Vega/Navi Rumors (Updated)

[Mopetar]

Not recently, there was the initial rumor then there was an event held in Asia somewhere that had systems running Vega playing Prey set up (I can't find the images right now since I'm at work). There were banners that had AMD logos and an "Optimized for Vega" which all looked pretty legit. That has been literally all I've seen though so it could have just been a really early tech demo and we're hyping the correlation a bit too much.

It probably is legitimate then. I couldn't remember if it was something "official" or was a leak like someone had taken a picture of a poster or cardboard cutout.

While that is true, it makes absolutely no sense. "Vega is coming, so prepare yourself by finishing Prey ahead of time!" I guess the message must be that Vega is so awe-inspiring that you would either want to be staring into your case or at just a blank desktop, as playing a game might ruin the feeling of awe? Yeah, I'm not quite buying that.

The only way it makes for effective marketing is if Vega knocks the socks off of a Titan in this game. They could have no card, but show some performance results and Vega beating a Titan Xp by 15% and people would probably be pretty hyped. Even if that does happen, I probably wouldn't draw general-case performance from such a benchmark unless AMD made official comments to that effect.

Now, while all of that is certainly possible, but I'd bet my money on inept marketing by AMD. That's a far more likely outcome.

 

[Crumpet]

Considering how many gaming pc components, including graphics cards that I buy... I'm starting to feel like I should spend more time playing games and less time building things to play them on...

...i'm still buying Vega though

 

[swilli89]

While that is true, it makes absolutely no sense. "Vega is coming, so prepare yourself by finishing Prey ahead of time!" I guess the message must be that Vega is so awe-inspiring that you would either want to be staring into your case or at just a blank desktop, as playing a game might ruin the feeling of awe? Yeah, I'm not quite buying that.

If you are planning on playing Prey and you are doing a new build it doesn't really matter which you buy first if they don't come out the same day. Yeah people that play other games are in a different boat but in that case they can just enjoy Prey earlier on their older hardware before upgrading.

 

[Valantar]

If you are planning on playing Prey and you are doing a new build it doesn't really matter which you buy first if they don't come out the same day. Yeah people that play other games are in a different boat but in that case they can just enjoy Prey earlier on their older hardware before upgrading.

I disagree. The subset of people either building their first gaming PC or upgrading their PC to be able to play Prey among people who want to/are going to play Prey is bound to be tiny. As such, the vast majority will prioritize finisihing the game with their current hardware - unless their current hardware either sucks horribly and would make the game look like a**.

I wouldn't be surprised if AMD actually has a public sit down/stand-up event on Prey release day and, at the very least, goes over Vega tech slides, actual numbers, and performance--at least as it relates to Prey. It seems like a standard dual-marketing opportunity to use a real piece of software that is specifically designed for optimal performance on the new and mysterious hardware within Vega.

That could be. But I have to agree with Mopetar here:

The only way it makes for effective marketing is if Vega knocks the socks off of a Titan in this game. They could have no card, but show some performance results and Vega beating a Titan Xp by 15% and people would probably be pretty hyped. Even if that does happen, I probably wouldn't draw general-case performance from such a benchmark unless AMD made official comments to that effect.

Having a pre-release marketing event alongside a game launch simply to say "hey, we're here too" would not be a good move. They would need to have a good point to make with that presentation. And it could definitely be argued that "teasing" beating Nvidia without launching the card would be as likely to piss people off as excite them. AMD has to tread a fine line here. There are plenty of pitfalls, and not many productive ways forward for them, regardless of Vega's performance.

 

[Mockingbird]

Prey was suppose to launch alongside Vega as a launch title to demonstrate the new features in Vega.

 

[happy medium]

ok guys instead of reading 94 pages , I'll just ask the 3 most important questions....

Is there a date for the Vega launch?
How fast is it going to be?
How much does it cost.

or is this 94 pages of guessing so far, and we know nothing.
thanks,

 

[Mockingbird]

ok guys instead of reading 94 pages , I'll just ask the 3 most important questions....

Is there a date for the Vega launch?
How fast is it going to be?
How much does it cost.

or is this 94 pages of guessing so far, and we know nothing.
thanks,

Even AMD doesn't know the answers to those questions because the product isn't finalized yet.

 

[Arachnotronic]

Prey was suppose to launch alongside Vega as a launch title to demonstrate the new features in Vega.

I think this was an assumption made up by people on the internet rather than anything verifiable.

 

[IEC]

Assuming rumors of a late May/early June launch turn out to be true, the product was finalized a while ago. You cannot change anything significant in production this late in the game.

 

[Arachnotronic]

Assuming rumors of a late May/early June launch turn out to be true, the product was finalized a while ago. You cannot change anything significant in production this late in the game.

Silicon is done, no way they can make significant changes there at this point. I bet you it's about stockpiling chips for the launch & finalizing drivers.

 

[Cookie Monster]

If June launch is correct, thinking the fine tuning of clocks vs power consumption is already complete (or doing very last minute tune ups clocks/voltage wise).

 

[.vodka]

We'll see what happens in ten days or so with the Prey thing. May 5 IIRC.

If nothing happens, we still have May and June in Q2. Big events coming are Computex (May 30 - June 3) and E3 (June 13-15). It should be launched in any of those two, if not there's still another two weeks until June 31... but then it'd be a mostly lost opportunity IMO. Silicon is finalized, it's mostly tweaking and final refinements at this point for a confirmed Q2 launch.

Computex is the rumored release stage for X399 and 16C32T HEDT Ryzen, so at least there's that to cure the itch for the moment...


For f***'s sake I hope AMD gets the voltage dialed in right this time. There's no need for another RX580 situation where most chips can be undervolted by about 0.1v or more with a corresponding >20-30w power drop and therefore be run at top boost clocks all the time with efficiency nicely going up... save for the balls to the wall 1450MHz out of the box models that actually need the voltage. I understand this is done to maximize usable chips at x frequency, but come on. It hurts them more than it helps when it's review time.

 

[Bacon1]

For f***'s sake I hope AMD gets the voltage dialed in right this time. There's no need for another RX580 situation where most chips can be undervolted by about 0.1v or more with a corresponding >20-30w power drop

Yeah AMD has been terrible about that. Hawaii, Fiji and Polaris are all overvolted out of the box and it really kills their efficiency.

 

[Paratus]

Since clock speeds, memory and die sizes have all been theorized to death what outputs are we hoping for?

 

[IllogicalGlory]

Since clock speeds, memory and die sizes have all been theorized to death what outputs are we hoping for?

I'd hope for Fury X (at good utilization) * clock speed increase at least. If we take the recent Dawn of War benchmarks as a sign of what Fiji is really capable of (it's the only AMD card that does unusually well in that game), that's about 1080 Ti level, maybe a bit higher. Holding out a bit of hope for arch increases to pull it further. Best case (and I mean best) it's a GV104 competitor.

I feel like I'm going to be sad looking back on this post after it releases tho.

 

[Cookie Monster]

For f***'s sake I hope AMD gets the voltage dialed in right this time. There's no need for another RX580 situation where most chips can be undervolted by about 0.1v or more with a corresponding >20-30w power drop and therefore be run at top boost clocks all the time with efficiency nicely going up... save for the balls to the wall 1450MHz out of the box models that actually need the voltage. I understand this is done to maximize usable chips at x frequency, but come on. It hurts them more than it helps when it's review time.

It would probably hurt them more if they forgo their yields in return for tightening the tolerances.

Imagine having major OEMs calling up due to stability issues.. ugh would be a nightmare to be honest. They've probably done their analysis (monte carlor monte carlo monte carlo!) and a comprehensive one on clocks vs voltage vs power consumption. Something like >100,000s of units or more. Not just <20 users in a forum reporting back their success in under-volting their GPUs.

If they want to fix their power consumption problems, they should go and look at the root of the issue which is the GPU architecture itself. And quite frankly, they are falling behind in this area vs the competition. Hopefully VEGA will turn that around or narrow the gap at the least.

 

[alcoholbob]

Silicon is done, no way they can make significant changes there at this point. I bet you it's about stockpiling chips for the launch & finalizing drivers.

If they are surprised by 1080 Ti performance, the only real tweak they can do is up the clocks on the top SKU via binning and sell the lower clocking cards as a lower sku/model.

 

[tential]

ok guys instead of reading 94 pages , I'll just ask the 3 most important questions....

Is there a date for the Vega launch?
How fast is it going to be?
How much does it cost.

or is this 94 pages of guessing so far, and we know nothing.
thanks,

We know nothing lol. Come on man!

 

[GodisanAtheist]

If they are surprised by 1080 Ti performance, the only real tweak they can do is up the clocks on the top SKU via binning and sell the lower clocking cards as a lower sku/model.

- Absolutely no reason AMD should be surprised by the 1080ti's performance. NV has been following the same playbook since the 600/700 series.

 

[sandorski]

I suspect that at least AMD will leak something with the Prey release. Leaks should start to come out in May, even if Vega will be released in June.

 

[3DVagabond]

What about 95-105% of GTX 1080 Ti performance for 499$?

Do you think they can bring us HBM2 for $500?

 

[Arachnotronic]

Yeah AMD has been terrible about that. Hawaii, Fiji and Polaris are all overvolted out of the box and it really kills their efficiency.

They needed to meet a certain performance spec and they needed a large % of dies to be able to achieve that spec, that meant upping the average voltage on shipping parts.

 

[Glo.]

Do you think they can bring us HBM2 for $500?

What I have heard in retail rumors is that if the GPU die is small enough, and it does not require big interposer, HBM2 tech is able to go to 300$ market as of today. When production will ramp up, and the volume will be higher - the prices for technology package(Interposer, TSV, HBM stacks, assembly...) will also go down, and HBM2 will be able to penetrate lower price markets.

As of right now 300$ is the goal, for the tech. Of course, we are talking about two stack tech. Design with 4 memory controllers, and 4 memory stacks may be more expensive and not available to anything less than 600$ at this very moment.

Once again, HBM2 is slated to go not only to GPUs, but also AMD APUs.

It does mean simple thing. Do not expect HBM2 this year to go into market that has prices lower than 349$. Anywhere, regardless if it is APU or GPU.

 

[lobz]

If they are surprised by 1080 Ti performance, the only real tweak they can do is up the clocks on the top SKU via binning and sell the lower clocking cards as a lower sku/model.

how could have they possibly been surprised by 1080ti performance, when the pascal titan x (GP102) was already out for ages?

 

[Paratus]

I'd hope for Fury X (at good utilization) * clock speed increase at least. If we take the recent Dawn of War benchmarks as a sign of what Fiji is really capable of (it's the only AMD card that does unusually well in that game), that's about 1080 Ti level, maybe a bit higher. Holding out a bit of hope for arch increases to pull it further. Best case (and I mean best) it's a GV104 competitor.

I feel like I'm going to be sad looking back on this post after it releases tho.

I know. I was actually asking about what versions of Display Port and HDMI we should expect. Along with how many of them.