Vega/Navi Rumors (Updated)

[PeterScott]

Given that GV100 is right in line with the same perf/w improvement that Pascal was over Maxwell and Maxwell was over Kepler, you look to be probably off on your prediction.

Pascal and Maxwell are equivalent on the architecture side.

The performance improvement came entirely from clock speed boost.

Edit:

There appears to be zero architecture improvements in GV100 at the Cuda core level.

GV100 is has ~40% increase in performance and from the huge die about ~40% increase in Cuda cores.

It also looks like there is negligible increase in transistor density from 12 nm FFN. 12 nm FFN might improve power usage, but that doesn't directly translate into performance for gaming.

All in all I seen nothing in GV100 that points to any significant boost for gaming in Volta.

NVidia may choose to increase the die size in each segment, to squeeze in more Cuda cores, but they could do that with a Pascal refresh and have much the same effect.

 

[stuff_me_good]

I vonder if volta will have proper asyn support so we would see more games supporting it.

 

[Qwertilot]

This is drifting way off topic, so apologies.

~40% more cores for ~50% more performance, yes, but at the same power draw. So they can make gaming volta bigger than gaming pascal without breaking the relevant power budgets. They have to keep moving perf/watt because of their huge gaming notebook market if nothing else.

 

[PeterScott]

This is drifting way off topic, so apologies.

~40% more cores for ~50% more performance, yes, but at the same power draw. So they can make gaming volta bigger than gaming pascal without breaking the relevant power budgets. They have to keep moving perf/watt because of their huge gaming notebook market if nothing else.

Typo? It's 40% more performance, not 50%.

Also, equal TDP does not mean equal power draw. It remains to be seen how this plays out.

What relevant power budgets. It hardly matters unless you are on the Titan End of the scale, or the laptop end. Desktop GPUs in between those, it remains of distant importance compared to perf/$ which really won't be affected.

The question is how much larger (and more expensive) is NVidia willing to make the die in each segment (GTX 2060, GTX 2080), that will govern how much performance increases, if they are sticking with 12 nm FFN. If they go for 10nm they should get a nice transistor density boost as well, enabling a nice performance boost at the same die size for a similar production cost.

 

[Despoiler]

1. Video cards aren't a service.

Gee really? I was pointing out a common area where bundles are used. An area that you'd likely have seen marketing for.

2. You seem to have missed the rest of my post.

No I read your whole post. You and several others seem to be under the impression there is room to debate this point. Do you have any product development experience? I do. I was part of product development for a Fortune 500 company on the technical side for over 5 years. I am telling you that bundling is a common practice. It's used because the more products and services someone has the less likely they are to switch vendors or providers. AMD has CPUs, GPUs, and feature support in monitors. That is a big chunk of a full computer build. More offerings means higher customer retention. More importantly they have more offerings than their competitors. Being that is an advantage that is why they are leveraging it. They would be stupid not to.

 

[EXCellR8]

not terribly upset over the LED tach (looks like you can turn it off), but those two 8-pin connectors = heavy power draw

 

[coercitiv]

Looking forward to the weeks after the Vega launch when the product simply cannot stay on the shelves even after underwhelming reviews and even with inflated pricing.

 

[PeterScott]

Looking forward to the weeks after the Vega launch when the product simply cannot stay on the shelves even after underwhelming reviews and even with inflated pricing.

Miners FTW?

Without mining AMD would be in so a so much deeper mess on the GPU side. GTX 1080 would be $399 by now and GTX 1070 would be $299, and NVidia could easily throw another $50 price cut on top of that.

As is, it is totally pointless to cut prices as it would just be more profiteering for the rest of the supply chain, so AMD gets a tiny bit of breathing room.

 

[coercitiv]

Without mining AMD would be in so a so much deeper mess on the GPU side.

And yet, even with mining in full bloom some consider the bundle offer is some kind of desperate move. It just shows how extremely dissonant such interpretation is with the market reality.

 

[tviceman]

And yet, even with mining in full bloom some consider the bundle offer is some kind of desperate move. It just shows how extremely dissonant such interpretation is with the market reality.

The bundles are a way to offset the costs of what will be low-volume Vega and move some of their lower selling Ryzen CPUs too. On it's own, Vega is at way more of a disadvantage versus the competition than any other AMD release since the HD 2900. It's not faster, it's not cheaper, it's not at all efficient, it's 15 months late, and its costs to manufacture are significantly higher than the competition.

The bundles are AMD's way to offset this laundry list of negatives and is only being portrayed as a way to discourage miners. AMD and Nvidia can easily cripple the mining capabilities of their cards via drivers if they wanted; don't be duped for a moment that AMD is trying to do gamers a favor by forcefully bundling AMD hardware. Vega is way less efficient at mining than Polaris to begin with, so by this same logic where are the Polaris bundles to help out the gamers? If Vega was competitive on all fronts and/or was 9 months sooner there wouldn't be any bundles.

 

[PhonakV30]

The bundles are a way to offset the costs of what will be low-volume Vega and move some of their lower selling Ryzen CPUs too. On it's own, Vega is at way more of a disadvantage versus the competition than any other AMD release since the HD 2900. It's not faster, it's not cheaper, it's not at all efficient, it's 15 months late, and its costs to manufacture are significantly higher than the competition.

The bundles are AMD's way to offset this laundry list of negatives and is only being portrayed as a way to discourage miners. AMD and Nvidia can easily cripple the mining capabilities of their cards via drivers if they wanted; don't be duped for a moment that AMD is trying to do gamers a favor by forcefully bundling AMD hardware. Vega is way less efficient at mining than Polaris to begin with, so by this same logic where are the Polaris bundles to help out the gamers? If Vega was competitive on all fronts and/or was 9 months sooner there wouldn't be any bundles.

according to some user reddit , If Only ETH , Yes but IF Dual ETH with Decred/Siacoin/Lbry/Pascal , then Polaris can't beat Vega

 

[Yakk]

according to some user reddit , If Only ETH , Yes but IF Dual ETH with Decred/Siacoin/Lbry/Pascal , then Polaris can't beat Vega

Yes I'm very curious when AMD will update their drivers for Claymore to release their money with VEGA optimizations.

 

[tviceman]

Looking forward to the weeks after the Vega launch when the product simply cannot stay on the shelves even after underwhelming reviews and even with inflated pricing.

And also looking forward to the Mercury, J. Peddie, and Steam surveys which will show Vega having a negative impact on discrete GPU market share for AMD.

EDIT: I should have said UNFORTUNATELY instead of looking forward to. I don't want a one horse race, but reality is that a one horse race is exactly what we are getting.

 

[crashtech]

And also looking forward to the Mercury, J. Peddie, and Steam surveys which will show Vega having a negative impact on discrete GPU market share for AMD.

As if anyone was in doubt regarding your sentiments?

 

[Stuka87]

Miners FTW?

Without mining AMD would be in so a so much deeper mess on the GPU side. GTX 1080 would be $399 by now and GTX 1070 would be $299, and NVidia could easily throw another $50 price cut on top of that.

As is, it is totally pointless to cut prices as it would just be more profiteering for the rest of the supply chain, so AMD gets a tiny bit of breathing room.

You can't be serious? If AMD went away, nVidia prices would go up, not down. The only thing that drives prices down is lack of demand or competition. Lack of either will cause prices to stay high, or go higher.

 

[crashtech]

Participants who want to bash AMD every chance they get don't seem to grasp that a strong AMD is good for Nvidia fans, too. The only one it's bad for are Nvidia stakeholders, which makes one wonder about some of the posts here.

 

[Elfear]

And also looking forward to the Mercury, J. Peddie, and Steam surveys which will show Vega having a negative impact on discrete GPU market share for AMD.

Why would you be looking forward to this?

 

[Snarf Snarf]

The bundles are a way to offset the costs of what will be low-volume.

This. However Vega is actually getting quite a bit of hype from the professional community as it seems to be tailor-made for certain workloads that aren't currently being addressed as efficiently as possible. My buddy over at ConocoPhillips has been talking up the entire Radeon Pro ecosystem for the past year or so as they prepare to get away from any workloads that require CUDA to get out of vendor lock in. For the enormous data sets they use to search for oil deposits I've been told the Radeon SSG based on Vega is exactly what they need hardware wise, and they've bought several of the Fiji based SSG's to start testing performance for life after CUDA. These are the types of customers that AMD need right now, and as sad as that makes me as an enthusiast it's probably the smart play long term for AMD. Here's to hoping they can milk some fat enterprise customers for a healthy chunk of R&D for Navi and beyond.

 

[tviceman]

Why would you be looking forward to this?

I'm not, at all. I was being sarcastic. It's the reality of what Vega is and my prediction on the impact it will have. I wish Vega was either 20% faster or 50+ watts more efficient because then it'd be in a way more competitive position.

 

[amenx]

Participants who want to bash AMD every chance they get don't seem to grasp that a strong AMD is good for Nvidia fans, too. The only one it's bad for are Nvidia stakeholders, which makes one wonder about some of the posts here.

Eh? I think we all know the obvious. In fact thats why I'm upset at AMD. Weak competition stagnates the market and allows one side to dictate pricing for non-existing tiers from other side. Pretty sure the high price I paid for a 1070 was due to this. So kind of hard now to grumble at the mention of AMD as of late.

 

[railven]

Participants who want to bash AMD every chance they get don't seem to grasp that a strong AMD is good for Nvidia fans, too. The only one it's bad for are Nvidia stakeholders, which makes one wonder about some of the posts here.

As someone who bought too many Radeons for one person, what exactly do you recommend someone like me do outside of venting my frustration with AMD's decisions. Even when looking at the bundles, it seems they left so much to be desired. $100 more for two games and coupons for products that aren't as enticing? Imagine how much more positive can be taken if they let you choose a 1080p or 1440p monitor with 100+ hz, where it seems Vega would really shine. Or R5 processors that are more than enough for gaming, cheaper, and allows the buyer to swallow a smaller up front cost - $1700+ vs <$1000 up front ($400 monitor (+$200 off), $500 GPU, $230+$120 CPU+MOBO (with $100 off) and 2 games.

I supported AMD/ATI for years when they had products I wanted. It seems outside of just buying AMD products to keep AMD healthy, there isn't much AMD offers someone like me (who upgrades regularly, and tries to push higher resolutions+refresh rates).

But yeah, I'm in it to see AMD fail, because I love paying $700+ for an NV GPU. You got me there. The funny thing about the bold part is with AMD's rise in stock value and cheap entry price, I'm starting to feel the most vocal to accepting AMD's own faults are their stockholders. Some of these posters are riding this train into early retirement. Kudos to them. Just want AMD needs, cheap and greedy user base that will drop the product the moment it doesn't satisfy their financial means.

 

[ozzy702]

Eh? I think we all know the obvious. In fact thats why I'm upset at AMD. Weak competition stagnates the market and allows one side to dictate pricing for non-existing tiers from other side. Pretty sure the high price I paid for a 1070 was due to this. So kind of hard now to grumble at the mention of AMD as of late.

Agreed, strong AMD CPU/GPU divisions are good for the consumer. I have several friends that waiting for Vega because they're fairly solid AMD fans and bought into the hype that Vega would beat a 1080ti for 1070 money. In the meantime I bought a 1070 a year ago and it's paid for itself several times over, and I've enjoyed flawless gsync and an overall great experience. Sorry, but AMD needs to get their act together and hopefully their success with Zen will open up a significantly larger R&D budget. By the time Volta hits AMD will be two full generations behind with manufacturing costs that exceed NVIDIA. Not a good place to be.

 

[PeterScott]

Participants who want to bash AMD every chance they get don't seem to grasp that a strong AMD is good for Nvidia fans, too. The only one it's bad for are Nvidia stakeholders, which makes one wonder about some of the posts here.

I don't think being critical of Vega is looking to bash AMD every chance they get. I love Ryzen, I say good things about Ryzen, I am thankful that Ryzen can keep AMD afloat, after this Vega debacle.

But the execution of Vega is so bad, it boggles my mind, so I feel the need to vent. And when people try to say with a straight face that Vega is a good product, well I am dumbfounded wondering what they see.

They used 3.9 Billion transistors to boost the clockspeed over Fiji? NVidia boost the clockspeed of Pascal way over Maxwell and Vega, seemingly at the cost of ZERO transistors. I don't like using picture memes, but imagine the largest facepalm of all time.

They are using a Titan Sized die, and HBM memory, to equal a nearly year and half old GTX 1080 running on much smaller die with cheaper memory. AMD has to destroy it's own margins, and sell at fraction of NVidias margin, to present something that only a minority would really consider as a reasonable purchase.

Vega is astonishingly bad. And it comes not out of the blue, but emerges from what appears to be a deepening downward spiral. I thought Fury X was a desperation play, and this just looks like Fury X redux, or worse.

I really can't see how the AMD GPU development turns around without a massive overhaul including lots of new blood.

I don't expect a turnaround anytime soon, and that makes be very sad.

 

[Elixer]

And this is the monitor AMD is promoting as a good example of freesync by putting it in the bundle... There is no quality control or min standards - you can label pretty well anything a freesync monitor.

Yeah, that is really mind boggling.
At the very least, they should have had a Freesync-2 monitor in the bundle, not that crap monitor they are bundling.

In fact, I bet the whole reason they are bundling that monitor is because samsung wants to EOL it, and they got a sweet deal on them.

 

[raghu78]

I don't think being critical of Vega is looking to bash AMD every chance they get. I love Ryzen, I say good things about Ryzen, I am thankful that Ryzen can keep AMD afloat, after this Vega debacle.

But the execution of Vega is so bad, it boggles my mind, so I feel the need to vent. And when people try to say with a straight face that Vega is a good product, well I am dumbfounded wondering what they see.

They used 3.9 Billion transistors to boost the clockspeed over Fiji? NVidia boost the clockspeed of Pascal way over Maxwell and Vega, seemingly at the cost of ZERO transistors. I don't like using picture memes, but imagine the largest facepalm of all time.

They are using a Titan Sized die, and HBM memory, to equal a nearly year and half old GTX 1080 running on much smaller die with cheaper memory. AMD has to destroy it's own margins, and sell at fraction of NVidias margin, to present something that only a minority would really consider as a reasonable purchase.

Vega is astonishingly bad. And it comes not out of the blue, but emerges from what appears to be a deepening downward spiral. I thought Fury X was a desperation play, and this just looks like Fury X redux, or worse.

I really can't see how the AMD GPU development turns around without a massive overhaul including lots of new blood.

I don't expect a turnaround anytime soon, and that makes be very sad.

Its obvious that GCN has hit its limits and a new architecture is badly needed for AMD to stand any chance of competing in the high end GPU market in the future. AMD does not need new blood as they have the talent. They need strong leadership and bold decisions. Going for a clean sheet design is the only option for the long haul. AMD needs to try and emulate what they did with Zen.