[Rumor (Various)] AMD R7/9 3xx / Fiji / Fury

[JDG1980]

I don't see getting 30% improvement over Hawaii for Grenada, that would indicate a massive performance per clock improvement. Nothing we've seen (outside of that link) indicates that much performance improvement and I don't think there's anything to say that the link is even legit. Even if it is legit, I think it is much more likely to be Fiji Pro. People have already got their hands on 390(x) cards and at best have shown a 5-10% improvement in Firestrike and lower power consumption. That's all I was expecting, tbh, and much more doesn't leave a lot of room in the performance stack for a Fiji Pro.

I think we're going to see the R9 390/390X outperform the R9 290/290X by roughly 10%. Looking back, the R9 270X had a 5% clock boost over the 7870 (in fact the transition from 1000->1050 MHz was exactly the same as they're doing on Hawaii now), and it had 16.7% faster RAM (1200->1400 MHz). For these changes, it got just under 10% performance improvement over the 7870. The 290X->390X transition has the exact same clock speed improvement (5%), and a 20% improvement in memory speed (1250->1500 MHz). So 10%-12% performance improvement over the 290X seems reasonable, given past precedent. Regarding power consumption, I'm waiting to see what TechPowerUp finds, as they have the best records of GPU power usage (AnandTech can't seem to separate the GPU from total system usage, which screws up their numbers).

 

[DiogoDX]

About 24 hours for the presentation and no solid leaks. :/

 

[RussianSensation]

Ah now we have a pickle. Is the extra 4GB useful on the 390X, and the 4GB limit on Fiji going to be a problem. Or is the 4GB limit on Fiji enough, and the 8GB 390X is just a pointless cost addition on 4GB 290X. Going to be difficult to eat both cakes.

8GB is worthless for Hawaii but 4GB isn't. The key videocard that has been devastating AMD's 290/290X has been the 970. The 3.5GB VRAM resulted in lower sales in February but since then most gamers forgot about the issue. With AMD positioning R9 390 with similar performance to a 970 but having 8GB of GDDR5 will really drive the message to the type of gamers who buy things at places like BestBuy. People have been asking AMD to improve their marketing, here it is. If 970 and 390 are similar in performance, it makes it a lot more difficult to recommend a 3.5GB card vs. an 8GB card because at that point 8GB becomes a 'free' bonus.

I have that monitor. It's huge. It took a bit to get used to it. If you have multiple monitors, this can effectively replace a couple 27 monitors or 3 24 inch monitors. The DPI is almost identical to the 27 inch 1440p monitors. The contrast is WAY better than IPS. Viewing angle is great. Input lag is low. I like it. But, it's a PWM monitor. That might turn some off. From my experience, a single GTX Titan X overclocked to 1550 still isn't enough to drive 4k. You'll need 2 top end GPUs to do 4k justice.

That means you'll still need 2 FIJI gpus to do 4k with good eye candy.

:thumbsup: Thanks for sharing your views on the monitor and your experience. You also keep highlighting a recurring theme - PC gamers who actually own 4K monitors continue to comment how one really needs dual flagship cards to really enjoy it vs. other PC gamers using 1080P-1440P monitors on a mid-range 290X/970/980 card saying that modern games can be played at 4K at high quality settings 40-60 fps, clearly a theme that contradicts every professional review site and every 4K owner's experience on this site.

The 290X cards with 8gb didn't make any waves:

290X with 16GB and $199 price wouldn't make any waves. R9 200 series has a tarnished reputation among the mainstream media/gamer for being hot and loud. These 2 aspects will no longer apply with 390/390X. 290 just needed about 5-6% more performance to match a 970 and 390 should be close to that but have at least true 4GB of VRAM against 970's 3.5GB, and now the 390 will run cool and quiet - addressing 2 major concerns of the 290 series tarnished image.

OK, AMD is dead now... and HBM is the last blow from them.

AMD could sell 0 video cards for the next 18 months and not go bankrupt. You might want to look up how the firm actually works.

That could turn out to be a critical mistake. What is better, having a low volume set of GPUs or the rest of the lineup updated and competitive in all metrics.

