It's just good marketing anyways. If they stuck with the 2xx series naming, it would just give that 'old product' feeling, which in the tech business is probably very unwanted.
AMD Radeon Fury X Series | HBM, Small Form Factor And Water Cooling | June 16th
- Thread starter wachie
- Start date
Unknown Soldier
Member
Yeah, AMD is trying to copy Nvidia by creating a halo-class naming with the Fury in the same way that Nvidia names their halo card Titan. Their attempt to create a non-numbered tier in Fury causes problems in the numbered line because the 300 series are forced to be same-tier rebrands. The normal strategy of moving rebranded GPUs down a tier can no longer happen if you create a new tier on top.
CassidyIzABeast
Member
Did AMD forget that the only reason they still exist is through bang for buck pricing?
I hope the rumored performance numbers and/or prices are different because if they end up being like this my best option will be R9 290X that I can currently get for 309. Here the cheapest GTX 970 is 399.
I was looking to get a new card for Arkham Knight and MGSV but I might end up just playing them with my 280X and waiting until next year for bigger upgrade.
Oh well we will know for sure in two days.
I was looking to get a new card for Arkham Knight and MGSV but I might end up just playing them with my 280X and waiting until next year for bigger upgrade.
Oh well we will know for sure in two days.
Nikodemos
Member
Comparing TDP between the 'new' cards and the ones listed on Wikipedia, it seems the '300' series has somewhat lower TDP.
I'd keep the 280X if I were you. It's still a very capable card and spending >€300 now for only a "minor" bump in performance doesn't seem worth it. I bet that spending the same amount on whatever next gen card that falls in that category would end up quite a bit faster than a 290X/970. It's still up you though.
Coulomb_Barrier
Member
Lol come on man. You are exceptionally negative on AMD - your reasoning how the new naming scheme is bad is really stretching. IMO it makes perfect sense for them to have their top card differentiated completely in product name like how Nvidia markets theirs.
God forbid, I think it's one decision that may help them sell more of their flagship cards in the future, you know...
How is his post "exceptionally negative"? It just describes a scenario of what could well have happened. And I don't think it's a stretch at all -- if they were not introducing the Fury line then obviously they would be moving tiers down as usual, otherwise they wouldn't have any left for the cards now called Fury.
Whether or not it will help them sell more high-end cards doesn't really have anything to do with it.
Unknown Soldier
Member
What will help them sell more flagship cards in the future is the cards being good. Period. Creating a new name tier won't magically make them sell more cards. My explanation is quite rational and it is the simplest reasoning for why the 300 series are 200 series cards in a new box.
Best to ignore.
While we wait for reviews, here are a few rebrand reviews from the past to peruse
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_770/1.html
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_760/1.html
Brad Grenz
Member
His posting history on anything related to AMD is exceptionally negative, and often illogical.
You mean the 770, a rebrand of the 680?
Ah, sorry, I thought we were discussing arguments and scenarios, not people.
Coulomb_Barrier
Member
He seems to be exceptionally negative in general, not in just this one post.
And whether it helps them sell more high-end cards has nothing to do with it? Of course it has. The one and only reason they have changed the naming scheme is in an attempt to sell more GPUs. Whether that means they'll be moving tiers down or not is immaterial.
Coulomb_Barrier
Member
Yeah, I'm not defending the rebrand. It's really shitty and personally disappointing.
But I do think naming their top cards something other than 3--x is a good move.
But I do think naming their top cards something other than 3--x is a good move.
michaelius
Banned
GTX 770 was overclocked 680 moving down $100 in price
GTX 760 was new sku created from existing chip
At least they usually drop prices when they make rebrands.
Yup, like I said it will come down to prices.
ThoseDeafMutes
Member
It would be analogous to if Titan X was the only card released, and then they brought out a 1080 that was the 980, 1070 that was the 970, etc. Traditionally rebrands occupy a lower tier on the rung, e.g. 7970 -> 280X, the 3rd GPU tier.
