Vega/Navi Rumors (Updated)

[HurleyBird]

I still don't see why Polaris is failure? Today we are paying $200 (RX 480 4GB) to get the same performance we needed to pay $330 last year (GTX 970). What's the problem with that? Just because nVidia offers similar (better?) products at the same price doesn't mean it's a failure

The RX 480 is a great deal relative to everything else on the market, but that doesn't make Polaris a success, it just makes it priced really low. What would be even better is if AMD was competitive and consumers actually responded to that the way they should. Then you'd be paying around $500-600 for a fully unlocked GP102 or AMD equivalent.

 

[HurleyBird]

Just that Vega 10 won't be a really huge die. It should be somewhere between 350-400 mm². In between GP104 and GP102 and there we should expect its performance. Then we have Vega 11 below it, which is mostly important for notebooks, because amd is loosing there against pascal. The real big Vega is Vega20, but this gpu won't come in 2017. None of the Vegas for next year have 2:1 Double Precision capability, but Vega 20 should have it and probably also more shader as it will be the next HPC offering (MI25 is just for Deap Learning).

While I'm not ready to call Vega 10 the bigger of the two, I'd be surprised if the bigger die was less than 450mm2. You're increasing the costs so much with HBM2 that anything less doesn't really make sense. Assuming small Vega still uses or at least has the capability to use HBM2, I expect it will be used for high margin, low power ultrabook parts.

On the other hand, this is AMD we're talking about. It's not like you can count on management doing the obviously beneficial thing. Eg. shooting themselves in the foot with the 290X heat sink, or not making anything larger than the tiny RV770, basically surrendering an entire generation when they had a truly massive architectural advantage.

 

[Elfear]

Vega 10 is rated at what, 12.5TFLOPs? Polaris 10 is 5.8TFLOPs. So Vega 10 = 2.16*Polaris 10.

2.16 * Polaris 10 should give you a pretty reasonable idea of how Vega 10 will be positioned performance wise relative to competition and AMD's current offerings.

So if Big Vega scales perfectly (doubtful), it would end up ~10-12% slower than Titan XP. That's actually not bad at all but there will have to be some solid improvements to the architecture for that to happen. Keeping my fingers crossed.

 

[Head1985]

If it doesn't compete with titan xp, with all the extra bonuses like Hbm2 it's a massive failure.



If it's that speed what's the point of calling it flagship if it can't compete with flagship? That's what I mean why isn't this more relevant? If big Vega is that weak it's in the range of a 1080 isn't that a massive issue? Either ncu is a massive failure or amd didn't show off the fastest gpu.

I just don't want to find I waited 8 months to find out the fastest gpu I can buy from amd is not even close to a titan xp.

Edit :
I'm not ready to mentally consider that amd on a generation past the competition can't compete at the high end. I'll believe it when I see it. Then cry myself to sleep if it happens.

4096SP at 14/16nm is around 300-350mm2 SKU(we have 2304sp rx480 at 232mm2 and HBM2 will save alot of space so i think it will be close to 350mm2).Thats 7970 successor/GP104 competitor.It will be similar to 7970Ghz vs GTX680 in kepler era.
If its faster than GTX1080 then it is ok.
It should be like 7970ghz vs gtx680-7970ghz was little bit faster with little bigger SKU.

If you wat TITANXP competitor wait for 420mm2+ SKU(290x successor)
This is why i think 4096SP is small vega, because its only 7970 successor(300-350mm2 SKU)
It is also possible AMD dont have money to release multiple SKU so they fastest SKU will be only 7970 succesor(right now they have fastest SKU on 14nm only rx480 which is 7870 successor)

 

[flopper]

So if Big Vega scales perfectly (doubtful), it would end up ~10-12% slower than Titan XP. That's actually not bad at all but there will have to be some solid improvements to the architecture for that to happen. Keeping my fingers crossed.

30% shader efficiency vs 300 series possible.

 

[zinfamous]

I don't think it's absurd to think that a huge die with super expensive memory tech coming out almost a year later should be near GP102 given AMD and Nvidia's historical competitiveness.

That said, I have a feeling that Vega could end up being another Polaris level disappointment. Too many subtle warning signs at this point. I don't doubt that AMD eventually wants to aim for the jugular again, especially with Zen looking so good. Good things would happen to the AMD brand if they were very competitive in both spaces at the same time. Hopefully AMD starts trying again with Navi. Well, hopefully I'm wrong about Vega altogether...

what is a Polaris level disappointment? It is a fantastic chip at it's tier, and already 6 months out, the 480 now outperforms its direct competitor, the 1060 6gb overall with the latest drivers. It's at least $50 less, too. what the hell more do you want from Polaris, other than it meeting unrealistic expectations and assumptions from fans and media that were never in-line with what AMD advertised?