True, but in this case all the GPU design work has been done on R9 300 series which means if you amortize the R&D costs, and consider how little time if left on this 28nm node for these cards before next year, I don't think the cost-benefit analysis would have justified creating new GPUs top-to-bottom on 28nm node for AMD. Their priorities are the future - R9 300 series is not it. If it costs $250 million - $1B to design a new GPU, do you honestly think AMD would have been able to make $ if they designed 3 new ASICS? I think Lisa realized from a financial perspective, it would have been a less profitable venture. Pretty sad state of affairs from both AMD and NV in the $300-500 range right now.

HDMI 2.0 certainly has enough bandwidth to do 4K/60Hz/ 4:4:4. I've tested it myself on one of the Samsung 4K sets.

However , it cannot do 4K/60Hz/ 4:4:4 if HDCP 2.2 is active due to added overhead. No current video cards support HDCP 2.2, but it will be required for 4K Blu-ray.

Thanks for clarifying that. Sounds like most 4K TVs and all 4K capable videocards aren't future-proof then for 4K HTPC. Hopefully by the time Pascal launches, we will have DisplayPort 1.3 and next gen 4K hardware HEVC capabilities top-to-bottom.

This specific forum on Anandtech is Video Cards and Graphics. People have been asking for a Display forum for quite a while. This forum has it's niche, just like others on the net. I'd recommend looking into the AVSForum.com for broader display discussions. HardOCP display forum is pretty good as well. Head-fi.org is great for audiophiles. But I'd understand if you want to maintain the Anandtech community.

:thumbsup: Good thing they are now adding a Display section. It never made sense to me to have so much focus on videocards but so little focus on monitors when a lot of gamers entering the PC gaming industry (younger generations) need a new monitor too. Even the traditional review sites place too little emphasis on monitor reviews vs. videocard reviews. When we use our PC, we interact with a monitor in both 2D and 3D work and it outlasts any GPU. A monitor buy imo is a more important component since a $650 videocard with a crappy $200 monitor is still a crappy gaming experience imo.

Look again, you'll find plenty of games that are playable. Check HardOCP, TPU, AT, TechReport, etc... I know, they are American sites and they're run by typical Americans and blah blah blah, spare me the drama. Why are we only including games within the last 8 months? Who made up that random timeline?

I am not going to waste my time. 4K is not playable on a 970/980/290X or therefore 390/390X at good setting and FPS and all 4 sites above confirm. Go read more reviews and come back with an objective response.

Face it, people are using the HDMI 2.0 ports on the Maxwell cards. Whether you think they should be or not (like that opinion matters to them at all), they are. And to say it is all just a gimmick is laughable at best, trolling at worst.

Way to take my posts out of context. Yes, HDMI 2.0 is useful but to imply that R9 300 series is DOA because it doesn't have HDMI 2.0 is stretching things. You keep saying how HDMI 2.0 is useful for gaming but the reality is for 4K gaming, none of these cards in question are sufficient.

And before my phone dies, I didn't upgrade out of necessity, I upgraded because I had to sell my 770 before the resale value tanked on it. We went over this before... At that time, I needed a new card. And the 970s were both faster and cheaper than the 290, so guess which one I went with! :awe:

1. The fact that you bought a 770 during R9 280X generation shows you don't care about price/performance at all. 280X was $299 vs. $380 for 770 2GB (obsolete) and $450 against 770 4GB. Also, the fact that you waited that long to buy a 770 and skipped 1Ghz 7970 cards that were $300 8 months before 770 even released shows you had no interest in any AMD card.

2. 970 was only cheaper than a 290 for 1 month at best. You aren't discussing how an after-market 290 was $350-375 5 months before 970 launched or how 290 was $400 10 months before 970 launched. Again, there was plenty of time to sell a 770 and buy a 290 but you didn't do that either. Instead you waited 10 months to get a card 5% faster. Again, shows you had no interest in buying an AMD card. You aren't fooling anyone trying to spin things as if your 970 purchase was objective. This entire forum already knows you only buy NV cards and the fact that you owned GeForce 5 and 7 and Fermi puts you in a very special group of GPU owners, which I won't name as I'll get an infraction.

Just look at 1440p displays. Even in 2015 almost none of them support 1440p over HDMI, which is a HDMI 1.3/1.4 feature. Most 1440P monitors are still limited to 2005/2006 HDMI refresh rates and resolutions.

