ASUS Republic of Gamers Announces Strix RX 480
Click here to post a comment for ASUS Republic of Gamers Announces Strix RX 480 on our message forum
Dygaza
Suprisingly big performance upgrades they promise with so little clock boosts. Just shows how inferior cooling reference design has. Small boost in clocks, no throtling from thermals and powerlimit, and we're almost up to 20% in performance.
Denial
Interestingly enough, the FE model of the 1080 is the same way. The stock fan profile tanks the frequencies after like 10 minutes.
http://media.bestofmicro.com/W/3/581763/gallery/01-Clock-Rate_w_600.png
I don't get how both companies screwed this up, it's not like it's their first GPU launch. Especially Nvidia, who is charging a premium.
Anyway the Strix stuff is cool -- I usually go with EVGA for my GPU's but decided to try ASUS this time and I'm impressed with the cooler. It feels solid and the RGB stuff is pretty neat, although i ended up turning it off. Going to be interesting to see this compared to the Nitro and stuff.
DarkQuark
I am desperately trying to decide between 480 and 1060. It seems to me that the 1060 has the edge overall. But one thing I wonder about is that extra 2GB of memory on the 480 (6GB vs 8GB).
Does anyone see that making a difference?
Dygaza
Quess it's only logical for them, as investing more money to design better cooling for reference might be waste, as AIB's would still go, and make their upgraded models. Quess the real question is, why bother with reference model in the first place?
Denial
Currently? No. In the future? Maybe.
Honestly my opinion is that if you plan on using the card as your main GPU for more than a year, get the RX480. If not, get the 1060.
Dygaza
I'm on the same line. Both cards offer good value.
Fox2232
This is purely nVidia's fault. People did not care about power efficiency much. Then Maxwell came as power efficient and nearly everyone started to think that power efficiency means better fps or huge financial savings.
And as with 14/16nm transitions, both companies overestimated final power efficiency gains and underestimated stable clock.
So in the end both of them could clock higher and AMD/nV decided to clock them higher for marketing purposes even while they knew about throttling.
And both manufacturers used higher Voltage than needed which is part of TDP problem.
All I needed was to see DX12 AMD/nVidia's titles and Doom Vulkan.
Considering that all big game engines support at least one of those, we can expect this to be pretty standard situation soonish.
Koniakki
Well, if going by HH latest GTX 1060 FE review both perform about the same @1080p with the GTX 1060 having a slight edge overall here and there.
That could change in future as many suggested if AMD history is any indication.
About the VRam, I wouldn't sweat on it. Both are fine for 1080p for some time to come.
I would wait for AIB RX 480 reviews before deciding just to get a fuller picture.
waltc3
I've heard that the 1060 has no SLI capability, whereas the RX 480 series does do x-fire, so between more 25% more very fast Vram and the capability of going x-fire if you want (if it's true that nVidia left it off just to lower the price of the card)--it would seem a no-brainer for the RX 480, imo.
Noisiv
And it does mean better fps. At the very least at high-end and in mobile.
Power ceiling being fixed at 250-300W for high-end desktop GPU, better perf/W is the only way to get more performance from the next GPU generation.
Look at Polaris vs Pascal to see how perf/W translates in pure perf:
If AMD wanted to challenge GTX 1080 on pure performance basis, they would need pretty much double RX 480, ie 400mm2+, 300Watt+ just to challenge 314mm2 171W GTX 1080.
Which would be a disaster of epic proportions, and hence it's not happening(!)
Instead we'll have to wait for HBM2 Vega improving 50% on perf/W for AMD to challenge 1080 on pure performance basis.
Is that good enough example to explain how Efficiency ~ Performance?
Where Efficiency does not matter too much is mid-range desktop.
120 or 160W is not that much of a big deal. Although I would take 120W part every time, particularly in this hot time of year, and simply because low-watt gaming is kewl 😀
No one is giving a hard time to 480 because it uses 40 watts more. Because 40 watt difference is rather unimportant when it comes to these two products.
Where this difference becomes important is when you look at the underlying arch and prospects of future/missing products.
Orthello
If considering an AIB card , the 480 vs 1060 looks to be a real gain fort the 480 , not so much the 1060.
