Next-gen AMD graphics to be codenamed 'Caribbean Islands'

[scaryjim]

... we use to have flagship single core products superseded on rougly a 12 month cycle?

4870 - June 2008
5870 - September 2009
6970 - December 2010
7970 - December 2011/January 2012
R9 290X - October 2013

So the R9 290X is the only real anomoly in that, with the other cards all being about 15 months apart (note the 7970 paper launched in December, retail parts weren't available until January). Then again, when you look at nvidia...:

GTX280 - June 2008
GTX480 - March 2010
GTX580 - November 2010*
GTX680 - March 2012
GTX780 - May 2013*
GTX 980 - September 2014

They've been on a > 18 month cycle largely - remembering that the GTX 580 was basically a respin of the GTX 480 - the design was almost identical, they just tweaked it to get better yields - and the GTX780 was basically a rebranded Titan, so they messed around a bit (a lot?) with the cards and chips throughout the 600/700 series. The difference could be explained by AMD largely being behind in the single GPU stakes for much of that time so accelerating their development cycle?

[donald4prez]

Well new nvidia 970 and 980, barely beats 290 and 290x and sometimes even looses

It's all about the power consumption though. Here's hoping that AMD's GCN Hawaii refresh in early 2015 brings them up to par with GTX 970 and 980 performance per watt.

All about power consumption!? Maybe if your on laptop or any other mobile device... but on a desktop it makes no sense to worry about power consumption. Even so the 290X is old lol The fact that it can keep up even is impressive and not to mention under $300! So yeah if you can afford paying $5 extra on your electric bill(as a baller/shot caller I can assure you I can) then power consumption isn't a big deal...

At the end of the day NO ONE picks out a DESKTOP GPU for power consumption. If your a true enthusiast you look at performance. The fact is the 290x can keep up(sometimes out perform) with the 980. Bottom line... not which card has better TDP.

My setup is Intel/Nvidia so I'm no fan boy..

[watercooled]

Yeah, Nvidia come out with new codenames and a lot of marketing for architecture increments, but things like the execution pipeline/functional blocks remain much the same just as they do with AMD for incremental improvements. I agree that a major way in which AMD falls behind is marketing.

GCN has seen layout/feature changes from 1.0>1.1>1.2 similar to Kepler>Maxwell. Parts like Hawaii (290/X), Bonaire (260/X) and Tonga (285) are entirely new silicon for the 200 series. Granted, a lot of the lineup uses existing 7000 series silicon, but this has been the norm for a long time. In Nvidia's 700 series lineup, everything bar the 750-series and 780/Titan-series was reused from the 600-series. And the GK110 die used in the 780/Titan was already in use for Tesla compute cards long before that.

[Plasmastorm]

Screw the power consumption. I don't spend >£300 on a GPU to worry about what it uses.
Get back to the old days of stonking great FPS and less worrying about the power bill imo.
If you are that worried your not the kind to buy one anyway.

[dfour]

This ^^^^

I think all this ho-ha about power consumption is blinding people to what really matters

Frames per second


Its just spin and marketing about saving what could amount to a couple of quid a year when the rest of your rig is chewing lots of watts and your 1500 watt psu is being used inefficiently and therefore wasting more power than your choice of low power GPU is saving you.


Do people think that basically AMD are waiting for the 20mn (or what ever die shrink) process to be available before debuting there next GPU?

Yes so it means they wont have anything new out for quite a while but its not as if they arent competitive on performance, I would guess we wont see any nvidia cards out until the die shrink process is mature enough for them as well to release the higher level cards.

[watercooled]

It's hilarious the amount of attention 20W here or there gets these days. Rewind just a few years and it was unheard of, at least on the desktop. You had/have people disabling the idle states on their CPU, Northbridges with their own heatspreaders (LGA 1366), and >100W idle power was pretty commonplace, and no-one bat an eyelid. Now power has gained some marketing value and it's suddenly the most important thing in the world.

And it's hardly like power prices have increased to the extent worrying about 20W under load (especially if you're paying more to gain that saving) suddenly becomes massively relevant. If power usage is such an issue, you have chosen a bad hobby TBH.

[george1979]

Really hoping AMD pull another 4870 out of their hat next year.
While I'd agree with some points regarding power consumption in desktops these companies also have a responsibility to consider the bigger picture - that 20W saving may not be much in one system but multiply that out to the hundreds of thousands sold and it has wider environmental benefits. Not to mention the improved cooling (also quieter), longevity and higher clockspeeds that can be wrung out of chips with lower power consumption (which also has obvious benefits for mobile systems as well).

Although for the high end, low volume parts, yeah go crazy - its always nice to read the reviews of those cards even though I'd likely never buy one

[loccothan]

Pirate Islands will be in 20nm/16 for sure cuz GlobalF. have ready chips in 20/16 process already ;-)

[watercooled]

Really hoping AMD pull another 4870 out of their hat next year.
While I'd agree with some points regarding power consumption in desktops these companies also have a responsibility to consider the bigger picture - that 20W saving may not be much in one system but multiply that out to the hundreds of thousands sold and it has wider environmental benefits. Not to mention the improved cooling (also quieter), longevity and higher clockspeeds that can be wrung out of chips with lower power consumption (which also has obvious benefits for mobile systems as well).

Power consumption is not the only factor in temperature (which in itself is largely irrelevant unless it leads to throttling), noise, clock speed or longevity. A GPU with a power draw of 30 watts can easily do badly on all of those things if it's not installed in a suitably designed platform.

[wtfitskeagan]

Unnf. I want.