That's true but if we are talking PC monitors and not TVs, as long as they have DisplayPort 1.2, that's enough to drive 1440P-4K. I got a BL3200PT and I don't care if it has HDMI 2.0 since it has DP1.2. HDMI 2.0 in general is inferior since it can't do FreeSync/GSync which means it's soon going into legacy status for PC monitors once adaptive sync gains more traction or manufacturers release HDMI 2.0 monitors that can do adaptive sync. The problem is the 4K TVs don't have DP connectors. That's a way bigger deal than for a PC monitor to not have an HDMI 2.0 connector. We can use DP for a PC monitor but if one gets a 4K TV, they are stuck using HDMI 2.0.

Sapphire's R9 290X Tri-X was clocked at 1040 MHz core and 1300 MHz RAM. It draws a maximum of 316W in FurMark, and a peak of 253W during normal gaming. I think we'll be seeing similar figures here, since it's the same chip with maybe a new stepping if we're lucky. But it's not going to go to 375W.

Friendly advice: stop using FurMark as a measurement of a videocard's real world power usage. Just like no one would use LinX + FurMark simultaneously on a PC system to figure out the maximum power usage when choosing a power supply, we do not use FurMark to extrapolate a GPU's real world power usage in games. As has already been explained to you many times, an identical GPU (GM204) will draw 350W of power in Furmark but only 200-210W in games because the PCB is designed for that (dual 8-pin, 8 power phases, etc.). One can easily design a 980 that is capable of drawing 500-600W of power in Furmark. You need to be able to understand how Furmark works both on the ASIC and the PCB/power circuitry components. It's a power virus, not a game.

Ugh, straight rebrands. How could anyone think this was a good idea?

It's not a good idea for us consumers this generation but for AMD's long-term future and hopefully a solid 14nm HBM2 line-up, this is probably a smart decision. Also, don't forget that AMD's GPUs tend to drop in price quicker than NV's. If R9 390/390X launch at $329/389, in about 5-6 months it should be possible to find them for $260/300 with rebates.

Agreed. Given AMD's R&D limits, they would have been better off putting the money into two smaller chips instead of Fiji.

Would have been true if this was the beginning of an all new 28nm node, if we were at the bidding phase for new laptop design wins and if AMD had a clear strategy of getting a lot more notebook OEM design wins. None of these apply. We are at the end of the 28nm node gen, all major laptop GPU design wins have been won by NV already. AMD did get GCN into Macs so that was a major marketing win. Fiji can be reused on a 14nm shrink since it was designed to use HBM. It's probably cost prohibitive to redesign Tonga, Bonaire, Hawaii, Tahiti to 14nm HBM2. At that point AMD will need an entirely new stack of chips -- and that's hopefully where they invested the money after choosing to refresh R9 300 series like that.

Also, your argument is inconsistent. You criticize AMD should have redesigned the entire stack top-to-bottom this round but then you say NV can readily drop the prices of Maxwell. That's exactly why Lisa Su would have been suicidal to spend $250M-1B to redesign 28nm low- and mid-range GPUs because NV could have just dropped 960 to $149, 970 to $249-269 and 980 to $399. Do you see now why what you are saying is completely inconsistent with the other point you keep making that NV's high profit margins could easily allow them to make AMD's new line-up obsolete? Don't you think Lisa Su knows this very point which is why she probably figured it's better to sell the aging 200 series at small profits than take massive $500M investment risk on redesigning the entire stack, and then not even win because nV could drop prices?

Also, it seems you keep using market share and sales numbers to drive your point. I mean if we look at the performance of 750/750Ti/960, it's nothing special. So if someone was a budget gamer and had a 300W PSU, would you recommend them a 750/750Ti or a $20 more expensive 30-45% faster R9 370? You seem to have not addressed this point.

Maybe a lot of less knowledgable gamers have no clue how Pitcairn pounds 750Ti into the ground in games? R9 300 series helps AMD reset this idea.

http://www.computerbase.de/2015-05/...ergleich/2/#abschnitt_leistungsratings_spiele

The negativity on the rebrands is understandable but declaring Fiji a failure before it ever comes out is silly at best.

There was hardly any negativity on 560 and 770. 770 2-4GB was $80-$150 overpriced compared to the 280X but it was cheered by professional reviewers. Also, many gamers don't remember (or are ignorant) but 680 sold for $320-340 when GTX770 came out for $379. That meant 680's replacement had worse price/performance to fire sale 670/680 cards. Don't remember this forum bashing a 560 or 770 though. Also, it's quite ironic how price/performance of 390/390X is being attacked now by the same people who ignored price/performance of 290/290X cards because to them after-market 290/290X cards didn't exist for the last 1.5 years.

Fury is a new design, not a rebrand. But it will sell miniscule compared to other chips. Specially the 370 and down. Chips(250/270) that desperately needed to be moved away from GCN 1.0.

It's longer term thinking. Regain flagship GPU performance, keep investing into future GPU tech, and Fury designs can be used as a foundation for next gen mid-range 14nm HBM2 product. People actually expect NV to release R9 400 and 500 and 600 series. I know this concert is hard to grasp for you since you've been waiting for decades now for AMD to declare bankruptcy. That's why it's hard for you to grasp why AMD's GPU division keeps investing into tech of tomorrow.

How confident are you with this prediction?

You have to ask? According to him AMD won't improve perf/watt this generation at all. He is in for a shock when Fiji is > 30% faster than Hawaii and it won't be just from HBM1.

Please explain this to me. I always thought if a card uses 275 watts it always consumes 275 watts.

Power usage is also impacted by temperatures and dynamic voltage/power load balancing. With lower temps, power usage falls. With fast switching dynamic voltage/load balancing, a reference 980 can use 165W of power in games but in compute, that can easily go over 200W. Because compute is a constant workload, there power load balancing doesn't work and dynamic voltage doesn't help since the GPU is pegged at 100% usage the entire time. Games do not do that which is why high frequency dynamic voltage switching and power load balancing help GPUs use way less power in games than more strenuous compute workloads.

For the sake of the argument. Lets just say that AMD release the 390X and its performance is pretty much equal to the Titan X but is significantly cheaper, do you think Nvidia might change the price of the Titan X at all? I'm keeping an open mind though I'd like a Titan X and would wait a few weeks to see if the price will budge at all.

No. Even after R9 290X came out at $549 and even after 780 Ghz editions beat the OG Titan, NV's $1K price remained. Even if Fiji is magically 15% faster than the Titan X, the TX will still cost $1K. Recall that NV sold slower, hotter, louder Titan Z for $3K when R9 295X2 cost $1.5K and beat it in nearly every key metric.

There is really no point to buying a TX anymore unless you are running 3 4K monitors in surround.

Don't say that! You might hurt someone's e-peen.

Site blocked at work, is that:
8% over GTX 980 power consumption for that 20% perf
or
8% over R9 290X power consumption for that 20% perf?

I'm assuming the former, which would be a huge boost to AMD. Reviews would have to point that out.

Chances are Simple Jack was either a made up BS/slide, or it was an early version of Fury X / Fury tested at lower clocks early in the development cycle.

Well that is sounding awesome for Fiji.

I haven't lost hope on it though.

This is good to hear

It's one of the reasons 300 series and Fury are marketed differently. You have 'legacy' GDDR5 products and all-new HBM1 products. With rumoured 4000 shaders, 128 ROPs, 512GB/sec bandwidth, Fury should be substantially faster than the 390X. In fact, if you look at AMD's new gen flagship cards built from the ground-up, besides HD5870->6970, AMD's next gen flagship is at least 30% faster. That means Fury X should be at least 30% faster than the 290X.

Even if Fury Pro is 15% faster than a 980 and Fury X only matches a 980Ti, it's still good to see 2 strong players in the GPU market this close again. Remember than NV's flagships tended to outperform AMD's by 10-18% over the years. If Fury is identical in performance to a 980Ti, it would mean AMD would have completely closed the flagship gap NV enjoyed with 280/285/480/580, etc. Stronger competition is good for the consumer since we could see more game bundles, rebates and price drops and lower 2nd tier SKUs as NV/AMD desire to win sales. Think about it, less than 1 months ago we had $550 980 and today we have a $650 980Ti that's way faster. Add Fury to the mix and the flagship $500+ landscape will change dramatically putting 970/290X/390X/980 strictly into mid-range category.

At the same time, we have to be realistic with our expectations. For example, after-market 980Ti cards when max overclocked are 24-35% faster than a reference 980Ti. As a result, no one should expect Fury to "blow away" GM200. It's not going to happen simply because GM200 overclocks and subsequently scales very well with increased clock speed.

About 24 hours for the presentation and no solid leaks. :/

If you got time to burn, there is a lot of E3 gaming content.

Where to Watch E3 2015 Press Conferences
http://www.playstationing.com/ps4/where-to-watch-e3-2015-press-conferences/2002

I am actually more excited about the new games coming out than Fiji -- Fallout 4, Doom, Forza 6, Rise of the Tomb Raider, Halo 5 Guardians, Xbox one gaining BC with 360, FF7 remake, Dishonored 2, etc.

 

[LTC8K6]

https://twitter.com/dankbaker/status/609445259597213696

Dan Baker's Tweet

Tweets like that about only needing 1 GPU to "play well" at 4K, will have people expecting Fury to blow the 980Ti out of the water like Bismarck vs Hood.

I hope it's true, because if it is, then the GPU war is on for sure.

 

[bryanW1995]

Couldn't this stuff be talked about in another thread? Off topic conspiracies about nvidia surely don't have a place in the Fiji/300 series thread

Naw...what else is there for AMD people to talk about lately? It's been all smoke and mirrors unfortunately. Hopefully fiji offers some respite from that.

Would love me some troll on troll action.

LOL

 

[AnandThenMan]

Fury isnt a new tech besides HBM.

Fury is a new design, not a rebrand.

I don't know how to reconcile these two. Old tech new design? How can it be old tech when it uses a brand new memory architecture/controller and very likely an update to GCN.

 

[tential]

Quite sad how few people understand amd right now. The company has had some major personnel changes and things will change. The 390x with 8gb is vram like rs said is uninteresting to us. But to casuals? 8gb > 4gb so purchase. Too many of you are too informed (and not even that we'll informed to be perfectly honest) and don't actually understand how markets work.
Which is easily proven by your comments on "lol amd bust bankrupt next year!" that's not how business works....

You'd have to try stupid hard to run your company into the group and completely destroy it to the point where it's no longer an ongoing concern. When amd is worried that they will no longer be an ongoing concern im sure we'll know.

 

[bryanW1995]

Yes, that's true - if they can pull that off. But will it still use >300W in GPGPU and FurMark?

<3 me some FurMark

Looks really small.

That's what she said. Um, er, to that other guy, not to me ofc.

You need FurMark figures for the video card to determine the required power supply size for a build. You also need this to help determine how much case cooling is required (and, if you plan to replace the stock cooler on the video card, what coolers you can get away with using).

What if Furmark is throttled by driver detection but GPGPU apps do not?

Shouldn't one base their rig around gaming if that's the purpose of the rig?

I'm hosting a multiplayer Furmark party next Thursday. IT WILL BE EPIC.

 

[3DVagabond]

Unless the new series throttles Furmark like the old days, that's a massive efficiency gain. But the clockspeed is 980mhz flat (& GPU load maxed) so it does not look like major throttling is occurring?

As said and probably something most of us can agree to, for the rumored price, its worth it if its 10% faster with some efficiency gains, otherwise is a major fail. Here's hoping AMD is competitive.

Looks like more of a TDP limiter than throttling.

 

[ShintaiDK]

I don't know how to reconcile these two. Old tech new design? How can it be old tech when it uses a brand new memory architecture/controller and very likely an update to GCN.

Its most likely GCN 1.2 based. And unlike the complete rebrand of the 300 series. Its actualy a new chip.

 

[bryanW1995]

Im between 390 and Fiji pro, lets see which one will get my money.


Unless NVIDIA surprise us with a nice GTX980 price reduction.

That is NV's ace in the hole. They have a much greater ability to slash prices and maintain good profit than AMD at this point...yet another benefit of not being 9 months late to market. I recommended to someone the other day to hold out for gtx 980 price cuts and caught a bunch of crap from Keys about it...but I think that my advice was sound. The best thing that could happen (to those of us who aren't on red/green payroll ofc) is for 390(x) and Fiji to be awesome, prompting Green team into aggressive price cuts.

If the Fury is a dissaster, AMD is dead.. Also... NVIDIA kiled their own products with an expirable 980ti. It will be the original Titan all over again

I must have read something similar to this 10 times so far in this thread...AMD is (probably) dead regardless of how great this launch is. Think about it, what's the best case scenario for them?

390 > 970 in performance and power usage
390x > 980 in performance and power usage
fury pro > 980 ti
fury x > titan x

Say they win every single category by 10%, and the power usage is the same. Wouldn't that be a home run for AMD? Unfortunately, a LOT of people who would have bought AMD over the past year gave up and bought NV. A lot of people who had NV and wanted to upgrade but were on the fence about which camp to go with also bought NV. AMD literally skipped an upgrade cycle, or at least have come into the middle of one. And, if they hit this theoretical home run, NV has proven in the past that they will ruthlessly cut prices to maintain sales.

NV has the name recognition and they have 3/4 of the gpu market, they could survive being ~ 10% lower at each price point and still maintain 50%+ market share or drop prices a bit and keep 60% + (recent 70+% is more of a function of AMD's lack of anything competitive than it is a realistic barometer of true market sentiment).

Oh, and even if none of that works, NV could quickly come out with Titan X PLus, GTX 980TI Plus, GTX 970 TI, all with 20% higher clocks.

Don't get me wrong, I'm still excited to see what AMD brings to the table here tomorrow, but I'm not a foolish romantic thinking that whatever they bring out is somehow going to save the company. Or kill it for that matter, expectations are so low for them these days that they can't help but outperform them.

I read the last few pages and I'm just shaking my head here...

I wont go into details because I don't want to type out a rant but I can't honestly believe that there are people who are in favor of dropping the press in favor of bribed social media advocates that literally do nothing but spam covert AMD advertising on forums and twitter.

I mean holy crap.

They are just playing the game by the rules dictated by the semi-monopoly that has a death grip on discrete gpu sales...

 

[sandorski]

<3 me some FurMark



That's what she said. Um, er, to that other guy, not to me ofc.





I'm hosting a multiplayer Furmark party next Thursday. IT WILL BE EPIC.

Frag 'til you slag!

 

[Udgnim]

24 hours before AMD reveals 4096 SP Fury which we already know and no released benchmarks

and then we riot

 

[crashtech]

Ugh, straight rebrands. How could anyone think this was a good idea?

It's not a good idea for us consumers this generation but for AMD's long-term future and hopefully a solid 14nm HBM2 line-up, this is probably a smart decision. Also, don't forget that AMD's GPUs tend to drop in price quicker than NV's. If R9 390/390X launch at $329/389, in about 5-6 months it should be possible to find them for $260/300 with rebates.

I can't totally disagree, but there was a hope that the rebrands would at least move one notch down the stack, for example, the 290X would become the 390, and the 390X would be something a least a tad better than just an optimized and OCed Hawaii. At least then it wouldn't smack of expediency so much. I get that Fiji probably consumed all their available resources, but it's not clear that virtually abandoning the rest of the product line is a smart move, imo.

 

[crisium]

Tonga was their chance for a new tech. All they had to do was improve on it. Imagine if we had a 3072 Shader, 384-bit 6GB Tonga as the 390X? More shaders, even if it stayed 32 ROPs it'd be better than Hawaii 64 ROPs and use less power, and 384-bit with color compression should equal 512-bit Hawaii with less power. Even if it were 2816 shaders, it'd still be an improvement over 290X. Just give us 2 new cards (2560 and 3072 Tonga) along with the 285 and Tonga XT:

1792 Tonga to fight the 960 as it does now
2048 to fight the 960Ti and/or the gap between 960 and 970
2560 and 3072 to battle vs 970 and 980

Was it just not worth it to invent new cards since they were so late and 970 already sold so much?

Instead, they won't even give us the full 2048 Tonga.

 

[3DVagabond]

I can't totally disagree, but there was a hope that the rebrands would at least move one notch down the stack, for example, the 290X would become the 390, and the 390X would be something a least a tad better than just an optimized and OCed Hawaii. At least then it wouldn't smack of expediency so much. I get that Fiji probably consumed all their available resources, but it's not clear that virtually abandoning the rest of the product line is a smart move, imo.

The reality of it is besides Fiji there really isn't much to gain for the investment.

Before the "Kepler effect" the 780 ti was ~6% faster than the 290X. The 980 is ~10% faster than the 290X. The 970 trades blows. I realize GM204 uses less power but actual game performance hasn't changed at all. If ~50W really matters to someone then they can buy it. AMD is going to try and offer 2x the RAM for less money @ the same performance level. Seems like a reasonable counter.

Sometimes being on the leading edge of tech has pitfalls when things are late or yields are bad.

 

[swilli89]

I just wish AMD could have offered a 4gb "390" for $250. Would have been killer. Regardless, tomorrow is a day tons of us have been looking forward to and will bring with it some juicy new information.

 

[Despoiler]

24 hours before AMD reveals 4096 SP Fury which we already know and no released benchmarks

and then we riot

We actually don't know Fury is 4096. That's been one of the longest running guess of a rumor out there.

Tomorrow is supposedly just an introduction not necessarily the full reveal. Supposedly more info on the 18th (specs). Rumored reviews and availability around the 24th.

 

[Elixer]

Does that mean this thread can finally die at 12PM EST time when they finally do whatever they are going to do on their live stream?

 

[BryanC]

Who thinks we'll actually get benchmarks tomorrow? I'm thinking it will be like Hawaii launch, where we heard only vague ideas about performance.

 

[Stuka87]

I just wish AMD could have offered a 4gb "390" for $250. Would have been killer. Regardless, tomorrow is a day tons of us have been looking forward to and will bring with it some juicy new information.

Yes, an end to the rumors at last!

 

[monstercameron]

Yes, an end to the rumors at last!

but we wont have all these flamewars anymore...

 

[JDG1980]

We actually don't know Fury is 4096. That's been one of the longest running guess of a rumor out there.

Tomorrow is supposedly just an introduction, not necessarily the full reveal. Supposedly more info on the 18th (specs). Rumored reviews and availability around the 24th.

This is dragging on way too long. AMD really needs to crap or get off the pot. A June release is much later than expected anyway, and almost everyone was expecting a Computex reveal. Then AMD specifically said the reveal would be at E3. If they now say "just kidding, we're not going to show you anything but a few worthless marketing slides" then a lot of people are going to decide it just isn't worth the wait. The GTX 980 Ti is already out there and selling like hotcakes. AMD doesn't have the luxury of time. The longer they stretch this out, the less confidence I have in the performance of their product. These repeated delays smell like they're trying to patch things up at the last moment. According to Hardware Canucks, "Drivers are still in their pre-beta form and clock speeds are undergoing final tuning." At this late date, that's incredibly alarming. It basically sounds like Fiji is a disaster and they're desperately trying to whip it into competitive shape.

 

[Silverforce11]

Who thinks we'll actually get benchmarks tomorrow? I'm thinking it will be like Hawaii launch, where we heard only vague ideas about performance.

They are launching Fiji at E3, per their statements earlier at Computex. We'll probably get in-house benchmarks, which is the best case scenario, 4K results where Fiji shines with its 128 ROPs and faster vram.

Review site NDA should lift after. Rumored on the 24th.

 

[Grooveriding]

Who thinks we'll actually get benchmarks tomorrow? I'm thinking it will be like Hawaii launch, where we heard only vague ideas about performance.

It's E3, a gaming convention. When do we ever get worthwhile benchmarks direct from either vendor ? Usually the best you can hope for are the usual benchmarks you get straight from AMD/nv that show the performance increase against their last generation or the competition; 1.2x the perf of Y, 1.4x perf/w of Z etc. This is just a marketing hypefest. They both do this where someone gets on stage and shows a product and says look at the shiny. It's usually the day they give reviewers a sample as well, so my guess is 1 or 2 weeks from tomorrow is when the card will launch.

We will probably get core count, memory amount/bandwidth and some other spec information. My guess is temperature and noise level information as well. This card has no fan and is fully watercooled, so they'll have both those metrics won and in their pocket. There will be no worthwhile performance benchmark information until reviewers put up their reviews the day the card goes on sale.