AMD VEGA official details

[leadeater]

2 hours ago, TidaLWaveZ said:

Jet fuel doesn't burn hot enough to melt steel beams.

Funnily enough jet fuel doesn't naturally burn. You have to aerosolize it before it will ignite. You can test this easily by pouring some on the ground and try lighting it, it'll only burn where your flame is and will go out once you remove the ignition source. This is what makes jet fuel, and diesel, so safe. Petrol on the other hand does readily burn and is very unsafe. 

 

Now is it hot enough to melt steel beams, I have no idea what you are referring to .

[TidaLWaveZ]

1 minute ago, leadeater said:

Funnily enough jet fuel doesn't naturally burn. You have to aerosolize it before it will ignite. You can tested this easily by pouring some on the ground and try lighting it, it'll only burn where your flame is and will go out once you remove the ignition source. This is what makes jet fuel, and diesel, so safe. Petrol on the other hand does readily burn and is very unsafe. 

 

Now is it hot enough to melt steel beams, I have no idea what you are referring to .

My comment was mainly about the Flat Earth Theory.

[Silentprototipe]

Im just hoping that tech makes it to the next generation of APU's. I need a new one to replace my aging A6 3650 

[Prysin]

1 hour ago, Notional said:

WRONG! Fake news!

you are fake news. Very fake news notional.

[Prysin]

20 minutes ago, Humbug said:

source?

AMD presentations and past experience with AMD presentations, AMD hype and AMD marketing.

 

You want to bet if i'm right?

[Humbug]

7 minutes ago, Prysin said:

AMD presentations and past experience with AMD presentations, AMD hype and AMD marketing.

 

You want to bet if i'm right?

LOL no, didn't say you were wrong.

just asked for a source.

 

personally I have no idea, wouldn't be surprised either way.

 

I did notice that they have not talked about performance though. Still maintain that gtx 1080 performance will be underwhelming when you are this late to market.

[Vode]

3 hours ago, Humbug said:

1080 performance is a given...

The question mark is about scaling upto 1080ti and Titan XP performance levels.

 

if they stop at 1080 or slightly higher performance it can still be a good product and sell well. But it will be very underwhelming after all this time.

 

Yes, I agree. That's the interesting question.

 

If it were "GCN 5" or "GCN 1.4" I would expect GTX 1080 performance for the XT variant at 225W-250W power draw, which would be terrible honestly.

 

But since it's NCU (or supposedly a new architecture ( + HBM2 instead of GDDR5(X) I would at least expect 15% faster than GTX 1080 at ~225W power draw OR GTX 1080 performance at same or lower power consumption -> ~185W.

 

I think TDP, power consumption will be the determining factor if VEGA will be a success or not. Enthusiasts will roast AMD if they feel it's inferior technology to Nvidia again; efficiency will be a major factor in how we'll it can scale up to be closer to the GTX 1080Ti and not be overshadowed by the GTX 1080 like the Fury and Fury X were.

[juri-han]

2 hours ago, Prysin said:

Vega (pro) will be slightly faster than the 1070. Vega (xt) will be around the 1080, at the very best 10% faster 

 Vega at CES using alpha drivers and taped up vents with 10% above the 1080

 

honestly I'd be socked if it wasn't a month after launch its 20-25% above the 1080 and an stock 1080ti is probably about the same (that 35% was only with a aggressive OC) it will trade blows with the ti but what I'm interested to see is it's performance on ryzen vs the ti

[MoonSpot]

4 hours ago, Grinners said:

I think you underestimate the effect of performance on sales. 

 

If HBM2 outperforms gddr5x in overall fps (on an outright and a per cost basis) then the sales will instantly follow.

 

Numbers talk. 

Numbers do talk. Especially when a marketeer is behind the message, like 16GB of vram vs 8GB.  Nvidia doesn't have to flex any of their brand power muscle to make that pitch work.

 

Not saying everyone blindly nods at bigger numbers, just most people and by the time they realize it its too late.

