AMD Announces The Radeon RX 480 Starting At Just $199

I've never really seen any real bias. Kyle's just more blunt. You should have seen him tear into nVidia back in the Radeon 9800 pro vs GeForce 5800 ultra days.

I don't always agree with him, but don't get your knickers in a twist just because he calls a spade a spade.

Is their focus on the $200 pricepoint a bit exaggerated?

Steam Hardware & Software Survey

Steam survey shows the 970 ($300) in the #1 spot, approximately 50% MORE than the GTX 960 ($200).
The 750 Ti and 960 combined slightly edge out the 970. There's also almost twice as many 980 Ti's as GTX 950's.

970 + 980 + 980 Ti = 7.07%
750 Ti + 950 + 960 = 6.61%


AMD's figures are harder to compare since it looks like drivers/vbios don't report GPU names accurately... still.
The top performant cards are less fragmented, since they are from the latest gen. The bottom cards span 100s of different products. The $200 price point is a solid choice. But there are also first time PC gamers this card is targeting to bring more folks into PC gaming there is a whole untapped market out there. Console gaming has been growing steadily year after year, and PC gaming is showing healthy growth as well. A solid GPU will do well at $200.
 
Well AMD is now first in line with their $199 VR Premium card to partner up with any HMD who wants to mass produce. Now we will see if any HMD manufacturer wants to go after the mass market, not just the enthusiast.

AMD seems to be buddy buddy with Microsoft and Samsung, isnt sammy building chips for them?
 
How is this not a disaster? it's trying to put lipstick on a pig, apologies to all the pig lovers. it's not going to work for the same reason it didn't work the last time, and the same reason it doesn't work in CPU land. You can make a good profit as a budget brand, so long as brand remain segregated. In most markets this happens, Porsche doesn't make a budget car, most brand feel going for that level devalues their own brand. that's not the case in CPU or GPU land, you can buy an intel processor or Nvidia processor for slightly more than the budget price and judging by history many people choose to pay a little more so they can say they have "intel" or "NVidia" inside. In my opinion a budget card like this only works, if 1. you are the only choice in the market, or 2. you build up a large enough cache that people don't like they are buying a budget card, but instead getting a great deal from a premium manufacture. So the choice is I can buy this budget AMD card, or I can buy the little brother of the currently all mighty 1080GTX and get this GP106 card. I agree with the above poster that this is the RV680 all over again. I don't think this is what they intended. Ever since gave up doing what NVidia does, by trying to design the biggest baddest chip they can and work down, it has been all down hill, which is what I said all those years ago. Mindshare is important in the electronics world, being cheap can only take you so far.
That's not how it works with GPU production. If Porsche designed an 8 cylinder engine but some of them only ran on 6 cylinders or 4 cylinders at the end of the production line and they were forced to sell them as cheaper cars, it would be comparable.

NVIDIA, AMD, Intel, etc all design usually 1 or 2 chips, and then bin them. The best parts are sold fully enabled and the ones with small defects have those sections shut off and are sold as budget components. This maximizes their ROI and reduces waste.
 
AMD seems to be buddy buddy with Microsoft and Samsung, isnt sammy building chips for them?

They are with the Samsung/Global Foundries deal.

Also, don't forget AMD is working with Sony who is activity working on their affordable HMD using their own in-house high quality optics which nobody else seems to have access to.

Basically AMD positioned themselves as a HUB for everyone.
 
They are with the Samsung/Global Foundries deal.

Also, don't forget AMD is working with Sony who is activity working on their affordable HMD using their own in-house high quality optics which nobody else seems to have access to.

Basically AMD positioned themselves as a HUB for everyone.
I think if anyone can pull of an affordable VR headset it's Sony. They've been kicking butt in digital cameras as of late.
 
sub $300 VR capable gpu, and a sub $300 HMD to go with it!... might be dreaming, but i have been waiting to build a new rig because of VR, and its getting closer every day!
 
sub $300 VR capable gpu, and a sub $300 HMD to go with it!... might be dreaming, but i have been waiting to build a new rig because of VR, and its getting closer every day!

Not dreaming, it's coming... only question is how proprietary is Sony going to make their HBM. Will it just be for their PS4, or will they release a version for everyone?

Either way AMD is accumulating huge amounts of experience and relationships in VR with all the biggest and best developers in the world as well as hardware manufacturers. Most relevant here is AMD seems quite willing to leverage this into the PC arena.
 
The interesting thing is that this was always a single card vs single card match up. What they did in this presentation is show a CF solution for < $500 that matches or beats a $700 solution. I am not sure nVidia would want to compete against itself by having a $199 1060 card that when SLI'd outperformed their 1080 for $200 less.
Sure, many of US knew about the benefits of CF/SLI mainstream cards , but this is the first time I have seen it put forward to the masses.
Interesting times.

I bet that PowerColor is lobbying an AMD engineer for knowledge on how to put two of these cores on a single card right now. I bet that card would be $500 or less. Call it the Budget Duo.
 
I bet that PowerColor is lobbying an AMD engineer for knowledge on how to put two of these cores on a single card right now. I bet that card would be $500 or less. Call it the Budget Duo.

It would make a great 300w mGPU-dependent 1080 competitor, at $499

The interesting thing is that this was always a single card vs single card match up. What they did in this presentation is show a CF solution for < $500 that matches or beats a $700 solution. I am not sure nVidia would want to compete against itself by having a $199 1060 card that when SLI'd outperformed their 1080 for $200 less.
Sure, many of US knew about the benefits of CF/SLI mainstream cards , but this is the first time I have seen it put forward to the masses.
Interesting times.


GP106 will presumably feature 1280 ALUs on a 192bit bus probably clocked very high. When SLI works I imagine it will give a 1080 a run for its money, much like SLI 960s vs 980.

Actually scratch that, it seems there's a 64-bit memory bus for each gpc, so two gpcs (1280 ALU) would be 128bit

no idea

maybe 1060ti will be a 1070 with a 64bit bus chunk and and 16 rops removed 1920:120:48:192-bit bus: 6GB G5

the 1060 will be 1280:80:32:128bit:4GB
 
Last edited:
This is more of a competitor to the 1060. People thinking it will be almost as fast as the 1070 are setting themselves up for disappointment. It's looking more in-between the 390 and 390x, which is a good 25% slower than the 1070. That's hardly almost as fast. It's been a great job already by AMD timing it alongside the 1070, and doing very vague marketing about "premium VR" and "VR ready" but the specs and leaked benches are far closer to a 1060 and miles from a 1070. But if you're ok playing at high or medium on 1080p, it's not a bad value.

I also think the idea of trying to run VR on a 4GB $199 version is pushing it to the edge of truthful marketing claims. It might run it...horribly...but technically run it, which is probably enough to keep the lawyers happy.
 
I bet that PowerColor is lobbying an AMD engineer for knowledge on how to put two of these cores on a single card right now. I bet that card would be $500 or less. Call it the Budget Duo.

XFX also did an odd slip up about a dual GPU card, so I'd guess there's been big discussions behind the scenes to possibly make one.
 
Not dreaming, it's coming... only question is how proprietary is Sony going to make their HBM. Will it just be for their PS4, or will they release a version for everyone?

Either way AMD is accumulating huge amounts of experience and relationships in VR with all the biggest and best developers in the world as well as hardware manufacturers. Most relevant here is AMD seems quite willing to leverage this into the PC arena.

Sony already said that they would love to bring their HMD to the PC, but they need time to see if feasible.

Also there is already a sub $300 VR headset that is open source. OSVR is the name. Razer is selling it's developer version in the Razer Store. ACER is joining in as a manufacturer for the OSVR.

cptnjarhead

So for $500 you can have a drop in VR kit for a base PC from Wal Mart. :)
 
Not dreaming, it's coming... only question is how proprietary is Sony going to make their HBM. Will it just be for their PS4, or will they release a version for everyone?

Either way AMD is accumulating huge amounts of experience and relationships in VR with all the biggest and best developers in the world as well as hardware manufacturers. Most relevant here is AMD seems quite willing to leverage this into the PC arena.

True, and with this leak of "xbox scorpio" makes me wonder(hope) is MS is making an HMD that might be cross-platform? Especially if this new xbox has polaris/zen inside. Makes sense to me, devs would be able to create content for console and pc easier if the hardware is basically the same. Going way out on the speculation limb here though.
 
  • Like
Reactions: Yakk
like this
... If 480 is a "price" replacement for 380, this price makes sense. ... NVIDIA doesn't have a counter for it yet, except for the 960 4GB with current gen. I have a feeling there will be a 1060 in the future to compete directly with it. That's 'going to be the real fight. Until then, I have no doubt that 480 will "win" over NV's last gen 960 4GB, In my opinion.

Nvidia can always do like AMD has done and simply rename the GTX 960 as the GTX 1060. The GTX 960 is currently available new on Newegg for $165 before a $25 rebate, net cost $140. Will the Rx 480 be worth $60 more than a rebadged GTX 960 for the people in that part of the market? I don't know, I'm not in that market myself.
 
Nvidia can always do like AMD has done and simply rename the GTX 960 as the GTX 1060. The GTX 960 is currently available new on Newegg for $165 before a $25 rebate, net cost $140. Will the Rx 480 be worth $60 more than a rebadged GTX 960 for the people in that part of the market? I don't know, I'm not in that market myself.

Why would they do that when they have the GP106/GTX 1060 coming?
 
Sony already said that they would love to bring their HMD to the PC, but they need time to see if feasible.

Also there is already a sub $300 VR headset that is open source. OSVR is the name. Razer is selling it's developer version in the Razer Store. ACER is joining in as a manufacturer for the OSVR.

cptnjarhead

So for $500 you can have a drop in VR kit for a base PC from Wal Mart. :)

Good to know about Sony, will follow that one.

You're right, completely forgot about Razer's open source HMD! AMD is already working with them for their external GPU box... might be a good reason for this...hmmm! Would be nice to see some kind of official announcement.
 
True, and with this leak of "xbox scorpio" makes me wonder(hope) is MS is making an HMD that might be cross-platform? Especially if this new xbox has polaris/zen inside. Makes sense to me, devs would be able to create content for console and pc easier if the hardware is basically the same. Going way out on the speculation limb here though.

Seems a very likely scenario to me also in some form or other. Microsoft, right now, seems very willing to do whatever they can to add value and options to their Xbox to try and catch-up to the PS4 in sales.

Polaris looks to be the perfect GPU for this, and also could be a good reason why Microsoft made a presence at AMD's reveal.
 
Another thing I want to point out, if VR is such a focus for Polaris 10, then it will inevitably suffer (relatively) from the absence of single-pass stereo which GP106 will most certainly have. So if we're talking about perf/$, as far as VR is concerned, I'd imagine the 1070 outmatches it slightly. Then again, I'm still unsure as to whether the developers need to plan for or add the feature into their games. Afaik you do stereo projection anyway for VR, so I'm guessing *please correct me if im wrong* it will be trivial to enable

Polaris 10 (judging by the firestrike benchmarks) appears to have roughly doubled the geoemtry performance of comparable gcn1.2 gpus, this puts it at maxwell levels of geo xput

pascal has higher geo performance + single pass stereo

980 @ 1240 (stock boost) = 5.07Tflops, 224 GB/s bw
Pol 10 =5.5tflops, 256 gb bw
10% advantage

xAMD-Radeon-R9-480-3DMark11-Performance-e1464104004571.png.pagespeed.ic.FvUQGMX9gN.jpg
http://www.kitguru.net/wp-content/u...1464104004571.png.pagespeed.ic.FvUQGMX9gN.jpg
 
I hope it has good benchmark numbers (and really, good performance as benchmarks are not always indicative of real-world usage). Still, I remember AMD's launch last year. It was very underwhelming. I hope this is not the case here. Having a decent card at $199 is great for consumers.
 
Good to know about Sony, will follow that one.

You're right, completely forgot about Razer's open source HMD! AMD is already working with them for their external GPU box... might be a good reason for this...hmmm! Would be nice to see some kind of official announcement.

Yes, it would be nice to see something official.
 
If Nvidia rebadges anything, it should be the GTX980. Rebadge as the GTX1060, price at $189, and you're good to go.

Rumors are that the actual Pascal GTX1060 is a third binned GP104 chip, and it remains to be seen if they have enough of them binned that low for a mass-market solution. I guess if they do, that's probably a bad thing.
 
Part of the reason that this card is getting so much hate is due to the audience here. At [H], the GTX 970/R9 290/390 were the minimum cards. It's practically entry level to be a member here it seems. This card supposedly offers performance between the 390/390x, which means that it will slot between the 970/980 as well (you're kidding yourself if you think it sniffs the 1070). So unless it's priced at $10-$20, most here see no need to "upgrade" to it. It's a sidegrade or downgrade, regardless of price, for many in this community.

But the card is still amazing for what it is. At $199/$229, it's the same MSRP as the GTX 960 2/4GB. It's the same as the R9 380 2/4GB. There was no price increase over the same product name like there was with the 970 to 1070, or 980 to 1080 ($50 each, not counting FE pricing). This card isn't aimed at 970/390+ owners like us. It's aimed at the 660/760/960 owners (and their AMD equivalents).

If you bought a GTX 660, you probably didn't upgrade to the 760. If you bought a 760, you probably didn't upgrade to a 960. There just wasn't enough improvement there to warrant to single-cycle upgrade. But even 960/380 owners should be taking a long hard look at this card. Using the performance summary from TPU's GTX 1080 review, we see the following (1080p results):

Going from a GTX 760 to a 960 meant a gain of roughly 26%. But, going from a 960 to a 970 is roughly 62%. And this card, the RX 480, is supposedly faster than the 970. What that means is that GTX 960 owners can see a greater-than two-generation performance bump in this segment. GTX 760 owners see an even larger jump, potentially doubling performance.

The RX 480 is great for what it is, a $200 card with better than $200 performance. Hopefully it does well for AMD so that the revenue generated can be allocated to R&D for future higher performance cards. The sooner that AMD can compete with NV across the board, the better. If AMD decides to only attack the mid-range, well, we saw how that worked out in their battle with Intel.
 
Another thing I want to point out, if VR is such a focus for Polaris 10, then it will inevitably suffer (relatively) from the absence of single-pass stereo which GP106 will most certainly have. So if we're talking about perf/$, as far as VR is concerned, I'd imagine the 1070 outmatches it slightly. Then again, I'm still unsure as to whether the developers need to plan for or add the feature into their games. Afaik you do stereo projection anyway for VR, so I'm guessing *please correct me if im wrong* it will be trivial to enable

Polaris 10 (judging by the firestrike benchmarks) appears to have roughly doubled the geoemtry performance of comparable gcn1.2 gpus, this puts it at maxwell levels of geo xput

pascal has higher geo performance + single pass stereo

980 @ 1240 (stock boost) = 5.07Tflops, 224 GB/s bw
Pol 10 =5.5tflops, 256 gb bw
10% advantage

Uhem, i will soon be tired of repeating this but this chart uses outdated numbers for every other GPU on the list (except 1080 obviously). Here's your set of canned benchmarks for unreleased card: Result
 
If Nvidia rebadges anything, it should be the GTX980. Rebadge as the GTX1060, price at $189, and you're good to go.

I don't see this as profitable for NV. Granted, there's always some markup, but you don't take a $549 card and price it at $189. If that was profitable, we'd never see new product range. Every release would be like AMD's 300/Fury lineup, where every part moves down a notch and the only new part is the top one. Even if the new node means that it's more expensive per transistor, you still get more chips per wafer, which means overall lower cost per chip. The GTX 980 is too large to be priced at $189 and generate any real profit (if any at all) for NV.
 
