Gemini (FuryX2) looms near

[moonbogg]

:thumbsup: Thank you, and that's max overclocked. People over-exaggerate power usage and consequently the PSU requirements so much, it's ridiculous.

It's not correct to estimate power usage by taking maximum CPU TDP + maximum GPU TDP x2. That is not how games scale or how real world power usage works because you aren't going to have 100% CPU load + 100% GPU 1 + 100% GPU 2 load unless you are running some synthetic combination of Intel IBT/Prime95/LinX + FurMark

Yeah, I was freaking out about my PSU when I got my new cards, but I bought a kilowatt meter and discovered about 800 watts coming from the wall with 3930k @ 4.6 and 980ti's at 1400-1500.
Some stupid website that estimates your PSU needs for you told me to buy a 1600 watt PSU or something ridiculous like that. First I freaked out, then I bought a killawatt meter and used it, then I lol'd.

 

[Silverforce11]

It actually doesn't make that much difference in reality if you CF Fury, Nano or Fury X. This is because the performance difference among those 3 cards isn't that large and CF (and SLI) scaling is not 100%.

CF vs SLI at 1440 and 4K is entirely different ballgame to single GPU comparisons.

Even pre-OC 980Tis in SLI cannot keep up with stock Fury X CF. Read the Techspot review over a wide range of games. They got one of the better OC 980Ti models and it was slightly slower overall with ~20% worse frame times.

They had to max OC the 980Ti SLI for it to end up ~4% faster.

So Fury X2, running cool and quiet dumping heat out of the case for $999 would stomp on SLI 980Ti.

You'll need hybrid water cooled 980TI SLI to compete with the stock performance on Fury X CF. If you're ballsy, you can push Fury X another 10% with the vcore OC support now.

980Ti OC is only king in single GPU mode.

 

[RussianSensation]

Yeah, I was freaking out about my PSU when I got my new cards, but I bought a kilowatt meter and discovered about 800 watts coming from the wall with 3930k @ 4.6 and 980ti's at 1400-1500.
Some stupid website that estimates your PSU needs for you told me to buy a 1600 watt PSU or something ridiculous like that. First I freaked out, then I bought a killawatt meter and used it, then I lol'd.

I know right. And the other thing is the @ the wall reading is less than what the actual load on the PSU is. Also, I guess from the past when a lot of PSUs were poor quality, we are told to have headroom on our PSU, usually it's too much headroom. The top PSU brands rate their PSUs at actual operating load. There is no need to make this adjustment. Essentially your 1050W PSU is actually rated to run at 1050W. No need to think that a 1050W Enermax, Corsair, SeaSonic, EVGA PSU can only safely run at 800W.

His point about requiring 1000W PSU would make a lot of sense if it was 7-10 years ago when they cost $300-400. Today, you can find a modular Platinum 1000W PSU for $100.

So Fury X2, running cool and quiet dumping heat out of the case for $999 would stomp on SLI 980Ti.

You'll need hybrid water cooled 980TI SLI to compete with the stock performance on Fury X CF. If you're ballsy, you can push Fury X another 10% with the vcore OC support now.

980Ti OC is only king in single GPU mode.

Ya, but the problem is people have been buying 980Ti SLI for 6 months now and soon we are into December. When is AMD launching this Fury X2? I also do not believe they will price it at $999. That's just me throwing that out there. Also, it's about perception too. The perception is 980Ti OC > Fury X therefore 980Ti SLI > Fury X CF.

Don't forget that in this market segment, a lot of gamers are now making the choice of G-Sync vs. FreeSync monitors and G-Sync has a big edge with 3440x1440 100Hz, 2560x1440 165Hz and native HDMI 2.0. For people spending $1000-1500 on GPUs, these things start to matter a great deal.

and I think it's kinda hard to overlook the 4GB vs. 6GB even if today 4GB is enough for most games. That's why I think Fury X2 has to be $999 to make a splash. If it's $1299-1499, it's going to be priced out of the market for most imo.

 

[Silverforce11]

Surely you jest if you think 4 vs 6GB will matter, even at 4K, no game under playable settings are limited by vram.

They are limited by GPU grunt, which CF and SLI of this caliber still cannot handle maxed out settings + 4x MSAA for vram to be an issue at 4K, which is a niche still in its infancy. At resolutions below that, not a chance.

The 2GB transition happened at a odd time when next-gen consoles are out with devs working on those specs, this is why we're seeing 2gb limitations, as they can't handle 4K textures. Now that we're well into this console generation, the ceiling is reached. I'm sure someone out there will release 8K texture mods but besides that, nope.

Re the perception, if its out, reviewers will be comparing Fury X2 vs SLI 980Ti.

 

[MagickMan]

I think a pretty good rule of thumb now is to add up the value of all the other components in your system, then buy the best PSU you can find for ~10% of that.

 

[RussianSensation]

Surely you jest if you think 4 vs 6GB will matter, even at 4K, no game under playable settings are limited by vram.

For most games you are right but what about Fallout 4 modded, GTA V modded? We haven't seen those scenarios.

Now that we're well into this console generation, the ceiling is reached.

You think 4GB will be enough until the end of PS4/XB1 generation though? Remember, a lot in the market is about perception too. When 2 cards are roughly neck-and-neck in performance at stock but one is much faster OC and has 6GB of extra VRAM, the second card will sell much better - plus it has the NV brand name.

Think about it, Fury X CF isn't selling at $570x2 so what makes you think Fury X2 at $999 would sell against $550x2 after-market 980Ti SLI?

Re the perception, if its out, reviewers will be comparing Fury X2 vs SLI 980Ti.

See the market already voted and they voted 980Ti / SLI > Fury X/CF. Once the perception of a card is instilled in the marketplace, it's going to be hard to change it. Remember 680 SLI vs. 7970 CF or 780Ti SLi vs. 290X CF or worse yet R9 295X2 vs. 980?

As crazy as it sounds, even if Fury X2 was $799, 980Ti would still outsell it. I think AMD needs to launch with good performance out of the box on day 1 and beat NV. Only then will the average consumer even start considering their flagship cards at launch. I mean just look at HD7970 vs. 580, or HD5870 vs. 285. Even when AMD was flat out dominating, it hardly mattered.

 

[lavaheadache]

For Amd to sell cards well, it doesn't just take making the best product possible. It also takes Nvidia making the worst product possible at the same time. When this happens, Amd will sell half as many units as they should meanwhile still being outsold by the green giant.

 

[Teizo]

Advertising a busted cooling loop? Probably not the best way to introduce speculation into an unreleased GPU.

 

[3DVagabond]

It actually doesn't make that much difference in reality if you CF Fury, Nano or Fury X. This is because the performance difference among those 3 cards isn't that large and CF (and SLI) scaling is not 100%.

Nano + Fury X CF is just 16% faster than R9 295X2. Think about $520 R9 390 CF vs. $1140 Fury X CF. AMD's problem is that Fury/Nano/Fury X do not scale linearly vs. 390/390X. Prices on 390/390X are so much lower, ensuring that Fury X2 would be still be meh value even at $999 against R9 390/X CF. That's why AMD needed this card June-July 2015 while prices of R9 390/390X were still high, while the demand for whatever few Fury cards were available was also higher than today.

The overclocking of 980Ti is a killer feature for NV this generation in this market segment.

Right now it's possible to buy an after-market 980Ti for $560 and that's without taking $15-25 off with various AMEX?Visa deals. Would you buy a Fury X2 for $999 when GTX980TI SLI when the latter has 25% overclocking headroom and 6GB of VRAM as a bonus, but costs just $1120? I wouldn't.

Haven't read a TPU review recently. Do they now use a full game suite that scales with crossfire? Or do they still keep a few ringers in their that don't scale?

