Soldato
- Joined
- 2 Jan 2012
- Posts
- 12,410
- Location
- UK.
Exactly !!
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD to Skip 20 nm, Jump Straight to 14 nm with "Arctic Islands" GPU Family
- Thread starter tommybhoy
- Start date
Exactly !!
It's shrinking because the 290 and 290x are old and not in the same gen as Titan X and the 9x0 cards. It has absolutely nothing to do with the fact that they don't have that one single card that will compete with Titan X.
Many companies get by just fine selling stuff without being competitive.
At the budget end AMD are still extremely competitive and their APUs are pretty much unmatched.
Put it this way. If AMD go down I reckon Nvidia will follow. IE - if no one is buying AMD then no one will be buying Nvidia.
Times are bad for every one right now. We've still not recovered from the recession and that means items like the Titan X are even more ridiculous than ever.
I pointed out elsewhere that so far around 56 members on OCUK have posted saying they have bought a Titan X. Given that the Titan X is a braggart's dream (IE if you buy one you're not going to be quiet about it) it could be that OCUK have not sold many more than just that 56.
OCUK are one of the bigger companies who do this, so let's assume a lesser competitor has sold 40.
Do you really think that's going to keep Nvidia in business?
The bread and butter comes from lower down, man. Without a bread and butter product you're screwed, not the other way around.
Right now until Nvidia make back every penny that they invested into Titan X each one sold is being sold at a loss. It would only be a profit if they made their money back very quickly and they won't. They'll do that with sales of their lower end models, given the ridiculous pricing of the Titan X.
So as I said I would stop worrying solely about AMD and get on with worrying about the bigger picture. What we do, IE - building these computers and then posting on here is a shrinking market.
If it shrinks enough then it won't just be AMD going down the swanny, trust me on that. Nvidia would be soon to follow.
You want to talk stagnant and so on? how about one company making one product and rarely bothering to update it so you have to go years on that one product?
Kinda like in the mid to late 90s when Intel came up with the Pentium and in the space of about three years only really sold it clocked about 30mhz higher than the first model, yet still charging a very high price for it.
If AMD go then there will be nothing to look forward to, and that means there will be no sales. If the Titan X did not, for example, beat the 290x as much as it does it would be nowhere near as desirable. Companies can not compete with themselves.
/rant over.
Nice to see you making assumptions again Andy, I never suggested AMD just need a Titan X competitor to survive. Everyone knows the market share has little to do with the top end cards.
Nice rant though.
I didn't suggest AMD's lack of market share was because of innovation, nor did I suggest the Titan X was anything groundbreaking, although it is nearly twice as good as AMD's current single GPU flagship whilst using around the same amount of power on the same die size.
The high end cards make little difference to market share and as you've correctly said it's because nVidia release more cards over a shorter space of time and basically create their own low/mid/high end tiers whilst AMD's only response seems to be keep lowering prices and thus make very little off their current cards anymore.
I think you will find that Gibbo has sold a lot more than 56 and most of the people who buy them are not forum members.
Soldato
- Joined
- 20 Apr 2004
- Posts
- 4,410
- Location
- Rugby
GDDR3/GDDR5/HBM - How is using a JEDEC standard (both AMD and Nvidia are members) "innovation". Its just a new standards which everyone will end up using and have a hand in in its development. Sure AMD might of been the first to uses such thing but hardly a "innovation". It be like saying who ever has a 16/14nm chip out first "innovated" 14/16nm chips, its just where the nature of chips are going.
As for Eyefinity, Matrox Parhelia brought multi screen gaming to the mainstreem some 10 years early, it was just a shame the card it self was underpowered next to the ti4600 of the time and so got little traction
Caporegime
- Joined
- 8 Jul 2003
- Posts
- 30,063
- Location
- In a house
Don't think anyone gives a about the Artic Islands, 3s are nowhere to be seen yet. 
ATI brought tesselation (trueform) long before it became a DX standard or mainstream. They also helped develop GDDR3 with JEDEC and were first to use GDDR4 and GDDR5. The point is that AMD/ATI as well as Nvidia do innovate and to suggest otherwise is disingenuous.
Soldato
- Joined
- 17 Jun 2012
- Posts
- 5,951
Are we not just within the realms of diminshing returns? It's easy to make big leaps when the tech is quite poor, but I expect big gains are really difficult and expensive to make now.
This happens with everything. I suspect the next true paradigm shift will be a completely new technology.
You suggested that Nvidia were being held back by AMDs lack of competition and Nvidia couldn't release trully "innovative" stuff because of monopoly laws. The implication being Nvidia don't even need to try. I countered that with the fact that the lack of die shrink for years has held back both vendors, AMD more so. Nvidia would be nowhere near releasing Pascal, even if AMD were more competitive and had the fastest single GPU.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
At no stage has there been massive performance bumps every 6 months, revisionist history, nothing more or less.
The average length between gpu's back from a 9700 pro through to 5870 is somewhere around 18 months.
Too many people looking at all the past releases like all were huge performance boosters, they weren't.
9800pro, x850, x1900/1950, 3870, 6970, none of these offered a huge performance boost.
9700pro, x800, x1800, 2900xt, 4870, 5870, 7970, 290x all offered huge bumps in performance, the smallest amongst them probably the 290x because it was within the same process.
The only reason the 3870 then the 4870 were so close together was because the 2900xt was so late, it was also on an 'old' process node of 80nm(but originally designed for 65nm, which didn't work great for gpu's timing wise more than performance afaik). The 3870 was a relatively simple shrink with the 4870 being the main 55nm chip.
