• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Vega/Navi Rumors (Updated)

Page 181 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
OatisCampbell said:
My re-spin theory is the only thing that makes sense to me here. Even if that's true the Frontier launch seems more damaging than positive to me.
Click to expand...

Unless they realized it late and already started mass-production and now had 20k "defective" chips before they could stop it. Instead of trash-bin they came up with the idea of the FE. But this doesn't sound very plausible. Why then not release the FE 2 month before RX launch?
 
Veradun said:
That's also my theory. Using Financial Edition as name would have sound a bit shady, so they went for Frontier.
Click to expand...
Or perhaps they picked "Frontier" thinking of the pioneers- lost in the heartland, sometimes going in circles, getting there slow in ox drawn covered wagons?

This is all starting to remind me of the ill fated FX5800 launch. Wasn't that partly due to process issues, the part released originally planned to be a developer design, and the consumer version released 6 months later?

(although obviously a Vega at 1080 level is a much more viable part than the FX5800 was)
 
beginner99 said:
Unless they realized it late and already started mass-production and now had 20k "defective" chips before they could stop it. Instead of trash-bin they came up with the idea of the FE. But this doesn't sound very plausible. Why then not release the FE 2 month before RX launch?
Click to expand...

To be honest, I think the PR damage done by selling defective chips as frontier editions would be a lot worse than the cost of manufacturing another 20K chips. And AMD is not so stupid that they would not realize this.

Imagine if word got around that a Frontier Edition was 20% slower than RX Vega even when using exactly the same drivers (assuming that is possible). Customers who bought Frontier Edition cards would never buy AMD again.
 
WX4100 is slower than RX 560 and costs 2.5x more. That is normal. You are buying support for professional workloads, not pure performance
 
Ancalagon44 said:
To be honest, I think the PR damage done by selling defective chips as frontier editions would be a lot worse than the cost of manufacturing another 20K chips. And AMD is not so stupid that they would not realize this.

Imagine if word got around that a Frontier Edition was 20% slower than RX Vega even when using exactly the same drivers (assuming that is possible). Customers who bought Frontier Edition cards would never buy AMD again.
Click to expand...
Exactly, I would hope AMD is smart enough to realize that selling a "defective" card will lead to lost sales and bad PR in the long run. I don't think it's a respin issue.
 
cytg111 said:
New Wait a minute... this thing is perf wise ~vanilla-1080@180watts ... and this thing is 300watts+ ? Something must have gone horrible wrong...
Click to expand...

It's not like Fury had perf or efficiency advantages either. It is business as usual for RTG.

JoeRambo said:
Looks like a shrink of Fiji on 14nm @ ~1.25Ghz clock with updated GCN architecture from Polaris and 16GB of HBM2 memory?
Click to expand...

Kudos to AMD for beating my clock estimate by 400Mhz. Sure while it is way out of sweet spot for clock/power, it is still huge advance for company that had 40-50% clock deficit vs Nvidia.

No worries tho, "wait for Navi" train leaves on July 30.
 
,
thilanliyan said:
Exactly, I would hope AMD is smart enough to realize that selling a "defective" card will lead to lost sales and bad PR in the long run. I don't think it's a respin issue.
Click to expand...

Correct me if I'm wrong, but aren't re-spins often process improvement revisions that bring about more chips per wafer, less leakage, chips capable of running at higher speeds?

What if Vega was designed for a 1600MHz base clock and that plus some variable boost is the card they want to release? (and the card that beats 1080, maybe sometimes competes with 1080Ti for less cash?)

That would make this strange launch make more sense. Low take rate on $1000/$1500 card that seems slower than NVs $500 card, but cash grab to recoup cost of low yield wafers.

Veradun's "Financial Edition" if you will.

Wasn't it the Thermis that had a similar launch back in the day? If you bought a rev 1 (or whatever launch was) Thermi, it was one hot, inefficient beast.
But by the time they stopped selling them they were a different animal altogether.

That kind of thing wouldn't be AMD's "fault", it would be "business as usual" but on a slower timetable because they were behind on development.
 
OatisCampbell said:
YES
This is what amazes me about this "launch".