EDIT: OK, just looked and they still have broken games in there. Then they're still using different drivers for different cards so it's impossible to compare scaling even between 295x2 and Fury cards.

They always include broken games to pull down the average crossfire scaling percentages. Plus most of the suite is CPU bottlenecked at anything less than 4K.

Not your fault, but TPU is complete BS with Crossfire results.

 

[Silverforce11]

@RS

If its as you say that AMD won't even be able to sell Fury X2 for $799 and the hordes will default to 980Ti SLI.. then why bother? Why even make any GPUs anymore, right? Just quit, sell all the assets, fire all the engineers and marketing, exit the market.

You know what I thought when I found out about the 390/X prices? I thought "No way they are gonna be able to sell Hawaii at that prices, they are effing nuts!"... I was wrong.

390 is even voted 2nd best GPU by gamers for 2015!! The best going to the 980Ti, well deserved.

http://www.game-debate.com/awards/2015/best-graphics-hardware

How did that happened? Custom R290s for dirt cheap for months before the 390 debuted... but they sold very poorly, but the 390 manage to sell well? Even JP Morgan has the last quarter, AMD gaining back some dGPU marketshare, which is not what we were expecting.

So at this point, rather than assume Fury X2 won't be competitive (it will, it will demolish reference 980TI SLI and match custom OC 980TI SLI) at $999, let's just wait and see.

IMO, the biggest reason to not get these top end SKUs now though is the looming next-gen over the short horizon.

 

[RussianSensation]

For Amd to sell cards well, it doesn't just take making the best product possible. It also takes Nvidia making the worst product possible at the same time. When this happens, Amd will sell half as many units as they should meanwhile still being outsold by the green giant.

hahah :thumbsup: :wub:

WCCFTech speculates December 2015 launch, 1Ghz clocks for each Fiji chip, 4GB HBM1 per chip and 375W TDP. R9 295X2 had 500W TDP. If this card has 375W TDP limit, I don't see how it will be able to compete with two stand-alone Fury Xs or especially the overclocked 980Ti SLI setup.

 

[Silverforce11]

hahah :thumbsup: :wub:

WCCFTech speculates December 2015 launch, 1Ghz clocks for each Fiji chip, 4GB HBM1 per chip and 375W TDP. R9 295X2 had 500W TDP. If this card has 375W TDP limit, I don't see how it will be able to compete with two stand-alone Fury Xs or especially the overclocked 980Ti SLI setup.

Powertune +50%. Good water cooling like on the R295X2 can handle 500W dissipation.

Stock config should definitely aim for power efficiency. Basically 2x Nano on a board, water cooled.

Did you recall Nano's review? Reviewers who just up the power limit in CCC got 10-15% better performance.

 

[boozzer]

AMD's problem is marketing. The 290X was competitive.

290x reference for 1 month without 3rd party ones was the stupidest move ever made.

 

[96Firebird]

You know what I thought when I found out about the 390/X prices? I thought "No way they are gonna be able to sell Hawaii at that prices, they are effing nuts!"... I was wrong.

But they aren't selling at MSRP, they got sale prices almost immediately. AMD knew they couldn't sell the 390X at $429, nor the 390 at $329. The 390 dropped below $300 pretty quickly, and 5 months later it has deals weekly that hover around $250.

Note I am talking about the US market, it may be different down under.

 

[Silverforce11]

But they aren't selling at MSRP, they got sale prices almost immediately. AMD knew they couldn't sell the 390X at $429, nor the 390 at $329. The 390 dropped below $300 pretty quickly, and 5 months later it has deals weekly that hover around $250.

Note I am talking about the US market, it may be different down under.

Yeah but they were selling for the months before the sale prices. Down here, the MSRP is what it is, we don't get discounts, and the etailers down here report good sales for the 300 series, much better than custom R290/X despite being more expensive by ~25%! Which is nuts.

I dont know why it sold well for them. I recommended people to get a custom R290/X instead and pocket the difference. There's just no justification for paying $100 extra for 8GB vram vs 4GB when even CF/SLI 980Ti/Fury X can't take advantage of the vram due to running out of performance faster.

 

[3DVagabond]

Surely you jest if you think 4 vs 6GB will matter, even at 4K, no game under playable settings are limited by vram.

They are limited by GPU grunt, which CF and SLI of this caliber still cannot handle maxed out settings + 4x MSAA for vram to be an issue at 4K, which is a niche still in its infancy. At resolutions below that, not a chance.

The 2GB transition happened at a odd time when next-gen consoles are out with devs working on those specs, this is why we're seeing 2gb limitations, as they can't handle 4K textures. Now that we're well into this console generation, the ceiling is reached. I'm sure someone out there will release 8K texture mods but besides that, nope.

Re the perception, if its out, reviewers will be comparing Fury X2 vs SLI 980Ti.

With AMD working hard on driver memory management they are using less VRAM than nVidia too.

 

[3DVagabond]

Advertising a busted cooling loop? Probably not the best way to introduce speculation into an unreleased GPU.

It's not an advertisement.

 

[3DVagabond]

hahah :thumbsup: :wub:

WCCFTech speculates December 2015 launch, 1Ghz clocks for each Fiji chip, 4GB HBM1 per chip and 375W TDP. R9 295X2 had 500W TDP. If this card has 375W TDP limit, I don't see how it will be able to compete with two stand-alone Fury Xs or especially the overclocked 980Ti SLI setup.

Use the power to +50% and it won't throttle. Of course it will use more than 375W, but that won't change the typical gaming power, which is how I think most cards are rated these days. Most sites won't show that though. Also, when you get less than ideal scaling the cards won't have to throttle to stay inside the 375W rating. Rarely does Crossfire/SLI use 2x the power of the single cards.

 

[3DVagabond]

Yeah but they were selling for the months before the sale prices. Down here, the MSRP is what it is, we don't get discounts, and the etailers down here report good sales for the 300 series, much better than custom R290/X despite being more expensive by ~25%! Which is nuts.

I dont know why it sold well for them. I recommended people to get a custom R290/X instead and pocket the difference. There's just no justification for paying $100 extra for 8GB vram vs 4GB when even CF/SLI 980Ti/Fury X can't take advantage of the vram due to running out of performance faster.

Most customers are clueless. They are perceived as newer tech and they have 8GB of RAM. That's enough to jump start sales. They should have released these cards when the 980/970 came out. They likely would have lost little market share instead of losing half their market share.

 

[RussianSensation]

Powertune +50%. Good water cooling like on the R295X2 can handle 500W dissipation.

Right but I mean what's the reason they are rumored to use 375W TDP on the Fury X2 but had 500W on the R9 295X2? That tells me they are after efficiency balance, not max performance.

Stock config should definitely aim for power efficiency. Basically 2x Nano on a board, water cooled.

But what's more impressive, 375W Fury X CF performance on 1 PCB or 10% faster than Fury X CF on 1 PCB? I think in this market segment, efficiency is not a selling feature no matter how some people try to spin it.

Most after-market 980Ti cards are using 260-290W of power and that's without being max overclocked. I don't think 375W power usage will be some big selling feature for AMD. If it was an NV card, sure.

Did you recall Nano's review? Reviewers who just up the power limit in CCC got 10-15% better performance.

Ya but remember most people will look at stock performance, especially since not all sites will show max performance. Just recall R9 290/290X reference reviews. If the card could get 10-15% more performance with max PowerTune limits, AMD needs to have Uber mode like the old days. That way reviewers will have to test it under 2 conditions. Leaving this as a choice rather by manually changing the power limit etc. is more likely to have some reviewers exclude such a scenario.

Most customers are clueless. They are perceived as newer tech and they have 8GB of RAM. That's enough to jump start sales. They should have released these cards when the 980/970 came out. They likely would have lost little market share instead of losing half their market share.

They should have, same with R9 380/380X. For months, R9 280X/290/290X sold alongside R9 285/380/390/390X. Their excuse that they needed to clear inventory doesn't even fly. Since they priced R9 390 and R9 390X at higher prices than R9 290/290X, it was possible to have 4 cards occupying the $249-429 price levels. So what that they are close in performance? They could have prevented the massive damage 970 did to them since at least January 2015. Not to mention they could have released R9 285 4GB at the very least.

 

[Mondozei]

A few Years ago the entusiast-tier GPU was 500€ (GTX 680).

This again? The 980 launced at $549 MSRP. The GTX 680 was GK104, not GK110. It amazes me to this day how hard it is for people to come to understand that the proper way to measure things are architecture by architecture. So GK104 should be compared to GM204. The price went up by $50, not $250.

Big Kepler cost the same as Big Maxwell. The reason why we have small and big dies within the same generation is because Moore's law is slowing down. If we would still have node shrinks every 2 years, instead of being stuck on 28 nm for 4 years, then we wouldn't see this big/small die strategy from a company like Nvidia.

It has a lot less to do with AMD. And as others pointed out, the Hawaii cards were competive, but AMD did fumble their launch(reference coolers were atrocious, supply was limited at launch which can't be blamed on bitcoin mining, etc).

People who keep insisting that all x80 GPUs should cost $500 seem to live in fantasy land where Moore's law hasn't slowed down and where you could do a de facto tick-tock strategy for GPUs. Those days are long gone. The price for x80 GPUs at the same arch(x104) has gone up 10%, or $50, not more.

Anyway, that's a tangential rant.

 

[Silverforce11]

@RS
Efficiency is the current metric. AMD needs to show they can provide outstanding perf/w, Nano achieves that.

Fury X with 2 x Nano and 375W TDP will certainly achieve that. Since NV has no plans for Titan X2 that we know of, it will assuredly give them the performance crown as well as perf/w crown.

 

[RussianSensation]

This again? The 980 launced at $549 MSRP. The GTX 680 was GK104, not GK110. It amazes me to this day how hard it is for people to come to understand that the proper way to measure things are architecture by architecture. So GK104 should be compared to GM204. The price went up by $50, not $250.

Big Kepler cost the same as Big Maxwell.

Everyone already compares GK104 to GM204 but that's not the reason why your post is flat out wrong. How many times does this need to be repeated? It's not about Kepler -> Maxwell. It's about 20 years of GPU development and NV's history prior to Kepler. You are using 2 wrong generations to prove a point -- the two generations that are responsible for the change in NV's strategy and pricing. This is like saying if apples were genetically modified in 2012 and their prices increased (Kepler) and then oranges were genetically modified in 2014 (Maxwell) and prices increased even more, then all fruit for decades prior to 2012 must have also been genetically modified and cost a lot and the small price increases we are seeing are just inflation and accounting for the cost of new modified fruit to come in 2016, etc. No, you need to analyze a longer series of history to have a proper assessment of reality and dive deep into the earnings/gross margins of companies that sell fruit. If you do all of that, everything many of us have been telling you will become crystal clear.

So obviously you used two generations that incorporated the new NV strategy to prove a point that nothing has changed. You have to use generations prior to the new change of strategy. What generations are those? Don't have to go too far. Fermi is already a good start.

$199-239 GTX560/560Ti are spiritual predecessors to $399 GTX670/$499 GTX680
GTX560/560Ti are spiritual predecessors to $330 GTX970/ $550 980.

$349 GTX470/570 & $499 480/580 are spiritual successors to $650 GTX780/ $699 780Ti.

980Ti is not a fully unlocked flagship chip so that's the current generation's GTX570. The Titan X is the old gen's GTX580 3GB that cost $549.

What's the Fury X? Well it's simple. HD6970. HD6970 (Fury X) competed well against NV's 2nd best GTX570 (980Ti today), while it lost to the fully unlocked flagship GTX580 (Titan X).

So yes, absolutely, both AMD and NV raised prices. AMD raised prices to ATI's historical levels but NV went even further.

You realize that 6800GT cost $399 and X850 Pro cost $399? In the NV/ATi line-up, those cards were at least as fast as a GTX980 relative to today's 980Ti/Fury X. Guess how much GTX980 cost when it launched? $550.

What about even further? 9700Pro cost $399 - that's the flagship ATI card. What about 9800Pro and 5900U? $499. Those were not $649 Fury X or $699 GTX780Ti or $649 980Ti.

Like how many examples of the past do we have to show you to get the point?

Since this thread is about Fury X2, let's focus on AMD/ATI. I'll use ATI's/AMD's historical dual-chip flagship cards to prove it to you that prices skyrocketed and that your assertion that GPU prices have always been this high and it's normal is flat out wrong.

HD3870X2 = $449
HD4870X2 = $549
HD5970 = $599
HD6990 = $699
HD7990 = $999
R9 295X2 = $1499
Fury X2 = $449 like the HD3870X2? $599 like the HD5970? $699 like the HD6990 flagship? Sure thing! :whiste:

Dude, just admit it, GPU prices have increased 50-100%. If you are new to PC hardware in relative ters, that's OK. However, don't try to obfuscate the reality of what has happened in GPUs in the last 3-3.5 years. We've never had this type of pricing before.

Right now a 980Ti isn't at all like the 6800Ultra/UE, or the GTX285 or the GeForce 4 Ti 4600/4800 or the 5950U. It's not even a full chip, similar to a GTX570, a chip that cost $349.

People who keep insisting that all x80 GPUs should cost $500 seem to live in fantasy land where Moore's law hasn't slowed down and where you could do a de facto tick-tock strategy for GPUs. Those days are long gone. The price for x80 GPUs at the same arch(x104) has gone up 10%, or $50, not more.

Sure thing, like GTX970 for $330 was a fantasy 10 months after NV milked us with a $699 780Ti - on the same node? That's why NV/AMD are struggling selling $500 Fury, $570 Fury X, and $550-580 980Ti cards / sarcasm. Why don't you go and study the financials of NV from 2009 until today and come back with better facts. The only person living in fantasy that GPUs have to cost $700-1000 and mid-range cards were like GTX680/980 at $500-550 at launch of a new generation is YOU. You have bought into the AMD/NV marketing and the entire costs are skyrocketing so we must increase prices 50-100% to survive PR. That's exactly what they want you to believe so that for 2016 Pascal, NV can release 980's successor for $599-649 and you won't even blink cuz you'll defend it as the 'new' flagship as it will beat 980Ti by 20-30%.

No, you won't even remember how NV's next gen mid-range card whipped the last gen's flagship, for instance, how a $199 6600GT beat $499 9800XT / 5950U. Instead, you'll just accept it as the normal status quo that mid-range Pascal should cost $500-600 just because it's "better" value than the last gen's Fury X and 980Ti. Congrats, NV/AMD have won.

 

[AtenRa]

hahah :thumbsup: :wub:

WCCFTech speculates December 2015 launch, 1Ghz clocks for each Fiji chip, 4GB HBM1 per chip and 375W TDP. R9 295X2 had 500W TDP. If this card has 375W TDP limit, I don't see how it will be able to compete with two stand-alone Fury Xs or especially the overclocked 980Ti SLI setup.

I told ya, CF 2x Nano (175W TDP each) will be 40-50% faster than NVIDIAs $1000 Titan-X.

Im expecting Gemini at $1500 and it will be the undisputed champion (single card) in performance for the entire 2016 as well.

 

[Silverforce11]

Sorry there's no way they are gonna sell many at that price.

$999 or go home.