Vega/Navi Rumors (Updated)

[Erithan13]

I dropped a lot of cash on a 144hz/1440p freesync monitor in August last year. If you'd told me then it would be a full year before Vega arrives I'd suddenly see the G-Sync premium as not such an unreasonable thing after all. I don't even want to consider that small vega (my most likely upgrade path) may be coming even later, and at goodness knows what price/performance. Add to that the great Polaris drought and I'm really hoping my 290 doesn't croak anytime soon. I suppose I should consider this penance for all those times in April/May when I earnestly recommended waiting for Vega. What's a person to do when it's been 'just around the corner' for so long?

I'm trying to remain optimistic, there's still time for AMD to deliver something compelling, but there's that horrible feeling that all I'm seeing is a truly legendary disappointment in the making. I'd like to think Nvidia's marketing department has a suitably scathing response to 'Poor Volta' ready. If you can't cry might as well laugh?

 

[cm123]

Whats to say that much like Intel with i9's, Nvidia doesn't have a Titan Volta just waiting in the wings already... Of course that's only if AMD pulls off a tremendous feat with whatever it is they decide to release and actually be back in the race on the middle high end and high end.

Personally my opinion on this Graphics side for AMD, they have their hands extremely full if they feel they can release a product faster and more advanced than Nvidia right now - Though to AMD's credit, if you count the Radeon Pro Duo, it wasn't until this latest Titan Xp release that Nvidia really took the single card performance card across the board.

 

[tential]

Considering that AMD's Market cap is roughly 1/9 of Nvidia, I think you're overestimating this. Remember, GPUs are the core of everything Nvidia does these days. Anything not GPU-driven (Denver ARM cores, for example) has been put aside/canceled. All their AI stuff, all their car stuff, it's all based on GPUs. All the while, AMD produces high end consumer, workstation and server CPUs too. I'd bet the RTG R&D budget is closer to 1/20 of that of Nvidia's GPU arch group.

With this logic, because tesla has the largest market cap they should easily be able to produce cars on the same level as competitors.

Market cap is simply the valuation of the company and isn't how "large" the company is on a scale of getting things done.

Not disagreeing with you that amd is a smaller company with less resources.

Market cap just isn't the correct metric to use for the context of your quote since you're talking about what amd has to work with which would be assets, and not their market cap which fluctuates based on investor sentiment.

For reference market cap definition is
"Market capitalization (market cap) is themarket value at a point in time of the shares outstanding of a publicly traded company, being equal to the share price at that point of time times the number of shares outstanding"

I'd say compare assets, and compare working capital personally as a good spot to start in the broadest general sense.

Otherwise we could be saying amd has at least twice as much resources to operate with since last year which isn't the case it's just that their stock price has been bid up which increased their market cap.

So zinfamous is also correct to use the terminology he used since operating budget would also be a good way of determining what amd has to work with in comparison to its competitors.

 

[Magic Hate Ball]

I totally get that, but they can relatively easily build a pro card and then do consumer mid-high to low end with the same chip. The thing that is constantly just out of their reach is the very top enthusiast card. That's the one part that screws their execution. That being the case I'm making the case that they just don't allow chasing a top card to cause a delay to products you could be selling competitively. They literally cannot afford to do what they are doing.

It really depends on how much driver work is left to do for gaming.

They had hard enough of a time getting Polaris going, it sounds like Vega is an even bigger change so, so I think dumping a half assed pro card with poor, buggy drivers is a PR nightmare and a permanent stain on the Vega brand.

Look at it this way from their perspective: delays give them more time to get HBM2 stock at potentially lower prices as time goes on with the benefit of additional driver development time.

Sucks that they're going to miss 2 months more of business but if the product comes out swinging it'll be for the best in the long run.

 

[ElFenix]

i just watched that demo and i think there's a few things we actually know:
they have xfire working
they have threadripper working
launch dates


everything else seems like 4chan finding the boston bombers.

 

[OatisCampbell]

I totally get that, but they can relatively easily build a pro card and then do consumer mid-high to low end with the same chip. The thing that is constantly just out of their reach is the very top enthusiast card. That's the one part that screws their execution. That being the case I'm making the case that they just don't allow chasing a top card to cause a delay to products you could be selling competitively. They literally cannot afford to do what they are doing.

I think you are sort of trivializing the whole "just build the most powerful graphics processor on the planet" thing, no offense intended.

AMD has been running on a shoestring their entire history, the fact they hang in there and bring (nearly) competitive parts to market at all is pretty amazing.

Wasn't there a leak a while back that went something like "~1080 performance, but $400-$450"? That wouldn't shock me, and I also think that would sell.

They do need to get the things out the door as every guy that says "F them, buying Pascal" is likely a sale lost this gen as people won't sell 1070/1080/180Ti to buy AMDs 1080 because it's AMD for the most part.

Would be really interesting to know where the bottleneck is- respin in process? HBM2 drought? New arch hard to code drivers for and driver staff depleted by layoffs or exodus? We won't know for a long time is the only thing I'd bet on.

 

[zinfamous]

Considering that AMD's Market cap is roughly 1/9 of Nvidia, I think you're overestimating this. Remember, GPUs are the core of everything Nvidia does these days. Anything not GPU-driven (Denver ARM cores, for example) has been put aside/canceled. All their AI stuff, all their car stuff, it's all based on GPUs. All the while, AMD produces high end consumer, workstation and server CPUs too. I'd bet the RTG R&D budget is closer to 1/20 of that of Nvidia's GPU arch group.

well, I was trying to be generous.

 

[Magic Hate Ball]

I think you are sort of trivializing the whole "just build the most powerful graphics processor on the planet" thing, no offense intended.

AMD has been running on a shoestring their entire history, the fact they hang in there and bring (nearly) competitive parts to market at all is pretty amazing.

Wasn't there a leak a while back that went something like "~1080 performance, but $400-$450"? That wouldn't shock me, and I also think that would sell.

They do need to get the things out the door as every guy that says "F them, buying Pascal" is likely a sale lost this gen as people won't sell 1070/1080/180Ti to buy AMDs 1080 because it's AMD for the most part.

Would be really interesting to know where the bottleneck is- respin in process? HBM2 drought? New arch hard to code drivers for and driver staff depleted by layoffs or exodus? We won't know for a long time is the only thing I'd bet on.

At least they'll get all those juicy cryptocurrency sales still I bet.

 

[Despoiler]

I think you are sort of trivializing the whole "just build the most powerful graphics processor on the planet" thing, no offense intended.

You clearly didn't read, or if you did, understand my point.

 

[Face2Face]

Found this over on Reddit. Was probably made by an NVIDIA fanboy, but thought it was pretty funny...

 

[tential]

well, I was trying to be generous.

You were actually right. Like I explained, you would not use market cap for a scenario like this it's not the appropriate metric and doesn't make sense since market cap can fluctuate wildly and isn't a measure of what amd has to work with in relationship to other companies.

Operating budget would be better and there are for sure even better ways of figuring this out.

Otherwise like I explained, you'd be saying tesla and gm are the same size companies. Tesla has a larger or equal market cap with much less to work with.

 

[Maverick177]

At this point AMD really needs to beat the Ti out of the gate with RX Vega, just to make it worth the wait. The only other way it isn't a disappointment is if the prices are really aggressive.

It's AMD we're talking about here. Be realistic.

 

[zlatan]

I honestly don't understand what RTG is doing. Have they ever lead with their professional products? If they want to change up their product stack launch order then fine. However, launching Vega in late July, early August puts them really close to where Nvidia could launch Volta consumer variants.

I don't know why is this a problem. Vega feature set is much closer to Volta. Vega will be the first dedicated GPU where the hardware can access the CPU page table directly, this is a big feature. Volta will be the second, assume that Intel will allow them to do this. I'm not sure about this, but AMD has the licenses.
There are several other important features comes in Vega and Volta, so I don't think that Polaris and Pascal can compete with these new architectures.

 

[swilli89]

Raja Koduri ""It was running comfortably about 60 and comfortably above any single gpu and we aren't done yet with Vega rx optimizations for prey""

http://imgur.com/a/aqvL7

 

[tential]

I don't know why is this a problem. Vega feature set is much closer to Volta. Vega will be the first dedicated GPU where the hardware can access the CPU page table directly, this is a big feature. Volta will be the second, assume that Intel will allow them to do this. I'm not sure about this, but AMD has the licenses.
There are several other important features comes in Vega and Volta, so I don't think that Polaris and Pascal can compete with these new architectures.

So this goes back to the post of mine and someone else posted recently. Vega is updated Polaris, so a full highend Vega should do better than if we expanded Polaris.
Vega looks to be a future looking gpu and amd even references Volta.

So shouldn't Vega blow pascal out the water if it's going to compete with pascal?

Could you explain why Vega wouldn't be a large jump now but might be competing with Volta later?

My guess is it has to do with feature like the ones mentioned in your post as well as others that may not see full utilization in games or Applications until the future.

So similar to Polaris Vega would be very future feature driven? Could this be another Hawaii repeat? Competitive forever because there is just so much potential the chip has to unlock?

That's the one thing being discounted way too much here is drivers and amds ability to continue to extract performance out of s gpu far agree release. That's why I don't get why anyone cares about amd showcases before they have final or close to final drivers. That's the bare minimum starting point for amd and even still for all we know Vega currently could be down volted for even more performance.

 

[zlatan]

Look I really don't want to get in trouble. But Vega is not an updated Polaris. It is the biggest change since the 1st GCN.

There will be a lot of features that will just works. For example HBCC. And some others, but I don't know now what features are publicly anounced.

Personally I think the main advantage of Vega and Volta will be the packing strategies. With Vega you can get half the VGPR usage with structure of arrays strategy compared to the unpacked method. And this will speed up the shader execution twice as fast. Volta will be the same. So these new high-end GPUs will be 20+ TFLOPS montsters in a lot of new shader codes. Impossible to beat them with a hardware that is forced to the traditional unpacked method. Even GCN3/4 won't be enough. These will support packing, but the potential speed up is around 10-15 percent. So nowhere near what is possible with Vega and Volta.

The drivers are far from "final".

 

[guachi]

If they only had given it an fps graph that Prey demo would have been useful. You can guess fps by the amount of screen tearing depending on what you assume the fps of the stream was.

At least the side-by-side Blender demos provided some kind of information about Ryzen even without stopwatches timing things.

 

[thilanliyan]

Raja Koduri ""It was running comfortably about 60 and comfortably above any single gpu and we aren't done yet with Vega rx optimizations for prey""

http://imgur.com/a/aqvL7

Is he talking about XFire Vega there? Hard to say from the post. It isn't impressive if he is as I would hope that XFire Vega would be handily beating Titan/1080Ti.

 

[Phynaz]

Raja Koduri ""It was running comfortably about 60 and comfortably above any single gpu and we aren't done yet with Vega rx optimizations for prey""

http://imgur.com/a/aqvL7

Two Vega's are faster than a single 1080Ti.

So are two 1060's. Is he saying Vega is as fast as a GTX1060?

 

[Jackie60]

Vega release issues could end up being due to most of AMD's money & labor being focused on Ryzen / TR / Epyc

I don't remember when exactly (definitely was not this year), but I remember Lisa Su saying something along the lines of AMD being primarily a CPU company

Vega release issues could end up being due to most of AMD's money & labor being focused on Ryzen / TR / Epyc

I don't remember when exactly (definitely was not this year), but I remember Lisa Su saying something along the lines of AMD being primarily a CPU company

If I was running a company that was OK if decent at GPUs and OK if decent at CPUs then put the two together you rtards and give us something compelling. I don't know about anybody else here but if AMD aren't producing a 6/8core Ryzen CPU WITH an integrated fury X or better GPU for 20/30% less cost than buying them separately from the competition very soon (they should already be doing this) then they need to ask themselves seriously wtf they are doing in this game. They can do it with consoles and get decent performance with jaguar cores, they have sufficient intellectual property to do it for the desktop. If they can't then they should sell themselves to Samsung or someone. However I understand if potential buyers are hesitant given how dismal execution has been in the last 4 years excluding Ryzen. I hope its driver development but I suspect it's a much deeper more profound malaise that has infected the corporation.

 

[crisium]

Two Vega's are faster than a single 1080Ti.

So are two 1060's. Is he saying Vega is as fast as a GTX1060?

You mean two 1080s, right?

Agree that there's no point in that Raja statement. 1080 SLI should be comfortably ahead of even the Titan Xp when scaling properly.

 

[Despoiler]

I don't know why is this a problem. Vega feature set is much closer to Volta. Vega will be the first dedicated GPU where the hardware can access the CPU page table directly, this is a big feature. Volta will be the second, assume that Intel will allow them to do this. I'm not sure about this, but AMD has the licenses.
There are several other important features comes in Vega and Volta, so I don't think that Polaris and Pascal can compete with these new architectures.

Look I really don't want to get in trouble. But Vega is not an updated Polaris. It is the biggest change since the 1st GCN.

There will be a lot of features that will just works. For example HBCC. And some others, but I don't know now what features are publicly anounced.

Personally I think the main advantage of Vega and Volta will be the packing strategies. With Vega you can get half the VGPR usage with structure of arrays strategy compared to the unpacked method. And this will speed up the shader execution twice as fast. Volta will be the same. So these new high-end GPUs will be 20+ TFLOPS montsters in a lot of new shader codes. Impossible to beat them with a hardware that is forced to the traditional unpacked method. Even GCN3/4 won't be enough. These will support packing, but the potential speed up is around 10-15 percent. So nowhere near what is possible with Vega and Volta.

The drivers are far from "final".

It just seems a bit crazy to think that AMD could skip competing vs Pascal and then come out and competitively go after Volta. Although they did throw a bit of smack talk with the "poor volta" thing. I didn't personally think they were being serious, but you have a strong technical understanding so perhaps they were being serious. Very interesting. Your insight is always much appreciated.

 

[swilli89]

Is he talking about XFire Vega there? Hard to say from the post. It isn't impressive if he is as I would hope that XFire Vega would be handily beating Titan/1080Ti.

Considering two 580's in CF can get to around 60FPS in 4k.. I think its a safe bet that he was referring to a single Vega. However I do agree choice of wording is poor. Hopefully AMD will clarify this soon.

 

[crisium]

Do you think Raja would casually reveal that a single Vega beats the Titan Xp like that "ahead of any single gpu"? I want to believe, but considering it was a dual GPU setup he was referring to... I see room for hope but I think it's reaching.

 

[Jackie60]

well, I was trying to be generous.

I get the AMD lack of resources thing, I don't expect miracles but Fury X was the 980TI competition in summer 2015. The 980TI is roughly equivalent to the 1070. So if VEGA (heaven forbid) is at 1070 levels of performance then the entire idiocracy of AMD's GPU division has been almost back pedalling for two years. OK they went forward a tiny bit but it would be such a puny pathetic incremental improvement that if they are really GPU engineers they need to find another job as they lost their way.