• 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.

[Rumor (Various)] AMD R7/9 3xx / Fiji / Fury

Page 56 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Fury pro will sell well if under $500. Better if significantly under 500. the e3 conference has a chance of turning things around on the marketing front as well. AMD could make themselves the manufacturer to have for dx 12 and the future of PC gaming.

$500 for a fury pro I can dream and it be the card to own the last 20 years for sure if that happens, omg only 1 day to go soon
 
where on earth are these assertions coming from? 😵 😕 😵

I doubt people are taking you seriously knowing your apocalyptic leanings when it comes to AMD and the GPU market.

Fury pro will sell well if under $500. Better if significantly under 500. the e3 conference has a chance of turning things around on the marketing front as well. AMD could make themselves the manufacturer to have for dx 12 and the future of PC gaming.

If they can price Fury Pro at $499 and be just 10% slower than the GTX 980TI, that would be a game changer. 2 Fury Pro for under $1000 will make smooth 4k gaming very 'affordable' relatively speaking. To me, if it the 4GB doesn't become a problem at 4k, it would be hands down the GPU to get.
 
No no no, see, you can't just focus on low end development, that's when you get your ass kicked the next generation. You have to focus on building the next generation of tech that will trickle down. If AMD thinks they can drag another year out of their current range (while remaining competitive perf/$) while developing new tech they will use at 14nm, then that's the logical thing to do over a multi-year period.

Also, how many gens has AMD owned the midrange and lower end market? It didn't do them any good.
 
You figure people know about marketing by now to not say what the max draw is by specs but actual use, I know its not 375w BUT that wont matter as all I see is 375w and that will soon be stated to be with the 390 series even worse than last time.
do the same as nvidia state a different tdp how hard can it be?
its writing a few different numbers ffs.

they need to fire a few at amd camp and hire someone who knows more about perceptions.

That's Sapphire's own site. We don't know what AMD will claim for TDP.
 
Absolutely. The reason must be related to Apple.

The number of imac's apple sells is a tiny fraction of the number of airs and pro's they sell. The number of 27" imacs is a fraction of the number of 21.5" imac's they sell. The number of upgraded 27" imacs they sell is a fraction of the number of number of 27" stock imacs (295m is a BTO option).

Imac is not the reason they don't have tonga. Likely there is some minor but significant flaw in tonga (or production problems/yields) that is holding tonga back.

Dont try to reason with him, if AMD would spend money for the low end cards, he would say AMD should of spend money on the high end where the big profits are. 😛

Fury is looking like, at the company level (not consumer level), a major disaster. HBM1 is troublesome and yields are low and costs are high. AMD is eating the early adopter fee hard. Limited to 4 GB the chip is a tough sell for 90% of the professional market. IMO with only 4 GB AMD should have done what nvidia did and stripped out a lot of DP (drop down to 1/4 or 1/8 rate from tahiti and Hawaii's 1/3 and 1/2 rate) as professional apps that need DP tend to need a lot of RAM. Fiji isn't going to sell very many cards being at the extreme high end; fixing the mid range and getting a good laptop die out there would have worked wonders.

In constast Nvidia is playing it cool and quiet. By putting emphasis on efficiency they have driven AMD out of the notebook market. Going for bandwidth compression enables their cards to do well with smaller buses allowing them to easily wait out HBM2 (still have wiggle room: they could conceivably release a 16nm 512 bit 7 ghz GDDR5 card and get performance scaling on the order of 33-40%). Nvidia doesn't need HBM now.

Fiji could have been possible with a beefed up 512 bit bus at slightly higher clocks (Hawaii is not bandwidth limited) and tonga colour compression. Scaling would still be good and power consumption would be 15-25W higher. AMD could have put 8 GB GDDR5 on it and called it a day.

As much as people seem to be hating on Pitcarin what exactly is the problem with the chip? It performs fine (though its large and uses more power than maxwell). None of the newer features really help too much at the budget level.
 
Considering what little we know about VRAM usage (especially on GDDR5) and the largely inflated numbers we've seen up until now, it's preemptive to whine about 4 GB.
 
From what I've read it sounds like the 390x is a 290x with 8gb of memory, nothing more nothing less.:'(

It might be? I was only stating his card doesn't look the same as the cards Sapphire is showing. All we can see are pics, but the 3xDP, 1xHDMI, 1xDVI is an update over the card the poster has.
 
$500 for a fury pro I can dream and it be the card to own the last 20 years for sure if that happens, omg only 1 day to go soon

If Wccftech has accurate prices for the 3xx series, then we should have a Fury pro around $500.

If Fury > Titan X in performance

Fury WCE = more that 980Ti = $799-849 ?
Fury X = a bit less = $ 599 ?
Fury Pro = $ 499 ?

We have to remember that GM200 is 601 mm2 [larger than Fiji die]and very few will sell in the professional high price market. Nvidia also has higher margins than AMD and can sell at $649, so don't let anyone state as a fact that we can't see a $500 Fury model.

Anyone here stating they know the costs of producing Fury is guessing. The assumption is, new must always be more expensive. That is a fair assumption but not some law of the universe. The art and science of engineering always tries to lower costs by introducing new designs, materials and techniques.

A bit less than 2 days for official info.
 
Last edited:
Can you provide a confirming source for the text in bold?

Fury is looking like, at the company level (not consumer level), a major disaster. HBM1 is troublesome and yields are low and costs are high. AMD is eating the early adopter fee hard. Limited to 4 GB the chip is a tough sell for 90% of the professional market. IMO with only 4 GB AMD should have done what nvidia did and stripped out a lot of DP (drop down to 1/4 or 1/8 rate from tahiti and Hawaii's 1/3 and 1/2 rate) as professional apps that need DP tend to need a lot of RAM. Fiji isn't going to sell very many cards being at the extreme high end; fixing the mid range and getting a good laptop die out there would have worked wonders.

In constast Nvidia is playing it cool and quiet. By putting emphasis on efficiency they have driven AMD out of the notebook market. Going for bandwidth compression enables their cards to do well with smaller buses allowing them to easily wait out HBM2 (still have wiggle room: they could conceivably release a 16nm 512 bit 7 ghz GDDR5 card and get performance scaling on the order of 33-40%). Nvidia doesn't need HBM now.

Fiji could have been possible with a beefed up 512 bit bus at slightly higher clocks (Hawaii is not bandwidth limited) and tonga colour compression. Scaling would still be good and power consumption would be 15-25W higher. AMD could have put 8 GB GDDR5 on it and called it a day.

As much as people seem to be hating on Pitcarin what exactly is the problem with the chip? It performs fine (though its large and uses more power than maxwell). None of the newer features really help too much at the budget level.
 
Fury isnt a new tech besides HBM. And tech havent trickled down. The 250 and 270 aka 370 is a prime example of this. And this is one of the reasons AMD is bleeding marketshare at an alarming rate. 35% to 22.5% in 1 year.

Tonga introduced GCN 1.2. The only GCN capable of 4K video decode as well. Almost a year later and what products got it besides Tonga? Unreleased Fury and unreleased Carrizo.

And dont expect any new GPU tech at 14nm from AMD. It looks to be GCN 1.2 shrinks at best when listening to their financial analyst day.

We don't know what if any improvements were made to Fury's version of GCN.
 
I'm curious if the ASUS versions of these cards will be using auto extreme. 390x or fury. Every since I read about the automated process and realized people were man-handling my GPUs during manufacturing, I figured I would have to get asus for my next GPU.

I feel bad people will lose their jobs, but other manufacturers need to get on this tech.

https://www.youtube.com/watch?v=aA7U6AcJCK8
 
Last edited:
Can you provide a confirming source for the text in bold?

Its an assumption on my part. I apologize.

