AMD VEGA confirmed for 2017 H1

humbug · 23 May 2017 at 14:12

JediFragger said:
I think their were a few minute caveats from what I remember, but it was probably AMD just protecting themselves a little.

Yes, that none of it would be use to convert into blackbox IP to benefit any one only vendor.

A very sensible caveat, i think.

JediFragger · 23 May 2017 at 14:13

humbug said:
Yes, that none of it would be use to convert into blackbox IP to benefit any one only vendor.

A very sensible caveat, i think.

God yes!! Would be the first thing nVidia would do!!

humbug · 23 May 2017 at 14:14

Right

Rroff · 23 May 2017 at 14:35

spoffle said:
Oh come on. Even you aren't really this disingenuous. Mantle was handed off to the Khronos group, which became Vulkan. I think anyone that actually believes the point of Mantle was to be a vendor specific API is in denial. AMD are very specific about being open standard supportive. They don't do this because they're the good guys, they do it because they understand that it's in their best interests for their long term viability that all and any standards are open.

My point was about vendor specific features - whether AMD or nVidia there is a long history that they don't take off and take a lot of pushing and resources by the vendor to make any traction at all. PhysX is just as much an example of this.

Mantle is just another example - while it was AMD exclusive it went pretty much nowhere it was only when opened up to all that it picked up any wider adoption.

So these Vega features that sound great but increasingly its becoming apparent require specific work to function to their best don't really stand a chance on the balance of history.

harney · 23 May 2017 at 14:37

Rroff said:
Bursons are over-rated IMO - I found I prefer the AD8066 (though not sure it can be used in the STX as a straight replacement due to needing a gain of 5+) - not quite the "musicality" of the Burson but better micro-detail and imaging. The 2114 stock op amps I don't rate at all and the LM4562 while quite decent when used right in combination with other op amps just gets plain nasty at the slightest hint of power supply noise.

A lot can be redeemed on the STX just sticking a 10uf audio grade AE cap in the right place and enhancing decoupling with 0.1uf polypropylenes (ideally high speed WIMAs).

Thanks for the info Rroff

Rroff · 23 May 2017 at 14:42

harney said:
Thanks for the info Rroff

I have mixed feelings slating the Bursons as a lot of people genuinely enjoy how they sound but IMO they are over hyped because they are sometimes the first time someone experiences actual music focused audio hardware design and far from exclusive at delivering that.

The 2114 opamp is an interesting one - from an engineering perspective it does very well objectively on all the fancy graphs hence probably why it turns up in the O2 amp and is relatively inexpensive but anyone who discerningly listens to hardware using it as per your description quickly becomes disenchanted by the sound. Hence partly I suspect why they changed to Muses for the STX II.

Mtom · 23 May 2017 at 14:51

JediFragger said:
God yes!! Would be the first thing nVidia would do!!

Nah we all know now that they have the moral highground.

Oh an interesting snippet:

It seems Volta has a similar feature to HBCC, but intel seem to think you need an x86 license to use it. (GPU acces to system memory).
I wonder if it's true or not.

harney · 23 May 2017 at 15:17

Nexus18 said:

Love it

JediFragger · 23 May 2017 at 15:40

Mtom said:
Nah we all know now that they have the moral highground.

Most of nVidia's hierarchy would slip their Grandma's throat for a dollar

humbug · 23 May 2017 at 15:42

Mtom said:
Nah we all know now that they have the moral highground.

Oh an interesting snippet:

It seems Volta has a similar feature to HBCC, but intel seem to think you need an x86 license to use it. (GPU acces to system memory).
I wonder if it's true or not.

I'd be interested to know the guts of why Intel think this?

D.P. · 23 May 2017 at 16:13

JediFragger said:
God yes!! Would be the first thing nVidia would do!!

yeah, because they did that with OpenGL.

D.P. · 23 May 2017 at 16:24

humbug said:
I think he's right, again....

I don't think its a bad thing, AMD don't have a share among the top 2%, that all belongs to nVidia so what is the point?

As long as AMD make GTX 1080 to < GTX 1080TI level cards at a more competitive price its all good as far as i'm concerned, i would much rather AMD concentrate their efforts and R&D where the 95% buy their GPU's, you can Keep your £700 1080TI's and £1000 Titan Xp's, i'm not interested.

No complaints from me AMD, just do what you do well.

This has been said numerous times but its worth repeating for those with short term memory loss.

The point of a Halo product is far more important than the high profit margins. People wonder why AMD's market share is so low, well perhaps if they ha the same marketing devices and appeared on the top of every benchmark year after year things would be different.

Then there is the fact that it doesn't actually cost that much R&D to make the bigger halo products. A 1080ti is basically just a 1080 scaled by 50%. Moreover, any design requirements to get the halo products working effectively is R&D work saved for the future. Today's Tian Pascal will be tomorrows 2080 and th next days 3060 etc. It is the same reason why car manufactures can sell ridiculously priced luxury models filled with high tech. You can prototype and develop fancy technologies like ABS/traction control, rear view camera, lane assist, air bags in the luxory model where margins are high and volumes low. Everything you learn at the very high end is a useful lessons to bring to mainstream products in the future.

The bottomline line is if AMD can make a good competitor to the 1080 in turns of performance and efficiency then they should certainly be able to scale that up be a Titan competitor. If they can't it points to some fundamental design flaw in the architecture or horrific power efficiency.

Wrinkly · 23 May 2017 at 16:41

AMD Confirms 7nm Products Will Tape Out This Year – Zen 2 & Navi
http://wccftech.com/amd-confirms-7nm-products-will-tape-year-zen-2-navi/

Looks like AMD are continuing the push. I read within another article that they are also working on Vega 2 (14nm) to compete with Volta.

humbug · 23 May 2017 at 16:41

D.P. said:
This has been said numerous times but its worth repeating for those with short term memory loss.

The point of a Halo product is far more important than the high profit margins. People wonder why AMD's market share is so low, well perhaps if they ha the same marketing devices and appeared on the top of every benchmark year after year things would be different.

Then there is the fact that it doesn't actually cost that much R&D to make the bigger halo products. A 1080ti is basically just a 1080 scaled by 50%. Moreover, any design requirements to get the halo products working effectively is R&D work saved for the future. Today's Tian Pascal will be tomorrows 2080 and th next days 3060 etc. It is the same reason why car manufactures can sell ridiculously priced luxury models filled with high tech. You can prototype and develop fancy technologies like ABS/traction control, rear view camera, lane assist, air bags in the luxory model where margins are high and volumes low. Everything you learn at the very high end is a useful lessons to bring to mainstream products in the future.

The bottomline line is if AMD can make a good competitor to the 1080 in turns of performance and efficiency then they should certainly be able to scale that up be a Titan competitor. If they can't it points to some fundamental design flaw in the architecture or horrific power efficiency.

AMD did try this, it didn't workout too well for them, as a result they are no longer in a position to throwaway R&D for a "Halo Product" that they then cannot sell enough of.

AMD have had a higher return's with just lower and mid class GPU's than they have had in years, the years they had those Halo Products.

Most pepole, actually, do not look at nVidia's £1000 GPU and think "wow, what an awesome company" they actually look at that and think "wow... what a bunch of #### ######## #### money ###### garbing ###### ####!"

silversurfer · 23 May 2017 at 16:46

benjii said:
Rumour has it that at a JP Morgan conference, today, Lisa Su said RX Vega will launch a couple of months after Frontier. I guess we'll find out if that's true when the transcript from that conference gets uploaded.

Here's where the transcript/recording will be uploaded.

latter half of June. You will see the enthusiast gaming platform, the machine learning platform and the processional graphics platform very soon thereafter. And so, we will be launching Vega across all the market segments over the next couple of months. "

See this seems contradictory. Very soon after means weeks at most. But 2 months is not soon

Maybe they just said the order in which they'll be released, July 1st for rx vega confirmed

http://vocaroo.com/i/s0Qz6Mlo7wHn

It seems Volta has a similar feature to HBCC

When we see them copying major features we know its a good idea

LoadsaMoney · 23 May 2017 at 16:57

humbug said:
they actually look at that and think "wow... what a bunch of #### ######## #### money ###### garbing ###### ####!"

Before hitting the buy button on a one

Steampunk · 23 May 2017 at 16:59

Rroff said:
Your point? neither DX12 or Vulkan are vendor specific features.

Infact kind of proves my point - Mantle itself went pretty much nowhere while AMD specific and dropped with minimal support by AMD.

They were until both Microsoft and Nvidia were forced to support it because AMD didn't keep it to themselves to the detriment of the market and the gamers.

Steampunk · 23 May 2017 at 17:03

Wrinkly said:
AMD Confirms 7nm Products Will Tape Out This Year – Zen 2 & Navi
http://wccftech.com/amd-confirms-7nm-products-will-tape-year-zen-2-navi/

Looks like AMD are continuing the push. I read within another article that they are also working on Vega 2 (14nm) to compete with Volta.

1. Wccftech link, so probably total fantasy.

2. Tape out can easily be six months before a product launch even if there is only one revision. There's no way we are going to see 7nm Navi six months after Vega.

Journey · 23 May 2017 at 17:07

Steampunk said:
1. Wccftech link, so probably total fantasy.

2. Tape out can easily be six months before a product launch even if there is only one revision. There's no way we are going to see 7nm Navi six months after Vega.

I posted this last night in the Zen thread (CPU forum), source is Lisa Su - AMD CEO.

Oh, and yeah incase you are wondering, I watched the whole JP Morgan event live.

Rroff · 23 May 2017 at 17:10

Steampunk said:
They were until both Microsoft and Nvidia were forced to support it because AMD didn't keep it to themselves to the detriment of the market and the gamers.

Neither MS or nVidia were forced to support it though - despite the rhetoric used in support of Mantle developers were hardly jumping on it - there are some good points to Mantle but it is a far more complex case than is relevant to the point I was making - one potential case in contrast to a history that goes the other way on both sides.