[cj09beira]

3 hours ago, Prysin said:

Vega (pro) will be slightly faster than the 1070. Vega (xt) will be around the 1080, at the very best 10% faster 

they get that perf with a 1500mhz fury x (50% increase), and ipc has increased a bunch since then, plus improvements from vega + fp16 is going to be added to vulkan api which could also bring huge perf gains

[Jito463]

Well, I heard from a friend of a friend of an enemy of a friend of my father's, brothers, nephew's, cousin's former roommate, that Vega will have three times the power of 6 Titan Xp's in Sextuple SLI.

 

Truth.

[cj09beira]

58 minutes ago, Vode said:

 

Yes, I agree. That's the interesting question.

 

If it were "GCN 5" or "GCN 1.4" I would expect GTX 1080 performance for the XT variant at 225W-250W power draw, which would be terrible honestly.

 

But since it's NCU (or supposedly a new architecture ( + HBM2 instead of GDDR5(X) I would at least expect 15% faster than GTX 1080 at ~225W power draw OR GTX 1080 performance at same or lower power consumption -> ~185W.

 

I think TDP, power consumption will be the determining factor if VEGA will be a success or not. Enthusiasts will roast AMD if they feel it's inferior technology to Nvidia again; efficiency will be a major factor in how we'll it can scale up to be closer to the GTX 1080Ti and not be overshadowed by the GTX 1080 like the Fury and Fury X were.

Vega is a still GCN, the thing is that they have completely redone one of the most important units: the compute unit, its like how ryzen is still x86 while being a "new arch" the way the components work are still somewhat the same (some of the cache is now closer to the compute units) but the components themselves are different.

[Prysin]

34 minutes ago, cj09beira said:

they get that perf with a 1500mhz fury x (50% increase), and ipc has increased a bunch since then, plus improvements from vega + fp16 is going to be added to vulkan api which could also bring huge perf gains

1500MHz Fury X. So many of them around.

[Humbug]

1 hour ago, Vode said:

 

I think TDP, power consumption will be the determining factor if VEGA will be a success or not. Enthusiasts will roast AMD if they feel it's inferior technology to Nvidia again; efficiency will be a major factor in how we'll it can scale up to be closer to the GTX 1080Ti and not be overshadowed by the GTX 1080 like the Fury and Fury X were.

i really don't think power consumption will be the major factor in buyer decisions. Anyway so far all we know is that it will be more efficient than polaris.

[Organized]

37 minutes ago, Humbug said:

i really don't think power consumption will be the major factor in buyer decisions. Anyway so far all we know is that it will be more efficient than polaris.

Probably depends on where you are living and how much you have to pay for electricity. Since my roommate and me had a bill over 6.000kWh (because we wasted electricity like nobody else before, lol, just plug it in forever) I really have a look on the power consumption of things. Also, consumed power results in heat, and if you live under the roof of a house you REALLY want to produce as less heat as possible in summer time

 

But of course, if it just consumes a tiny bit more than its competitor, it doesn't matter. But more than 10% difference definately influences my decision whether to buy product A or B.

[cj09beira]

48 minutes ago, Prysin said:

1500MHz Fury X. So many of them around.

the point if a fury x core was at 1500 it would be at 1080 levels of perf which means that vega should be higher than that by a good margin

33 minutes ago, Humbug said:

i really don't think power consumption will be the major factor in buyer decisions. Anyway so far all we know is that it will be more efficient than polaris.

polaris was made for lower clocks which is why it consumes more in higher clocks, why they did that is beyond me (mobile/apple required a certain spec) 

[cj09beira]

2 minutes ago, Organized said:

Probably depends on where you are living. Since my roommate and me had a bill over 6.000kWh (because we wasted electricity like nobody else before, lol, just plug it in forever) I really have a look on the power consumption of things. Also, consumed power results in heat, and if you live under the roof of a house you REALLY want to produce as less heat as possible in summer time

my rx 480 helped me by not having to turn on the heating at night because i would leave it running folding@home :-)

[Vode]

1 hour ago, cj09beira said:

Vega is a still GCN, the thing is that they have completely redone one of the most important units: the compute unit, its like how ryzen is still x86 while being a "new arch" the way the components work are still somewhat the same (some of the cache is now closer to the compute units) but the components themselves are different.