But basically, reading between the lines.

http://www.pcworld.com/article/2922...ill-power-its-next-gpus-pokes-nvidia-too.html

Macri said HBM is new, but that doesn’t mean people should assume yield issues. Macri wouldn’t elaborate on yield from its chief partner in the project, Hynix, but said AMD wouldn’t adopt it for a consumer part if it didn’t think it could get enough HBM RAM to make GPUs.
If things were going well you say things like, "Everything is on track, yields are high, [positive things]". Company speak for "things are not so good" is 'no comment'.

Later on in the video he 'refuses to elaborate about the 4 GB issue'. Well, at this time we pretty much know Fiji and its derivatives are 4 GB only (does not preclude an 8 GB down the road but launch is strictly 4GB).
 
Question, the migration path from hbm 1 to hbm2 is it just new chips connected via an interposer to a Fiji or comparable gpu? Doesn't hbm shift the memory controller logic to the memory dice?
 
I'm curious if the ASUS versions of these cards will be using auto extreme. 390x or fury. Every since I read about the automated process and realized people were man-handling my GPUs during manufacturing, I figured I would have to get asus for my next GPU.

I feel bad people will lose their jobs, but other manufacturers need to get on this tech.

https://www.youtube.com/watch?v=aA7U6AcJCK8

There cooler needs to fit Fiji properly, unlike Hawaii, before I would consider an Asus Fury. Also not ever heard anything good about Asus GPU RMA procedures. I would consider XFX first, if their implementation is good. Their support is top notch.
 
I'm still very curious as to why they are calling 390/390X "Grenada" when according to all current evidence is just a rebrand?

Looks like a rebrand to me. Who was it that said "do you really think AMD will just rebrand a card and charge more for it, if it didn't have better performance"?

H66ZrcT.gif
 
Its an assumption on my part. I apologize.

But basically, reading between the lines.

http://www.pcworld.com/article/2922...ill-power-its-next-gpus-pokes-nvidia-too.html

If things were going well you say things like, "Everything is on track, yields are high, [positive things]". Company speak for "things are not so good" is 'no comment'.

Later on in the video he 'refuses to elaborate about the 4 GB issue'. Well, at this time we pretty much know Fiji and its derivatives are 4 GB only (does not preclude an 8 GB down the road but launch is strictly 4GB).

First you apologize for stating assumption as fact then you do the same thing with your [positive things] comment. 😕

AMD does not comment on unreleased products in general, not just when things aren't good.
 
This post from ocn is similar to one that influenced my thinking on AMD and dx12

The main difference between Maxwell and GCN is that GCN is a massively parallel architecture with higher compute power and the ability to perform more tasks asynchronously from the graphics queue. This difference is not going to be noticed until DX12 and Vulkan take foothold. Then people are going to realize just how gimped of compute power Maxwell 2 is.

Hell, even AMD's years old cards support feature level 12_0 in DX12 and nothing before Maxwell 2 supports it. Nvidia's claims that Maxwell 2 is the only one that supports full DX12 is also FUD since they don't support resource binding tier 3, it does however support feature level 12_1 which so far none of AMD's cards have been confirmed to support.

This is good article explaining an architectural lead AMD has over Nvidia going into the DX12 and Vulkan era: http://www.anandtech.com/show/9124/amd-dives-deep-on-asynchronous-shading


Looks like a rebrand to me. Who was it that said "do you really think AMD will just rebrand a card and charge more for it, if it didn't have better performance"?

http://i.imgur.com/H66ZrcT.gif[/IMG[/QUOTE]

PCB can be similar or almost identical. We care about the GPU
 
Fury is looking like, at the company level (not consumer level), a major disaster. HBM1 is troublesome and yields are low and costs are high. AMD is eating the early adopter fee hard. Limited to 4 GB the chip is a tough sell for 90% of the professional market.

I know its a safe guess to call it expensive and poorly yielding, but do you have any solid information on the yields and cost for HBM?

Also there is nothing stopping AMD releasing an 8GB version for the pro market. If Fiji performs well in compute then they will want to take advantage of nvidias lack of maxwell options in that market.
 
Back
Top