Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Polaris architecture – GCN 4.0
- Thread starter Boomstick777
- Start date
same way aliens might knock on your door and say we going to beam you up and do some probing? at least they give a fair warning first right?
wont be enough.
Only one card to rule all other cards and its called Rx 480 now
E3 going to be interesting as partners are really hyped for this card.
Hyped!
You make me lol on every post.
The best thing about the AMD Live stream was this ...
Re MS rep
53:18 How many here are running windows 10 about half the room place there hands up ... how many people like windows 10 about 4 or 5 put there hands up EPIC FAIL..... after the poor response he steps back and he as to resort to saying "if you find it not making your expectations file in those feedbacks" what a mess M$
I love the fact also that he is using the old trick that if i hold my hand up you will too !
https://youtu.be/ZwlQvjwYFEM?t=3194
Very interested in the P10 and the price it will be ...
LOL loving the Windows 10 comments - I've had feedback on critical bugs filed for 9 months or more that still haven't been addressed never mind quality of life/experience feedback.
why nobody is talking about the elephant in the room, perf per watt, down to x1.7 instead of x2.5 in the roadmap slide
ashes bench shows that 1080 is about 70-80% faster, although it's hard to say thats dx12, need dx11 games to have more accurate read on performance, and i think performance will be lower than we expected, slightly higher than 970, the 480 might beat the 980 in dx12 games, but could be consistantly beaten by 980 in dx11 games by a fair 10% margin, although with that price i dont really care 199$ is perfect pricing for that perf at the moment.
now i just hope they have a higher card than what they showed with extra 20-25% perf for about 299$, unlikely but one can only hope.
Yea, without a whole bench review about power/performance and improvement we rely on to much guesswork.
a few weeks to go until E3 and then we will know more.
I know right...
Do not be alarmed they will get back to you soon enough re feedback & bugs....the problem for the delay is that there still sorting out all the telemetry data they have gathered which is up for sale soon...
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
I somewhat assumed that $300 puts a top pricing limit on AIB/custom/overclocked versions, but we will see I guess.
Yea i can see this.. the 7950 was 2x 6-pin but my Sapphire 7950 had 1x 6-pin & 1x 8-pin.. AIB versions should add some good perf with factory overclocks etc, 'if' they overclock well.
This may be a good upgrade for me coming from a 7950, at least i'd be able to sell it quite easily when Vega comes, my 7950 is only worth a few peanuts now
Wait, I've skimmed through the Polaris bit just now from the stream. Ashes bench, 1080 98% utilization and 59 fps, 480 cf 51% utilization with 62 fps. In theory, 480 cf at 95% utilization would be what, 115 fps? Isn't that like, double the performance of 1080 for $200 less? Even if utilization:fps is not linear, 480 cf with proper drivers should hit 100 fps, and again I'm assuming, that cf scaling is a least 70% a single 480 beats 1080? What? Or have I messed up my maths somewhere? I probably did, only up for 10 minutes or so.
Yeah, no one's really sure what's going on there.
Since that also suggests a single 480 is the same speed as the 1080 (in ashes anyway) for $199.
Also it occurs to me, if they're now claiming a massive 2.8x perf/W increase (in max ideal conditions of course). Doesn't that mean they're saying they could do 390X performance at ~95W? If so, maybe the card people are saying around 970 performance is actually the larger Polaris 11. Since that would allow them to get 970 performance at ~80W. And largest Polaris 10 could be faster than people are assuming.
Last edited:
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,459
- Location
- Planet Earth
PCB looks very short indeed - the cooler overhang seems similar to what Nvidia have done in the past.
Can we start shouting 'nVidia is DOOOOOOMED!!!!!' yet or not?
Since P11 is going into laptops, I can imagine £800 laptops pulling off 1080p60 (maybe not ultra) easy if it will have 970 perf at 80W or less.
Back to 480. 1x 6pin means 150W tdp, add 2.5x perf/W and you get 375W Fiji ~ Fury/FX? Way to rain on nVidia's parade.
PCB looks very short indeed - the cooler overhang seems similar to what Nvidia have done in the past.
Slightly longer than the nano which was hinted at before, they more than likely had an aftermarket or some kind of custom loop for the capsaicin demonstration since you could not see a reference style cooler from the back case shots.
Can we start shouting 'nVidia is DOOOOOOMED!!!!!' yet or not?
Since P11 is going into laptops, I can imagine £800 laptops pulling off 1080p60 (maybe not ultra) easy if it will have 970 perf at 80W or less.
Back to 480. 1x 6pin means 150W tdp, add 2.5x perf/W and you get 375W Fiji ~ Fury/FX? Way to rain on nVidia's parade.
It just means the board can draw up to 150W. Since once you need over the 75W the PCI-E slot can provide, the smallest auxiliary power connector for 12V is a 6pin connector. The card will more than likely use far less during normal use.
Last edited:
Can we start shouting 'nVidia is DOOOOOOMED!!!!!' yet or not?
Since P11 is going into laptops, I can imagine £800 laptops pulling off 1080p60 (maybe not ultra) easy if it will have 970 perf at 80W or less.
Back to 480. 1x 6pin means 150W tdp, add 2.5x perf/W and you get 375W Fiji ~ Fury/FX? Way to rain on nVidia's parade.
Don't assume the 150w TPD is the actual power P10 will use. That is simply the power supplied by the PCIe slot and the 6 pin connecter added to get 150 (75+75).
I suspect from the leaks that 480 will be around R9 Nano performance (at least in DX12). So a great rpice/perf GPU that will get close to 1070 speeds for a lot less money.
If 480 ends up ~10%-15% slower than 1070 in DX11 and trades blows at DX12 then 1070 will need to be cheaper IMHO. This is where I'm excited for the price/perf of 480.
I understand that. I just hope AMD don't skimp on power delivery components and there's no power throttle when you reach 10% OC. Thermals we can take care of. That ref cooler looks sexy tho.
True of course. It does also mean though, it gives them a lot of room to offer a lot of performance even at low power (depending on how well this 2.5-2.8x scales).
Potentially they could make a 200W card (as in draws 200W) which would be faster than an overclocked 1080.
Pretty good mainstream launch from AMD.
Also this seems like believable now.
The 480 is between the 970-980.
I'm inclined not to believe any of those scores.
GTX 970 @ 1550/1950 Graphics score: 14,000 http://www.3dmark.com/fs/8573432 < thats a top score for any GTX 970
They cite 15,000 @ stock? @ stock they bench at around 11,000 - 12,000
Last edited:
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,493
The 2.8x figure was "using AMD technologies", its similar to Nvidia's VR performance claims
The real figure is 1.7x, AMD dont exactly have a massively good history with market penetration on their AMD specific gubbins
No, what they were saying there was the move from 28nm to 14nm was 1.7x. So if they just shrunk the 390X etc. it'd be 1.7x. But then through their architecture changes they bumped it to 2.8x.
Similar to how Nvidia managed a ~1.5x perf/W increase on the same node going from Kepler to Maxwell.
EDIT: Also this means they're claiming GCN1.x to GCN4 is a ~1.64x increase, since 1.64*1.7 is ~2.8x.
Last edited: