I think it was the only move they could make to gain back some market share. $199 is good marketing and will make headlines. People will recommend it for every budget build. It'll be the card that provides near $400 worth of performance for $200. The only question now is how fast can Nvidia get a 1060 out the door.
And the 1060 will have to provide GTX 980 level performance at ~$200. It will probably have to be faster than nVidia had planned for it. Their x60 series cards have been underwhelming for generations now. If this card is everything they say it is this is good for everyone. $200 performance mainstream cards are back, and nVidia will be forced to stop ripping everyone off and provide good performance for the price again. Thank god some competition is returning to the video card market. Prices were getting insane. And nVidia seemed almost entirely focused on super high-end cards almost no one can afford.
So these are the latest Apple rumors. I guess Apple locked up the original high-end Tonga chip in 2014 so some people are worried they did it gain with the missing 480X which is most likely a shrunken high-end Tonga. I guess we'll know more on the 13th at E3. MacBook - Intel/Polaris 11 MacBook Pro - Polaris 10/11? iMac - Polaris 10 Mac Pro - AMD Fire Pro http://fudzilla.com/news/graphics/40835-all-apple-2016-laptops-come-with-amd-gpus http://fudzilla.com/news/graphics/40840-amd-apple-gpu-polaris-11-based
AMD announced the 460 and 470 cards. Estimated MSRP are $99 and $150. Leaked benchmarks show the 470 performing around a 290 level. Should be a great value for people looking for a 1080p card.
AMD is releasing their own OC tool for the 400s and for the first time ever they will be building in voltage controls. Neither Nvidia or AMD has ever done that before because adjusting voltage is a super easy way to ruin your card. I'm guessing this means AMD wants people to have fun and OC as much as they want. The reference $200 boards will have a single 6 pin power connectors but there will also be $300 (beast versions) boards with 6+8 pin power connectors. Boost on the standard boards is 1266Mhz and 1500Mhz on the beast version. There are a couple of reports of people hitting 1570Mhz and 1600Mhz. Supposedly 1500Mhz will match stock 980ti/1070 performance in DX12 games. That probably means 5 to 10% slower in DX11 games. Lots of cards seem to be leaking out so I'm guessing AMD is ready to go. Sadly the miners are expecting this card to be very effective at mining so they might be sucking up the stock early like when it happened to the 290. http://wccftech.com/amd-rx-480-1500mhz-overclocking-tool-voltage-control/
The NDA end tomorrow. Most early benchmarks show the 480 just slightly behind the the 980 which isn't bad for $200/$240. Several Micro Center stores complained that they got less than 10 of 1070s and 1080s. I don't think the 480 will have the same inventory problem.
Well the performance is not where I'd hoped it'd be but it's still going to be a performance per dollar monster for people wanting to break into high-end 1080p PC gaming.
HardOCP and Toms Hardware have reviews up, and Anandtech has some benchmarks. The 480 offers very similar performance to a GTX 970. It's generally a little slower. In a few games that favor AMD it's slightly faster. So the people thinking this would offer 980 or 980 Ti level performance were off the mark. But 970 performance for $200 is still an excellent value and this should be a great card for the masses who game at 1080p. Polaris also seems right for the console refreshes. It's cheap and does 1080p and VR pretty well and would get the consoles up to modern levels of performance.
So is HBM dead? It looked good on paper, but didn't seem to do much for the Fury cards other than jack up the price, and Fury was awhile ago and it flopped. Pascal and Polaris are GDDR5/X. Seems kind of like RAMBUS vs DDR all over again, just with VRAM.
Vega 10 and GP100 are supposed to have HBM2. Both might pop up in January but Vega is rumored for an October date as well. My understanding is the GPUs have to be designed to use it so GDDR5 and HBM2 version of the same cards wont co-exist. Could be wrong though. HBM is still expensive and Fury X demonstrated the increased bandwidth is pointless without a much faster GPU. The GP100 might be the first card that really needs it. If the next Vega is set to match the 1080 I doubt it would need the bandwidth but from the looks of Polaris it might need it for the ~70% reduced power reduction over GDDR5. I'm betting Vega will just be a shrunk Fiji with 20 to 40% performance increase and a heat and power bump to match. Also... Hardware Canucks broke the 480 performance up into DX11 and DX12 games. There are only about 4 DX12 games right now but if someone is looking towards the future and DX12 the 480 might be a good leap over the 970. http://www.hardwarecanucks.com/foru...9-radeon-rx480-8gb-performance-review-24.html
It looks decent enough for the price. I expected it to either outperform Maxwell and consume the same power or perform the same and consume less. The difference is there but not enough to matter for either performance or power use - so that is a little disappointing. Maybe it has nothing to do with the polaris arch and is only global foundries 14nm process vs TSMC 16nm? I'd really like to replace my 290X with something that is not so power hungry for around $300-350. 1070 performance is about what I want as a minimum but the price is artificially inflated because of the stupid Founders Edition. I'm thinking 1060TI will be a bit too cut down. Maybe if Vega is more like 7970-290 in die size and not Fury there will be a version just cheap enough. If they go all out to compete with the next Titan/1080ti/whatever then what the hell do people who like high-ish end hardware but don't love to throw money away buy?
