AMD VEGA confirmed for 2017 H1

Gregster · 29 Dec 2016 at 17:51

drunkenmaster said:
7970 was a 350mm^2 core, the 680 was 300mm^2, Fury X was 600mm^2, had an AIO and more expensive memory yet cost about the same as a 980ti without the AIO or more expensive memory but was also about 600mm^2.

AMD price based on a sensible/fair margin and the actual production costs. nvidia just look out and decide how much they can get away with ripping off their customers and when AMD don't let them(680gtx) they price with a fair margin to their production costs.

Nvidia is charging you 600mm^2 core pricing for 300mm^2 cores. AMD won't suddenly charge £600 for a ~350mm^2 core.

I don't think Joe Public care for what size the die is in truth and buy on performance figures. This is where AMD really need to nail it from the off. Release a card that consistantly beats the opposition and you have a winner. No good a year down the line and has to be from the off.

Silent_Scone · 29 Dec 2016 at 18:09

I don't think AMD care what size die they're charging you for either, that's not how these metrics work. The Fury X was priced based on it's performance. There's no voodoo magic there.

DarrenM343 · 29 Dec 2016 at 18:13

LoadsaMoney said:
Not like AMD wern't ripping their customers off either, when both the 980 and the much faster Ti were cheaper than it when it was released, with the Ti also having an extra 2GB vram

But HBM was the new tech(mostly marketing gimmick) people paid extra for

vapor matt · 29 Dec 2016 at 18:53

If AMD can nail it with vega like the old 9800pro did in the day, it sure would mix things up in the GPU market. I would love to see it happen

LoadsaMoney · 29 Dec 2016 at 19:35

Theres no way Vega' going to be cheap, as its their new Premium brand, which they started with the FuryX, and look how that was priced, the HBM2 won't be cheap either.

jay0906 · 29 Dec 2016 at 20:07

Given that it's going to be the first consumer card to use HBM2, i'm guessing there will be supply issues which is going to bump up the prices initially.

maonayze · 29 Dec 2016 at 20:11

Griffildur said:
good luck with that. Even if it does beat 1080 there's no way AMD can get away with charging more. The 1080 has already been out quite a while.

Plus this doesn't quite stand right since a few people here seem to regard AMD as the savior and champion of amazing value. Having higher prices will surely be a slap in the face to everyone who's waiting for Vega.

As for better gaming experience and better drivers you know very well that is is likely to attract opposite answers and then you'll complain why people talk about this in a Vega thread etc etc. We've all seen this already, you'd think by now people would know a little better, unless of course you're trying to create that kind of atmosphere yourself.

Calm down Griff, it's only Flopper. You know...the guy with the wild grey hair and beard that lurks in the last cell of the Asylum, who's been there far longer than anyone can remember....Yes that's the one, the one who spouts crazy like it's going out of fashion. He's totally harmless unless provoked.

maonayze · 29 Dec 2016 at 20:16

Gregster said:
I don't think Joe Public care for what size the die is in truth and buy on performance figures. This is where AMD really need to nail it from the off. Release a card that consistantly beats the opposition and you have a winner. No good a year down the line and has to be from the off.

Have to agree with you there fella. However, it hasnt taken as long for the 480 to equal and start to overtake the 1060 in some areas. Not like the older gen cards etc...

That is obviously down to the driver guys getting on with....well getting on with sorting the drivers out I guess after Raja promising that things will improve in that area and to be honest I think it's come on leaps n bounds this last 12 months compared to previous. I dont think anyone can take that away from them. Now if they can turn a bit more attention to MGPU and Linux on the driver front then I think that will start to turn heads a bit and if they can keep it going and Vega is indeed the new tech were were hoping for then I think it will be more of a level playing field for the coming generations.

Mauller · 29 Dec 2016 at 23:11

Something i heard recently is that Vega's NCU's will be capable of processing multiple multi-length waveforms in each 16xSIMD block. Instead of just a single variable length waveform as current GCN CU's process.

It will be capable of scheduling multiple into a single 16 block as long as their total length does not exceed 16 lanes.

And if no work can be processed it is capable of powering down the SIMD's for the cycles they are unused.

Don't have the original source, i only just remembered about this if it was not already posted here.

Anyway, the above should improve throughput and allow graphics workloads to work far closer to the theoretical performance than with past iterations of GCN.

Rroff · 29 Dec 2016 at 23:22

You mean wavefront? unless AMD goes down a narrower, higher tick, approach (which they did a bit with the 480) it would probably bring bigger power savings than performance.

JediFragger · 29 Dec 2016 at 23:43

Mauller said:
Something i heard recently is that Vega's NCU's will be capable of processing multiple multi-length waveforms in each 16xSIMD block. Instead of just a single variable length waveform as current GCN CU's process.

It will be capable of scheduling multiple into a single 16 block as long as their total length does not exceed 16 lanes.

And if no work can be processed it is capable of powering down the SIMD's for the cycles they are unused.

Don't have the original source, i only just remembered about this if it was not already posted here.

Anyway, the above should improve throughput and allow graphics workloads to work far closer to the theoretical performance than with past iterations of GCN.

