AMD: Pirate Islands (R* 3** series) Speculation/Rumor Thread

Kaotik said:
Wait what? There's Oxide benches somewhere on Fiji?
Click to expand...
Not to my knowledge, I meant to discuss the relative position of the two vendors.
The API overhead results in the picture are not a departure from what came before.

Titan X matches the 980, and Fury puts up numbers similar to Hawaii.
 
silent_guy said:
I don't think many people think the AIO cooler is a bad deal for the consumer. It may be a bad deal for AMD.
Click to expand...
It may be a bad deal how exactly? It's a premium cooler for a premium product. There isn't going to be a bad deal anywhere with this, especially if the performance really is even better than Nvidia's comparable offering.

You're just looking for an excuse to rain on AMD's parade, and I think that is ridiculous.

3dilettante said:
It's still a power-limited device at stock, perhaps more so since bandwidth has been significantly improved.
Click to expand...
How do you figure? It's watercooled. It has dual eight-pin power connectors. It's not going to be power limited, and saying that it is is just wholly unfounded. Quite the opposite - AMD repeatedly said the card is built to handle overclocking, so it follows it can't be power limited even at stock speed.
 
Grall said:
How do you figure? It's watercooled. It has dual eight-pin power connectors. It's not going to be power limited, and saying that it is is just wholly unfounded. Quite the opposite - AMD repeatedly said the card is built to handle overclocking, so it follows it can't be power limited even at stock speed.
Click to expand...
Power going into the board is power dissipated out the cooler, unless the board is planning on exploding.
I'm working on the assumption that at stock Fury X is trying to stay within standard board spec, as the 390 cards also try to do.

If at stock it is targeting 300W+, then it needs that water cooler and the comparisons become less favorable from an architectural perspective.
 
Grall said:
It may be a bad deal how exactly? It's a premium cooler for a premium product. There isn't going to be a bad deal anywhere with this, especially if the performance really is even better than Nvidia's comparable offering.
Click to expand...
It's a bad deal for AMD if it eats into their margins.

How hard is it to understand that some people like talk about hardware from an engineering/business point of view?

I'm sure the FuryX will be just as wonderful to own as a GTX 980 Ti. Be happy about that! I'd like to point out that I've commented plenty of times that, at their low prices, an R9 290 is too much of a good perf/$ deal to ignore, even if it's efficiency stinks.

But I'm not in the market for a GPU at all, and since this subforum is supposed to discuss architecture, technical aspects and their consequences, it's fair game to bring this up.

You're just looking for an excuse to rain on AMD's parade, and I think that is ridiculous.
Click to expand...
I will absolutely rain on AMD's parade if their only efficiency improvements can be attributed to using HBM and nothing else. And nobody with brand security issues is going to stop me.
Right now, the jury is still out on it.
 
According to PCPer:
http://www.pcper.com/reviews/Graphics-Cards/AMD-Exposes-Fiji-World-HBM-Enthusiast/Fiji-GPU

Fiji also improves upon what we first saw in Tonga. It can do as many theoretical primitives per clock (4) as Tonga, but AMD has improved the geometry engine so that the end result will be faster than what we have seen previously. It will have a per clock advantage over Tonga, but we have yet to see how much
Click to expand...

The chips features 4 shader engines each with its own geometry processor (each processor improved from Tonga)
Click to expand...

Fiji includes improved clock gating capabilities as compared to Tonga. This allows areas not in use to go to a near zero energy state. AMD also did some cross-pollination from their APU group with power flow. Voltage adaptive operations only apply the necessary voltage that is needed to complete the work for a specific unit. My guess is that there are hundreds, if not thousands, of individual sensors throughout the die that provide data to a central controller that handles voltage operations across the chip. It also figures out workloads so that it doesn’t overvolt a particular unit more than it needs to to complete the work.
Click to expand...
 
The application of Carrizo's adaptive clocking is one of the items I thought could be included in Fiji, which was modest for the APU in percentage terms, but a decent chunk for a device operating an order of magnitude higher in power consumption.
I don't think the description of voltage adaptive operation in the PC Perspective piece is accurate.
 
UVD 6
UVD 6.0 will be used by the upcoming AMD Radeon Rx 300 Series (Pirate Islands GPU family) and AMD Radeon Rx 400 Series (Arctic Islands GPU family), providing the Video Coding Engine 3.0 (VCE 3.0) technology and featuring a new high-quality video scaling.[15] The UVD version in "Tonga"-, "Iceland"- and "Carizzo"-based graphics controller hardware is also announced to provide support for High Efficiency Video Coding (HEVC, H.265) hardware video decoding;[16] however, as of May 2015 there are no announcements about the VP9 video codec support.[17][18][19]

https://en.wikipedia.org/wiki/Unified_Video_Decoder
 
pharma said:
Best case? Meaning comparing an OC'd Fury X against a reference 980ti?
Click to expand...

No, it's reference to reference.

The "best case" simply refers to the practice of benchmarking your new product with basically every game out there, and only releasing the benches that show it in the best possible light. That's something everyone does.

(edit) wait how quickly is this thread filling?
 
http://www.forbes.com/sites/jasonev...-new-fiji-graphics-card-beats-nvidias-980-ti/

Looking good for AMD :)
Fury-X-4K-Gaming.png
 
Last edited:
UniversalTruth said:
UVD 6
UVD 6.0 will be used by the upcoming AMD Radeon Rx 300 Series (Pirate Islands GPU family) and AMD Radeon Rx 400 Series (Arctic Islands GPU family), providing the Video Coding Engine 3.0 (VCE 3.0) technology and featuring a new high-quality video scaling.[15] The UVD version in "Tonga"-, "Iceland"- and "Carizzo"-based graphics controller hardware is also announced to provide support for High Efficiency Video Coding (HEVC, H.265) hardware video decoding;[16] however, as of May 2015 there are no announcements about the VP9 video codec support.[17][18][19]

https://en.wikipedia.org/wiki/Unified_Video_Decoder
Click to expand...
Rx 300 Series doesn't have UVD6, that's for sure. Fiji's might have, though, and that's what I'm trying to find out.
 
3dilettante said:
Power going into the board is power dissipated out the cooler, unless the board is planning on exploding.
I'm working on the assumption that at stock Fury X is trying to stay within standard board spec, as the 390 cards also try to do.
Click to expand...
...Yes? And? AMD claims 50C at typical loads, and 500W cooling solution. How would the card possibly be power constrained if either of those are true (much less both; and why would AMD lie about it?)

If at stock it is targeting 300W+, then it needs that water cooler and the comparisons become less favorable from an architectural perspective.
Click to expand...
AMD states same power dissipation as good ole Hawaii (275W), which aircools just fine, seeing as the product is over a year and a half old already. What's there to be concerned about, and why the need to make up hyperbolic speculation about 300+ watts and whatnot? :p

silent_guy said:
It's a bad deal for AMD if it eats into their margins.

How hard is it to understand that some people like talk about hardware from an engineering/business point of view?
Click to expand...
Margins schmargins. You have the AMD execution woes thread to talk business; this is the GPU board thread. That said, even if their margins take a (likely minor; these water coolers aren't exactly made of unobtanium) hit, having a premium product with cutting edge performance is still beneficial, as it builds their presence in the high performance market and builds goodwill with hardware enthusiasts; both of which is sorely lacking at the moment. Amongst other things. You're being too myopic here, try looking at the bigger picture instead.

I will absolutely rain on AMD's parade if their only efficiency improvements can be attributed to using HBM and nothing else.
Click to expand...
What does it matter how they obtain their efficiency improvement as long as they do obtain it?

Even so, your concern would seem to be unfounded, as HBM alone could not possibly represent a full 50% increase in perf/watts. That would mean GDDR bus and chips in Hawaii ate up more than half the board power, which is simply absurd.

And nobody with brand security issues is going to stop me.
Click to expand...
PFFFFFTTTTT!!!

Take that kind of bullshit someplace else. You can find plenty threads of me absolutely lambasting AMD, and not even very long ago either for that matter.
 
Grall said:
...Yes? And? AMD claims 50C at typical loads, and 500W cooling solution. How would the card possibly be power constrained if either of those are true (much less both; and why would AMD lie about it?)
Click to expand...
There's no rule that a GPU must use the full dissipation capacity of its cooling solution at stock. If the card limits itself to 300W of supplied power, then it remains power constrained because power in is power out. Dissipation concerns are one of the reasons why designs have their power ceilings, but since watts are watts thanks to conservation of energy, a design is power limited if the value is fixed on the supply or dissipation side.

AMD states same power dissipation as good ole Hawaii (275W), which aircools just fine, seeing as the product is over a year and a half old already. What's there to be concerned about, and why the need to make up hyperbolic speculation about 300+ watts and whatnot? :p
Click to expand...
The static leakage component of a chip is heavily influenced by temperature, when Hawaii's standard chip temps is 95C. I provided an example where the exact same Hawaii chip drew 28W less thanks to a drop in chip temps due to a change to a closed-loop cooler.
Static leakage scales in terms of the number of transistors, which Fury has 40% more of.
That means, assuming relatively constant process parameters, that if a 40% larger chip were pegged at 95C, even at the same dissipation level, it would have less actual power budget to play with than the smaller chip.
It could have cost the chip a speed grade or two, and likely would have tripped thermal throttling since Hawaii already has notable instances where its clocks get nowhere near the "up to" portion of its spec.

And then we see benchmark wins from AMD where some of them are about a speed grade or two away from being a loss.
I am curious to see where Fury's clock consistency goes with the air-cooled variant.
 
3dilettante said:
The static leakage component of a chip is heavily influenced by temperature, when Hawaii's standard chip temps is 95C. I provided an example where the exact same Hawaii chip drew 28W less thanks to a drop in chip temps due to a change to a closed-loop cooler.
Static leakage scales in terms of the number of transistors, which Fury has 40% more of.
That means, assuming relatively constant process parameters, that if a 40% larger chip were pegged at 95C, even at the same dissipation level, it would have less actual power budget to play with than the smaller chip.
It could have cost the chip a speed grade or two, and likely would have tripped thermal throttling since Hawaii already has notable instances where its clocks get nowhere near the "up to" portion of its spec.

And then we see benchmark wins from AMD where some of them are about a speed grade or two away from being a loss.
I am curious to see where Fury's clock consistency goes with the air-cooled variant.
Click to expand...


Yep, and also with increased leakage, to compensate the chip needs to use more power.
 
You must log in or register to reply here.

Similar threads

U
AMD: Southern Islands (7*** series) Speculation/ Rumour Thread
236 237 238
Replies
5K
Views
1M
BrynS
B
Back
Top