AMD VEGA confirmed for 2017 H1

nashathedog · 9 Feb 2017 at 09:56

JediFragger;30485843 said:
Big Vega will have 6144 shaders yes? I'm expecting it to be a bit of a beast tbh and eclipse the TXP (at least in DX12 titles)

If it has 16GB of HBM2 I can see it being approx £1k. I'd be happy-ish with that, esp considering it'd last me a goooooooooooood long while (3840x1600 res)

I don't want to spend that much on a graphics card but if that's what it costs I might as I believe it'll age better than the equivalents and if I do spend that much I'll want a card that will last for the next 5 years still giving acceptable performance

/\ndy · 9 Feb 2017 at 09:57

drunkenmaster;30484629 said:
We're on a forum so no one is spending half an hour proof reading every post, it's not a formal presentation. In fact the only time anyone cares about grammar/typos on a forum... is when they are completely and utterly wrong and can't respond to the content of someone's reply to them. Also, the first thing in IT isn't learning to proof read, I honestly can't even wrap my head around you stating that, it's not a good comeback, it's not true, it doesn't mean anything at all. If I wrote a technical manual and my boss found a mistake he might say hey, your job is writing tech manuals, typos aren't acceptable..... but then every manual has mistakes regardless so meh, just try not to make any more.

You won't let a typo go, but you're ignoring that you tried to claim that NVlink bridge maybe used on cards that don't support NVlink. It's pretty obvious you won't discuss any of my points so why do you bother coming on here.

Gaming, again, and I quoted you before for easy access, you said the quadro had the same number of active cores as a Titan, not the same number of FP32 cores OR the same gaming performance, there was no quantifier, you stated they have the same number of active cores, compared to either GP100 or P6000(because you said quadro, not a specific quadro) neither has the same number of active cores, so you are wrong on this, flat out.

Second, I pointed out that the same number of FP32 cores doesn't mean the same gaming performance, GP100 has the same number of FP32 cores, but an extra 1792 power leaking FP64 bit cores, meaning in any relatively similar power bracket, there is no way GP100 can run at the same clock speeds as a GP102, so it won't match it's performance, ever, so GP100 has zero relevance as a gaming card. GP102 also has a further what 200+ FP32 cores that can be used for a gaming card without the downside of the Fp64 cores to take up a large chunk of the TDP.

SO your post had no merit, Nvlink is only available on GP100, so any idea of NVlink helping gaming could only come on GP100 and GP100 has no relevance to gaming, less performance for gaming than GP102.... but significantly higher price... wooo.

TTTRRRRRIIIGGGERRRED

Freddie1980 · 9 Feb 2017 at 10:24

Kaapstad;30484441 said:
And you still have not conceded that your addition is terrible.

LOL

As if anyone on these forums would ever have the strength in character and swallow their pride and admit to making a fool of themselves in the middle of fanboy war. We've all been there at some point. At most what you will get is the other person will realise their wrong and rather then continue the debate just go away and leave the thread until the dust settles.

Sargatanas2511 · 9 Feb 2017 at 10:25

Rroff;30485831 said:
Just incrementally increasing shader count doesn't necessarily equate to anything like a linear increase in performance - infact you can end up with quite a lot of utilisation problems depending on the architecture.

I understand that, but being that they have already tested the waters with Fury X for 4096 shaders and the architecture was similar, I don't think it would be too unrealistic to assume it would be quite close to the 1080.

Now factor in (probably) higher clock speeds and many design changes and optimisations. I would be extremely disappointed if it wasn't beating a 1080 comfortably personally.

Gregster · 9 Feb 2017 at 10:52

Besty;30485646 said:
i would expect that the chip that well held up and shown to the press was big vega and the card running the doom demo was the same chip. I expect that, more than anything it was to re-assure investors rather than gamers that AMD had a working prototype and not just a chip

Yer, Raja did say in an interview that it was big Vega. They will be streamlining as well, so I expect more performance than what that Doom demo showed.

Freddie1980;30486029 said:
LOL

As if anyone on these forums would ever have the strength in character and swallow their pride and admit to making a fool of themselves in the middle of fanboy war. We've all been there at some point. At most what you will get is the other person will realise their wrong and rather then continue the debate just go away and leave the thread until the dust settles.

That happens a lot here unfortunately. I am one of the first to hold my hands up when I am wrong but others will argue black and blue or like you say, vanish til the dust settles. I never get it and just say "Yer, I was wrong" Nothing hard

TheRealDeal · 9 Feb 2017 at 11:10

What was Raja's exact words. Did he say this is Big Vega or did he say Vega 10? From what i have heard the number only depicts what is coming first i.e. Polaris 10 Rx480/70 came before the Polaris 11 Rx460. So if Raja said it was Vega 10 then it would still not tell us what chip it was. It is pretty common up to this point for AMD to release the big chip first as it was with Nvidia up until the gtx680.

If he did say the big chip then that's enough proof.

Gregster · 9 Feb 2017 at 11:48

Well he did say Vega 10, so I take it that is the big chip.

Lokken86 · 9 Feb 2017 at 11:49

JediFragger;30485843 said:
Big Vega will have 6144 shaders yes? I'm expecting it to be a bit of a beast tbh and eclipse the TXP (at least in DX12 titles)

If it has 16GB of HBM2 I can see it being approx £1k. I'd be happy-ish with that, esp considering it'd last me a goooooooooooood long while (3840x1600 res)

Another £1000 card please don't AMD. I find it difficult to justify £500 as it is.

Gregster · 9 Feb 2017 at 11:55

Another thing I remember is Raja talked about cutting the price of 4K - $2000 gaming machines to half ($1000) and with Ryzen and Vega being mentioned in the same sentence. This might be wishful thinking but hopefully not.

TheRealDeal · 9 Feb 2017 at 11:57

Gregster;30486354 said:
Well he did say Vega 10, so I take it that is the big chip.

I thought it worked that way as well but when i read what the number meant in there internal naming, it only means the chip that's being developed first.

Okay, let's understand the naming of polaris/vega 10/11.
the use of 10/11 is NOT related to performance, unlike countrary belief. It is related to when it was developed (or development started), with the lower number being the first. So this means polaris 10 was started before polaris 11. And Vega 10 was started before the rumored vega 11. Furthermore they are NOT product names, but rather development names that the press took to heart, and that caused confusion.
Again, they are internal naming used by the company. They will NOT be used on the actual products (okay, the press probably will at first)

So if the small chip was in development first it would be Vega 10.

Lokken86 · 9 Feb 2017 at 12:03

Gregster;30486390 said:
Another thing I remember is Raja talked about cutting the price of 4K - $2000 gaming machines to half ($1000) and with Ryzen and Vega being mentioned in the same sentence. This might be wishful thinking but hopefully not.

That would be ideal.

Gregster · 9 Feb 2017 at 12:08

TheRealDeal;30486398 said:
I thought it worked that way as well but when i read what the number meant in there internal naming, it only means the chip that's being developed first.

So if the small chip was in development first it would be Vega 10.

The way they do that is confusing as hell. Why not just say "Big Vega" or "Small Vega". Originally the big chip was supposed to be 10 no? I know I am easily confused but I am sure that was confirmed?

TheRealDeal · 9 Feb 2017 at 12:15

Gregster;30486448 said:
The way they do that is confusing as hell. Why not just say "Big Vega" or "Small Vega". Originally the big chip was supposed to be 10 no? I know I am easily confused but I am sure that was confirmed?

I think it was only ever confirmed that there would be a Vega 10 and 11. The press put what it assumed as we have done and said Vega 10 is big Vega. It might well still be the big chip but nothing was confirmed by AMD as far as i am aware. Even by AMD standard this one is being kept pretty close to the chest.

TNA · 9 Feb 2017 at 12:17

Gregster;30486448 said:
The way they do that is confusing as hell. Why not just say "Big Vega" or "Small Vega". Originally the big chip was supposed to be 10 no? I know I am easily confused but I am sure that was confirmed?

My guess would be that at this point they do not want us to know what is what yet. But would be nice if they did as you said. We will find out at most in 4 months time

Gregster · 9 Feb 2017 at 12:21

TheRealDeal;30486470 said:
I think it was only ever confirmed that there would be a Vega 10 and 11. The press put what it assumed as we have done and said Vega 10 is big Vega. It might well still be the big chip but nothing was confirmed by AMD as far as i am aware. Even by AMD standard this one is being kept pretty close to the chest.

Well hopefully we don't have long to wait to find out. Something solid at Computex would be sweet.

TheRealDeal · 9 Feb 2017 at 12:34

Gregster;30486497 said:
Well hopefully we don't have long to wait to find out. Something solid at Computex would be sweet.

Yea hopefully they give us something else to talk about.

LoadsaMoney · 9 Feb 2017 at 13:57

TNA;30486476 said:
My guess would be that at this point they do not want us to know what is what yet. But would be nice if they did as you said. We will find out at most in 4 months time

The only ones who know, are themselves (obviously), and Nvidia

Baboonanza · 9 Feb 2017 at 14:41

Gregster;30486390 said:
Another thing I remember is Raja talked about cutting the price of 4K - $2000 gaming machines to half ($1000) and with Ryzen and Vega being mentioned in the same sentence. This might be wishful thinking but hopefully not.

But what Marketing decide is a 4k capable gaming mnoachine is probably a lot weaker than what I would consider a 4k capable machine.

Look at the Scorpio for example. It will be marketed as 4k capable but 6 Tflops is barely scraping the barrel as far as I'm concerned, even on a console.

meldarthx · 9 Feb 2017 at 15:58

Well from Guru3d that did interview with Raja;

Vega 10 is the smaller chip; Vega 11 is the bigger chip - everyone assumed they were the other way......

http://www.guru3d.com/news-story/amd-vega-10-vega-20-and-vega-11-gpus-mentioned-by-cto.html

Gregster · 9 Feb 2017 at 16:20

meldarthx;30487295 said:
Well from Guru3d that did interview with Raja;

Vega 10 is the smaller chip; Vega 11 is the bigger chip - everyone assumed they were the other way......

http://www.guru3d.com/news-story/amd-vega-10-vega-20-and-vega-11-gpus-mentioned-by-cto.html

Got a link to that interview? I wouldn't mind reading/watching that. The only one I know was Raja and PCWorld interview at CES but if there is another, sweet.

Baboonanza;30487008 said:
But what Marketing decide is a 4k capable gaming mnoachine is probably a lot weaker than what I would consider a 4k capable machine.

Look at the Scorpio for example. It will be marketed as 4k capable but 6 Tflops is barely scraping the barrel as far as I'm concerned, even on a console.

Very true and in the PCWorld interview, he does state that settings don't need to be on Ultra and games are designed for 1080P with default settings in mind. Take from that as you will.

Here is that interview: