Fury Nano: First results in!

[redzo]

http://videocardz.com/57438/amd-rad...ainst-geforce-gtx-970-in-mini-itx-form-factor
30% faster than GTX 970 SFF.

Looks like AMD marketing tries to sell the nano by comparing it to a mini 970. Unfortunately for them, reviewers are probably going to place it price wise against the 980.

 

[ShintaiDK]

Looks like AMD marketing tries to sell the nano by comparing it to a mini 970. Unfortunately for them, reviewers are probably going to place it price wise against the 980.

And using 4K with very selected settings

 

[Alatar]

Pretty much every test with 0xAF again?

just why?

 

[zlatan]

I am not sure if this is an error or a fake slide or the Nano does have some flawed/disabled ACE parts other than the traditional shaders, TMUs and ROPs. Otherwise, it's hard to explain how a fully unlocked 1Ghz Nano is only 5-10% faster than a 290X.

All newer GCN 1.2 cards have this configuration. There are 4 core ACEs. The two HWS units can do the same work as 4 ACEs, so this is why AMD refer to 8 ACEs in some presentations. The HWS units just smarter and can support more interesting workloads, but AMD don't talk about these right now. I think it has something to do with the HSA QoS feature. Essentially the GCN 1.2 design is not just a efficient multitask system, but also good for multi-user environments.

Most GPUs are not designed to run more than one program, because these systems are not optimized for latency. They can execute multiply GPGPU programs, but executing a game when a GPGPU program is running won't give you good results. This is why HSA has a graphics preemption feature. These GCN 1.2 GPUs can prioritize all graphics task to provide a low-latency output. QoS is just one level further. It can run two games or a game and a GPGPU app simultaneously for two different users, and the performance/experience will be really good with these HWS units.

 

[Glo.]

Fury, and Fury X have different layout, at least from AMD slides. And they are GCN 1.2

 

[redzo]

And using 4K with very selected settings

Yes. Marketing folks hide product cons all the time.
It looks like they are trying to justify a probable premium price, but price wise is the way it should have been. It used to be the amd way of things.
It is a tough job to sell a product like this. AMD's marketing only hope achieving anything is if the mini high performance gpu is a niche market and if the supply was meant to be on the low side in the first place.

 

[Erenhardt]

"Premium industrial design" <facepalm>
Get out!

Industrial and premium is like boiling icecream, ot doesn't fit!

Another one of nvidia gimmicks that was tounted from the rooftops as an extra feature that ends in amd products.

Give me premium Laguna design, or premium Octoberfest design. Industrial? Industrial is practical, rough, suited for the task without any weight put into looks, feel etc.

<flip table>
</rant>

 

[KingFatty]

Is there a design term to capture a design based on the design elements of matte black, full metal, and brushed aluminum? Those are premium elements, but not sure what to call a design based on them.

 

[Erenhardt]

Is there a design term to capture a design based on the design elements of matte black, full metal, and brushed aluminum? Those are premium elements, but not sure what to call a design based on them.

I have a goldsmith in my family, I guess our point of views differ quite a bit when it comes to "premium elements" and "premium finish".

Actually, stone encrusted GPU would be quite premium I guess... depends on the stone.

 

[RussianSensation]

Looks like AMD marketing tries to sell the nano by comparing it to a mini 970. Unfortunately for them, reviewers are probably going to place it price wise against the 980.

That's not how I am reading it. They are not discussing price but are addressing power in a certain form factor. In their own slides they use R9 380 and 970 as points of reference but it doesn't mean the Nano will be priced at 970/380's levels.

What they are talking about is how they've been able to cram so much GPU power in such as mall form factor.

Since the mini GTX970 is the fastest miniITX card out today and the Nano is aimed to compete in this space (i.e., mini ITX), that's why they are comparing it to the best product available in that form factor.

This is going to be downplayed as usual like R9 295X2/Fury X CLC, how HBM is downplayed -- all because NV didn't do it first. Everyone objective knows if GTX980Ti/Titan Z came with factory CLC, if GTX980Ti had HBM1, if 980/980Ti could be made in this miniITX form factor, it would be more beneficial for PC gaming community but of course it's AMD so it doesn't count.

And using 4K with very selected settings

Pretty sure if NV made a card this powerful and this small this generation, you would break the F5 key on your keyboard.

Even has an actively cooled VRM section, something not present on some premium cards. They really paid attention.

It's also interesting to note that the Gigabyte minITX GTX970 is getting destroyed for user satisfaction due to running hot, loud and having stability/crashing issues.

That means the only true direct competition in this space is the Asus miniITX 970 that costs $355 and a rebate.

If the Nano is only 5-7% faster than the R9 290X, it will have no competition in its intended space.

If it comes in at 192-195 on this chart, it'll be uncontested at $429-449.

For someone with a very small case, having the flexibility of full Fury X performance at 1.05Ghz clocks and yet the flexibility of running a cool and quiet card is unique. Waiting for overclocking, temperature and power usage results.

 

[RussianSensation]

Personally, I am not interested in miniITX GPUs but I can see how there is demand in this space as some cases cannot accommodate 980/980Ti/Fury style cards.

As long as GPUs keep getting smaller and prices of PCIe M.2/U.2 drives start coming down to earth in the next 5 years, case makers will have more incentive to produce smaller chassis. Once HBM2/3 is ubiquitous, there will be little reason to make 10-12" cards outside of flagships or dual-GPUs. In the beginning though, this is an early adopter product so I expect a price premium since there is no competition in this space.

Slowly the market is moving into the direction of smaller desktop PCs. Silverstone already introduced a 700W Platinum small-form factor PSU.

All newer GCN 1.2 cards have this configuration. There are 4 core ACEs. The two HWS units can do the same work as 4 ACEs, so this is why AMD refer to 8 ACEs in some presentations. The HWS units just smarter and can support more interesting workloads, but AMD don't talk about these right now. I think it has something to do with the HSA QoS feature. Essentially the GCN 1.2 design is not just a efficient multitask system, but also good for multi-user environments.

Ok, thanks for clarifying. That's interesting. I wonder what changes to GCN AMD will introduce with GCN 2.0 in 2016 considering the existing GCN 1-1.2 are already good at compute/GPGPU tasks. I wish more games took advantage of compute shaders to help produce more realistic physics/game effects.

I would love to hear from a Naughty Dog developer/programmer on how they took advantage of Asynchronous Compute Engines in PS4 to make Uncharted 4. That game looks stellar given how much weaker the GPU in PS4 vs. say a Fury X/Nano.

 

[ShintaiDK]

Pretty sure if NV made a card this powerful and this small this generation, you would break the F5 key on your keyboard.

No for several reasons.

Its a pointless product in SFF because its directly unfit. Same reason why I wouldnt touch a 970 mini or anything like it. If you owned a SFF setup you would understand.
I am not emotional about it as you.
Cards with the same performance and power usage already existed for a year.

 

[dark zero]

And AMD did seppuku by pricing it at $649...
What a dissaster from AMD and a clear advise:

HBM might make the cards way more expensive than expecting... Maybe the idea to go HBM is not good after all.

Intel has the golden chance to take AMD and Nvidia down since HBM will be hella expensive

 

[x3sphere]

$649 - not the price most were hoping for but I can see why they'd charge a premium. That price is more justifiable than the Fury or Fury X in my opinion. Unlike the Fury X which is clearly beaten by the 980 Ti at same price point, the Nano has no direct competitor really if you want this kind of power in a small package.

 

[tential]

Where are people seeing this $650 pricing from?
WHAT?
Lol AMD....

 

[SPBHM]

$650 is to high, no way it's the correct price.

 

[happy medium]

Cards with the same performance and power usage already existed for a year.

Just as small for about 300$ too.
I don't think people will say "hey look a new card 12% faster for 479$" and dump their gtx970 mini's just yet.

 

[tential]

$650 is to high, no way it's the correct price.

Ya, if AMD misprices this, they're going to misprice it by a reasonable amount. Not just yolo price it.

 

[shady28]

Where are people seeing this $650 pricing from?
WHAT?
Lol AMD....

Seriously, where is that price coming from?

That's way too much. At 450-500 it would make people who are in the market for a good 970 think twice (which is a big group), but 650....

 

[casiofx]

Many people have forgotten that GTX 980 is only 15% faster than GTX 970 yet willing to pay 50% more for it.

 

[happy medium]

At 450-500 it would make people who are in the market for a good 970 think twice (which is a big group),

I think the people who wanted a mini 970 already bought one. It has been out for quite a while.

 

[tential]

I think the people who wanted a mini 970 already bought one. It has been out for quite a while.

It's a massive problem for AMD when they're the second to market. Hard to make a sale when someone already bought something. How many people couldn't wait any longer when the 980Ti came out and just bought a reference cooler? I saw a ton of people say "Can't wait for Fury X 980Ti yayayayay finally a new high end card!"

They need to come out with a quick release with a good performance jump that entices people to upgrade before NV can respond, or have enough of a performance boost to leak the performance as being a massive jump over NV, and have it actually be believable enough for people to wait.
After Fury X, "Hey, it's an overclocker's dream.", that thought will get trashed all over the internet.

So ya, they need to either be first to market, or hit a home run.

 

[AnandThenMan]

It's a massive problem for AMD when they're the second to market.

Fermi was 2nd to market and went on to outsell Evergreen. Brand image>being first to market (see HBM).

 

[boozzer]

Many people have forgotten that GTX 980 is only 15% faster than GTX 970 yet willing to pay 50% more for it.

Bro!!! that only applies to NV gpus don't you know!

amd gpu? please, it is trash!

 

[tential]

Fermi was 2nd to market and went on to outsell Evergreen. Brand image>being first to market (see HBM).

How out of touch are you with the market that you consider AMD to be first to market this gen because it was the first to have HBM?

HBM isn't some feature I can utilize as a consumer. It's the memory technology. What did HBM do for me this gen? It brought me Fury X, a card that underperforms the 980ti in anything but fringe cases.
R9 Fury, an oddly priced/performing card that while faster than GTX 980, offers no actual benefits. There isn't a resolution/setting the R9 Fury magically opens up that wouldn't work on the 980. In fact, the 980 would have more in that respect due to gameworks. It sucks, but this game works for Nvidia (get the pun?).
Then R9 Nano and the yet unreleased card aren't relevant to the vast majority of buyers.

Now tell me why that changes ANY of my opinions when you go "Oh, it was first to have HBM!". Ok...

Great the R9 390 is a better value than a GTX 970. It only took god knows how long to do so. In reality, the GTX 970 was a better deal. Because you could have been PLAYING GAMES THAT WHOLE TIME.

The problem is, AMD has minds similar to yours running the company. That care about being first with new technologies, new engineering feats, new APIs, etc. This doesn't convince customers to purchase your product, if the underlying product takes over a year before it can see it's full potential realized, and even then it's waiting for DX12 (see R9 290x).

Customers aren't GPU hobbyists. They buy GPUs around holilday seasons, when they get money, or when they can no longer play their games. GTX 970/980 launched before Christmas. AMD had no new card to compete. Now Nvidia captures a whole holiday season. It sounds glorious really when I think about it from a business perspective.
AMD better hope holiday buyers still remember the Fury X launch.