Question 'Ampere'/Next-gen gaming uarch speculation thread

[Ottonomous]

How much is the Samsung 7nm EUV process expected to provide in terms of gains?
How will the RTX components be scaled/developed?
Any major architectural enhancements expected?
Will VRAM be bumped to 16/12/12 for the top three?
Will there be further fragmentation in the lineup? (Keeping turing at cheaper prices, while offering 'beefed up RTX' options at the top?)
Will the top card be capable of >4K60, at least 90?
Would Nvidia ever consider an HBM implementation in the gaming lineup?
Will Nvidia introduce new proprietary technologies again?

Sorry if imprudent/uncalled for, just interested in the forum member's thoughts.

 

[Glo.]

Wasn't Kopite adamant GA102 taped out in March? Is it even possible for Nvidia to have gotten silicon back so quickly?

Considering they will release RTX 3080 first, cut down chip, yes its possible they can rush this GPU.

Yields, you know?

 

[Tabalan]

What rumors? Any link?

There were some rumours about Ampere RTX 30 series that it will include DLSS 3.0. Making DLSS as common as possible is very reasonable, will boost Nvidia performance a lot. Tho, rumours are only rumours till RTX release, so don't take it for granted.

Nvidia GeForce RTX 3080 Ti specs and performance figures leaked, AMD "Big Navi" performance comparison also included

Thanks to the information compiled by Tom from Moore's Law Is Dead, we now have an almost clear picture of the RTX 3080 Ti, complete with specs and estimated performance gains over Turing GPUs. Furthermore, Tom reveals additional info on AMD's upcoming RDNA2-based "Big Navi" high-end GPUs, which...

www.notebookcheck.net

The GeForce RTX 3000 Would Come With An Improvement In The IPC Of Up To 20% And DLSS 3.0 - Cernisoft Gaming

New rumors are coming around Ampere and the Nvidia GeForce RTX 3000 Series, and there are more and more rumors that next week we will see a new high-end GPU

cernisoftgaming.com

 

[Glo.]

Fine, not your speculation, someone elses. The crux of my point left unanswered: what is below a 3080 102 that serves as the mid-range and not cost an arm and a leg?

It appears also its not Kopite's speculation .

Do you NEED an Nvidia card? There will be alternatives, you know, maybe even faster and more efficient, you know?

Well the plus of this situation is that Nvidia can maximize the SKU numbers from dies, so they can release for example 3070 Ti, based on 102 or 104 die, depending on the competitions performance. The same thing can happen with 3060 Ti SKU and it can be released using both: 104 or 106 die. ETC, etc, etc...

I don't believe we, consumers have to worry about anything. This generation upcoming might end up better from perf/dollar perspective than previous generation.

However, Highest end will not be cheaper than 1000 bucks...

 

[uzzi38]

Wasn't Kopite adamant GA102 taped out in March? Is it even possible for Nvidia to have gotten silicon back so quickly?

Taped out = come back from the fabs.

Unless you work in AMD/Intel/Nvidia itself as an engineer or high up in management, you likely don't even know when chips get sent over to the fabs, only when they come back.

 

[Glo.]

https://twitter.com/x/status/1270303185194770432

Also, kopite's latest info/speculation.

 

[uzzi38]

There were some rumours about Ampere RTX 30 series that it will include DLSS 3.0. Making DLSS as common as possible is very reasonable, will boost Nvidia performance a lot. Tho, rumours are only rumours till RTX release, so don't take it for granted.

Nvidia GeForce RTX 3080 Ti specs and performance figures leaked, AMD "Big Navi" performance comparison also included

Thanks to the information compiled by Tom from Moore's Law Is Dead, we now have an almost clear picture of the RTX 3080 Ti, complete with specs and estimated performance gains over Turing GPUs. Furthermore, Tom reveals additional info on AMD's upcoming RDNA2-based "Big Navi" high-end GPUs, which...

www.notebookcheck.net

The GeForce RTX 3000 Would Come With An Improvement In The IPC Of Up To 20% And DLSS 3.0 - Cernisoft Gaming

New rumors are coming around Ampere and the Nvidia GeForce RTX 3000 Series, and there are more and more rumors that next week we will see a new high-end GPU

cernisoftgaming.com

Sorry to say but MLID is spouting BS. He doesn't even believe the rumours himself.

I have a quote from another one of these types of Youtubers who I know is friends with MLID stating so.

 

[sontin]

It appears also its not Kopite's speculation .

Do you NEED an Nvidia card? There will be alternatives, you know, maybe even faster and more efficient, you know?

Intel, right?

 

[rainy]

Why is it so hard to believe, that RTX 3080 could use 300W of power, and RTX 3090 375W's, considering even the person who has got legit, correct info touts that 102 based GPUs draw anywhere between 300-375W of power?

For someone like sontin (some other guys too) who is worshiping Nvidia, it's simply unthinkable.


You can't call people worshippers of nvidia anymore than you can call them fanboys.

AT Moderator ElFenix

 

[FiendishMind]

Taped out = come back from the fabs.

Unless you work in AMD/Intel/Nvidia itself as an engineer or high up in management, you likely don't even know when chips get sent over to the fabs, only when they come back.

Has Kopite clarified that as their meaning? I mean those are two very different things. I know of other "tape out" leaks that were clearly leaks of the actual tape out, the original Ampere/GA100 tape out leak for instance.

Considering they will release RTX 3080 first, cut down chip, yes its possible they can rush this GPU.

Yields, you know?

That seems so fast though, I've always heard it's like six months from initial tape out to getting silicon back and then three to six more months before product launch.

 

[uzzi38]

Intel, right?

Pfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffft

 

[uzzi38]

Has Kopite clarified that as their meaning? I mean those are two very different things. I know of other "tape out" leaks that were clearly leaks of the actual tape out, the original Ampere/GA100 tape out leak for instance.


That seems so fast though, I've always heard it's like six months from initial tape out to getting silicon back and then three to six more months before product launch.

Like what? Tapeout always refers to getting chips back from the fabs.

 

[Glo.]

Im surprised also nobody posted this article from VCZ, based on Igor's info: https://videocardz.com/newz/nvidia-geforce-rtx-3080-cooler-could-cost-up-to-150-usd-alone

So RTX 3080 after all will use 300W of power?

 

[uzzi38]

Im surprised also nobody posted this article from VCZ, based on Igor's info: https://videocardz.com/newz/nvidia-geforce-rtx-3080-cooler-could-cost-up-to-150-usd-alone

So RTX 3080 after all will use 300W of power?

Weird, I thought I did.

 

[Glo.]

Nope, Gaming chips are 7nm not 8nm from Samsung.

https://twitter.com/x/status/1237687606869188608

You were saying? They are in fact on Samsung's 10 nm node that is called 8 nm.

 

[Glo.]

Weird, I thought I did.

Oops, sorry, I missed your post entirely .

 

[uzzi38]

Oops, sorry, I missed your post entirely .

No worries, I missed it as well when I went to go double check XD

 

[sontin]

GA100 doesnt use EUV. And Samsung supports EUV only with 7nm+.

Im surprised also nobody posted this article from VCZ, based on Igor's info: https://videocardz.com/newz/nvidia-geforce-rtx-3080-cooler-could-cost-up-to-150-usd-alone

So RTX 3080 after all will use 300W of power?

A little bit obsessed with power consumption, huh? RTX2080 Super FE has a 250W TDP.

 

[FiendishMind]

Like what? Tapeout always refers to getting chips back from the fabs.

Are you sure? I could be mistaken but I lurk Beyond3D and RealWorldTech forums a lot and I've seen "tape outs" explained numerous times, tape out leaks are commonly discussed and the discussions always to refer to tape outs and getting silicon back as very separate things.

 

[Ajay]

Oh hell, people will care. The last last bastion of rhetorhic against AMD GPUs will fall. Bashing AMD for their GPUs being inefficient, and buying Nvidia cards, despite AMD GPUs being better value, will make people only look like complete hypocrites if they will not bash Nvidia, for the same thing, and buying Nvidia GPUs if the only thing Nvidia can do to compete with AMD is to lower prices below AMD offerings.

Its the same dilemma between RX 570 vs GTX 1050 Ti all over again but in reverse.

Will people buy more RX 6500 XT's, than direct competitor from Nvidia, if it is more efficient, than GTX 3050 Ti, despite being slower and costing the same?

It's not the last bastion, there's still the driver issue. AMD has to hit the market with better day one drivers or people will be turned off (as happened with Navi 10).

 

[Ajay]

Taped out = come back from the fabs.

Unless you work in AMD/Intel/Nvidia itself as an engineer or high up in management, you likely don't even know when chips get sent over to the fabs, only when they come back.

Tape out means design implementation is sent to the fab. Mask are made etc. First silicon is when the company gets the first chips back from the factory.

 

[xpea]

Tape out means design implementation is sent to the fab. Mask are made etc. First silicon is when the company gets the first chips back from the factory.

^^This is correct ^^

 

[Hitman928]

^^This is correct ^^

x3.

Depending on how high priority you are and how busy the fab is, it's possible to get a small volume, rushed order of chips back from the fab within a couple of months so you can start testing. Then the rest of the order will come later during the normal delivery window.

 

[Tup3x]

Because the numbers, yes theoretical, show that Ampere is actually less efficient in FP32/64 per watt.

Also, so FP32 Load is not "full" load, now. Got it.

Also, its actually on you to prove that that 400W TDP rating is not done on maximum load of FP32 numbers... And that it is not full load.

Wait what... So you can pull numbers out of your hat but I can't? That isn't fair. In any case, it's meant to do totally different thing that gaming cards. Also gaming Ampere likely has different tensor cores too so their efficiency and power consumption is likely very different. If I'd had to guess a random number, I doubt that we'll see much difference in power consumption compared to Turing (vs. same price range cards). They'll push 'em further if they think that's what it takes... there's not much point at worrying about that.

According to rumours DLSS 3.0 is supposed to work with every game with TAA implemented. They'll just suggest using those games and Ampere GPUs will run circles around Turing/Pascal.

I don't see how that's possible...

 

[Glo.]

Wait what... So you can pull numbers out of your hat but I can't? That isn't fair. In any case, it's meant to do totally different thing that gaming cards. Also gaming Ampere likely has different tensor cores too so their efficiency and power consumption is likely very different. If I'd had to guess a random number, I doubt that we'll see much difference in power consumption compared to Turing (vs. same price range cards).

First, I haven't pulled those numbers out of thin air, every human being that has IQ level even 5 points above 90 pts is capable of calculating them.

Secondly: Other sources quote even worse numbers than I do, for RTX 3080 power consumption. https://videocardz.com/newz/nvidia-geforce-rtx-3080-cooler-could-cost-up-to-150-usd-alone

So deal with it my friend. Next gen Nvidia GPUs will not be as efficient as you, and others, hope and dream, but that was expected since they are made on 10 nm Samsung process.

 

[sontin]

So let me repeat: Ampere will bring less performance than Turing to the market while using a lot more power.

At the same time nVidia has released a GPU which is more than twice as efficient as Volta.

Makes sense. Cant argue against it.