[DanceswithUnix]

If power usage is such an issue, you have chosen a bad hobby TBH.

lol, I like that

[crossy]

It's hilarious the amount of attention 20W here or there gets these days. Rewind just a few years and it was unheard of, at least on the desktop. You had/have people disabling the idle states on their CPU, Northbridges with their own heatspreaders (LGA 1366), and >100W idle power was pretty commonplace, and no-one bat an eyelid. Now power has gained some marketing value and it's suddenly the most important thing in the world.

And it's hardly like power prices have increased to the extent worrying about 20W under load (especially if you're paying more to gain that saving) suddenly becomes massively relevant. If power usage is such an issue, you have chosen a bad hobby TBH.

Agree to an extent - yes, the eco lobby has gained some mindshare, so you're encouraged to feel guilty if you've got a tower system burning hundreds of watts to display a Windows desktop. On the other hand 20W additional on the graphics, another 40W on the cpu, etc. This means heat. And that heat has to be dispersed, so you're talking about audible fans. I'm also going to suggest that the current fad for really small systems (mini-ITX and smaller) also means cooler running is a big issue.

On a personal note, I'd like cooler running - not for any financial aspect, but more to stifle the comments from SWMBO along the lines of "phew, it's a bit 'tropical' in here ... I guess you've been playing computer games again". Although that does mean I can turn the radiator down in that room - so maybe it all balances out?

My current XFX 7970 is a bit on the vocal side when pushed hard, the Gigabyte 460SOC that it replaced was noticeably quieter. So, yes, I'd like a new AMD graphics setup to run cooler, but only because this'd presumably mean quieter running.

[CAT-THE-FIFTH]

If AMD came out with a card which was 20% faster and consumed 10% less power than the GTX980,I bet it would be good effort AMD,but a little too late. If Nvidia then came out with a card 20% faster but consumed 40% more power,it would be AMD is doomed.

When the GTX580 was out - it was all about performance. When the GTX460 was out and was no faster than similar AMD cards and consumed more power - it was all about CUDA,tessellation and drivers.

When AMD had much better general purpose compute performance in many of their cards,the whole CUDA/compute thing was not important anymore.

When AMD started putting in extra features - its all half baked.

When Nvidia started putting in extra features -its the dogs bollocks.

When Nvidia was ahead of AMD in frametimes in a number of games -it was very important.

When AMD was ahead of Nvidia in frametimes in a number of games - it is not important,since Nvidia is good enough.

When AMD is cheaper and has games - they are fail.

When Nvidia is similar price or cheaper and has games - they are giving uber value.

When AMD tries to charge a similar price to Nvidia for a faster card - they are overpriced.

When Nvidia keeps its slower cards at the same price as a faster AMD card - its AMD fault or the quality is better.

Its nothing to do with being eco-minded,its perception and the same thing can be seen with Apple.

[DanceswithUnix]

On a personal note, I'd like cooler running - not for any financial aspect, but more to stifle the comments from SWMBO along the lines of "phew, it's a bit 'tropical' in here ... I guess you've been playing computer games again". Although that does mean I can turn the radiator down in that room - so maybe it all balances out?

I just had thermostatic radiator valves fitted to the bedrooms and computer room to get around that sort of thing. It makes the house much more comfortable with the heat better and automatically balanced.

[watercooled]

Sadly, what CAT says above seems to be the case a lot of the time. I don't remember people screaming about efficiency back when the AM2 platform was more power efficient than 775 i.e. chipsets/motherboards generally using less; AMD tended to use more advanced process nodes for their chipsets whereas Intel would use fairly old nodes, which probably had a bit to do with it. And the original Atom platforms were laughable for reasons like this - the CPU itself probably struggled to exceed a few watts, but the rest of the platform was drawing like 20W. But since around the SNB timeframe when Intel started paying attention to that, it's suddenly the most important thing in the world.

Same on graphics, 4870, 5870, 6970, 7970 IIRC were all more efficient than what Nvidia had on the market at the time. But it wasn't important then.

One area where people can generally save a few watts is display brightness, and you might be surprised by how much. Monitors often come factory set to an eye-burning 100%, but my Dell looks more than far more comfortable at 25%, though OFC it will depend on room brightness etc.

Look at this one for example, just the first one I came to:
http://www.tftcentral.co.uk/reviews/benq_gw2765ht.htm
100% out of the box, 44.6W.
Calibrated 18%, 21.5W - less than half.

[MustardCutter]

And that matters why? Increasing IGP performance negates the need for many lower-end discreet cards, which make up a large part of the market, especially in mobile. Whether it's part of the same die as the CPU, or connected through PCIe, it's still a GPU at the end of the day. As I said in another thread, you can set cut-off points in various places to make the figures look better or worse for a given company. E.g. why not include the GPU in all three consoles for AMD? That would significantly change things...

Sorry, the point I was trying to (albeit poorly) make is that if you made it all about discreet cards then the original statement being commented on of NVIDIA pulling away may ring true. Discrete GPUs seem to be the only items you can compare directly between the AMD and NVIDIA to give an idea of who's best and bringing the proportion of intel integrated graphics (that people who care about graphics are unlikely to be using) sold into the equation isn't much use to who's doing better in the NVIDIA v AMD stakes.

Edit: just want to point out that I'm not a fan boy of either side; I run a system using an AMD APU but for the gaming system I currently use intel/nvidia and have had all possible permutations of intel/amd/nvidia for cpu and gpu combinations over the past decade.

And yes the whole AMD v NVIDIA thing is usually a silly argument skewed in NVIDIAs favour; as mentioned it's probably down to marketing and historical perceptions of the brand rather than reality.