 

[Glo.]

Guys disputing about Doom performance and Vega. The problem is that Doom with Vulkan incredibly well responds to compute performance of the GPU. For example, RX 480 is 10% slower in Vulkan than GTX 1070 FE. What is TFLOPs nomber for both GPUs? 5.8 TFLOPs and 6.5 TFLOPs. How does RX 470 with 4.4 TFLOps compare to 4.4 TFLOPs GTX 1060? Its on par with it.

12.5 TFLOPs GPU will not be only 10% faster than GTX 1080. It will be faster than GTX Pascal Titan X.

 

[tential]

It's pretty simple, set absurd expectations from vega and call it a failure later it it doesn't meet any.

when amd competed they didn't got any reward so why now ? i actually applaud amd for not giving anything for 1080 from their stable. you treat amd as they are only here as a catalyst for low prices and still badmouth it. well guess what ? no candy for you to waste anymore as no matter what you won't buy that candy.

you people don't really deserve it. those who do, they will appreciate and buy it. just wait and watch.

So because of amds history, I shouldn't be able to purchase a high end gpu in a timely manner from amd? Good for amd... Guess they'll continue to suffer at the high end

 

[kraatus77]

They suffered even when they had something. and that makes it even worse. at least this won't be because they don't need to spend on developing something no one will buy anyway because people constantly keep changing goalposts to stay away from it.

You treat someone like crap even when they are treating you nice, well guess what ? they will do same with you one day. and AMD has learned it. they cannot attract those no matter what so why waste resource on chasing them ?

(By you i mean as in general, not you. so please don't take it as offence)

 

[Dave2150]

Fury X did not cause a drop in price of its direct competitor.

True, though at the end of the day the Fury X was a terrible product - too late, too slow, too power hungry, forced AIO cooler/radiator, loud/faulty pumps, no overclocking, to name a few.

I'm hoping Vega will be a much better product, if it's not then 1080TI here I come.

 

[Arachnotronic]

True, though at the end of the day the Fury X was a terrible product - too late, too slow, too power hungry, forced AIO cooler/radiator, loud/faulty pumps, no overclocking, to name a few.

I'm hoping Vega will be a much better product, if it's not then 1080TI here I come.

Largely agree, Fury X's direct competitor was the better buy, no question.

IMO, AMD found itself in a bit of a pickle with the part. It was designed as a balls-to-the-wall high end part -- no partially disabled die, HBM, premium card, etc. but when the Fury X's direct competitor dropped at $650, AMD might have found itself unable to price much lower due to the cost structure of the card, especially at the beginning of the cycle (brand new HBM, huge die on 28nm, etc.)

AMD seems to be making ~$299-$390 Fury X cards work now though. This can't be inventory draining because the card has been on the market for 1.5yrs...it doesn't take that long to clear inventory of a card that didn't really see much demand early on to begin with. So I suspect it is still being manufactured so that AMD has some presence in the >$250 GPU market (even if it is, for all intents and purposes, a token presence). The fact that AMD probably doesn't sell too many of these means that even if AMD's margins on them are very low, it doesn't dilute the company's overall margin results too much.

That's the curse of using bleeding edge technologies that have very little industry scale -- if they don't pan out and deliver the required results, you end up in a financial jam.

But I should caution, I don't think this is a case of AMD necessarily executing badly. AMD did a good job of getting a big die + HBM GPU out in the timeframe that it did, but it was simply outmaneuvered by a competitor.

 

[Carfax83]

Guys disputing about Doom performance and Vega. The problem is that Doom with Vulkan incredibly well responds to compute performance of the GPU. For example, RX 480 is 10% slower in Vulkan than GTX 1070 FE. What is TFLOPs nomber for both GPUs? 5.8 TFLOPs and 6.5 TFLOPs. How does RX 470 with 4.4 TFLOps compare to 4.4 TFLOPs GTX 1060? Its on par with it.

12.5 TFLOPs GPU will not be only 10% faster than GTX 1080. It will be faster than GTX Pascal Titan X.

Sorry but I just had to point out that's an old benchmark which is using NVidia drivers that weren't properly optimized for Vulkan. This is what Doom performance looks like now with proper Vulkan drivers:

 

[kraatus77]

^ literally pointless. different website. probably different scene/location.

you can't compare like that. regardless this isn't let's update nvidia's vulkan result thread.

