[Benchlife] R9 480 (Polaris 10 >100w), R9

[Adored]

AMD buys around $1bn worth of wafers from GF yearly. Compared to Nvidia they have vastly more supply, and they need it because of the consoles - but it's only good so long as GF isn't messing things up.

Going forward though, GF will be a major benefit to AMD assuming their strategic alliance with Samsung continues.

 

[raghu78]

Price points are of course driven by wafer yield. When you consider that A9 chip at Glo Flo is 96mm as compared to 104.5 at TSMC you can extrapolate that out to a 232 mm2 chip from Glo flo is comparable to a 252 mm2 TSMC chip. Then depending on each process yields from each foundry on a 300mm wafer this will lead to lower pricing on AMD products. Also consider this that a 610 mm2 GP100 would be a 558 mm2 Glo flo chip.
On a 300mm wafer 232 mm2 can yield 246 dies as opposed to 252 mm2 providing 225. I believe this is a significant advantage.

You are talking 14LPE vs 16FF+ . 14LPP is expected to bring further reduction to die size over 14LPE due to process maturity and transistor improvements.

http://www.eetimes.com/author.asp?section_id=36&doc_id=1328711&page_number=5

"What should we expect for their 14 nm Low Power Plus (LPP) process? Samsung’s recent press releases suggest that the LPP process will feature a 15% increase in the transistor switching speed and a 15% power consumption decrease. These are being achieved by increasing the transistor’s fin height and enhanced strain engineering. For me, I am anticipating a bit of a process shrink as well to bring the transistor and 6T SRAM cell sizes closer to that of Intel."

 

[tviceman]

RS, 390X + 10% is within range of Fury X, especially at 1080p.

I don't agree with your regarding the 290X failing to deliver. Are you thinking straight? It beat Titan as it was designed to do, at nearly half the price. Even with that crap reference cooler.

390X + 10% at 1080p is a gtx980. While I personally wouldn't consider that a disappointment for a ~230mm2 chip with a 120 watt TDP, it wouldn't be exciting by any means.

 

[AtenRa]

If Polaris 10 is not 20-30% faster than Hawaii at 1080p then it will be a disappointment.

Polaris 10 will have more transistors than Hawaii, it will have a lot of architectural enhancements and it should have higher clocks as well. If it will not reach performance close to Fiji at 1080/1440p, now that a lot of games are coded for GCN, then they better increase at least the perf/watt more.

 

[swilli89]

390X + 10% at 1080p is a gtx980. While I personally wouldn't consider that a disappointment for a ~230mm2 chip with a 120 watt TDP, it wouldn't be exciting by any means.

Possibly true if you are talking about 2013 games, but for games released in 2015 and 2016 they are tied in lower resolution and the 390x beats the 980 at high resolutions.

For pretty much any DX 12 game this is going to look more like 980 + 15% = 390x.

 

[flopper]

If Polaris 10 is not 20-30% faster than Hawaii at 1080p then it will be a disappointment.

Polaris 10 will have more transistors than Hawaii, it will have a lot of architectural enhancements and it should have higher clocks as well. If it will not reach performance close to Fiji at 1080/1440p, now that a lot of games are coded for GCN, then they better increase at least the perf/watt more.

amd already said it be a nice surprise.
question to be answered what performance delta they went for.

 

[jpiniero]

If Polaris 10 is not 20-30% faster than Hawaii at 1080p then it will be a disappointment.

Seems unlikely this will happen if they are really calling it the 480(X).

 

[bradsh]

Seems unlikely this will happen if they are really calling it the 480(X).

The product naming is all speculation based on speculated performance from speculated TDP... it's all in the air.

 

[el etro]

If Polaris 10 is a 120W ACP card, having 2,5 times performance/watt means that it is faster than even the Fury X.

 

[IEC]

We'll see when it launches. I'm inclined to believe Polaris 10 will be closer to 390 performance levels, but with some driver optimizations and the new hardware features may approach 390X levels.

Personally, I think equivalent performance to a Fury X is overselling the card before any reliable benchmarks have been seen.

 

[el etro]

I would still prefer top dog Polaris is a 100W card, that would make the GPU a Notebook champ. Anyway ~120w of ACP and 2,5 times of perf/watt would mean monstrous performance...

 

[Kenmitch]

I would still prefer top dog Polaris is a 100W card, that would make the GPU a Notebook champ. Anyway ~120w of ACP and 2,5 times of perf/watt would mean monstrous performance...

Didn't they state something like console like performance in a thinner, lighter laptop? Thin laptops don't really have the best thermal capabilities.

 

[railven]

These recent posts/rumors have got me excited that team red is going to bring good game. I don't want to get on this hype train, damn you fury - you broke me!

Be nice to sport a Radeon in my main rig again. Got nothing against my 980 Ti, but it just isn't a Railven rig without a Radeon.

 

[Madpacket]

Polaris 10 will replace 390/X cards and hopefully at a lower price point. Fury-X continue to be AMD's top card until Vega arrives.

 

[R0H1T]

If Polaris 10 is a 120W ACP card, having 2,5 times performance/watt means that it is faster than even the Fury X.

Even with GDDR5, I seriously doubt that! They'd high to clock it super high, both the GPU & memory, for it to lead the Fury X & 390(x) even at 1080p. However as & when the Polaris' come with GDDR5x then even the 980Ti or Titan are well within reach, they'll just have to clock them high enough & stick that Pro Duo cooler if need be

 

[Qwertilot]

Rather hope they have the self confidence to not pull that sort of stunt and wait for Vega. Just not what these cards are really about.

If the desktop card is 120w, then there will be notebook version. NV have just pushed something derived from a ~160W desktop card into some 'notebooks'!
(Luggables more like but its a lucrative market with seemingly very real margins.).

The 2.5 times performance (as with anything of this sort) will be measured in the most optimistic manner possible

 

[topmounter]

So either 390/x performance at 380/x prices or Fury/X performance at 390/X prices? It sounds like nVidia is targeting the latter price and performance targets based on the 1070/80 rumors.

 

[dark zero]

It's reminiscent of Zen hype in the other sub-forum -- 8 core Zen with Haswell IPC with 4.2-4.4Ghz boost, and while at it in a 95W TDP and priced under $400. This is how AMD products are set up to fail on this forum. A certain minority will keep hyping something up and when data comes out that shows it may not live up to their expectations, they disregard it and hype up the expectations even more.

Thus far, I have not seen 1, not 1 reliable source, that has been able to produce any evidence to support the claim that Polaris 10 will = Fury X @ 130W TDP.

It's all like statements that someone makes online, the statements themselves aren't verifiable. But then it magically becomes the truth and people go with it.

For example, in AMD's own slides, they claim 2.5X perf/watt compared to 2014 GCN parts. What does that tell me as a reader? Well it 100% means the 2.5X improvement had nothing to do with it over Fiji. And yet, here we are with people going along with a 2560 shader 130W chip ~ 280W Fiji. So this forum magically went from having 390X +5-10% performance in Polaris 10 (280W vs. 130W), to now claiming that nope, it's not good enough. Let's just bump that up to Fury X. And then others on other forums (WCCFtech or Videocardz) even go on as far as to say Polaris 10 is a fail if it cannot easily beat 980Ti/Fury X. It's because people cannot accept that NV's GP104 will smash Polaris 10 into the ground and that AMD is 6 months behind with Vega 11.

I personally will not believe that Polaris 10 with 8Gbps 256-bit memory bus and rumored 230-240mm2 die size will match the Fury X.

Like think about it, on April 26, AMD is launching Radeon Pro Duo with an expensive water cooler for $1500 US, but just 1.5 months later here comes Polaris 10, with 8GB of VRAM, using 1/2 the power and costing way less than a Fury X? So why wouldn't AMD just cancel that Radeon Pro Duo and release a dual-Polaris 10 card without expensive water cooling, more VRAM, better features (HDMI 2.0a, DP1.3, etc.)

Logically, this tells me Polaris 10 is being overhyped everywhere online. It's a staight up midrange product that's now being hyped way way up since it looks like AMD will have nothing to counter 1080. It also contradicts AMD's own strategy of bringing 290/290X spec to more affordable (sub-$349 price levels: AMD's own words, own presentation). So now I am supposed to believe that Polaris 10 will cost less than $349 and be as fast as the Fury X? I am just following along all of this logic.

Sorry, but I don't buy it. AMD's track record speaks for itself -- 290X, Fury X, massive hype and both failed to deliver. I'd rather be cautious than hype up another midrange AMD chip.

Even sites like Videocardz are starting to post pure garbage like Polaris 10 replacing Fury X, the same site that posted AMD's roadmap showing Vega as a Fury X replacement. Facepalm!

With the same logic, nVIDIA won't make it too... 980 Ti would be the best card for a long while...

Also that means that the hype is dimnishing very hard to the point that it would be a not as big improvement as expecting... the improvements over 980 Ti according to the leaks has the similar hype like this case.

In AMD case they won't deliver the Fury X performance since is not possible... they can match at best the Vanilla Fury instead. Similar to nVIDIA and their 1080. Both will be also so rare due supply and demand like the Fury Nano that it will be very hard to obtain one of these to fully test it.

And considering that the 1st generation of a shrinking die is not as great as expecting... is better to wait to the 2nd generation of 14/16nm cards, once the process becames mature.

 

[el etro]

Don't matter that much the limitations imposed the memory bus and clockspeeds. 980 didn't need much to be faster than 780TI, have 2/3 of it only and outperform Kepler comfortably.

Polaris/Vega may come with a major overhaul on the memory controller and the memory system(cache, registers and all other Sram available ln the GPU), making more intelligent/rational/efficient use of all the vRAM Bandwidth available. And once vRAM B/W is not a bottleneck, its perxormance can go high enough.

 

[Silverforce11]

Actually, the new pre-fetch patent is making Polaris use some very CPU-like features to save on bandwidth requirements.

Primitive Discard Accelerator removes scene complexity to improve performnace, but logically, it should also lower bandwidth requirements to render a scene as a result.

Cautious optimistic, but my biggest concern is whether Apple wants all the full Polaris 11 and 10 dies for their Mac Books leaving only cut chips for PC gamers.

 

[swilli89]

Actually, the new pre-fetch patent is making Polaris use some very CPU-like features to save on bandwidth requirements.

Primitive Discard Accelerator removes scene complexity to improve performnace, but logically, it should also lower bandwidth requirements to render a scene as a result.

Cautious optimistic, but my biggest concern is whether Apple wants all the full Polaris 11 and 10 dies for their Mac Books leaving only cut chips for PC gamers.

Nah, I have a feeling Apple will be content with Polaris 11 as it should easily be a nice bump in performance and a massive reduction in power from the 380x chips they were receiving now. A full Polaris 10 might end up being just a tad over the thermal design target for Apple's notebooks. Beyond that, I don't think their Mac Desktops really ship high enough volume to strain supplies of Polaris 10.

 

[Silverforce11]

Nah, I have a feeling Apple will be content with Polaris 11 as it should easily be a nice bump in performance and a massive reduction in power from the 380x chips they were receiving now. A full Polaris 10 might end up being just a tad over the thermal design target for Apple's notebooks. Beyond that, I don't think their Mac Desktops really ship high enough volume to strain supplies of Polaris 10.

You will be surprised to learn that Mac Pros ship in great volume, enough to propel AMD's Firepro line from ~20% marketshare to ~30% for workstation/servers.

Polaris 10 would be like Tonga, full dies goes to Apple's lineup and cut dies to PC... that's my concern. It'll be like we're second class citizens!

Whether this happens or not really depends on GF's ability to have 14nm FF with good volume/yields.

 

[tential]

These recent posts/rumors have got me excited that team red is going to bring good game. I don't want to get on this hype train, damn you fury - you broke me!

Be nice to sport a Radeon in my main rig again. Got nothing against my 980 Ti, but it just isn't a Railven rig without a Radeon.

Once people start adopting freesync/gsync I don't see how this will matter as much. The performance delta between cards will be less significant if you can't tell the difference in actual gaming. That's why I think the most important upgrade thing is to grab both the adaptive sync monitor and new GPU this gen.

 

[Silverforce11]

Once people start adopting freesync/gsync I don't see how this will matter as much. The performance delta between cards will be less significant if you can't tell the difference in actual gaming. That's why I think the most important upgrade thing is to grab both the adaptive sync monitor and new GPU this gen.

Eh, even if u got smooth gaming at 45 fps, PCMR would still prefer 60 fps or more.

Especially for online shooters or anything competitive.

 

[tential]

Eh, even if u got smooth gaming at 45 fps, PCMR would still prefer 60 fps or more.

Especially for online shooters or anything competitive.

None of that contradicts with getting a Gsync or Freesync monitor.

Online shooters or anything competitive run at high FPS quite easily so that's not even worth bringing up.