[BitsAndChips]390X ready for launch - AMD ironing out drivers - Computex launch

[Silverforce11]

Most likely "launch" at Computex in early June, but volume in retail won't be til later. If its this delayed, it stands to reason there's a good reason for it and its not unreasonable to expect volumes will be very low.

 

[Sequences]

Man, they really need to work on that execution. Never mind how the product actually performs. Shipping a product seems to be difficult for AMD.

 

[Rakehellion]

AMD ironing out drivers

AMD? Ironing out drivers? They would never do such a thing! :awe:

 

[Azix]

I'm wondering if the secrecy even helps them. All the people considering going 970 or 980 might hold off if AMD gave some promising info. By that logic, them not doing so might mean they have nothing good to show. It does not hurt to have some positive leaks/PR. and what is nvidia going to do at this point? Odds are they are done with maxwell high end after the 980ti.

So secretive to the point that we don't even know when they will launch. That just seems ridiculous. Holding information for the sake of holding it.

If its to sell older cards out... not happening unless they promote those. They'll just keep losing potential sales to nvidia.

 

[Cookie Monster]

Man, they really need to work on that execution. Never mind how the product actually performs. Shipping a product seems to be difficult for AMD.

Well.. we are talking about a product based on a potentially new but unproven 20nm technology for GPUs, new memory technology in the form of HBM1 and most likely a new architecture. So it could potentially take alot longer to wrap things up for launch than what you'd expect.

 

[JDG1980]

I'm wondering if the secrecy even helps them. All the people considering going 970 or 980 might hold off if AMD gave some promising info. By that logic, them not doing so might mean they have nothing good to show. It does not hurt to have some positive leaks/PR. and what is nvidia going to do at this point? Odds are they are done with maxwell high end after the 980ti.

So secretive to the point that we don't even know when they will launch. That just seems ridiculous. Holding information for the sake of holding it.

If its to sell older cards out... not happening unless they promote those. They'll just keep losing potential sales to nvidia.

AMD has always been secretive. I think they have some serious issues with corporate culture, including a belief in the sunk-cost fallacy, excessive concern with channel inventory levels, and a fear of the mythical "Osborne effect". (The events that actually caused Osborne's downfall are almost the opposite of what the traditional story says; some idiot VP didn't want to write off $150K of old parts, and wound up spending $2 million to remake old plastic molds and build them out, at which point no one wanted them.)

 

[tential]

I'm wondering if the secrecy even helps them. All the people considering going 970 or 980 might hold off if AMD gave some promising info. By that logic, them not doing so might mean they have nothing good to show. It does not hurt to have some positive leaks/PR. and what is nvidia going to do at this point? Odds are they are done with maxwell high end after the 980ti.

So secretive to the point that we don't even know when they will launch. That just seems ridiculous. Holding information for the sake of holding it.

If its to sell older cards out... not happening unless they promote those. They'll just keep losing potential sales to nvidia.

Or, people like myself aren't purchasing a GTX 970/980 because we don't want to see the R9 300 series undercut them on price and perform better. The GTX 970/980 doesn't even perform on a performance delta necessary for me to justify their prices. They only seem to do well on launch day/early in a game cycle. Late in a game cycle, when I actually play a game, both cards perform well so it even makes the Nvidia case worse for me, hence why I'm waiting. If I knew the R9 300 perf now, and I wasn't satisfied with it, I'd already have a gTX 970(even if it's not good enough for me just for the DSR and HDMI 2.0). But, I don't, so I wait.

 

[DiogoDX]

I'm curious to see the market reaction to these 300 series. Lauching ahead of Nvidia keep AMD with the 35-40% market along the years and now they are probably at 20%.

My bet is that AMD will not go back to 30% anymore unless nvidia is late again like was with Fermi.

 

[AdamK47]

Yeah, are those drivers ironed out yet?

 

[RussianSensation]

I'm wondering if the secrecy even helps them. All the people considering going 970 or 980 might hold off if AMD gave some promising info. By that logic, them not doing so might mean they have nothing good to show. It does not hurt to have some positive leaks/PR.

I don't know but this looks a bit too 'semi-pro' for a random person online to do. This could have been an intentional AMD leak right there:

^ Most GPU observers would never be able to come up with "up to 8GB of HBM". Most of our forum missed the dual-link inter-poser scenario despite it being linked for months. So for someone to know that the card will have up to 8GB of HBM already tells us whoever made this slide knew more than 99% of PC enthusiasts.