 

[happy medium]

Sorry but I just had to point out that's an old benchmark which is using NVidia drivers that weren't properly optimized for Vulkan. This is what Doom performance looks like now with proper Vulkan drivers:

Looking at that chart I would hope that the Vega 10 gpu is more than 10% faster than a gtx1080 at Doom . If not it will be slower in any other game vs the gtx1080.

 

[Carfax83]

^ literally pointless. different website. probably different scene/location.

you can't compare like that. regardless this isn't let's update nvidia's vulkan result thread.

No it's definitely the drivers. There was a bug in NVidia's drivers with Vulkan for a while, and it wasn't fixed until the 372.xx series. I made a post about it here in the Doom Vulkan thread.. The Computerbase.de review used the older 368.xx drivers which were bugged..

And yeah, I'm not trying to go off topic, I'm just trying to correct bad information..

 

[Arachnotronic]

No it's definitely the drivers. There was a bug in NVidia's drivers with Vulkan for a while, and it wasn't fixed until the 372.xx series. I made a post about it here in the Doom Vulkan thread.. The Computerbase.de review used the older 368.xx drivers which were bugged..

And yeah, I'm not trying to go off topic, I'm just trying to correct bad information..

The correction is appreciated as it helps us understand where Vega 10 could ultimately land competitively. Thanks.

 

[kraatus77]

It's irrelevant where it lands, because of obvious reasons. but whatever, it's not like anything is going to change.

 

[flopper]

Looking at that chart I would hope that the Vega 10 gpu is more than 10% faster than a gtx1080 at Doom . If not it will be slower in any other game vs the gtx1080.

That is with fury adapted drivers.
so...

 

[raghu78]

This thread is about Vega just to remind. Anyway a tweet about Vega. New Year New architecture. Hopefully AMD can bring a huge leap in architectural efficiency and performance with Vega.

https://twitter.com/Radeon/status/813561408612945920

 

[tential]

This thread is about Vega just to remind. Anyway a tweet about Vega. New Year New architecture. Hopefully AMD can bring a huge leap in architectural efficiency and performance with Vega.

https://twitter.com/Radeon/status/813561408612945920

The only good thing is at least I really didn't need a new gpu I realized. I can still use my 290 for another 6 months playing gamecube games.

Might as well just coast til Navi at this rate almost lol.

I just want to know why Vega is a significant enough to explain the new naming convention changes of things
Is Vega a significant architectural upgrade over Polaris or just hbm?

 

[tential]

http://wccftech.com/amd-vega-gpu-radeon-rx-500-series-teaser/

The consumer variant of the Vega GPU was demoed in Doom and put up some fiery performance against an GeForce GTX 1080 graphics card. The card was seen outpacing the NVIDIA solution but we don’t have exact metrics at hand. Regardless, it’s nice to see Vega GPU being faster than an enthusiast solution based on the NVIDIA’s latest Pascal GPU architecture.
Later on, we confirmed that this was the exact GPU that showed up in AOTS database earlier this month.This variant packs 8 GB of HBM2 and is more focused at desktop PCs rather than HPC market. Performance of this card was on par with GeForce GTX 1080 but drivers have not been optimized for it yet so we can expect even faster performance from the final variant.

That's not the high end chip then? I think the high end chip has 16 GB HBM2. So ya, looks like AMD will have a 1080 competitor and a 1080ti competitor.

Could AMD possibly move to a tick/tock style GPU release? They followed Polaris up with Vega a "new" architecture quite quickly. By only releasing half a lineup, and then switching to a newer architecture, perhaps it will allow AMD to innovate more quickly? Navi may not even be focused on the high end at all. We may need to wait til the next architecture after that.

Pretty much just comes down to performance and pricing, looks like the performance of each GPU class this generation will be quite close.

 

[happy medium]

16 GB HBM2.

I think that the 16gb card is a professional card with the same chip..

 

[flopper]

I just want to know why Vega is a significant enough to explain the new naming convention changes of things
Is Vega a significant architectural upgrade over Polaris or just hbm?

Yea its called NCU or Next compute Units.
expected shader efficiency with 20%+ for Vega vs the 300 series (my estimate is 30%)

 

[tential]

Yea its called NCU or Next compute Units.
expected shader efficiency with 20%+ for Vega vs the 300 series (my estimate is 30%)

Not your fault but I hate amd and their comparison to older generations that are not the most current. I mean Fiji and Polaris have occurred I don't care about a comparison to an architecture as old as Hawaii...

 

[Krteq]

Yea its called NCU or Next compute Units.
expected shader efficiency with 20%+ for Vega vs the 300 series (my estimate is 30%)

Huh, is there any document showing NCU structure? We know nothing yet to create such conclusion.