From Asus they are saying 15-19% from the strix 480 card at factory OC , from a link at HardOCP today "clocked at 1873MHz in OC mode, ROG Strix GeForce GTX 1060 delivers up to 5%-faster performance in 3DMark Fire Strike Extreme and 6.5%-faster gaming performance in Doom®."
Both cards can still be overclocked further , i would say by a similar amount of real world gain , 100 mhz on 480 will be equal or better to 200 mhz on pascal.
So early days without reviews but if that panes out you can see where its heading - DX11 parity or win for 480 , RX480 DX12 win , RX480 Vulkan win.
Dygaza
Way too overoptimistic ratio between frequency scaling. Keep in my that for AMD the biggest gain on that promised 15-19% increase, doesn't come from clockspeed itself, but from better powerlimit/cooling. The card doesn't throttle so it can actually maintain boostclocks, unlike reference models. Remember there were people pointing out almost 10% increases just by undervolting their reference 480.
Fox2232
Doubling transistors is not 100% necessity. There are parts of GPU which will not be doubled. Like PCIe IO, VCE, Scheduler, internal bus, ...
+RX-480 is not really that power hungry on GPU side, solid part of consumption is blower and gddr5.
At mobile, it sure is important. But we have not seen those yet. Secondly Polaris have seen nice improvement over Fiji on terms of higher performance per ROP, TMU, SP, Memory bandwidth at same clock.
And on top of this improvement, if ASUS's statement is solid and their 5% OC + proper cooling/power delivery brings 15~19% improvement in performance over stock card...
Back in the day, I have seen Fiji as more capable than Maxwell per transistor investment and same clock. But AMD as always picks denser manufacturing, higher leakage, and lower stable clock. Improvement from Maxwell to Pascal is much smaller than Fiji to Polaris. And as such AMD/GloFo has to fix that clock and leakage issue... and we have big improvement maximum achievable performance per single GPU.
Noisiv
it mostly comes from 1120 MHz base clock used for ref. card
1330/1120 = 119%
Noisiv
that's why I called it 400mm2+, and not 2x 232 = 464mm2 part
+5% OC =+19% performance? Perpetuum mobile? give me a break...
It's +19% over base clock locked card.
We've been over these high hopes before 480/Pascal launch.
("All AMD needs to do is... increase clocks")
And what did we get? MHz gap is bigger than ever. Performance gap is bigger than ever.
And every other Maxwell metric advantage only got increased. Additionally AMD's time-to-market just cratered, nothing, nada, MIA...
AMD is still missing perf/W work that Nvidia has done Kepler -> Maxwell. If they can do that AND cover anything Nvidia comes up in the meantime, they're back in game.
Dazz
I would wait and see how the RX 480 custom cards pan out as at the moment they are hitting power draw bottlenecks so with an 8pin power connector they may have very good overclocking potential RX480's have good performance gains from very modest overclocks no doubt this is due to the RX480 having higher IPC compared to Maxwell and Pascal. Must say the RX480 custom cards are taking quite a while to get to market and not seen a review on one yet.
Orthello
Asus included the benches , its 15-19 % over the ref cards 1266 boost. Of course the Ref card is throttling and its 1440 and 4k numbers they are quoting. Yes the 15-19% likely comes from holding the boost clock vs throttling it and more available board power. Either way it is a Ref 480 1266 mhz boost card they were comparing not a Boost disabled card which would be rediculous to include.
It means real world without touching any voltage / power sliders 15% - 19% in 1440p and 4k respectively , i think the test 3dmark in 1440p and hitman at 4k.
The ref card is the one in the reviews after all, and its looses by far less %s if at all in 1440p to a 1060 , so these will be significant gains for the AIB 480 cards in the reviews vs the 1060 ref / 1060 aibs.
SHS
No need for that just pick a 2nd card which is something that the 1060 can't do SLi
Noisiv
No need to disable boost.
Just heat up till ref. 480 hits 1120MHz, then bench. voila Up to +19%.
That's the only way to get +19% over ref. 480; to have at least +19% GPU clocks.
They are already throwing AMD under the buss mentioning PCIe power issue.
As they should. Why shouldn't they show their product in the best light?
Instead of being so naive as to put competing product in a great case so that it can maintain 1266MHz:
"Here is our awesome new product and it's almost 5% faster than ref."
:banana:
DarkQuark
The elephant in the room is that we can talk all day about both cards but you pretty much cannot buy either. The buzz kill is the total and entire lack of availability.