^ This slide just builds on the first slide. There is way too much technical data here for the casual PC enthusiasts. The slide shows maximum and minimum operating HBM voltages and for the first time from any leak ever shows a picture of dual-link interposing with TSV. The slide even goes further and says HBM 2.0 will eliminate the need for dual-link interposers. Do you know how hard it is to "fake" this type of knowledge? Does this look like something someone from a forum could put together? I am doubting that.

^ In this slide. Point #3. That shows an high depth of technical knowledge of GPU specs to specify Tier 3 of DX12. A 'casual' PC enthusiast would never be able to fake that or even think of putting something like that into the slide. He would just write something generic like "Full DX12 support." But NO, the slide specified not just Full DX12 support, but Tier 3 support. Also, Point #4 is an odd one for someone faking slides on a new AMD card. Most of us care about gaming performance and with Oculus delayed to 2016 and beyond, full VR immersion is the last thing on our minds for R9 300 series. This looks like more of a selling point for a GPU company than some random person trying to hype up an unreleased GPU.

and what is nvidia going to do at this point? Odds are they are done with maxwell high end after the 980ti.

Who knows. NV's GM200 can overclock to 1.4Ghz on air but ships with 1.075Ghz clocks. That means NV could release a 1.2Ghz 980Ti and then in 6 months a 1.35Ghz 980Ti Ultra Extreme.

So secretive to the point that we don't even know when they will launch. That just seems ridiculous. Holding information for the sake of holding it.

Could be good like Apple. Could be bad like Bulldozer.

If its to sell older cards out... not happening unless they promote those. They'll just keep losing potential sales to nvidia.

Sounds like R9 300 series isn't ready yet. I think if they were ready, AMD would have launched them already. At this point then, it makes sense just to keep your mouth shot and keep the competitor guessing and then if the product is actually good launch the entire mobile and desktop discrete line-up at insane price/performance values similar to what AMD did with HD4870, 5870 generations. With 980 at $550 and Titan X at $1000, AMD can make a huge impact. However, I have a feeling NV will try to spoil R9 390's launch with GM200 in May.

Well.. we are talking about a product based on a potentially new but unproven 20nm technology for GPUs

Highly unlikely.

most likely a new architecture.

By all indications, it's an evolution of GCN, not a new micro-architecture. AMD does GPU design differently from NV. They reuse the underlying architecture for many generations but instead upgrade the building blocks around it. HD2900->HD6900 were basically a similar underlying VLIW architecture. That's 5 consecutive generations on the same architecture that kept evolving. R9 300 series would only be a 3rd generation, which is way too early for AMD to switch the underlying architecture based on their GPU design methods.

I'm curious to see the market reaction to these 300 series. Lauching ahead of Nvidia keep AMD with the 35-40% market along the years and now they are probably at 20%.

My bet is that AMD will not go back to 30% anymore unless nvidia is late again like was with Fermi.

If AMD's almost entire focus was desktop cards with R9 300 series, they will be lucky to maintain 25% by the time Pascal launches. All we've heard thus far are rumours on desktop R9 300 series cards. Forums such as ours are very desktop GPU biased but most of the market is made up of laptop GPUs. NV probably has 90%+ market share in there!

 

[2is]

I'm curious to see the market reaction to these 300 series. Lauching ahead of Nvidia keep AMD with the 35-40% market along the years and now they are probably at 20%.

My bet is that AMD will not go back to 30% anymore unless nvidia is late again like was with Fermi.

IMO AMD can get 30% if they make some headway with developers. The 300 series from what I've read is an appealing card to me. But I don't want to buy the next Game Works game and have it run like crap when I enable God Rays

 

[digitaldurandal]

IMO AMD can get 30% if they make some headway with developers. The 300 series from what I've read is an appealing card to me. But I don't want to buy the next Game Works game and have it run like crap when I enable God Rays

Too much money put into those relationships from Nvidia. If this were to happen it would take years. The way this would happen is most likely in the opposite order you propose, if AMD took a large market share, developers would be less likely to box them out due to alienation of customer base. The utility of the Nvidia relationship would shrink proportionally to AMD's market share.

Of course at least a part of the purpose of the Game Works relationship is to keep and take market share so if AMD were able to approach 50% market share the venture has already failed anyways.

 

[2is]

Too much money put into those relationships from Nvidia. If this were to happen it would take years. The way this would happen is most likely in the opposite order you propose, if AMD took a large market share, developers would be less likely to box them out due to alienation of customer base. The utility of the Nvidia relationship would shrink proportionally to AMD's market share.

The end game you describe would be perfectly fine as far as I'm concerned. I know it wouldn't happen overnight, but AMD's current marketing tactic (which is practically non-existent) is obviously not very successful.

 

[5150Joker]

IMO AMD can get 30% if they make some headway with developers. The 300 series from what I've read is an appealing card to me. But I don't want to buy the next Game Works game and have it run like crap when I enable God Rays

You guys are asking the wrong question. It's not whether or not AMD can reach 30% again, it's whether they can turn a profit with such a small share. History has shown they have bleed money and ATi also had a similar track record. AMD can't afford to keep bleeding cash with a shrinking R&D budget. Even if R9 300 does extremely well, I don't predict more than a short term 5% bump in market share before Pascal and lower priced NVIDIA parts stomp on it again.

 

[boozzer]

You guys are asking the wrong question. It's not whether or not AMD can reach 30% again, it's whether they can turn a profit with such a small share. History has shown they have bleed money and ATi also had a similar track record. AMD can't afford to keep bleeding cash with a shrinking R&D budget. Even if R9 300 does extremely well, I don't predict more than a short term 5% bump in market share before Pascal and lower priced NVIDIA parts stomp on it again.

marketing is everything

nv fans would happily buy a 15% faster card for 250$ more. what can amd do against that? better and more marketing! :thumbsup:

nothing else matters as much as marketing

 

[raghu78]

You guys are asking the wrong question. It's not whether or not AMD can reach 30% again, it's whether they can turn a profit with such a small share. History has shown they have bleed money and ATi also had a similar track record. AMD can't afford to keep bleeding cash with a shrinking R&D budget. Even if R9 300 does extremely well, I don't predict more than a short term 5% bump in market share before Pascal and lower priced NVIDIA parts stomp on it again.

you wanna take a bet. I am confident that AMD gets back to 35 - 40% market share by Q3/Q4 2015 and that AMD will get to 14nm first due to their experience with HBM. Nvidia has 2 options - Maxwell shrinks and then Pascal with HBM and NVLink in early 2017 or jump straight to Pascal, 14nm, HBM, NVlink. The last time they tried a similar high risk transition was with GF100 Fermi and we know the results. I am guessing the earliest Nvidia can bring a GP204 is Q4 2016. AMD on the other hand has done all the hard engineering to transition to HBM on a mature and high yielding 28nm node. AMD has chosen the right time to transition to HBM even though that meant Maxwell GM204 went without competition for 9 months and GM200 for 3 months. The next 18 - 24 months will show that AMD's decision was the right one.

To remind you AMD made a similar transition to GDDR5. AMD made HD 3870 on 55nm first and then made the HD 4870 with GDDR5. By the time the HD 5000 series launched their experience with GDDR5 helped them execute flawlessly on HD 5970, HD 5870, HD 5850, HD 5770, HD 5750, HD 5670 and HD 5650. In fact AMD had the entire HD 5000 product stack out before GTX 480 launched. I suspect we will see something similar with Pascal and AMD Arctic Islands.

 

[Silverforce11]

I don't know but this looks a bit too 'semi-pro' for a random person online to do. This could have been an intentional AMD leak right there:

Those AIB presentation slides also have 390X as 65% faster than R290X.

If they want to sell it for $799 for the 8GB version, I would buy it. Take my money!

Hopefully the 390 4GB version comes in at ~50% faster than R290X for $499. Sweetspot for bang for buck for non 4K users.

 

[96Firebird]

If its to sell older cards out... not happening unless they promote those. They'll just keep losing potential sales to nvidia.

If they really want to clear out the 2xx series cards, they should bundle a game. They used to have bundles left and right, what happened to that?

 

[raghu78]

Those AIB presentation slides also have 390X as 65% faster than R290X.

If they want to sell it for $799 for the 8GB version, I would buy it. Take my money!

Hopefully the 390 4GB version comes in at ~50% faster than R290X for $499. Sweetspot for bang for buck for non 4K users.