The reason this is happening is pretty obvious, the new chip they do have is pricey and difficult to produce in large volume on account of HBM and new design, hence the lower tier Fury's rumoured to be delayed in general. The decision to make the lower tier Fiji cards (also rumoured) part of the Fury branding rather than the highest tier on the regular branding is just continuing the brand name inflation that Nvidia started with the introduction of the Titan. Pretty groanworthy, imo.
ThoseDeafMutes
Member
The 3XX series aren't even bad prices, although I would argue that the 8gb framebuffer is overkill for those GPUs and thus the 290x standard is going to be a better buy.
ThoseDeafMutes
Member
Wait for reviews. I've heard that there's a slight OC, but otherwise it's the 8GB 290/290X sku with a new name. But you won't have to wait long.
Unknown Soldier
Member
I was not aware that AMD was a person whose feelings I was hurting. I thought I was just offering criticism of things a corporation called AMD have done or are doing, I didn't know AMD was like someone's best friend.
Coulomb_Barrier
Member
No that would be a catastrophic disaster. They're not allowed to ever change their naming schemes, as that messes up their tiers and is an incon...
Joking I'm sorry Unknown Soldier 
RandomSeed
Member
I wonder if the price jump for the R9 290 on newegg is a temporary thing between sales, or if they are increasing the price because the 300 series is coming out soon. I was probably going to order one today (was $270 minus $30 rebate last night and I wanted to think it over), but pushing the price past $300 makes the GTX 970 a no brainer when I do upgrade. I'd rather stick with AMD, but price does matter.
fizzelopeguss
Banned
When nvidia was struggling with fermi we saw rebadge after rebadge of their tesla product. It sucks from a tech and competition standpoint.
With AMD rebranding this year and with fury being a low volume product... I'll be amazed if pascal releases in 2016 in a mainstream affordable form. They can milk the node for another year and shunt the line down a notch.
With AMD rebranding this year and with fury being a low volume product... I'll be amazed if pascal releases in 2016 in a mainstream affordable form. They can milk the node for another year and shunt the line down a notch.
I realy doubt this will happen.
or at least I hope.
even the 390x being the 290x with 8gb would be a joke.
opticalmace
Member
There already is a 290x with 8GB of vram, too. We'll see in a couple days.
They are just shifting the numbering. Remember how the 5870 was slightly faster than the 6870? That's because we got the 6950 and 6970. Their marketing team has to try something new since AMD sales are at an all time low. Hence we get Fury as the new top.
Some flipped a coined and chose a rumor to believe.
http://www.tweaktown.com/news/45835/amd-radeon-fury-rumored-limited-30-000-units-2015/index.html
http://wccftech.com/amd-fury-30k-units-2015-air-liquid-cooling/
Ah... those rumors. HBM is very inefficient tech if it has that low yields when fabricated and would make it very undesirable for AMD & NV both. I'm very surprised if there is any truth to that "30k cards manufactured during whole 2015" rumor.
Very interested to see and hear what AMD has on 16th.
Edit: HBM*
First Gen Technology usually has teething issues, so those assumptions aren't outside the realm of possibility.
So lets say AMD is able to manufacture only 30k Fiji based GPU's during whole 2015. Each card has 4 stacks of HBM (4x1GB). This means they can only fabricate 120k stacks over ~12 months meaning 10k / month. That would mean 2 500 cards / month if we use 12 months as time factor.
Even for new tech, HBM 1.0, that sounds very inefficient and waste of company resources. At global scale, in which they will want sell these cards, that isn't even drop in ocean of GPU sales.
IF we were to take rumors/leaks as gospel it would mean that 300 -series is rebadge with slightly better cooling solution and Fiji is rarer than unicorns in wonderland making that also basically dead at arrival. Thing like that would devastate AMD's market share and make any 2016 plays next to impossible.
imo
Edit: I don't mean that HBM has been waste for AMD to develop and it has whole industry excited, but to invest so heavily in its 1.0 to get it out on GPU in 2015 and with so little availability would be waste of resources.
Sir Abacus
Member
We saw a process node drop in the 9000 series which droves clocks up and gave us 20% more shader power on the top end along with massive gains in the mid range segment especially going from an 8600GT to 9600GT. We saw a massive increase in die size and another 50% more shader power on the 200 series. Keep in mind this entire transition happened over 18 months too.
If you're going to do a rebrand with the same architecture it should be an example to be followed. Find and replace 2 with 3 is not. Having an architecture for over 18 months and squeezing another 60MHz out of it is not.
Well those numbers are kind of extreme, but Fiji having volume difficulties (or at least lesser economies of scale) is quite possible.
12 months is extreme scale yeah and if we scale it down it will increase cards / month figure, but amount of acceptable HBM stacks for 2015 would still be same. 120k stacks of memory is basically nothing. If HBM 1.0 is giving so low yields I don't understand why AMD has invested so much R&D into getting it to market during 2015 in 28nm.
There's no "would be". It straight up is. Would you have preferred the 380X being the 290x?
It's all marketing either way
Likely hedging their bets. Without HBM they'd have what? Only thing I can conceivably think of is applying Tonga hardware improvements to the rest of the cards which would likely have been costly as well and not really have improved peak performance. In this instance they've taped out a new chip with all the improvements of Tonga (and perhaps more) and HBM at a mature node. Makes economic sense
A few thoughts on the pricing.
Based on this it’s starting to look like the rebrand most likely will have an MSRP of what the current 290 and 290X models sell at before rebate. These video cards whether they’re 290/x or 390/x are not attractive to me if I were purchasing a videocard. Personally, I believe even the 980 TI is not a good deal. These talks have been done to death. When NVidia launched the Geforce 2 Ultra, 8800 GTX etc.. they pushed pricing. NVidia figured it out for the 780, 780 TI, and 980 TI and succeeded where the others failed to hold their pricing. Release an even more expensive video, then release another one at the previous price people didn’t want to pay. “980 TI is such a good value!” Keep in mind ATI and Nvidia did the halo product branding years ago with the first Geforce to the TNT2 and Radeon to the Rage lines. The only thing new about Titan (and possibly Fury) are the even higher premiums.
Don’t get me wrong, I like reading about this new technology and I’m very curious to read about Fiji. However if I were looking to buy a video card, I'd be buying up a used 290/X or 970.
- On the subject of Best Buy pricing; Best Buy in general sells video cards at MSRP and keeps them there until they’re cleared out. (I say in general because they will price match Tiget Direct, Amazon, and Abes of Maine.) I have benefited from this just before the announcement of the 6800GT where the MSRP was raised from 300 to 400 USD. It was special online pre-order where you paid in full and received the card a week earlier. I purchased the 470 GTX at launch for MSRP of 350 USD when online retailers were selling them for 380 USD and up.
- On the subject of pricing leaks, MSI and Gigabyte’s pricing has been leaked. Their base models are 330 and 380 respectively while their models with their premium cooling are at a 50 to 80 USD premium. The XFX models follow the same pricing as far as custom premium cooling solutions go.
Based on this it’s starting to look like the rebrand most likely will have an MSRP of what the current 290 and 290X models sell at before rebate. These video cards whether they’re 290/x or 390/x are not attractive to me if I were purchasing a videocard. Personally, I believe even the 980 TI is not a good deal. These talks have been done to death. When NVidia launched the Geforce 2 Ultra, 8800 GTX etc.. they pushed pricing. NVidia figured it out for the 780, 780 TI, and 980 TI and succeeded where the others failed to hold their pricing. Release an even more expensive video, then release another one at the previous price people didn’t want to pay. “980 TI is such a good value!” Keep in mind ATI and Nvidia did the halo product branding years ago with the first Geforce to the TNT2 and Radeon to the Rage lines. The only thing new about Titan (and possibly Fury) are the even higher premiums.
Don’t get me wrong, I like reading about this new technology and I’m very curious to read about Fiji. However if I were looking to buy a video card, I'd be buying up a used 290/X or 970.
If we believe rumors about Fiji availability then it's basically same as not having it release at all during 2015. 30k cards (over whole 2015) is absolutely nothing and basically same as not having stock at all, which just increases AMD's bad rep and will have negative impact when moving into 2016. Is there any economical sense in "Well we released world's first GPU with HBM, glhf with finding any store on this planet that has one in sale after first 3 days for rest of the year" release? Backlog on orders would be mind blowingly big and people just would pick 980Ti instead.