16nm finfet is very new for Nvidia, while 28nm is extremely mature. Rumors are that 16nm is not a savings at all yet, contrary to previous process changes. Nvidia has been making GM204s for a very, very long time in huuuge quantities. Obviously I have no access to their cost sheets either, but I wouldn't be so sure about that.
 
16nm finfet is very new for Nvidia, while 28nm is extremely mature. Rumors are that 16nm is not a savings at all yet, contrary to previous process changes. Nvidia has been making GM204s for a very, very long time in huuuge quantities. Obviously I have no access to their cost sheets either, but I wouldn't be so sure about that.

Finfet is more expensive per transistor. But because the chips are smaller, you get more chips per wafer. Example: Wafer costs 10% more due to higher per transistor cost, but there are 25% more chips per wafer. This is a net reduction in the cost per chip. Note: Example was illustrative only.

We do know that the per transistor cost is higher. We also know that smaller chips = more per wafer. What we don't know is how well the gain/reduction cancel each other out. Is the overall cost higher, or lower? Cost per chip SHOULD be lower, but cost per wafer is higher.
 
Part of the reason that this card is getting so much hate is due to the audience here. At [H], the GTX 970/R9 290/390 were the minimum cards. It's practically entry level to be a member here it seems. This card supposedly offers performance between the 390/390x, which means that it will slot between the 970/980 as well (you're kidding yourself if you think it sniffs the 1070). So unless it's priced at $10-$20, most here see no need to "upgrade" to it. It's a sidegrade or downgrade, regardless of price, for many in this community.

But the card is still amazing for what it is. At $199/$229, it's the same MSRP as the GTX 960 2/4GB. It's the same as the R9 380 2/4GB. There was no price increase over the same product name like there was with the 970 to 1070, or 980 to 1080 ($50 each, not counting FE pricing). This card isn't aimed at 970/390+ owners like us. It's aimed at the 660/760/960 owners (and their AMD equivalents).

If you bought a GTX 660, you probably didn't upgrade to the 760. If you bought a 760, you probably didn't upgrade to a 960. There just wasn't enough improvement there to warrant to single-cycle upgrade. But even 960/380 owners should be taking a long hard look at this card. Using the performance summary from TPU's GTX 1080 review, we see the following (1080p results):

Going from a GTX 760 to a 960 meant a gain of roughly 26%. But, going from a 960 to a 970 is roughly 62%. And this card, the RX 480, is supposedly faster than the 970. What that means is that GTX 960 owners can see a greater-than two-generation performance bump in this segment. GTX 760 owners see an even larger jump, potentially doubling performance.

The RX 480 is great for what it is, a $200 card with better than $200 performance. Hopefully it does well for AMD so that the revenue generated can be allocated to R&D for future higher performance cards. The sooner that AMD can compete with NV across the board, the better. If AMD decides to only attack the mid-range, well, we saw how that worked out in their battle with Intel.

I'd would want one just for the H.265 4K decode / encode. As long as it's faster than what I have I'm not going to complain about a $200 card. Since it's only $200 so I can pass down into my next HTPC build and get a GTX 1080ti or Vega card when they come out. Heck at $200 I could buy two of them and run them for awhile!

My only issue is when do I get rid of my older "backup" cards. ;)
 
I scored P14 461 in 3DMark 11 Performance

Why don't you'll just run your stock 980's, 980ti, 390x, 390 against the benchmark and see how you compare? Stock because we ASSUME that the card tested above was stock. Everyone knows what happens when you make assumptions. ;) Then maybe we can pick an argument over something new. ;)

DETAILED SCORES
3DMark Score 14461
Graphics Score 18060
Physics Score 9109
Combined Score 8966
Graphics Test 175.65 fps
Graphics Test 284.93 fps
Graphics Test 3113.8 fps
Graphics Test 458.29 fps
Physics Test 28.92 fps
Combined Test 41.7 fps
 
It's all going to come down to what the actual performance shows at the end of this month. If the 480 really is 390/970 for $200, NV has no competitive product except two relatively high priced Pascal cards. AMD has finished wiping Maxwell off the map with the first card out of the gate and will get to enjoy that fact for a least a quarter before a 1060 has a chance to even sniff daylight.
 
Looks closer to 390x/980 for $200, actually. And yeah, NV needs to rebadge a GTX980 or release a GTX1060 to compete with this.
 
AMD has delivered value like this before. And they have made promises that turned out to be empty before. In recent history, more of the latter than the former. Perhaps we can all remember the promises of being 4K-ready that came with many recent products? Given that, I think proof will need to be established by testing before any VR-conquest scenarios are spun into the stratosphere.
 
Perhaps we can all remember the promises of being 4K-ready that came with many recent products?

It was beyond idiotic for them to claim 4k support, but not include HDMI 2.0 on their cards. It was even more true on the R9 Nano given that they were trying to target HTPC gamers. "Here's a 4k card that won't work on your 4k HDTV!"
 
why does everyone forget AMD put up a slide saying 100-300$

They then talked about the RX-480 being 199$(4gb) It's meant to replace the r9-380. It performance is 390+ that's a rather large leap in performance for the $$. I could see their 300$ card compete with the 1070. may just smoke it (p10 full die 40cu, gddr5x)
 
  • Like
Reactions: Yakk
like this
If the RX 480 is the "C7" flagship Polaris 10 then what is the $300 variant?
That would make 3 SKUs from the same chip.

Unless the C4 is Polaris 11 and the "RX 480X" has yet to be seen. It would be 2,560SP most likely, putting it around Fury X~980 Ti.
 
Finfet is more expensive per transistor. But because the chips are smaller, you get more chips per wafer. Example: Wafer costs 10% more due to higher per transistor cost, but there are 25% more chips per wafer. This is a net reduction in the cost per chip. Note: Example was illustrative only.

We do know that the per transistor cost is higher. We also know that smaller chips = more per wafer. What we don't know is how well the gain/reduction cancel each other out. Is the overall cost higher, or lower? Cost per chip SHOULD be lower, but cost per wafer is higher.

The cost per transistor is just a function of wafer cost + transistor density. If I'm not mistaken the 7% increase per transistor was from TSMC regarding 16nmFF; and it was assumed a certain transistor density. This does not take into account yields which are influenced by the size of each die on the wafer.

GP104 is not as transistor dense as 16nmFF+ allows, this means that 7% figure (which was in reference to FF not FF+ i believe) is low
 
I'm buying two of these for other rigs and getting a1080 for my main. 400 plus tax to update two machines is a fucking steal.
 
I'm disappointed. It's a pretty standard generational increase in performance, the mid range encroaching on high end territory. The Price is the same as launch last generation though. It just seems like a steal as they're launching after the Nvidia high end. It's all marketing, just like the Nvidia launch, of course. That being said, if it performs better than the eventual 1060, I might look into one for an HTPC setup I want to build.
 
why does everyone forget AMD put up a slide saying 100-300$

They then talked about the RX-480 being 199$(4gb) It's meant to replace the r9-380. It performance is 390+ that's a rather large leap in performance for the $$. I could see their 300$ card compete with the 1070. may just smoke it (p10 full die 40cu, gddr5x)

you still have an 8 gb version, and they are allowing AIB's to over clock these cards, so that is where that $300 a factory overclocked 8gb card.

The launch isn't disappointing its pretty much what was expected of them to do. Replace the current midrange product at its price range with the performance of a performance segment card.

Also the upside is AMD was able to cut the perf/watt gap down a good deal.

With the perf/watt numbers AMD was throwing around and initial previews of performance, I think many people expecting a bit more on the performance.
 
Last edited:
Back
Top