Gemini (FuryX2) looms near

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
I dont see any technical reasons here for proof that this slide is bs.

Of course you dont. Its AMD.

To get 8GB, the cards would have to work on completely different scenes. Something like a PIP mode for example.

Nobody can even show a practical example where you get say 4.5GB out of dual 4GB.

That is not what you said, you said DUAL 14nm, not in a 14nm GPU period.

Yes, the context is dual GPU. Why dual 28nm when you can get dual 14nm. Obviously one of them exclude the other.
 

Mondozei

Golden Member
Jul 7, 2013
1,043
41
86
So with Polaris' first new GPUs being low-power/budget - and possibly even being aimed at laptops as Ryan hinted - and with no real visibility beyond that, AMD's decision to launch Gemini in 2016 maybe wasn't so terrible after all.
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
So with Polaris' first new GPUs being low-power/budget - and possibly even being aimed at laptops as Ryan hinted - and with no real visibility beyond that, AMD's decision to launch Gemini in 2016 maybe wasn't so terrible after all.

It does depend on the performance of what they roll out. They said they had two chips for Polaris and if one of the two chips is that estimated 120mm2 chiplet, they are going to ride Fiji and company well into 2H 2016. Woof.

If the estimate of that chip running at 0.87v @ 850 clocks is true, I'm guessing there is more performance on the table waiting to be unleashed. Still wouldn't rule out a two card/GPU option of Polaris smacking Fury X2 offside the head.
 
Feb 19, 2009
10,457
10
76
We may not get 600mm2 monsters on the new node for awhile, but consider that a mid-range 300mm2 14nm ff chip with new uarch and HBM2 will smack down current high end stuff at much lower power usage. That already obsoletes current stuff.

Fury X2 is still DOA. They may have a niche in selling it for "VR Ready" rigs. But for gaming consumers it does not make much sense.
 

maddie

Diamond Member
Jul 18, 2010
5,116
5,462
136
We may not get 600mm2 monsters on the new node for awhile, but consider that a mid-range 300mm2 14nm ff chip with new uarch and HBM2 will smack down current high end stuff at much lower power usage. That already obsoletes current stuff.

Fury X2 is still DOA. They may have a niche in selling it for "VR Ready" rigs. But for gaming consumers it does not make much sense.
Trying to get a feel for some terms.

What does smack down mean in %?
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
We may not get 600mm2 monsters on the new node for awhile, but consider that a mid-range 300mm2 14nm ff chip with new uarch and HBM2 will smack down current high end stuff at much lower power usage. That already obsoletes current stuff.

Fury X2 is still DOA. They may have a niche in selling it for "VR Ready" rigs. But for gaming consumers it does not make much sense.

Keep in mind that people need and buy new stuff every day. It's not like video card sales stop ~6mos before the next gen drops. If, for example, the best option to run VR is FuryX2, then you will see the early adopter VR users buying them. Would it have been better to have launched it sooner? Sure. Better late than never though.
 
Feb 19, 2009
10,457
10
76
Trying to get a feel for some terms.

What does smack down mean in %?

I fully expect Pascal and Polaris mid-range chips of ~300mm2 to be 20% or more faster than Titan X & Fury X, obviously with much less power usage.

With the the node jump alone, a ~300mm2 chip should be able to field near the transistor count on the 600mm2 28nm chips currently. Add uarch gains, HBM2, it could end up much faster.
 

maddie

Diamond Member
Jul 18, 2010
5,116
5,462
136
Depends. If nVidia is ahead of AMD it's 7%-12%. If AMD is ahead then it's 40% minimum. /sarc.
Good one

I fully expect Pascal and Polaris mid-range chips of ~300mm2 to be 20% or more faster than Titan X & Fury X, obviously with much less power usage.

With the the node jump alone, a ~300mm2 chip should be able to field near the transistor count on the 600mm2 28nm chips currently. Add uarch gains, HBM2, it could end up much faster.
That performance estimate is around what I believe will happen early next gen.

Still, until a double star arrives, Gemini should rule, especially for VR.
 
Last edited:

moonbogg

Lifer
Jan 8, 2011
10,731
3,440
136
I don't think 4GB is adequate for a card like this, but this card does have one thing going for it: Its not a Geforce product. We should all know that doing math to predict future GPU performance does not work for Nvidia products because its hard to estimate just how bad the driver neglect for Maxwell is sure to end up. As soon as Pascal hits the scene, you can fully expect Maxwell to become a secondary piece of trash that Nvidia starts to neglect from day 1 of Pascal release and possibly even shortly before.
AMD's performance is more predictable and offers longer term value, but this particular product is lacking in Vram for its combined GPU power and I find it odd that people disagree with that point, especially since games today break the 4gb barrier and people will use the card for a year or more in many cases, at which point Vram limitations only get worse with 4GB.
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
I'm with Silver on this one. Especially now if rumors are true of a close to March 2016 release for Little Polaris. What they demo'd was a capped product using extremely low volts and clocks to perf/watt it's competitor.

Now imagine that chip with 8GB GDDR5(X) going for $300-350 (what I assume will be the new "mid-range" price bracket), suddenly buying two are probably within (if not equal) with Fury X2 performance, possibly less priced, less power, less heat, updated uarch and more VRAM.

It be GTX 980 Ti Vs Fury X redux, except this time AMD is cannibalizing it's self. They should have just launched Fury X2 the moment it was ready.
 

MrTeal

Diamond Member
Dec 7, 2003
3,910
2,669
136
I'm with Silver on this one. Especially now if rumors are true of a close to March 2016 release for Little Polaris. What they demo'd was a capped product using extremely low volts and clocks to perf/watt it's competitor.

Now imagine that chip with 8GB GDDR5(X) going for $300-350 (what I assume will be the new "mid-range" price bracket), suddenly buying two are probably within (if not equal) with Fury X2 performance, possibly less priced, less power, less heat, updated uarch and more VRAM.

It be GTX 980 Ti Vs Fury X redux, except this time AMD is cannibalizing it's self. They should have just launched Fury X2 the moment it was ready.

I don't think there's a snowball's chance in hell of two of these GPUs touching Fury X2 performance. What they've shown so far is an encouraging start, but you're still talking about a ~100mm^2 14nm die vs a 600mm^2 28nm die. Unless they can clock that thing up past 2GHz, they're not going to make up the lack of shaders vs Fiji.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
If only 100mm2 and 200mm2 Polaris are to be released this year then Fiji X2 will be fine.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Actually, a lot here were expecting its going to be a >400W monster, requiring water cooling just because its so power hungry and air cooling wouldn't be possible...

But if it had brought the smackdown, that would not have mattered so much...
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Now imagine that chip with 8GB GDDR5(X) going for $300-350 (what I assume will be the new "mid-range" price bracket), suddenly buying two are probably within (if not equal) with Fury X2 performance, possibly less priced, less power, less heat, updated uarch and more VRAM.

That means you think AMD will bring down Fury X level of performance from $649 to $299-349 in 12 months? Those days seem to be long gone. Welcome to the milking way of selling graphics cards => 25-35% performance increases trickled down slowly while milking each card as a flagship at $550-650.

January 4, 2016

"AMD’s Joe Macri stated, during our talks, that they expect this FinFET technology will bring a 50-60% power reduction at the same performance level OR a 25-30% performance increase at the same power."

Unless Mr. Joe Marci is completely wrong, it seems AMD the only way AMD is going to give us 75-100% increase in performance over Fury X is with major architectural advancements/core re-balancing and a breakthrough in memory bandwidth: HBM1 -> HBM2. What are the chances AMD is going to deliver on such a performance advancement by nailing all of these fronts, delivering HBM2 and a 550-600mm2 14/16nm FinFET die in 2016? imo, close to 0%.

What are the chances the competitor will do the same for 2016? imo, again close to 0%. It's really simple: why release a card 75-100% faster than Fury X/980Ti right away and wipe out huge profit margins across all the latter $100-500 chips when you can increase each level 25-35% (maybe a bit more) and then repeat in 2017?

> This way both companies offer more tangible performance in 2017
> This way both companies maximize profits
> This way both companies don't have to deal with terrible yields on immature 14nm/16nm node wrt to 550-600mm2 die
> This way both companies don't need to panic about HBM2 yields or supply in 2016

More importantly, gamers bought $500-550 680/7970 and again $550-700 290X/780Ti. That means both AMD and NV got the message loud and clear ==> gamers will upgrade 2 or even 3 times in the same generation and/or some don't care if the "flagship" is 300-350mm2 or 550-600mm2. MOAR performance => MOAR sales.

