just drop x amount of money on a CPU, motherboard, and graphics card at the same time. Their sales would be awful.

[MageTank]

 

 

I don't think you realize how many times motherboards have changed expansion slots...

remember PCI?

yeah, that old slot that is no longer on motherboards? yup, its becoming obsolete.

Of course if you want backwards compatibility there are still motherboards with BOTH TYPES OF SLOTS

like seriously, motherboards are already CPU specific

if you lived back when AMD and intel used the same socket you would have probably made the same argument

look how far we've come now

it seems like you're just afraid of change, like most people here

I've been around that long. But once again, you've completely ignored to read what i have said. Nowhere in my wall of text did i say i was afraid of change, nor am i worried about slots changing. I still have AGP boards and video cards old enough to occupy them. I have my fair share of long antiquated hardware. OP was the one worrying about it, as stated in my post. 

 

Be certain to have all of your information in order before jumping the gun. This marks the second time in this single thread that you've failed to follow the context in which people have been speaking. 

[WhiteSkyMage]

We don't have much information on NVLink, but it wouldn't make sense to abandon PCIe when most(if not all) will support NVLink(if it completely changes the standard for video cards). Really, NVIDIA would lose a lot of money because most people can't just drop x amount of money on a CPU, motherboard, and graphics card at the same time. Their sales would be awful.

 

That's the other thing I was worried for. It would be as failure as Mantle. I mean, Mantle wasn't a failure in a sense - it, in a way, "woke up" MS, but if AMD didn't do it, then we wouldn't have DX12. However in this situation of NVLink, we are taking a lot more money investing than for a new API (for mass production after all)... yes it doesn't make sense to leave out PCIe right now, but what is this aggressive advertising on NVLink? It's on every single article for about the Pascal generation ...Id agree for HBM, but that's just ridiculous...

[King_Tuna]

WTF is NVlink? Something like PCIe? Someone please help i dont know anything about NVlink.

[WhiteSkyMage]

WTF is NVlink? Something like PCIe? Someone please help i dont know anything about NVlink.

 

It's like PCIe but with 5 times higher bandwidth. Right now only super computers can benefit from it...you won't get much from it and we are just discussing on which sectors would Nvidia would implement it. But seems like the answer is clear that it would just be for scientific use...at least for this decade. That topic is really already answered...

[King_Tuna]

Even today's beefy cards can work on PCIe 2.0 without fps drops. So, who cares NVlink?

[Godlygamer23]

Even today's beefy cards can work on PCIe 2.0 without fps drops. So, who cares NVlink?

Getting the tech in place for future video cards.

[Glenwing]

Getting the tech in place for future video cards.

We have that. PCIe 3.0 Still waiting for things to make use of it.

It's only really been exceeded in the HPC market, which is where NVLink will step in.

[Stefan1024]

We have that. PCIe 3.0 Still waiting for things to make use of it.

That's very true. The high end SSDs have even surpassed the GPUs in PCI-e bus utillisation.

[Prysin]

NVlink is for servers and supercomputers. It is not for consumer desktops (unless your desktop is running 30+ GPUs)

[Godlygamer23]

We have that. PCIe 3.0 Still waiting for things to make use of it.

It's only really been exceeded in the HPC market, which is where NVLink will step in.

Perhaps, but NVIDIA may be thinking "hey, let's make sure we're ready for the next few generations." They might be thinking ahead to what they want to do with their products.