Vega/Navi Rumors (Updated)

[antihelten]

Async multi projection? No explanation so i take as similar nvidias tech.

Async reprojection, not multi projection, it's basically Valve's version of Oculus' async timewarp. AMD already supports async timewarp, so them adding support for async reprojection isn't really a big deal, and was probably just a way to show of their collaboration with Valve.

AMD does however also support simultaneous multi projection now, or as they call it multiview rendering (Bacon1 linked a video of it earlier).

 

[tamz_msc]

No tiled rendering?

I suspect this is the biggest thing that Nvidia has and AMD is missing, without that not sure they can catch Nvidia for efficiency.

Draw Stream Binning Rasterizer is what AMD calls it for Vega.

 

[Glo.]

I genuinely suggest looking at what Nvidia offered with GTX 1080 Ti, and asking yourself, why would they do this, when they have no competition?

Or do they...?

Fottemberg have said that Vega supposedly will have tremendous Performance/Dollar ratio.

 

[jpiniero]

I genuinely suggest looking at what Nvidia offered with GTX 1080 Ti, and asking yourself, why would they do this, when they have no competition?

You're right, but nVidia (unlike Intel) seems to be very proactive. Whatever AMD was planning on pricing the RX Vega is out the window now.

 

[blublub]

I genuinely suggest looking at what Nvidia offered with GTX 1080 Ti, and asking yourself, why would they do this, when they have no competition?

Or do they...?

Fottemberg have said that Vega supposedly will have tremendous Performance/Dollar ratio.

I agree. Nvidia would continue to milk the customer with at least 999$ for the TI if they knew there wouldn't be a competition.

We know that Vega was around 10% faster in the CES demo than 1080 so well let's say drivers and a little frequency boost until launch give another 10-15% then Vega will be too close to 1080TI for NVidia too charge 999$ if Vega will be priced between 400-600$.

So from a political POV NVidia cuts prices now so that they don't have to do it later because then it would be obviously because of Vega - now it's "free will"

 

[french toast]

I agree. Nvidia would continue to milk the customer with at least 999$ for the TI if they knew there wouldn't be a competition.

We know that Vega was around 10% faster in the CES demo than 1080 so well let's say drivers and a little frequency boost until launch give another 10-15% then Vega will be too close to 1080TI for NVidia too charge 999$ if Vega will be priced between 400-600$.

So from a political POV NVidia cuts prices now so that they don't have to do it later because then it would be obviously because of Vega - now it's "free will"

performance wise i agree, but 999$? I thought it was 699$? unless you titan x which is now a ripoff.

 

[tamz_msc]

There is a possibility that AMD might do a full product refresh with both Vega and Polaris parts in May. The partnership with Bethesda suggests to me that Prey will be a launch title when it releases on May 5th.

 

[zinfamous]

You're right, but nVidia (unlike Intel) seems to be very proactive. Whatever AMD was planning on pricing the RX Vega is out the window now.

Pretty much. 1080ti looks phenomenal for the listed price, which is drastically lower considering their recent, non-competitive history of "price gauging." This gives AMD very little room for pricing, but hey--you still look at the R&D budget AMD had to compete on two overwhelming fronts, and they seem poised for a heavy strike against desktop and server systems (CPU) this year (something that has never been on nVidia's radar), and is looking like they will be competitive here with Vega....but will likely lose significantly due to nVidia marketing.

AMD basically needs to retain that recent ~20% re-grab of market share in GPUs to not fail, and if they can grab an additional 5% on that, it can be a "win," albeit a very disappointing one. Still, it would be a huge 2017 overall for AMD, regardless. This is a far, far, far cry from the "AMD will be out of business" claims over the last 5 years. How soon everyone forgets.

 

[tential]

You're right, but nVidia (unlike Intel) seems to be very proactive. Whatever AMD was planning on pricing the RX Vega is out the window now.

So you're saying that the third year time in a row of high end ti cards would surprise amd?

Sorry but your theory is just not based in reality.

$650-700 price point for the ti card is nothing new.

780ti
980ti
1080ti now surprises you?

Please tell me how this price is a surprise...

 

[french toast]

You're right, but nVidia (unlike Intel) seems to be very proactive. Whatever AMD was planning on pricing the RX Vega is out the window now.

why?
If vega is 15% slower, offers a more modern future proof uarch and i would presume consume the same or a bit less power then why cant AMD charge 599$?

 

[Glo.]

Nobody considers a possibility that AMD can still have the faster and more power efficient GPU, at the same level of pricing, eh?

 

[french toast]

Nobody considers a possibility that AMD can still have the faster and more power efficient GPU, at the same level of pricing, eh?

Nope, but i would love to be proven wrong.

 

[Dribble]

Draw Stream Binning Rasterizer is what AMD calls it for Vega.

That doesn't sound like tiled rendering to me, which incidentally is an industry standard name not an Nvidia one for that technology as it's been in many brands of mobile and even some console gpu's since the late 90's. If Vega used tiled rendering then that is what they would call it.

 

[tamz_msc]

That doesn't sound like tiled rendering to me, which incidentally is an industry standard name not an Nvidia one for that technology as it's been in many brands of mobile and console gpu's since the late 90's. If Vega used tiled rendering then that is what they would call it.

It is the same thing, quoting AnandTech article.

For Vega, AMD will be doing something similar. The architecture will add support for what AMD calls the Draw Stream Binning Rasterizer, which true to its name, will give Vega the ability to bin polygons by tile. By doing so, AMD will cut down on the amount of memory accesses by working with smaller tiles that can stay-on chip. This will also allow AMD to do a better job of culling hidden pixels, keeping them from making it to the pixel shaders and consuming resources there.

As we have almost no detail on how AMD or NVIDIA are doing tiling and binning, it’s impossible to say with any degree of certainty just how close their implementations are, so I’ll refrain from any speculation on which might be better. But I’m not going to be too surprised if in the future we find out both implementations are quite similar. The important thing to take away from this right now is that AMD is following a very similar path to where we think NVIDIA captured some of their greatest efficiency gains on Maxwell, and that in turn bodes well for Vega.

 

[tential]

By the way is 8k vsr supported with Vega like 8k dsr is supported with Nvidia?

 

[sm625]

I genuinely suggest looking at what Nvidia offered with GTX 1080 Ti, and asking yourself, why would they do this, when they have no competition?

But what are they offering, really? 30% more TFLOPs than a 1080, for the same price, one year later. But this is what GPUs have been doing for years. 30% per year, more or less. Actually I think this is less than trend, so they look to be gouging even more. If they were actually facing better competition, I'd expect them to offer above-trend performance per dollar increases. And the 1070 pricing has barely moved after almost an entire year. The price cut is like 10%, so they're definitely gouging more in that bracket. If they were actually anticipating competition, we would be seeing 1070 reduced to $299 and 1060 reduced to $199. Even with those price cuts, the yoy GFLOPs per dollar growth rate would be rather unimpressive. Even Intel has been producing far better than 30% GFLOPs per dollar growth rates over the last several years.

Granted, those numbers are pure bs, but Intel has definitely been producing much more than 30% annual gains in raw GFLOPs per dollar or watt.

 

[Headfoot]

JHH isn't some kind of sales person or businessman or non-nerd. He's an electrical engineer and worked at AMD prior to starting NVidia. Just because someone is good at public speaking or likes presentations to be more professional doesn't make them non-nerdy.

He's a CEO, he's absolutely a businessman... a very fine and well respected one at that. In addition to his other experience

 

[Headfoot]

But what are they offering, really? 30% more TFLOPs than a 1080, for the same price, one year later. But this is what GPUs have been doing for years. 30% per year, more or less. Actually I think this is less than trend, so they look to be gouging even more. If they were actually facing better competition, I'd expect them to offer above-trend performance per dollar increases. And the 1070 pricing has barely moved after almost an entire year. The price cut is like 10%, so they're definitely gouging more in that bracket. If they were actually anticipating competition, we would be seeing 1070 reduced to $299 and 1060 reduced to $199. Even with those price cuts, the yoy GFLOPs per dollar growth rate would be rather unimpressive. Even Intel has been producing far better than 30% GFLOPs per dollar growth rates over the last several years.
Granted, those numbers are pure bs, but Intel has definitely been producing much more than 30% annual gains in raw GFLOPs per dollar or watt.