In such low yield scenario it would make more sense, to me at least, do light refresh on 200 -series for rest of 2015 and do 14nm Fiji release in 2016 with higher yielding HBM 2.0. Basically what NV is aiming to do with Pascal.
If some early rumours are to be believed, they were banking on 20nm processes being available but that process node turned out to be a dud for anything by LP processes.
Now, the issues with HBM weren't specified, so it could be problems with the interposer or even the memory itself. Intel, notably, was supposedly having yield issues with the crystalwell edram due to its constraints.
One other thing to consider is that the costs and research AMD has sunk in bringing HBM to market now may yield benefits on their next product which will be using HBM2. It could well be that Nvidia will run into similar issues with their HBM2 integration.
Their 14nm releases will be the Arctic Islands series (4xx), which will be based on GCN 2.0 as opposed to Fiji which is in all likelyhood GCN 1.2.
Sir Abacus
Member
In my uneducated opinion and purely speculating, I think the biggest problem is that the 4GB HBM limit kind of paints AMD into a corner.
In a perfect world the new 390X would have been a 4096 SP GCN 1.2 GPU running 1GHz backed by 8GB of 512-bit 7GHz GDDR5 VRAM that could go toe to toe with a 980 Ti at $600 or less. More FLOPS, more memory bandwidth, more VRAM than the Ti for less money. But then the Fury X would probably end up being only slighly faster or even slower at higher resolutions/detail levels than the 390X. How do you show off your new HBM tech when most consumers aren't even saturating current memory bandwidth at typical 1080p/1440p resolutions but still want more frame buffer for ever higher resolution textures? You can't make a bigger die with more SPs because with 4096 SPs you're already pushing the absolute limit of how big TSMC can make a die.
It's a massive quagmire. Where the hell are AMD supposed to go?
In a perfect world the new 390X would have been a 4096 SP GCN 1.2 GPU running 1GHz backed by 8GB of 512-bit 7GHz GDDR5 VRAM that could go toe to toe with a 980 Ti at $600 or less. More FLOPS, more memory bandwidth, more VRAM than the Ti for less money. But then the Fury X would probably end up being only slighly faster or even slower at higher resolutions/detail levels than the 390X. How do you show off your new HBM tech when most consumers aren't even saturating current memory bandwidth at typical 1080p/1440p resolutions but still want more frame buffer for ever higher resolution textures? You can't make a bigger die with more SPs because with 4096 SPs you're already pushing the absolute limit of how big TSMC can make a die.
It's a massive quagmire. Where the hell are AMD supposed to go?
The problem with that is that AMD generally doesn't like doing big dies. Their GDDR5 memory controller takes up a lot of die space (especially being 512 bit), this does not leave them with a lot of room for 4096 SPs, unless they allowed the die size to balloon.
AMD has invested a lot R&D into HBM for sure, but I just don't see how they have anything to gain from HBM 1.0 GPU if yields are pathetic meaning HW just won't be available after initial few days. I just don't see what meaningful they would have to gain from such GPU release, it would do more harm to their image than good.
That whole upcoming Arctic Islands put Fiji into even more awkward spot, so close to 14nm Islands and if limited supply...
Little off-topic, but apparently NV has successfully locked down Pascal designs meaning HBM 2.0 integration didn't delay platform; http://www.kitguru.net/components/g...ly-taped-out-on-track-for-2016-launch-rumour/
No matter what people think about AMD's current GPU's they and Hynix did great job by creating HBM for industry.
We don't know how HBM will perform in gaming environment and how many GB's of GDDR5 it may resemble. If leaked dias for Fury are right then FuryX with 4GB of HBM can do 6k gaming which would mean excellent 4k experience. With just 4GB of HBM. That also could mean that memory, because of its bandwidth, can handle more GDDR5 heavy games.
It will be interesting to see in benches how HBM behaves in gaming, especially if doing test where its amount is exceeded.
Edit:
Isn't Fiji largest die they have ever done?
The 6K stuff is pure marketing. 6k what? A static image? Current 4GB cards can do 4k gaming just fine, but only in games that don't require as much vram.
AFAIK, yes. They probably felt that they had no choice this time around.