No review samples is the equivalent of movies released directly to rental or theatres without prescreenings- not a good sign.

Phantom Edition availability and $1000/$1500 MSRP? Yeah that will thrill masses of gamers waiting for two years for FuryX follow up.

"Prosumer" card first? There just a "few" less programmers than video gamers. If you release a Phantom Edition gamer card that at least makes it into the hands of 20 review sites and a few hundred gamers the masses are whipped into a frenzy of anticipation and they will wait another month.

My re-spin theory is the only thing that makes sense to me here. Even if that's true the Frontier launch seems more damaging than positive to me.

I still believe a $499 RX with 1080 level performance and a Freesync monitor is a combination that must be considered in 2017, and that 1080Ti competition is not necessarily relevant.
Click to expand...

I'm not sure why you can't hold the belief that you have in bold, while still not believing EVERYTHING else you said. Everything you said in this post is exactly true. This is ridiculous. It's not like I'm buying a 1080Ti or a GTX 1080 now. If you do you're insane because of what you stated in bold. I'm tired of the remarks I've gotten when:
I literally wouldn't use a free GTX 1080Ti.

Which is a fact many posters just don't understand here. Because they're so focused on the Nvidia vs AMD thing, they assume that you must be for a side. That's not the point. The point is what you stated in bold. You just can't beat the combination. This is just the best combination for the money out there with the most cost effective upgrade path for 4K owners who intend to upgrade their 4K monitor and want to use adapative sync.

It just sucks it had to be like this.

There is nothing that makes me happy about buying this card.
 
To be fair, you shouldn't be surprised this is the impression people get when you make melodramatic posts at a breakneck pace over the last 2 months, alternating between Vega being the second coming my lord and savior to deliver us from evil, to Vega stole my first born and wizzed on my power supply.

Anyway, I'm not entirely convinced it's a huge problem that the FE is fairly uncompetitive with a Titan in gaming performance. But, I think AMD has really set themselves up for that comparison by using the Titan as their basis for workload comparisons. Even if nVidia wanted people to believe the Titan was a prosumer card at one point, the reality is it's not, and it's a bit disingenuous for AMD to be claiming victory in a segment the Titan isn't actually occupying. Furthermore, this sets AMD up for failure when the FE can't compete in gaming performance when the market views the Titan for what it really is--a halo early adopter tax gaming product.
 
notT1beriu said:
Ryan from PCPer, in their live FE review. said that AMD told him that Gaming Mode doesn't do anything. It's just a GUI change. 😱
Click to expand...

He also said expect up to an 8% boost with RX Vega according to reddit.

I really wish anyone else would have gotten this card card other than PCPer though.
 
tential said:
He also said expect up to an 8% boost with RX Vega according to reddit.

I really wish anyone else would have gotten this card card other than PCPer though.
Click to expand...
They bought a couple of them, I don't think AT would do that, nor would most of the other sites.

They rather sit back and wait for AMD (or...) to give them cards (or...), rather than buy them themselves.

As for the drivers, yes, it is a given that AMD's drivers get better performance over time, but, even with a 8% boost, things aren't adding up.
 
Well, I did say that Vega 11 could possibly be the new Polaris replacement to fill out the line and that they never intended to go for the high end...

Looks like that could be true and that's what Vega 11 really is releasing later this year.
 
Arachnotronic said:
8% is pretty solid. All we need to see now is pricing.
Click to expand...
Up to 8%, which ain't much. I think the main issue is that the GPU can't reliably maintain 1600 MHz... There's another 10-15% performance on the table that can't be extracted due to not hitting those peak clocks consistently. On the other hand, I am hard pressed to imagine that Vega is simply Polaris with higher clocks and higher core count. There's been so much hyped up in the Vega architecture that there has to be inconsequential IPC gains, but they aren't getting utilized effectively due to driver reasons or whatever. We'll soon find out in 1 month, I suppose.
 
The problem with blowers is... they are blowers. I don't know of one blower card that is a good cooler, and quiet.
I have a feeling that it just isn't performing well enough to maintain 1600MHz.
The RX version really needs a Tri-X type cooling, and we can then see if thermals play a big role or not.

I really think they should take apart the card(s), check everything, and then re-apply thermal paste and see where it gets you. If you are gonna test these cards, might as well go all out.
 
Saylick said:
Up to 8%, which ain't much. I think the main issue is that the GPU can't reliably maintain 1600 MHz... There's another 10-15% performance on the table that can't be extracted due to not hitting those peak clocks consistently. On the other hand, I am hard pressed to imagine that Vega is simply Polaris with higher clocks and higher core count. There's been so much hyped up in the Vega architecture that there has to be inconsequential IPC gains, but they aren't getting utilized effectively due to driver reasons or whatever. We'll soon find out in 1 month, I suppose.
Click to expand...

Yep, no sense worrying too much about it now. Wait for reviews and make a purchasing decision based on those.

Simple 🙂
 
Aircooled Vega FE was running around 1400MHz on almost every test PCper ran. Well cooled (Open air Tri-X style or AIO'd) gets that up to a fixed 1600MHz, there's 10-15% extra performance. That lands square right into 1080 territory.

The other 25% to match the 1080Ti (stock...), well.. we don't know how it overclocks. Yet it's already power throttling because it's hitting its 300W board TDP limit at 1400MHz, shown in the captured power data in the stream. The watercooled FE card has a 375W board TDP limit, so there you have it. Fury X wasn't this power hungry out of the box.

It sure as hell won't get a 25% boost from gaming drivers in a month, unless they're trying to bamboozle the entire internet while tarnishing their new brand in the process. Up to 8% is PCper's estimate for driver improvement. That may or may not be the best AMD can do with this new architecture for gaming.


I'm truly worried about a clock for clock comparison between Fiji and Vega. Both can be very easily be set to say 1100MHz core and memory clock speeds to have the same memory bandwidth and do this comparison, since they have the exact same number of functional units everywhere, we get to see exactly what all those architectural improvements (are they?) are doing.


Meh. Unless something outright magical happens to RX Vega, in the end pricing will make or break this. Zen was the product that couldn't fail or be this lackluster and it delivered more than it promised. At least AMD now has some actual incoming cash to fuel better R&D in Navi and future Zen iterations.
 
Last edited:
french toast said:
Up to 8℅? Oh dear, things have gone badly wrong here, if that's even remotely true raja has got to go.
Click to expand...
This "Raja has got to go" narrative is such bullshit. Maybe this is the best anyone could have done with the budget allocated to RTG, Zen was their main focus and it shows.
Also, if Vega performans like this (maybe with 10%+ due to drivers), they knew that months ago. He would probably have been fired months ago (see AMDs leadership change before Bulldozer launch), rarely is someone let go because of internet outrage or bad reviews. They know their product and can act accordingly long before anyone on the outside knows.
 
Peicy said:
This "Raja has got to go" narrative is such bullshit. Maybe this is the best anyone could have done with the budget allocated to RTG, Zen was their main focus and it shows.
Also, if Vega performans like this (maybe with 10%+ due to drivers), they knew that months ago. He would probably have been fired months ago (see AMDs leadership change before Bulldozer launch), rarely is someone let go because of internet outrage or bad reviews. They know their product and can act accordingly long before anyone on the outside knows.
Click to expand...
Although it is not looking good, the gaming Vega card with updated drivers might still show a greater then expected increase in performance (> 10-20%). Who knows, and if it still disappoints somewhat then it's not the end of the world. I still have high hopes for their new APU with Vega included.
 
I cannot believe it would take over 300W to match GTX 1080 performance. If so, it isn't any more efficient in performance per watt than Polaris is.

Think about it - crossfire 480's can match or beat a 1080, and use around 300W. So this is not any more efficient, in fact it may be less efficient.

Could be a few possibilities:
Process issues result in terrible leakage, which pushes power draw and heat up. But, AMD is no stranger to 14nm anymore, having made a few products on it by now.
Driver issues make it perform far worse than it should. Easily possible, but still doesn't explain the 300W power draw.
Design issues. In which case this is like 2900 XT vs 3870. Wait for the Vega respin which will cut power draw nearly in half.
 
Status
Not open for further replies.

TRENDING THREADS

Back
Top