I think the enthusiast were disappointed by the power but I'm guessing AMD wanted a cheap card and lower power wasn't a priority. They kept their R&D low, got a ton of usable chips to market, and can now gain back some market share by having the best value card we've seen in years. At the end of the day it's still uses less power than a 970 which was considered fantastic just a few months ago and I doubt anyone whose scrapping together $200 for a graphics cards gives a hoot that it's not any better. If there is a problem with the global foundries process AMD is going to have bigger issues with Vega and Zen. I hope that's not the case. As for the 1060ti I have no idea. It has to be a least as fast as a 980 and probably 5 to 10% more. I bet they wanted $300 for it but now we might see $250. I'm guessing the 490 will match the 1070. I'd be nice to see it at <$400 but who knows. From what I heard the Titan is coming soon but I can't imagine they'd release it before Vega.
lol nobody outside of hardware enthusiast forums gives a shit about power. All the average gamer is going to care about is ~GTX 970 for $200.
Right. Power matters for laptops but this is a freaking desktop card. Who cares about + or - 40 watts. That's like an extra $6 of power a year.
Personally I either wanted a slightly faster video card than what I had that was very low power to then move it to my gf's mITX pc when I upgrade to something properly faster. So yeah, it fell a little short on what I wanted for performance and for power use. Anyone who doesn't have a 970/290 and up and doesn't have a big budget should obviously get a 480 if they want to upgrade now. For the enthusiast forum discussion side of things though, lower end products can be indicative of their future high end products where performance is often limited by a power ceiling. Either it will mean nvidia will be faster or they will overclock better. That said apparently Polaris was designed for mobile first and foremost, so to reach the clock speeds it has in the 480 efficiency went out the window a little. Hopefully if this is true it will mean Vega will both clock higher and be more efficient.
Sounds like the 1060 might be what your looking for but who knows what the price will be yet. I wanted something at least at the Fury level and we might see that with the 1.5Ghz editions but if they are $300 a $380 1070 is probably a better deal. Of course that assumes one can find a 1070 for $380. If Vega is just a larger Polaris they are going to have heat and power issues that are going to look ugly for a 2017 card. Custom cooling will be mandatory and we might see another water cooled version for the high-end. My bet is Vega isn't a larger Polaris but that doesn't mean they haven't solved the heat problem. From what I've read Nvidia has about 10 times the R&D budget of AMD so they are more likely to have better designed and power efficient chips. I think Polaris was just a shrunk Tonga or Tahiti. If Vega is a shrunk Hawaii there are going to be big challenges to overcome. Huge performance improvements would be guaranteed. We'd get 1070 performance by default, maybe even 1080ti performance in DX12, but the heat would be unmanageable. I'm not sure Fiji would be any better. I hope it's something new.
http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699 It seems like AMD overcompensated with voltage for the 480. Performance is up to 4% more with this sample when undervolting and power use is dropping 10-30w at the wall. Sure this may be because some lesser quality chips need the extra voltage, but there are a few of these articles out there now (in other languages) that have found the same. Fortunately undervolting is now as easy as overclocking since the wattman software allows it from the control center. A aftermarket model with a custom cooler, 8-pin power and tweaked voltages by default could end up being a lot better than what we have seen in reviews. Something around GTX 980 performance with 10-20w less power draw than a 970 making it very quiet in even small cramped cases would be great. Of course other models will likely use the extra cooling and power on tap to overclock even more, throwing efficiency out the window, just like most 970 and 980 models did.
The whole PCI-e power draw thing is really odd. For some reason AMD seems to have linked the power draw between PCI-e and 6 pin rather than locking the PCI-e to a limit and then pulling the extra power from the 6 pin. The result is long term power spikes really hit the motherboard hard. Lot's of cards will spike the motherboard well beyond 75w PCI-e limit and that's okay. I think there is a version of the 960 that can peak at 225w but it's only for a fraction of a second. The difference is the 480 will draw over 75w from the board for long periods of time. In a worst case it might pull 85w. The 75w limit is supposed to be a very conservative number so 85w might not be an issue but the question is why did they do it? They could have easily pulled all the power from the 6 pin. Also why does undervolting the card increase performance when we'd expect the opposite? I have to think someone made a mistake. These guys explain it really well. Also it seems that the 4GB versions of the 480 actually have 8GB on the board. It's possible once BIOS tools are made the extra memory might be accessible .