Think I read the same article a couple of weeks ago. Hopefully they'll find the right balance of power savings + performance. They do still have some very talented engineers :cool:

Hoping we begin to hear more about Vega these coming weeks

nashathedog · 30 Dec 2016 at 07:59

Gregster said:
I would like to see some normalisation in pricing though and bring cards to around the £400-£500 for top end.

That would be nice.

I suppose having a single Halo product above that is fair if it has a genuine reason for costing more. We can give the Kepler Titans a pass as they had the extra compute performance or whatever it was plus a 6gb ram buffer which was a lot back in the day but, pricing now is terrible.

r7slayer · 30 Dec 2016 at 08:38

nashathedog said:
That would be nice.

I suppose having a single Halo product above that is fair if it has a genuine reason for costing more. We can give the Kepler Titans a pass as they had the extra compute performance or whatever it was plus a 6gb ram buffer which was a lot back in the day but, pricing now is terrible.

It would be extremely nice but a lot are more than willing to pay the price and that's what drives prices. I'm holding off tbh because i don't want to pay more than £500 if i can help it but willing to top out at £600 if the performance is there. Currently its not for me so ill wait.

drunkenmaster · 30 Dec 2016 at 11:03

Gregster said:
I don't think Joe Public care for what size the die is in truth and buy on performance figures. This is where AMD really need to nail it from the off. Release a card that consistantly beats the opposition and you have a winner. No good a year down the line and has to be from the off.

One, where did I say anything about Joe Public, two, where did you say anything about Joe public. I explained the pricing of the 7970 and the Fury. Your point was that the 7970 was more expensive than the 680, as if AMD priced it higher simply because it was out first, and ignored the fact that AMD priced the 7970 fairly based on it's die size and manufacturing cost on a new process and Nvidia was forced to price the 680 at a fairer price because the 7970 was already out.

Fury was more expensive, because it was a massive core, not just because it was pricing it against Nvidia or wanted to make more profit.

Nothing I said anywhere had anything to do with Joe Public's perception of cost, though if Joe Public was less stupid and actually factored in the cost to make the product they are buying and refused to be ripped off... prices would be better for everyone.

Kaapstad · 30 Dec 2016 at 11:34

drunkenmaster said:
One, where did I say anything about Joe Public, two, where did you say anything about Joe public. I explained the pricing of the 7970 and the Fury. Your point was that the 7970 was more expensive than the 680, as if AMD priced it higher simply because it was out first, and ignored the fact that AMD priced the 7970 fairly based on it's die size and manufacturing cost on a new process and Nvidia was forced to price the 680 at a fairer price because the 7970 was already out.

Fury was more expensive, because it was a massive core, not just because it was pricing it against Nvidia or wanted to make more profit.

Nothing I said anywhere had anything to do with Joe Public's perception of cost, though if Joe Public was less stupid and actually factored in the cost to make the product they are buying and refused to be ripped off... prices would be better for everyone.

Joe Public is more than 99% of the population so their opinion counts for everything.

Forum members however clever are a very long way short of 1% of the population and therefore their opinions count for very little if anything.

Gregster · 30 Dec 2016 at 11:42

drunkenmaster said:
One, where did I say anything about Joe Public, two, where did you say anything about Joe public. I explained the pricing of the 7970 and the Fury. Your point was that the 7970 was more expensive than the 680, as if AMD priced it higher simply because it was out first, and ignored the fact that AMD priced the 7970 fairly based on it's die size and manufacturing cost on a new process and Nvidia was forced to price the 680 at a fairer price because the 7970 was already out.

Fury was more expensive, because it was a massive core, not just because it was pricing it against Nvidia or wanted to make more profit.

Nothing I said anywhere had anything to do with Joe Public's perception of cost, though if Joe Public was less stupid and actually factored in the cost to make the product they are buying and refused to be ripped off... prices would be better for everyone.

Why all humpty? It was just a basic post and nothing more and from my perspective. It's all good.

Kaapstad said:
Joe Public is more than 99% of the population so their opinion counts for everything.

Forum members however clever are a very long way short of 1% of the population and therefore their opinions count for very little if anything.

This!

Mauller · 30 Dec 2016 at 11:52

Rroff said:
You mean wavefront? unless AMD goes down a narrower, higher tick, approach (which they did a bit with the 480) it would probably bring bigger power savings than performance.

Yeah, I meant wavefront, they are not fixed length under graphics workloads. So fitting multiple down the same 16 lanes should improve performance as shader utilisation increases. But under lighter workloads this and powergating simds will improve efficiency.

Yutang13 · 1 Jan 2017 at 16:12

http://ve.ga/

akarypid · 1 Jan 2017 at 16:28

Yutang13 said:
http://ve.ga/

In the beginning, at 0:07 seconds, green sticker reading "POOR VOLTAGE", which has a #BETTERRED sticker covering "GE"...

They're hyping up Vega to be better than even Volta???

humbug · 1 Jan 2017 at 16:32

Yutang13 said:
http://ve.ga/

Ooohhh.....