If you are NV/AMD, if just saw this type of consumer behaviour, the last thing you'd want is to launch a card 75-100% faster than your last flagship for $650 vs. $550-600 25-35% faster in 2016, then $550-700 25-35% faster in 2017. Milk that wallet baby.

If NV and AMD are only gonna release small chips this year, I'm not upgrading! lol

The way both companies set up a GPU generation now makes it almost impossible to "win." On the front end of a generation, you'll get milked for mid-range die at flagship prices, on the back end of a generation, you'll get milked for flagship die at flagship prices but then it's already almost half-way into the generation. I offer my 6 ways to upgrade with the new system:

1) Buy as close as possible to the launch of new next gen cards, then resell as close as possible to the launch of next cards to minimize loss of resale value. Rinse and repeat. This strategy entails frequent upgrading but allows maintaining top performance year over year. AMD/NV would love this! The huge risk of this strategy is mistiming optimal time to resell. For instance, if you bought a $700 780Ti and didn't sell it on time, your resale value bombed.

2) Buy 2 mid-range $350-400 cards and SLI/CF them. This would be akin to buying 290 CF/970 SLI as close as possible to launch date. Throughout the entire generation, you'd get that gen's flagship performance but with a caveat that some games won't work for months as far as SLI/CF goes.

3) Find GPUs with a severely tarnished reputation and pounce on them at the right time. Perfect examples of this would be something like $299 GTX480 weeks after $499 580 came out, or $280 R9 290X when $550 980 was the bees knees. For example, if in 2016 cards come out 20-25% faster than the Fury X for $650 and use less power, then suddenly Fury X sells for $280 since "no one wants it". That's an opportunity for a cheap upgrade. It's hard to predict which card of any gen will end up to be in this spot, but chances are a flagship AMD card is likely to be the most likely candidate for this position closer to the end of a generation of that series. This is because generally speaking flagship AMD cards aren't very popular as far as market share goes, which means last gen / close to EOL flagship AMD cards would be even less popular. We've actually seen this scenario play out with HD7970/7970Ghz/R9 290X and now Fury X will join them soon.

4) Focus on software, not hardware. That means stop worrying about hardware % increases and instead look if game X, Y, Z is actually (1) worthwhile to upgrade for because it's a great game you are likely to spend lots and lots of hours playing; (2) the performance advantage is beneficial as in going from 25-30 fps to 55-60 fps vs. 60-70 fps to 120-140 fps.

5) Go used. In some countries like Canada/Australia/Russia, etc., where the local currency got destroyed against the USD, the best values will be in the used market, while new cards will be more expensive than ever!

6) Get your gf/son/daughter into gaming, so this way you pass on your GPU upgrade down to this person as a gift; and suddenly your new GPU upgrade means your old/current GPU is their new FREE upgrade.

If only 100mm2 and 200mm2 Polaris are to be released this year then Fiji X2 will be fine.

Even if 350mm2 Polaris comes out for $650 and is 25-30% faster than Fury X, Fiji X2 can still sell for VR. If you think about the context of VR, a card 25-30% faster than Fury X is not a direct competitor to 2xFury Xs in a VR environment. As long as VR allows both Fury X GPUs to work seamlessly and independently of each other on a per eye basis, the Polaris card would have no chance of coming close to that level of performance. That means you could technically have a $600-650 Polaris 300-350mm2 and a $999 Fury X2 selling to completely different markets without directly competing with each other.

Also, AMD could offer some discount combo bundle with a VR headset of choice and Fiji X2 but such a deal wouldn't be available for a Polaris card. That would only cement the distinction between a gaming GPU vs. a VR GPU.
 
Last edited:

maddie

Diamond Member
Jul 18, 2010
5,116
5,462
136
Unless Mr. Joe Marci is completely wrong, it seems AMD the only way AMD is going to give us 75-100% increase in performance over Fury X is with major architectural advancements/core re-balancing and a breakthrough in memory bandwidth: HBM1 -> HBM2. What are the chances AMD is going to deliver on such a performance advancement by nailing all of these fronts, delivering HBM2 and a 550-600mm2 14/16nm FinFET die in 2016? imo, close to 0%.

What are the chances the competitor will do the same for 2016? imo, again close to 0%. It's really simple: why release a card 75-100% faster than Fury X/980Ti right away and wipe out huge profit margins across all the latter $100-500 chips when you can increase each level 25-35% (maybe a bit more) and then repeat in 2017?

> This way both companies offer more tangible performance in 2017
> This way both companies maximize profits
> This way both companies don't have to deal with terrible yields on immature 14nm/16nm node wrt to 550-600mm2 die
> This way both companies don't need to panic about HBM2 yields or supply in 2016

More importantly, gamers bought $500-550 680/7970 and again $550-700 290X/780Ti. That means both AMD and NV got the message loud and clear ==> gamers will upgrade 2 or even 3 times in the same generation and/or some don't care if the "flagship" is 300-350mm2 or 550-600mm2. MOAR performance => MOAR sales.

If you are NV/AMD, if just saw this type of consumer behaviour, the last thing you'd want is to launch a card 75-100% faster than your last flagship for $650 vs. $550-600 25-35% faster in 2016, then $550-700 25-35% faster in 2017. Milk that wallet baby.
If both companies had close to comparable market share I could see this scenario playing, but as I stated in the Polaris thread.

"IF AMD wants to regain market share, they can't dance with Nvidia. They have to be clear in front. Raja must know this."

If AMD plays Nvidia's game now, they remain stuck at 20% GPU sales.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
If both companies had close to comparable market share I could see this scenario playing, but as I stated in the Polaris thread.

"IF AMD wants to regain market share, they can't dance with Nvidia. They have to be clear in front. Raja must know this."

If AMD plays Nvidia's game now, they remain stuck at 20% GPU sales.

Profits matter more than market share long-term. It's better to be profitable with 18% market share than to have 50% market share by essentially giving away your product.

Also, as I already stated, showing up first is more important for AMD. Let's say there are 1000 PC gamers, of which 300 are 100% upgrading their GPU in 2016. Of these, 40% are loyal to team green, 50% brand agnostic, 10% loyal to team red. If AMD launches late, they will lose 90% of 300 customers. If you launch early, you can go for 50% of brand agnostic. Rory Read is the one who knew this about the GPU market. Objective/brand agnostic PC gamers won't wait 6-9 months for AMD's response and OEMs won't either. That's why for AMD, winning back market share has to be prioritizing launch timing, more than absolute max performance.
 

96Firebird

Diamond Member
Nov 8, 2010
5,737
334
126
I disagree. Plenty of people, a lot even on these forums, waited for Fury X to launch, even though the 980 Ti had already been released. I know this goes against your "buyers are sheep" narrative though. As far as the 6-9 month waiting period, who knows if people will wait that long. I don't see there being that large of a gap between releases...
 

MrTeal

Diamond Member
Dec 7, 2003
3,910
2,669
136
I disagree. Plenty of people, a lot even on these forums, waited for Fury X to launch, even though the 980 Ti had already been released. I know this goes against your "buyers are sheep" narrative though. As far as the 6-9 month waiting period, who knows if people will wait that long. I don't see there being that large of a gap between releases...

That's why you don't tell people it will be a 6-9 month wait. Look at the GTX480, it launched 6 months after the 5870. nVidia wasn't going around saying in Sept of 2009 "Just hang on 6 months, Fermi will be awesome." They just left it ambiguous, and it's a lot easier to get people to wait just another month six times than it is to have them hold off a half year. If AMD or nVidia is significantly delayed, I don't expect we'll know exactly how long that will be until the late card is just about ready to ship.
 

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
I disagree. Plenty of people, a lot even on these forums, waited for Fury X to launch, even though the 980 Ti had already been released. I know this goes against your "buyers are sheep" narrative though. As far as the 6-9 month waiting period, who knows if people will wait that long. I don't see there being that large of a gap between releases...

The ones that waited are the AMD loyalists, and objective types who were not in any hurry to make a choice. Everyone else just grabbed a 980ti.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |