[Rumor] RX480 AIB Card Leaked and Tested!

[Eroticus]

"PowerColor RX 480 DEVIL will overclock above 1.4 GHz, however the exact specs are not yet finalized"


http://videocardz.com/61769/exclusive-powercolor-radeon-rx-480-devil-pictured

 

[medi01]

3 coolers is... to much, imo, my 380 by Sapphire was near silent with 2.

Single connector (8 pin, apparently) looks right.

 

[RejZoR]

With more stuff on it you kinda defeat the purpose of the card unless the gains from overclock are significant. In that case, alright. If the card then comes to a price point of 280-300€ (that's how much they charge for reference models now), then I think it's still reasonable. But if gains are almost not there, 250-260€ versions with just better cooler and no other goodies, I can see that as viable option as well for those who value low noise levels.

 

[AhmedRehan98]

guys, this is the only available adapter near my place
but it doesn't say anywhere if its passive or active
would it be able to convert 480 to my vga monitor ?
https://www.hama.com/00054569/hama-hdmi-converter-for-vga

 

[Ungari]

guys, this is the only available adapter near my place
but it doesn't say anywhere if its passive or active
would it be able to convert 480 to my vga monitor ?
https://www.hama.com/00054569/hama-hdmi-converter-for-vga

Sapphire AIB 480 has a VGA connection port over the reference cards.

 

[Finners]

3 fan on the powercolor seems a little OTT. Sapphire is looking good, same as the early leaked version by the shroud is black which looks way better. 2 Fans and the PCB has cut outs in it for the air to pass through the heatsink and straight out the back of the card which is a nice little feature.

 

[AhmedRehan98]

Sapphire AIB 480 has a VGA connection port over the reference cards.

this ?
isn't that a dvi-d ? :\

 

[95Viper]

Single connector (8 pin, apparently) looks right.

Funny that the +2 is not connected... sorta looks like just using the 6 pins.

 

[Ungari]

Funny that the +2 is not connected... sorta looks like just using the 6 pins.

Yes. This must have been what Sapphire Ed meant when he said the Nitro would have an 8-pin, but wasn't necessary to use it.

this ?
isn't that a dvi-d ? :\

Yes, sorry.

 

[The Mac]

AMD no longer has analog on any card.

Its been retired.

DVI is close behind, but AIBs can add DVI back. DVI-D only. the TDMS streams are still there for HDMI.

 

[truth teller]

since this is also speculative, for now, i guess this fits better in there than on the rx480 pci express power news thread
maybe the new drivers will just apply an undervolt to the card?
http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699

10-30W power reductions
4% performance improvement
improved high clock stability/holding

and maybe a bit cooler too, thus not triggering the overtemp protections

 

[Loosenut]

guys, this is the only available adapter near my place
but it doesn't say anywhere if its passive or active
would it be able to convert 480 to my vga monitor ?
https://www.hama.com/00054569/hama-hdmi-converter-for-vga

I use that exact same passive adapter on my R9 290. It works great. I payed less than $10 CDN on fleebay

 

[medi01]

1080 at 314mm² is only 36.5% bigger 230mm² 480.
Stock vs stock it is 77% faster in TPU performance charts.

1060 is likely half of 1080, so around 160mm².
480 is 75% bigger, yet 1060 is expected to be 15% faster and consume less power.

So, we are in a worst situation than last gen. (960 sucked, no idea why anyone would buy that shit)

HOPEFULLY that is mostly caused by Samsung/GloFo process being inferior to TSMC.
Facts supporting this are:
a) nVidia had a choice and chose TSMC
b) Apple's TSMC chips run cooler than Samsung's

PS
Could anyone point me to a source claiming that Zen is manufactured at TSMC?

AMD no longer has analog on any card.

Its been retired.

Yep. And that was announced... back in 2011, I think. (not only by AMD, Intel and others were there too)

 

[Palladium]

1080 at 314mm² is only 36.5% bigger 230mm² 480.
Stock vs stock it is 77% faster in TPU performance charts.

1060 is likely half of 1080, so around 160mm².
480 is 75% bigger, yet 1060 is expected to be 15% faster and consume less power.

So, we are in a worst situation than last gen. (960 sucked, no idea why anyone would buy that shit)

Yeah, it's pretty clear GP106 is shaping to demolish the Polaris 10 at a per mm2 die basis. AMD can squeeze more cores per mm2 but NV more than compensates that with better IPC and clocks. 1060 @ 160mm2/980 perf will be somewhere like 60% better perf/mm2 than the RX480 going by TPU numbers. AMD was lucky that the terrible 960 was stuck at 2GB and kept the same IPC as 970.

 

[medi01]

with better IPC

Lol, no.
clock for clock Fury X would run circles around 980T/TX.

 

[the54thvoid]

Lol, no.
clock for clock Fury X would run circles around 980T/TX.

But the 'point' is the architecture of Fiji does not allow such clocks. The architecture of Maxwell (and Pascal) does. It's like saying a Ford Fiesta could go as fast as a Ferrari if the Fiesta could do 185mph.

EDIT: by 'point' I don't mean IPC, I mean the reality.

 

[cdawall]

But the 'point' is the architecture of Fiji does not allow such clocks. The architecture of Maxwell (and Pascal) does. It's like saying a Ford Fiesta could go as fast as a Ferrari if the Fiesta could do 185mph.

EDIT: by 'point' I don't mean IPC, I mean the reality.

It would be more like a suburban could go as fast as a Ferrari. Reason this analogy works is because unlike the festiva suburban would require more fuel produce more heat and be doing more work at 185 than the Ferrari.

 

[uuuaaaaaa]

It would be more like a suburban could go as fast as a Ferrari. Reason this analogy works is because unlike the festiva suburban would require more fuel produce more heat and be doing more work at 185 than the Ferrari.

Fiji is like a truck, can transport 20 tons in a single trip, while a maxwell is like a ferrari, it takes 40 trips, but in each trip it is way faster than the truck...

 

[cdawall]

Fiji is like a truck, can transport 20 tons in a single trip, while a maxwell is like a ferrari, it takes 40 trips, but in each trip it is way faster than the truck...

Right so you just repeated what I said

 

[uuuaaaaaa]

Right so you just repeated what I said

I felt that you focused too much on the power comsumption and heat side of things xD

 

[cdawall]

I felt that you focused too much on the power comsumption and heat side of things xD

Which the truck will need to reach those speeds lol

 

[xLegendary]

That Devil looks pretty good!
But if most reference are not doing more than 1350, I find it very hard if they can clock more than that at default!!

 

[medi01]

Just a reminder that you were responding to a "AMD GPUs actually have higher IPC than nVidia's".

 

[Dethroy]

Lol, no.
clock for clock Fury X would run circles around 980T/TX.

I wouldn't exactly call 8.3% more IPC "running around circles" (assuming the 980 TI could possibly hold its maximum boostclock at all times → best case Fury X scenario). Next time you claim something like this, better make sure you don't pull stuff out or your ass.

Edit: math

138,6 fps (OC results of MSI GTX 980 TI Lightning)/ 103,8 fps (base clock results of Fury X) = 1,33526
1518 MHz / 1,33526 = 1136,86 MHz
1136,86 MHz / 1050 MHz (Fury X stock clocks) = 1,0827 → 8,27%

or

(1202 MHz + 1493 MHz) / 2 = 1347,5 MHz (boost average) → 1347,5 MHz / 1050 MHz (Fury X stock clocks) = 1,283
Perf Summary @ 1080p: 100% / 81% = 1,235 (1,234567901234568)
1,283 / 1,235 = 1,0395 → 3,95%
Perf Summary @ 4K : 100% / 84% = 1,19 (1,19047619047619)
1,283 / 1,19 = 1,078 → 7,8%

 

[medi01]

I wouldn't exactly call 8.3% more IPC

Stock FX beats stock 980Ti at 1440p and higher if we trust TPU (for nearly a year already).
Stock Fury X runs at 1050 Mhz.
Stock 980 Ti runs at 1177 or 1202, I don't remember, let's go with lower: 12% higher frequency, than Fury X.

And modern example, 480 running at 1266 is (close in Futuremark benches that was leaked) to 1.7Ghz 1060 (34% higher clock)