Vega/Navi Rumors (Updated)

[guskline]

AMD just needs a new high end gpu on the market.

 

[Veradun]

We already know vega performance.They showed it in SE4.It was slower than aftermarket 1080TI.SE4 is best case scenario where furyx is 22% faster than GTX1070.So in average it should be somewhere between GTX1080 and 1080TI.

This actually depends on how much Vega arch is similar to Fury's

 

[w3rd]

Remember when they showed Ryzen beating the 6900K side by side? Months before launch? That was pretty cool. Remember when they showed Vega beating the 1080 Ti side by side? I don't either.

Didn't AMD debut a Ryzen chip 3 months ago, running Battlefront on ES Vega at 4k..? And some other video, at 8k?
Pascal is not a 4k chip.

What don't you understand?

 

[Valantar]

We already know vega performance.They showed it in SE4.It was slower than aftermarket 1080TI.SE4 is best case scenario where furyx is 22% faster than GTX1070.So in average it should be somewhere between GTX1080 and 1080TI.

Edit:my tip is this

Based off of a single non-specific (as in: no recorded frame times, frame rates or other real data), non-comparable (SE4 is open enough to necessitate repeated runs through the same path for comparability) benchmark of a single title off a pre-production card with pre-production drivers? Yeah, that's a stretch. I'm not saying your estimation is necessarily off given the information that we have (although that depends heavily on whether or not AMD has balanced out the architecture compared to Fiji and Polaris), but that the information we have isn't nearly enough to make any valid assumptions.

 

[T1beriu]

What don't you understand?

I sense a disturbance in the force of your reading comprehension.

 

[Valantar]

Pascal is not a 4k chip.

You keep repeating this statement, without arguing for it, as if it's anything but the complete nonsense that it is. (First off, Pascal is an architecture, not a chip, but let's leave that dead and talk about GP102, which is the most relevant chip here.) The 1080Ti is a card with a cut-down GPU, not really pushed to its limits in terms of clocks (at least in FE form), yet it handles 4k60 on Ultra settings with aplomb, even with the stock blower cooler. It is - still - the most powerful consumer facing GPU out there. And sadly, we don't really have any data showing that Vega can clearly beat it. Do we have indications that Vega can match it? Arguably, yes. But beat it? No. Not that we've seen. If you have information to contradict this, show it to us.

Now, before you start accusing me of being and Nvidia shill (as you have before, repeatedly, despite the fact that I've never owned an Nvidia GPU in my life), please try to not somehow take this personally, and look at the data we have, the information we have, and what reasonable estimations can be made from these. What is making you so sure that Vega must beat GP102 across the board, when AMD hasn't shown us this? Previously, your argument has boiled down to "it's a newer architecture, so it must be faster," which isn't even remotely logical when comparing different architectures from different manufacturers. That's like arguing that VIA's newest octa-core CPUs must have better IPC than Skylake since they're newer. And again, I'm not saying that AMD's position or level of advancement in the GPU space is comparable to VIA's in the CPU space, but simply pointing out that there is no relation whatsoever - logical, causal, architectural - between Vega and Pascal performance outside of the fact that they're meant to compete and as such ought to be in the same ballpark. If you have any actual arguments that contradicts this, I would love to hear them. No doubt about it. But I have yet to see any.

You have repeatedly stated that you're a "realist". If so, please show us, either through data or argumentation, how Pascal is not a 4k chip, and how Vega will beat Pascal. Please.

 

[CatMerc]

We already know vega performance.They showed it in SE4.It was slower than aftermarket 1080TI.SE4 is best case scenario where furyx is 22% faster than GTX1070.So in average it should be somewhere between GTX1080 and 1080TI.

Edit:my tip is this

It is a best case scenario for Fury X. Clearly not all AMD cards benefit equally, seeing as 480 is neck in neck with 1060 in that title.

Claiming anything is best case scenario for an architecture that we know is a big change from current GCN iterations is silly as hell.

It's like claiming that a Kepler best case scenario game represnts a Maxwell best case scenario.

 

[w3rd]

Well, whatever Vega's performance is, place TWO of those on one card using fabric....

With that formula, it does little to fuss over exactly where single Vega falls, because the entire product sku is going to rock the industry. Vega X2 is real. The bigger question is, what is it's Price? (Because that is what will set the product stack).


LOL @ Valantar.
Yes pascal is an architecture, that given the constraints of GP100 (GP102), is not as powerful of a design as Vega. Not sure we needed a whole post of you trying to understand Pascal. We already know how well Titen Xp & Ti handle 4K. So we already know Pascal's gaming limit, and it doesn't handle 4k gaming with aplomb. Because if it did, nobody would be looking toward Vega or Volta.

Vega exceeds Pascal's limits, because of fabric. GPU don't care as much about clocks, as it does parallelism. Infinity fabric has much better use in GPU design, than in CPU. So, if you are excited about what ThreadRipper brings, why aren't you going crazy for TitanRipper..?

Is it because many of you can't multiply x2..?
Or, just have not admitted to yourself, (I am right) that Vega X2 is real. But just can't bring yourself around to connecting all the dots & admitting it.


Jump on me all you want. You are the ones hiding from reality.

 

[jpiniero]

AMD's tried X2 cards in the past, and they haven't worked.

What you think Vega has isn't coming until Navi.

 

[Mopetar]

I could see an x2 chip if Vega has good thermals. However it only makes sense to me if a chip with a defective HBCC can piggy back on one with a working HBCC.

It might be a way of getting some experience with this for Navi and for developers to have a card to play with and code against.

I don't expect it to be a consumer card, though some consumers may buy one.

 

[Mopetar]

We already know vega performance.They showed it in SE4.It was slower than aftermarket 1080TI.

That was the FE though, was it not? I'm not expecting RX Vega to be significantly better, but Raja did say it was faster for gaming than the FE card they showed that day.

Hell, just matching the Ti is a big accomplishment considering where they were at with Fury.

 

[Head1985]

It is a best case scenario for Fury X. Clearly not all AMD cards benefit equally, seeing as 480 is neck in neck with 1060 in that title.

Claiming anything is best case scenario for an architecture that we know is a big change from current GCN iterations is silly as hell.

It's like claiming that a Kepler best case scenario game represnts a Maxwell best case scenario.

rx480 is 17% faster than GTX1060 6GB its not "neck in neck"

 

[OatisCampbell]

We already know vega performance.They showed it in SE4.It was slower than aftermarket 1080TI.SE4 is best case scenario where furyx is 22% faster than GTX1070.So in average it should be somewhere between GTX1080 and 1080TI.

Edit:my tip is this

We haven't seen anything Vega RX, but I would not be surprised if you are correct based on what we have seen if

Well, whatever Vega's performance is, place TWO of those on one card using fabric....

With that formula, it does little to fuss over exactly where single Vega falls, because the entire product sku is going to rock the industry. Vega X2 is real. The bigger question is, what is it's Price? (Because that is what will set the product stack).


LOL @ Valantar.
Yes pascal is an architecture, that given the constraints of GP100 (GP102), is not as powerful of a design as Vega. Not sure we needed a whole post of you trying to understand Pascal. We already know how well Titen Xp & Ti handle 4K. So we already know Pascal's gaming limit, and it doesn't handle 4k gaming with aplomb. Because if it did, nobody would be looking toward Vega or Volta.

Vega exceeds Pascal's limits, because of fabric. GPU don't care as much about clocks, as it does parallelism. Infinity fabric has much better use in GPU design, than in CPU. So, if you are excited about what ThreadRipper brings, why aren't you going crazy for TitanRipper..?

Is it because many of you can't multiply x2..?
Or, just have not admitted to yourself, (I am right) that Vega X2 is real. But just can't bring yourself around to connecting all the dots & admitting it.


Jump on me all you want. You are the ones hiding from reality.

Could you please link me to something reliable that specifically says the infinity fabric means AMD can put two GPUs on a card and have them share memory and work load as if they were one GPU?
Link it and quote it please.

That would be revolutionary, the kind of "lightning strike invention" they need.

 

[OatisCampbell]

rx480 is 17% faster than GTX1060 6GB its not "neck in neck"

????

Where do you get that? Techpowerup review of 1080 Ti gaming shows 1060 a percent or two above RX480 at every resolution.

If you're talking some sort of mining or DX12 maybe?

 

[T1beriu]

He's talking about Elite Sniper 4. You even quoted his chart that shows it.

 

[guachi]

Remember when they showed Ryzen beating the 6900K side by side? Months before launch? That was pretty cool. Remember when they showed Vega beating the 1080 Ti side by side? I don't either.

This is the biggest factor that tells me Vega won't be faster than a 1080ti.

If it's priced competitively (and I assume it will be) and performs at 1080 levels I'll get one. Freesync makes AMD a cheaper choice than nVidia overall.

 

[lobz]

AMD's tried X2 cards in the past, and they haven't worked.

What you think Vega has isn't coming until Navi.

it's not comparable at all. Not saying it's the holy grail, but not you, nor I have any idea exactly how effective will IF on a dual GPU card be.

Sent from my VTR-L09 using Tapatalk

 

[CatMerc]

rx480 is 17% faster than GTX1060 6GB its not "neck in neck"

In the Guru3D result yes, in the TechPowerUp result no.

While I don't know the origin of the above discrepancy, if memory serves, most sites point to TechPowerUp's results rather than Guru3D's.

Well, whatever Vega's performance is, place TWO of those on one card using fabric....

With that formula, it does little to fuss over exactly where single Vega falls, because the entire product sku is going to rock the industry. Vega X2 is real. The bigger question is, what is it's Price? (Because that is what will set the product stack).
LOL @ Valantar.
Yes pascal is an architecture, that given the constraints of GP100 (GP102), is not as powerful of a design as Vega. Not sure we needed a whole post of you trying to understand Pascal. We already know how well Titen Xp & Ti handle 4K. So we already know Pascal's gaming limit, and it doesn't handle 4k gaming with aplomb. Because if it did, nobody would be looking toward Vega or Volta.

Vega exceeds Pascal's limits, because of fabric. GPU don't care as much about clocks, as it does parallelism. Infinity fabric has much better use in GPU design, than in CPU. So, if you are excited about what ThreadRipper brings, why aren't you going crazy for TitanRipper..?

Is it because many of you can't multiply x2..?
Or, just have not admitted to yourself, (I am right) that Vega X2 is real. But just can't bring yourself around to connecting all the dots & admitting it.


Jump on me all you want. You are the ones hiding from reality.

Alright, can we just STOP thinking of Infinity Fabric as literal magic? Its getting on my nerves. Yes Infinity Fabric is an important and very impressive piece of engineering from AMD, but it's not the silver bullet you claim it to be. The problems with current Crossfire/SLI do not get solved by putting two dies on an interposer and connecting them with infinity fabric. It's a coherency and control standard protocol for AMD's products, not the second coming of silicon jesus. It CAN help Crossfire due to being far lower latency than traditional Crossfire, but it's nowhere near fast enough to just make it behave as one piece of silicon.

 

[IllogicalGlory]

Well, whatever Vega's performance is, place TWO of those on one card using fabric....

With that formula, it does little to fuss over exactly where single Vega falls, because the entire product sku is going to rock the industry. Vega X2 is real. The bigger question is, what is it's Price? (Because that is what will set the product stack).


LOL @ Valantar.
Yes pascal is an architecture, that given the constraints of GP100 (GP102), is not as powerful of a design as Vega. Not sure we needed a whole post of you trying to understand Pascal. We already know how well Titen Xp & Ti handle 4K. So we already know Pascal's gaming limit, and it doesn't handle 4k gaming with aplomb. Because if it did, nobody would be looking toward Vega or Volta.

Vega exceeds Pascal's limits, because of fabric. GPU don't care as much about clocks, as it does parallelism. Infinity fabric has much better use in GPU design, than in CPU. So, if you are excited about what ThreadRipper brings, why aren't you going crazy for TitanRipper..?

Is it because many of you can't multiply x2..?
Or, just have not admitted to yourself, (I am right) that Vega X2 is real. But just can't bring yourself around to connecting all the dots & admitting it.


Jump on me all you want. You are the ones hiding from reality.

Vega x2 is two Vega chips on a PCB whose performance is limited by the multi-GPU capabilities of the game and drivers. No different from the R9 295X2 or Radeon Pro Duo. Connecting them with fabric is not coming until Navi in 2019, and even if it is ready now, I don't know how they'd fit two 500mm^2 dies and four stacks of HBM on an interposer.

Vega is not designed in any way for that kind of paradigm. A chip that's meant to be connected through fabric would more likely be somewhere in the range of 100-200mm^2, so that it could be used to fill out the entire product stack. See: Ryzen. Two Ryzen dies for HEDT, four for server. If Vega were a Polaris 10/11 replacement, your argument might hold some weight.

Vega is a monolithic die in the same vein as GP102, GM200, Fiji, etc. We will be seeing one of them on an interposer and later on two on a PCB, same as always.

 

[Mopetar]

Vega x2 is two Vega chips on a PCB whose performance is limited by the multi-GPU capabilities of the game and drivers.

This is likely, but we don't know for fact. AMD may have baked in some forward capabilities as a means of testing them out.

It's highly unlikely but they could make a huge interposer to fit 2 Vega die and 2 stacks of HBM2 which allows salvage of parts with defective memory controllers or possibly HBCC.

This is really unlikely, but seems like the kind of things you'd want to try if doing something like Navi. As the wisdom goes it's not a good idea dor a product to try to many new things at once.

 

[Oddzz]

Could you please link me to something reliable that specifically says the infinity fabric means AMD can put two GPUs on a card and have them share memory and work load as if they were one GPU?
Link it and quote it please.

That would be revolutionary, the kind of "lightning strike invention" they need.

Raja mentioned that this would be possible in his AMA on reddit: https://www.reddit.com/r/Amd/commen...n_technologies_group_at_amd_and_were/dhqo06p/
But so far there aren't any signs that we will see this soon.

 

[Harmaaviini]

This is the biggest factor that tells me Vega won't be faster than a 1080ti.

If it's priced competitively (and I assume it will be) and performs at 1080 levels I'll get one. Freesync makes AMD a cheaper choice than nVidia overall.

Keep in mind the complexity of driver development for a new GPU architecture. Maybe they just didn't have a finished driver to show off a competitive product in a straight up comparison like they did with their CPU months before launch.

 

[diaaablo]

Even if Vega will be slightly slower 1080ti, I will choose it. I tired to change gpu every year. AMD always behind in the beginning, but in final their architecture potential is always higher so ati top gpu's staying competitive for ~2years. Or Jensen H. just don't want to continue improve performance of previous gen. architecture for better sales whatever )

 

[OatisCampbell]

Raja mentioned that this would be possible in his AMA on reddit: https://www.reddit.com/r/Amd/commen...n_technologies_group_at_amd_and_were/dhqo06p/
But so far there aren't any signs that we will see this soon.

OK, I read that, but what about that says it will allow two GPUs to act as one? I just saw more low latency interconnection stuff, and isn't the whole problem with multi GPU dividing and re-compiling the work? Whether it's checkerboard, scissors, or AFR what's being rendered has to be split between GPUs and output reintegrated and paced smoothly- how can infinity fabric possibly balance work between GPU resources AND be as fast as single die? It's counter intuitive.

 

[Topweasel]

OK, I read that, but what about that says it will allow two GPUs to act as one? I just saw more low latency interconnection stuff, and isn't the whole problem with multi GPU dividing and re-compiling the work? Whether it's checkerboard, scissors, or AFR what's being rendered has to be split between GPUs and output reintegrated and paced smoothly- how can infinity fabric possibly balance work between GPU resources AND be as fast as single die? It's counter intuitive.

IF scales upwards of 500Gb/s per second. Infact it's main development seems to be primarily aimed at a internal GPU core interconnect and scaled down for CPU's. It is inevitable that it gets used for multi GPU setups much like ThreadRipper and EPYC.