I dont think GFLOPS/TFLOPS are going to be a very accurate predictor here. It's never been a good predictor of performance between brands and architectures, but here we've got an architecture where they've specifically come out to say its got IPC/efficiency enhancements which would imply performance in real world scenarios will exceed what the raw TFLOP number would suggest. And it's known that making increases on the bottom end of the performance spectrum range is easier than making performance increases from the top, even in GPUs, so I dont think Intel's iGPU strategy is very illustrative here either.

 

[tential]

JHH isn't some kind of sales person or businessman or non-nerd. He's an electrical engineer and worked at AMD prior to starting NVidia. Just because someone is good at public speaking or likes presentations to be more professional doesn't make them non-nerdy.

Jhh isn't some kind of businessman? Get that out of here.
Look at Nvidia revenues, metrics, etc.
They are winning bigly in quite a tremendous way. You can't deny that it's irrefutable.

 

[Mopetar]

Jhh isn't some kind of businessman? Get that out of here.
Look at Nvidia revenues, metrics, etc.
They are winning bigly in quite a tremendous way. You can't deny that it's irrefutable.

You can't just dump out of product and expect people to buy it unless they have a compelling reason to do so. I can't just take a turd, buff it up a little, and expect to sell it on the side of the road for $5,000. No one is going to pay for it.

NV is in the place they are right now because they've spend years building a brand and much like Intel prior to Ryzen had a segment of the market all to themselves where there wasn't any direct competition. It will be only a year after the 1070 and 1080 launched that AMD will have any current generation competition. Why the hell would NV charge less than they can get away with when there's no one else with a competing product? You aren't going to go into work and tell your boss you'll take a pay cut out of the goodness of your heart, so why would you expect NVidia to behave differently?

If Nvidia is winning its because they've got a history of making good products and have built a good brand while continuing to engineer products that people want to buy. If they've got no direct competition in some high-value market segments where the margins are incredibly high, why would you expect them to be anything else?

I don't know how much JHH is directly involved in product design, but I wouldn't be surprised if he does more than just act as CEO. There are a few patents from as recently as 2013 that bear his name, so it's likely that he's still doing some work. Also, just because his company is doing well doesn't necessarily mean he's a great businessman either. He's got more a Steve Jobs vibe where he is a part of the brand and helps promote it, but it could be a case where there's a COO (someone like Cook) that's making a lot of the business decisions.

 

[darkswordsman17]

Why are people having senseless discussions about how much of a nerd or not JHH? Frankly I don't know why people are doing that at all, but definitely don't understand them doing it in this thread.

That doesn't sound like tiled rendering to me, which incidentally is an industry standard name not an Nvidia one for that technology as it's been in many brands of mobile and even some console gpu's since the late 90's. If Vega used tiled rendering then that is what they would call it.

Nvidia hasn't even admitted to using it have they? They didn't make a fuss about that when they originally made the switch and it took some in depth testing to show that was actually what they had done. Most likely I think it was due to their push into mobile where it's practically a requirement (and probably matters more on smaller GPU sizes that are limited).

Coming up with their own specific marketing buzzword/bulletpoint term is incredibly common. I wonder if some of it is specifically for legal reasons (i.e. if they use a common term that existed before, then these days that could open them up to patent infringement cases too often).

We'll see if it matters much when AMD implements it, and they might have held off because of other bottlenecks that would have reduced the effectiveness (much like how Fury cards' full utilization wasn't realized).

 

[Mopetar]

Looks like Vega will be shipping sometime in Q2: http://www.amd.com/en-us/press-releases/Pages/amd-vega-gpu-2017mar01.aspx

GPU products based on the Vega architecture are expected to ship in the second quarter of 2017.

 

[Dribble]

It is the same thing, quoting AnandTech article.

I stand corrected. Thanks.

 

[Erithan13]

Lurking in the Reddit AMA:

You will hear more from us on Ryzen 5, Ryzen 3 and Vega shortly. Thanks.

Think we can trust this source?

 

[Bacon1]

http://radeon.com/en-us/amdatprey/

Prey Releases May 5th