We don't have enough details to say it's GCN. I'm purely going by what they have said so far. It looks like it's significantly different; like the step from Kepler to Maxwell was. That's why I wrote "supposedly".

[MoonSpot]

16 minutes ago, cj09beira said:

the point if a fury x core was at 1500 it would be at 1080 levels of perf which means that vega should be higher than that by a good margin

What are you basing that on?  Are you saying "vega should be higher" because it needs to be, in which case I'd be inclined to agree.  Or for some reason you "know" its going to clock that much higher?  In which case I'd want to know details.

[Trixanity]

1 minute ago, Vode said:

We don't have enough details to say it's GCN. I'm purely going by what they have said so far. It looks like it's significantly different; like the step from Kepler to Maxwell was. That's why I wrote "supposedly".

People have a tendency to think that Nvidia makes huge changes because they use different names for each architecture instead of doing it like AMD who is just calling it GCN + generation number. Nvidia has a similar iterative approach; their marketing department just spends more money on coming up with clever names.

 

With that being said. AMD has stated that Vega will be the biggest architectural change since GCN launched in 2011.

[Vode]

58 minutes ago, Humbug said:

i really don't think power consumption will be the major factor in buyer decisions. Anyway so far all we know is that it will be more efficient than polaris.

I think it is for enthusiast buyers. Vega will be $500+ and most reviewers focus so much on TDP, noise and power consumption these days. Enthusiasts, even the less informed ones still do research and will notice this kind of stuff and it influences their buying decisions. Overclocking is also indirectly correlated with power consumption how far AMD pushed the chip from the factory. Most Radeon carsd could be way more efficien at -200Mhz but the chips are clocked so high so they can compete on performance. Their main competitor has more R&D hence better refined Architecture that's doesn't need to be pushed so hard and can operate at the designs voltage sweet spot.

 

I could be wrong but that's just my best guess.

[Taf the Ghost]

2 minutes ago, Vode said:

We don't have enough details to say it's GCN. I'm purely going by what they have said so far. It looks like it's significantly different; like the step from Kepler to Maxwell was. That's why I wrote "supposedly".

AMD's roadmap for their APUs is going from 4 CPU cores + 4 GCN cores to 4 Zen cores + 11 Vega cores.  My thought is the Vega cores are going to be smaller relative to the previous GCN cores.  Unless the APU's entire package is going to be much bigger in their 2018 revision, though I'd be surprised.  We'll obviously find out more relatively soon.

 

Though there is one hilarious little thing: if Nvidia hasn't figured out the Ryzen + 1080(ti) issue by the time Vega drops, Vega could show >30% performance gains over the 1080 series on Ryzen.  That'd be funny to see.

[cj09beira]

26 minutes ago, MoonSpot said:

What are you basing that on?  Are you saying "vega should be higher" because it needs to be, in which case I'd be inclined to agree.  Or for some reason you "know" its going to clock that much higher?  In which case I'd want to know details.

i scaled the fury perf by 50% which is the diference in clock speed between vega and fury x (by the mi25 specs), it should be higher because vega has 2 gens of ipc improvement on top of those numbers: fiji to polaris and polaris to vega.

1525 mhz is the frequency needed to acheive the 25 tflops of fp16 in vega while using 4096 compute units.

[cj09beira]

36 minutes ago, Vode said:

We don't have enough details to say it's GCN. I'm purely going by what they have said so far. It looks like it's significantly different; like the step from Kepler to Maxwell was. That's why I wrote "supposedly".

its a big change yes but its still gcn there isn't any reason to believe it isn't.

[Mira Yurizaki]

On 4/23/2017 at 6:55 AM, Kilobytez95 said:

Won't be going Vega since I just bought a G-Sync monitor =(

Honestly variable refresh rates aren't even worth it. It's only useful if your frame rate is going below the refresh rate. With every gamer wanting to be at or above the refresh rate, both technologies are just going to sit there twiddling their thumbs.

 

Hell, I haven't really used G-Sync on my monitor. I use ULMB because I'm getting motion blur sensitivity and ULMB practically eliminates that. But the side effect is I can't use G-Sync.