The 390x can do 5.9 TFLOPS and it's not faster than a 980.
Why would you attach a watercooler to a card with sub-150W TDP?
Why would you attach a watercooler to a card with sub-150W TDP?
But they still need a faster P10 to at least be within a stones throw away from the 1070.
No, they don't. Christ, it's a 480X. There's going to be a 490, 490X, Fury and Fury X above it and people still insist on matching it up to the wrong bracket. It's giving me fits.
I think they are doing it on purpose.No, they don't. Christ, it's a 480X. There's going to be a 490, 490X, Fury and Fury X above it and people still insist on matching it up to the wrong bracket. It's giving me fits.
I don't get why people are complaining so much, if this card is 980 performance at lower power, lower price, and more features, how is that not a win ?
I would be buying one as soon as it is released, and many others will as well.
AMD Radeon RX 480 specifications leaked?
http://videocardz.com/60752/amd-radeon-rx-480-specifications-leaked
Positioned against Geforce GTX 970 / GTX 980. We can probably rule out 90% Geforce GTX 1080 performance at half the price/TDP predictions.
No, they don't. Christ, it's a 480X. There's going to be a 490, 490X, Fury and Fury X above it and people still insist on matching it up to the wrong bracket. It's giving me fits.
This slide bolsters my suspicions that AMD was caught off-guard by Nvidia's early release of the GP104 cards. I think they expected them to be released circa October, and maybe that was the original plan, but Nvidia pulled out all the stops to beat AMD to market.
5.5 TFlops of computing performance would equate to 2304 shaders at ~1200 MHz. We know that GCN on 28nm has optimal perf/watt near 800-900 MHz, so this would fit in well with the ~40% clock rate increases that FinFET has brought for most designs.
I don't consider this to be a disappointment in any way. By most accounts, Polaris 10 is about the same die size as Pitcairn, which topped out at just under 2.6 TFlops. If Polaris 10 has more than double the raw performance at the same die size, bus width, and power consumption, plus a bunch of new features, that's about the most we can reasonably expect for a full node shrink.
Let's not forget that GTX 980 does "only" 4.6 TFlops at stock. The reason it contended with Hawaii was because AMD cards of the 28nm generation were less effective at using all their resources in DX11 titles. If Polaris fixes this using some combination of hardware and/or software optimizations, and we see shader utilization on par with Maxwell, then we could be looking at GTX 980 + 20% performance from Polaris 10.
What historical parallel to draw the P10 to? I compare it to Pitcairn, which at launch was about 10% slower than a 660 Ti. Compare the die sizes of Pitcairn vs GK104 to P10 vs GP104 - virtually identical ratios. It's debatable if 1070 is cutdown less than the 660 Ti, so I'm gonna be very generous and give P10 15% instead of 10%.
This is the new Kepler vs GCN 1.0 - new cards after a major node shrink. Nvidia made great strides with Maxwell but appear to have not improved Pascal IPC. Will AMD still be stuck in 2012-2013? Or can they make similar gains to Nvidia from 2014 and "reset" where Kepler and GCN 1.0 were at launch (which is pre-GCN improvements - 7870 matched 660 Ti later on in the gen - again I am being generous in my expectations).
You can compare P10 to 380X and 1070 to 970 and be happy. But continuing the performance-per-mm, performance-per-FLOP, and performance-per-watt of Maxwell vs GCN 1.0-1.2 is unacceptable if they truly intend to compete again and not suffer at 20-25% marketshare.
To be fair looking at the 1080 aftermarket review for Strix (in multiple reviews it does not OC better than the FE) the 980ti when overclocked is just as fast as the 1080 in games (due to throttling at 1.125V which cannot be changed supposedly.) Nvidia has not brought anything new in terms of performance only in terms of power usage.
Hardware.fr writes about Polaris : http://www.hardware.fr/news/14651/computex-polaris-doubler-pdm-amd.html
Would also add that partners did show custom designs on the Polaris Tech Event. They talk about refined products for midrange cards.
No it's not. OCed Strix is almost 50% faster than stock Geforce GTX 980 Ti. It can close the gap but won't match Geforce GTX 1080 when both are OCed.
I don't get why people are complaining so much, if this card is 980 performance at lower power, lower price, and more features, how is that not a win ?
I would be buying one as soon as it is released, and many others will as well.
Why? Simple, I want an high end card from AMD. Not an Ngreedia, GameWreck overpriced card. Actually I would buy 3 cards.
NV doesn't officially support 3-4 way SLI anymore. It smells like 3-4 way SLI will only work well in synthetics. You could always grab 2x1080 SLI and then 3x Vega in 2017.