I expect the R9 390X to be a monster GPU. AMD GPUs love high bandwidth and GCN is a low latency architecture. HBM would benefit GCN extremely well as it provides massive bandwidth at low latency.

http://wccftech.com/interview-amds-robert-hallock-future-prospects-amd-explored/

"It’s been something of a well-kept secret that the GCN Architecture is a low-latency powerhouse, which is ideal for VR"

I believe there will be substantial architectural improvements for improved IPC in the next gen GCN 1.3 or GCN 2.0 , whatever its named.

https://translate.google.com/transl...ande-chip-monolitico-anche-per-amd&edit-text=

"The third new feature is the micro-architecture. With Fiji design GCN should fully embrace the "tiled architecture" going to review the organization for ALU thread within Compute Units in order to improve workload management."

People who think Maxwell GM200 is unbeatable are probably underestimating AMD R9 390X. AMD R9 390X with the next gen GCN built on GF 28SHP will be impressive at both stock and overclocked. I have high expectations from GF 28SHP vs TSMC 28HP as its a less leakier process with better performance at same power or reduced power at same perf.

 

[RussianSensation]

you wanna take a bet. I am confident that AMD gets back to 35 - 40% market share by Q3/Q4 2015 and that AMD will get to 14nm first due to their experience with HBM.

Wow, either of those are some ballsy bets. I wouldn't take those if I were you. AMD can't just retain 15-20% market share in a matter of 3 months. Your estimate of Q3 2015 is completely unrealistic and even Q4 2015 sounds unrealistic. Even if R9 300 series is good, NV can easily release 960Ti, 970Ti and refresh 980, while at the same time dropping prices. Did you see NV's margins as of late? 54-56%. NV could probably drop the price of a 970 by $100 and 980 by $200 and still make more $ than AMD.

If NV wanted to, they could undercut the entire Maxwell series so much it would literally make it impossible for AMD's R9 300 series to compete. Why? Because NV has so much cash and almost no debt that they could practically make their market share go to 95% and bankrupt AMD. Luckily NV isn't interested in 15-20% margins and because they are aiming for > 50%+ margins that they haven't completely obliterated AMD by now. Realistically speaking, if NV really really wanted to, they could price GTX980 for $149 for 12 months straight and then GP204 for $149 for another 12 months, on purpose to just bankrupt AMD's GPU division. NV could survive this due to their cash reserves. Believe me, NV has all the cards in its hands, no matter how good R9 300 series can be. That's because financially, NV can withstand the price war - AMD cannot.

Those AIB presentation slides also have 390X as 65% faster than R290X.

If they want to sell it for $799 for the 8GB version, I would buy it. Take my money!

Hopefully the 390 4GB version comes in at ~50% faster than R290X for $499. Sweetspot for bang for buck for non 4K users.

Are you already at 4K, or moving to a 4K screen this year?

If they really want to clear out the 2xx series cards, they should bundle a game. They used to have bundles left and right, what happened to that?

There are no good games games in the near future to bundle - Project CARS, TW3, they are GWs titles. () AMD missed its chance to optimize for GTA V and bundle with that game. It would have sold them quite a bit of cards but I think it would cost way too much $ to make it worth it. In Canada the stock of R9 290/290X cards has almost dried up. I think it's a matter of time before the U.S. is hit with it. Our stores here aren't replenishing stock of R9 290 series at all. Once they sell out, they are done. The SKUs are completely removed from the online inventory of the stores.

People who think Maxwell GM200 is unbeatable are probably underestimating AMD R9 390X. AMD R9 390X with the next gen GCN built on GF 28SHP will be impressive at both stock and overclocked. I have high expectations from GF 28SHP vs TSMC 28HP as its a less leakier process with better performance at same power or reduced power at same perf.

Even if all of that is true, NV can respond with a price war the likes AMD has never seen. See my response to Silver above. Secondly, you can't take back that millions of people already upgraded to Maxwell and those customers are lost for another 2-4 years as far as their upgrade path goes. If R9 300 desktop cards are competitive but they are lacking in perf/watt, then R9 300 mobile won't get many design wins come Skylake/Windows 10 Q3-Q4 2015 and more sales to NV. For example, I fully expect the summer 2015 MacBook Pro laptops to be refreshed with Maxwell, not R9 300 series. It's hard to say what's going to happen with the Mac Pro because Maxwell is lacking double precision performance but the 12GB of VRAM would be a huge selling point for Apple to use Quadro GM200 cards in the Mac Pro. Reading on some Apple forums, the R9 M290X/295X got too much negative feedback which makes me think the next iMac refresh will also get Maxwell.

All the new cutting edge laptops coming out like the Asus UX501 - all Maxwell GPUs.

Even if R9 300 is great, if AMD can't execute with solid supply, it won't get design wins with OEMs. With NV, we know they can deliver millions of products at the customers demand. Winning new business isn't about just bringing out a new card - NV understands this better today than AMD does.

 

[raghu78]

Wow, either of those are some ballsy bets. I wouldn't take those if I were you. AMD can't just retain 15-20% market share in a matter of 3 months. Your estimate of Q3 2015 is completely unrealistic and even Q4 2015 sounds unrealistic. Even if R9 300 series is good, NV can easily release 960Ti, 970Ti and refresh 980, while at the same time dropping prices. Did you see NV's margins as of late? 54-56%. NV could probably drop the price of a 970 by $100 and 980 by $200 and still make more $ than AMD.

It took NV just one quarter to go from 65 to 75% market share when the competition had no reply and you feel AMD cannot do that even with a competitive product stack. btw AMD's market share numbers are misleading as they are not shipping much product so as to clear the channel of R9 2xx inventory. So thats why you saw Jon Peddie report that AIB market crashed in Q1 2015. My expectations are that Nvidia will be forced to cut prices and come out with new SKUs once the R9 390 and R9 380 series launch. I expect R9 380 at USD 300 - 350 to match or beat (5% faster) the GTX 980. R9 380X at USD 400 - 450 should be roughly 15- 20% faster than GTX 980. anyway I am willing to take a bet with you that AMD will be back to 35% by Q4 2015.

If NV wanted to, they could undercut the entire Maxwell series so much it would literally make it impossible for AMD's R9 300 series to compete. Why? Because NV has so much cash and almost no debt that they could practically make their market share go to 95% and bankrupt AMD. Luckily NV isn't interested in 15-20% margins and because they are aiming for > 50%+ margins that they haven't completely obliterated AMD by now. Realistically speaking, if NV really really wanted to, they could price GTX980 for $149 for 12 months straight and then GP204 for $149 for another 12 months, on purpose to just bankrupt AMD's GPU division. NV could survive this due to their cash reserves. Believe me, NV has all the cards in its hands, no matter how good R9 300 series can be. That's because financially, NV can withstand the price war - AMD cannot.

I think your logic is horrible. Nvidia cannot sell any chip at very low or negative margins as their shareholders are bothered about profitability and margins. The end result is Nvidia is there to make money and not gain marketshare at the cost of profits. Nvidia will do what it wants to maximize profits and cash flow. So I don't think Nvidia will enter a price war to keep the market share at 80% or even aim for 95% (which is frankly ridiculous)

Even if all of that is true, NV can respond with a price war the likes AMD has never seen. See my response to Silver above. Secondly, you can't take back that millions of people already upgraded to Maxwell and those customers are lost for another 2-4 years as far as their upgrade path goes. If R9 300 desktop cards are competitive but they are lacking in perf/watt, then R9 300 mobile won't get many design wins come Skylake/Windows 10 Q3-Q4 2015 and more sales to NV. For example, I fully expect the summer 2015 MacBook Pro laptops to be refreshed with Maxwell, not R9 300 series. It's hard to say what's going to happen with the Mac Pro because Maxwell is lacking double precision performance but the 12GB of VRAM would be a huge selling point for Apple to use Quadro GM200 cards in the Mac Pro. Reading on some Apple forums, the R9 M290X/295X got too much negative feedback which makes me think the next iMac refresh will also get Maxwell.

All the new cutting edge laptops coming out like the Asus UX501 - all Maxwell GPUs.

Even if R9 300 is great, if AMD can't execute with solid supply, it won't get design wins with OEMs. With NV, we know they can deliver millions of products at the customers demand. Winning new business isn't about just bringing out a new card - NV understands this better today than AMD does.

I think you are way too appreciative of Nvidia and way too critical of AMD. I may say that you exaggerate things to make it look as though Nvidia is invincible and can do no wrong. Let me say you are just not grounded in reality. Why do you think AMD started gaining market share in Q2 2014 once the bitcoin mining demand faded and supply met demand. Q3 2014 saw a big shift due to GM204. But still I wouldn't say AMD cannot gain back what they lost with a competitive stack. btw I expect mobile R9 380X with HBM to make some serious inroads in high end gaming laptops. You can come back in late Q3 and Q4 and say I was right.

 

[AtenRa]

Why would you design ROPs for Tonga that are 70% more efficient and then take that design that cost you millions of dollars and throw it all away and start over? I mean if you are going to go with the argument of using 128 ROPs, might as well use the same ones from Tonga and then you'll have overkill pixel fillrate.

Unless you are a GPU / electrical engineer, I don't see how you can state with such certainty that 128 smaller ROPs are better than "64 or 96 Tonga style ROPs". In fact, right now the amount of ROPs is one of the moving marks, far less certain than the consistenly leaked 4096-bit memory bus width, 4096 SPs and 256 TMUs specs over 1-1.05Ghz clocks. I am not even sure where some people on this board got 128 ROPs from?! I must have missed a rumour that even talked about 128 ROPs. When was this leaked?

Sounds like some of you guys are just chasing "paper specs" aka more is better.

AMD already doubled the number of ROPs with R9 290X but only increased shaders/TMUs by 37.5% over the 7970Ghz. With Tonga's colour compression, there is more than enough ROP performance with just 64 ROPs. AMD needs to focus on the shader, texture, L2 chace and geometry performance since by far those are the 4 biggest bottlenecks in the design. HBM1 takes care of memory bandwidth. ROPs should be the last area of focus for AMD with R9 390 series since there is going to be more than enough of pixel filtrate as is from Tonga's improvements alone. I am not saying I am right but I don't see why 128 ROPs is more credible than 64 or 96 ROPs. Where is the logic behind that claim?

First of all im not saying the R9 390 will have 128ROPs, it was just a number used to illustrate my point.

What im saying is that the number of ROPs alone mean nothing. You could have 128 ROPs and your throughput be the same with 64 faster ROPs.

 

[Silverforce11]

Are you already at 4K, or moving to a 4K screen this year?

Been waiting for a good 4K screen, if nothing comes up, I'll prolly go with a 1440p FS, but then FS not working with overdrive to reduce blur is a deal-breaker.

If I were gaming on 1440p, I would activate D/VSR because its AA is just flat out superior, given many games don't even have MSAA but instead blurfest post-AA. That would cripple 4GB GPUs, given that a 50-65% faster R290X can handle such settings, I don't want to be vram limited, best to be performance limited instead.

 

[RussianSensation]

Been waiting for a good 4K screen, if nothing comes up, I'll prolly go with a 1440p FS, but then FS not working with overdrive to reduce blur is a deal-breaker.

If I were gaming on 1440p, I would activate D/VSR because its AA is just flat out superior, given many games don't even have MSAA but instead blurfest post-AA. That would cripple 4GB GPUs, given that a 50-65% faster R290X can handle such settings, I don't want to be vram limited, best to be performance limited instead.

Don't you have dual 290/290Xs though? Sounds like you might need dual 390s at least to feel the upgrade as worthwhile.

I expect R9 380 at USD 300 - 350 to match or beat (5% faster) the GTX 980. R9 380X at USD 400 - 450 should be roughly 15- 20% faster than GTX 980. anyway I am willing to take a bet with you that AMD will be back to 35% by Q4 2015.

Wow, you are crazy optimistic, huh R9 380X 15-20% faster than a GTX 980 priced at $400-450? I was thinking R9 380X = 980 for $399 would already be amazing given the milking NV has been doing for so long. Even having 980 level of performance at $399 would already be a breath of fresh air but 15-20% faster than a 980? Sounds too good to be true!

 

[raghu78]

Wow, you are crazy optimistic, huh R9 380X 15-20% faster than a GTX 980 priced at $400-450? I was thinking R9 380X = 980 for $399 would already be amazing given the milking NV has been doing for so long. Even having 980 level of performance at $399 would already be a breath of fresh air but 15-20% faster than a 980? Sounds too good to be true!

My expectations are consistent with chiphell leaks and what I believe will be the actual chips.

http://wccftech.com/nvidia-gm200-titan-2-amd-fiji-380x-bermuda-390x-benchmarked/#ixzz3M5AGlIkF

Bermuda XT - 4096 sp
Bermuda Pro - 3840 sp

Fiji XT - 3072 sp
Fiji Pro - 2816 sp

R9 390X - 160-165
GM200 - 145
R9 390 - 145-150
R9 380X - 125-130
R9 380 - 115-120
GM204 - 110
R9 290X - 100

Fiji will be 15% faster than GTX 980. Anything above is all upside. btw I also believe there is improved perf/sp due to a improved microarchitecture which is why R9 380 will be significantly faster than R9 290X at the same sp count. Maybe you still believe AMD has not done anything to improve that . :whiste: