Fury Nano: First results in!

[Abwx]

They can't sell the Nano at too low price because the card is extremely fast. Much faster than an ASUS GTX 970 Mini, and the performance is not far from a Fury X. Nano will beat the normal Fury, so I'm pretty sure it will be higher priced ... $600 or more. If we just judge the raw performance $619 should be a reasonable price, but the 175 watt TDP can increase this.

They should price it as much as they can, this is a halo product that has no equivalent in the competition, downclocked 980ti cant match it because their memory subsystem power comsumption is higher than the Nano s and that it wont scale down if the GPU frequency is reduced, AMD is untouchable in this segment thanks to HBM.

 

[AtenRa]

Mantle is dead. DICE broke it's function awhile ago, now it runs slower than DX11 and broken for GCN1.2. It did its job (Vulkan/Metal/DX12), let it die gracefully.

What the hell, look post above.

Also, anyone playing BF4 with a GCN GPU and knows what hes doing they play with Mantle. There is no dx-11 substitute for Mantle in BF4 period.

 

[zlatan]

Why do you think Nano will beat normal Fury?

Because I've tested one at my friends office. And also saw the specs. It's a Fury X with a little lower core clock. I'm pretty sure there will be some leaks in the coming days.

 

[Glo.]

Guys, Mantle is in DirectX 12.
https://community.amd.com/community/gaming/blog/2015/05/12/on-apis-and-the-future-of-mantle

The Mantle SDK also remains available to partners who register in this co-development and evaluation program. However, if you are a developer interested in Mantle "1.0" functionality, we suggest that you focus your attention on DirectX® 12 or GLnex

In plain english this means that OpenGL Next and DirectX 12 are based on Mantle.
At least Mantle 1.0 Functionality, and low-level part is in those APIs. So it still lives, just in another form .

 

[zlatan]

They should price it as much as they can, this is a halo product that has no equivalent in the competition, downclocked 980ti cant match it because their memory subsystem power comsumption is higher than the Nano s and that it wont scale down if the GPU frequency is reduced, AMD is untouchable in this segment thanks to HBM.

Yes, they have a big advantage in this market. But I'm pretty sure that not just the mITX fans will buy this. I'm also want it to my big developer machine. Almost Fury X performance under 175 watt is a big deal for me.

 

[tential]

Zlatan, your pricing/performance numbers are vastly different than what AMD themselves has so far shown us....

 

[zlatan]

Zlatan, your pricing/performance numbers are vastly different than what AMD themselves has so far shown us....

I don't know anything about the pricing. I'm just saw the specs and the performance, and from that I think they will sell it at a really high price, $600+. But $559 is the minimum, because the normal Fury is much slower than Nano.

 

[tential]

I don't know anything about the pricing. I'm just saw the specs and the performance, and from that I think they will sell it at a really high price, $600+. But $559 is the minimum, because the normal Fury is much slower than Nano.

Can you explain the chart amd provided in the first post then?

 

[Silverforce11]

Because I've tested one at my friends office. And also saw the specs. It's a Fury X with a little lower core clock. I'm pretty sure there will be some leaks in the coming days.

Well, 850mhz Fury X would in theory be ~20% slower than Fury X. That's a big gap.

Particularly since Fury X isn't much faster than 980 at 1080/1440p, it really only shines at 4K.

 

[ShintaiDK]

Ehm no,

Thief
Fury X (Mantle) vs Titan X (DX-11)
https://www.youtube.com/watch?v=FLtvOfrV7fM


BF4
Fury X Mantle 1080p Ultra
https://www.youtube.com/watch?v=nL6XGc16WGg

Fury X Mantle 1440p Custom
https://www.youtube.com/watch?v=Q5vW1ZBVNSo

Fury X Mantle 4K Custom
https://www.youtube.com/watch?v=291c-eLfPyo

Oh yes, random YouTube. I thought you learned the last time.

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/13

Mantle for GCN 1.0 and GCN 1.1 cards (7970, 290X), but we’re using Direct3D for GCN 1.2 cards like the R9 Fury X. This is due to the lack of Mantle driver optimizations on AMD’s part, and as a result the R9 Fury X sees poorer performance here, especially at 2560x1440 (65.2fps vs. 54.3fps).

http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/11

Initially, I tested BF4 on the Radeons using the Mantle API, since it was available. Oddly enough, the Fury X's performance was kind of lackluster with Mantle, so I tried switching over to Direct3D for that card. Doing so boosted performance from about 32 FPS to 40 FPS. The results below for the Fury X come from D3D.

 

[zlatan]

Can you explain the chart amd provided in the first post then?

I don't know what games they test. With my workload the core clock was stable at the highest speed. There might be some workloads when the card reduce it.

 

[zlatan]

Well, 850mhz Fury X would in theory be ~20% slower than Fury X. That's a big gap.

It has higher core clock. Much higher with four numbers not just three in MHz. I can't say more.

 

[Silverforce11]

I don't know what games they test. With my workload the core clock was stable at the highest speed. There might be some workloads when the card reduce it.

Okay, that makes more sense. Dynamic clocks, with a forced 175W TDP. This basically the same behavior as anyone on R290X or onwards, in CC with -50% power limit. My R290X hoovers 850-880mhz in gaming load (from 1ghz), power use drop massively.

 

[Glo.]

IIRC, one of the engineers in visual interview, at the day of presentation of Fury X, have said that Nano will be in range of Fury X, but it will be dependent on thermal envelope we will see there. Sometimes it may be as fast, sometimes it may be slower.

What happens is, I think, AMD put really low voltage on the GPU and put really steep TDP limit for it. Thats what will make it jump between core clocks depending on the situation. It may be one time 850 MHz, and one time 950 or even 1000 MHz.

It really depends on power envelope here, thats my theory.
In similar slides that have been leaked few weeks ago we've seen that in Valley 4K benchmark Fury Nano was on par with Fury X while using 157W of power.

 

[zlatan]

Oh yes, random YouTube. I thought you learned the last time.

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/13



http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/11

I'm pretty sure that Thief was patched for GCN 1.2. Update 1.7 or 1.8, I don't know exactly.
Battlefiled 4 was not patched, but the engine get some update so the newer Frostbite games works well.

This will be an ongoing problem with all explicit APIs like D3D12 and Vulkan. If a new GPU comes (especially with new microarch) it may not have better performance than 980 Ti and Fury X because the games won't have the optimal memory allocation scheme for it. The devs may need to patch the application.

 

[Silverforce11]

Yeah. Now we go from "meh" to "woah"... if true ofc!

 

[AtenRa]

http://itcm.co.kr/review/348489

 

[ShintaiDK]

This will be an ongoing problem with all explicit APIs like D3D12 and Vulkan. If a new GPU comes (especially with new microarch) it may not have better performance than 980 Ti and Fury X because the games won't have the optimal memory allocation scheme for it. The devs may need to patch the application.

If that is the case, then DX12/Vulkan will damage the industry and DX11.3 will come out as a winner.

 

[AtenRa]

If that is the case, then DX12/Vulkan will damage the industry and DX11.3 will come out as a winner.

Or devs will patch the games

 

[ShintaiDK]

Or devs will patch the games

What have they done so far to inspire you that they do that along the road? There isnt any economic incentive to do it on already released games. Sure you will see a few doing it. But you will see the vast majority not doing it due to cost. On the other hand it could be a DLC

 

[therealnickdanger]

I can't say I have much faith in developers. Games are released half-baked nowadays. Some don't get patched for months even to fix crashes. To think they will patch a game a year later just to get performance improvements is wishful thinking.

The saddest part is that probably the only games that WILL get this kind of ongoing attention will be F2P MMO/MOBAs to ensure ongoing subscriptions and playability. You can forget about "seasonal" games like sports games or any other games that get annual sequels - they make all their money in the first couple months of release and then fade away...

 

[zlatan]

If that is the case, then DX12/Vulkan will damage the industry and DX11.3 will come out as a winner.

This is a far more complicated question. A PC graphics API can be implicit or explicit. Both approach has it's own disadvantage.
For an implicit API like D3D11, the biggest problem is the kernel driver, especially when it is heavily threaded. The runtime validation will serialize to workloads and with this a lot of free CPU time will be inaccessible to the engine. There might be a chance to get good graphics on PC with D3D11, but it will need an AI/physics/simulation downgrade which is not useful when the consoles have more usable CPU time. An explicit API like D3D12 will give a solution for this problem by cutting out the kernel driver and the whole runtime validation process from the system. This will grant access to all available CPU time, so the AI/physics/simulation, basically all gameplay element can be better and better, but the downside is that the kernel driver won't exist anymore so the GPUs must be controlled explicitly by the devs.

I think an explicit API is a better approach, because it will be easier to deal with the newer hardwares and patch the program for those. Dealing with the heavily threaded D3D11 driver is much harder. Also it is impossible to optimize a game when the drivers don't allow me to see what is the problem.

 

[Phynaz]

Well you can plainly see from AMD's slide that, at least they claim, it's close to 980 speed. And if you shave 50W off of it then it is about 960 power consumption.

What is it exactly you are meaning to say?

I never said anything about 980 performance. I only spoke about power consumption.

 

[Phynaz]

You paying a $50 premium for.

Smaller size
Lower power consumption ??
Better DX-12 performance
FreeSync
Mantle
TrueAudio
Better CF scaling with lower stutter

Better DX-12 performance - maybe.
Mantle - Dead - it's never going to see release. Even AMD says it was a proof of concept only.
TrueAudio - Stillborn. And good riddance, audio doesn't belong on a video card.

 

[Enigmoid]

I can't say I have much faith in developers. Games are released half-baked nowadays. Some don't get patched for months even to fix crashes. To think they will patch a game a year later just to get performance improvements is wishful thinking.

The saddest part is that probably the only games that WILL get this kind of ongoing attention will be F2P MMO/MOBAs to ensure ongoing subscriptions and playability. You can forget about "seasonal" games like sports games or any other games that get annual sequels - they make all their money in the first couple months of release and then fade away...

Sadly I feel the same way. While there are great games, there are enough mediocre and terrible games that I'm really not sure if I want devs to have more low level control over the game.

Patches for a SP game after 2+ years are pretty much non-existent.