We have a similar situation in more ways than one here. 390x being meant for 20nm but the process not being good enough so pushed back towards the previous process.
Yes, some generations were shorter, some were longer. This was really about process availability, delays in previous gen, waiting on a new technology to be available. 390x is likely waiting more on HBM production than GPU production IMHO. That is how technology goes. If you're told HBM is ready for July 2014... back in mid 2012, and you start making a product to use it, then when it comes around HBM is a year late... then you get caught out.
It's the same thing as always though, AMD are taking the normal amount of time between generations.... people want to jump down their throats and make up crap bout 6 months between generations for years before and pretend like Nvidia are on some 6 month schedule when they aren't. Same as this "why aren't AMD talking" crap people have been banging on about, making out like Nvidia or AMD have in the past always talked about products long before release... which is again simply not true.
The average length between gpu's back from a 9700 pro through to 5870 is somewhere around 18 months.
Too many people looking at all the past releases like all were huge performance boosters, they weren't.
9800pro, x850, x1900/1950, 3870, 6970, none of these offered a huge performance boost.
9700pro, x800, x1800, 2900xt, 4870, 5870, 7970, 290x all offered huge bumps in performance, the smallest amongst them probably the 290x because it was within the same process.
The only reason the 3870 then the 4870 were so close together was because the 2900xt was so late, it was also on an 'old' process node of 80nm(but originally designed for 65nm, which didn't work great for gpu's timing wise more than performance afaik). The 3870 was a relatively simple shrink with the 4870 being the main 55nm chip.
We have a similar situation in more ways than one here. 390x being meant for 20nm but the process not being good enough so pushed back towards the previous process.
Yes, some generations were shorter, some were longer. This was really about process availability, delays in previous gen, waiting on a new technology to be available. 390x is likely waiting more on HBM production than GPU production IMHO. That is how technology goes. If you're told HBM is ready for July 2014... back in mid 2012, and you start making a product to use it, then when it comes around HBM is a year late... then you get caught out.
It's the same thing as always though, AMD are taking the normal amount of time between generations.... people want to jump down their throats and make up crap bout 6 months between generations for years before and pretend like Nvidia are on some 6 month schedule when they aren't. Same as this "why aren't AMD talking" crap people have been banging on about, making out like Nvidia or AMD have in the past always talked about products long before release... which is again simply not true.
Caporegime
- Joined
- 8 Jul 2003
- Posts
- 30,063
- Location
- In a house
Stagnation of lung, that was it, thanks
(has nothing to do with this thread, its something else i was trying to think of
)780 - 780Ti ~6 months.
780Ti - 980 ~9 months
So 6-9 months between top end card releases. I hate the fact these cards were only marginally faster then the previous card but it is still keeping the customers appetite whetted.
290 - 390. No confirmed release date yet, though assuming June it will be ~19 months. Oct/Nov 2013 until June 2015. In the meantime while waiting on 390 AMDs top end is slowly but surely slipping drastically behind. It is now ~50% slower than a Titan X.
Sorry for cutting your post but this is the only part I take issue with. I personally didn't say that massive performance jumps came every 6 months. I do remember such jumps every die shrink and they were coming practically every year or so. I remarked that Nvidia seem to be releasing cards every 6 months or so with massive profit and marginal performance increases. While I disagree with this, it does work and keep enthusiasts coming back for more. Discreet market share numbers show how successful this approach by Nvidia has worked.
Last edited:
Soldato
- Joined
- 2 Jan 2012
- Posts
- 12,410
- Location
- UK.
No, lack of node change has held everything back. Architectural improvements can only do so much (Maxwell) for example. A die shrink would equal less power use / heat output, you could have more performance in the same TDP or the same performance at much less power use. It also reduces overall cost longer term, rather than becoming more expensive.
We've been on the same node for way to long. Once we get off of 28nm we should see real progression in performance per watt, and some monster GPU's, along with a price shake-up.
Last edited:
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
It doesn't really because AMD also made these small speed bump products over their history and yet market share hasn't changed hugely. It's only dropped a little further than normal during one of these dips where Nvidia has new products and AMD doesn't. The general trend is as always, Nvidia/AMD don't often release within a few months of each other and there are ups and downs as each company releases something new.
Throughout the past decade, even when AMD has had significantly better value for money or performance for a period(7970, 5870), Nvidia had the better market share.
This doesn't reinforce the 6 month strategy as when both had cards out that often nothing was fundamentally different.
nvidia is a cash rich business, AMD has almost no marketing budget. Companies that make crap products with great marketing can be industry leading. Marketing wins you market share, that is the only proof I see there. I'm not referring to Nvidia in regards to crap products, I was actually thinking of Beats by Stupid.... awful product, massive marketing budget, there is absolutely no competition with who leads the headphone industry(in sales) and how they achieved it.
This is a world where marketing wins over product quality and has done for some time in the majority of areas.
Last edited:
Soldato
- Joined
- 2 Jan 2012
- Posts
- 12,410
- Location
- UK.
Don't think anyone gives a about the Artic Islands, 3s are nowhere to be seen yet.
Ha, very true.
It seems Artic Islands is to the 3XX series what Skylake is to Broadwell..
I.e we'll brush over the 3s because better stuff is coming
This is killing AMD's credibility in the market, you can keep saying "ooh look at this great new tech that we'll have available at some time in the future" but as long as all they are actually selling is old tech people are going to get fed up of waiting and actually move on to a company who although seem to be slower at announcing new tech does actually deliver.