That guy Gibbo wins troll of the year awards if its complete opposite of what he's saying.
Printable View
That guy Gibbo wins troll of the year awards if its complete opposite of what he's saying.
Funny too since he's the admin at OCUK so he has a lot at stake to troll his boardQuote:
Originally Posted by kadozer
Would sign up at that forum just to tip my hat to him.Quote:
Originally Posted by zerazax
Read that again, he said on par with the 58xx series, NOT 5xx series... He's comparing the launches between the 69xx and the 58xx from AMD, not performance. :up:Quote:
Originally Posted by jvc
Well I am sure AMD would love for the release to be more like the release of 58xx than 5xx.Quote:
Originally Posted by DilTech
If the 6970 does have less shader units then thats very disappointing. And why are they still using a 256 bit memory bus? I'm not impressed at all.
Any reviews mention you can play 3D BluRays with a 6xxx? (with PDVD, Arcsoft etc.) I'm all set up and running, just waiting for the 6970.
Quote:
Originally Posted by Chrono Detector
what do you know about gpu or cpu architecture again to make those comments without knowing the rest of the picture ????
dittoQuote:
Originally Posted by Sn0wm@n
256bit*5700Mhz GDDR5 = 182.4GB/sQuote:
Originally Posted by Chrono Detector
384bit*3800Mhz slow GDDR5 = 182.4GB/s
With GDDR5 there's no reason to increase bus width unless you have an inefficient memory controller that can't handle high speed GDDR5 such as nvidia. Just with that a chip cost more-take more space in die --> bigger die- for the same performance.
I see you are new to AMD arq improvements. See Barts HD6800 for example. Cayman brings even more efficiency than Barts.
PaganII
You already can with any HD6000.
You forgot one minor tid-bit... The slower GDDR5 generally has tighter timings. As such, with both at the same bandwidth over-all the wider-bus will generally end up the faster, albeit by what amount this effects things remains to be seen.Quote:
Originally Posted by Nintendork
Sadly, there's no way do to do an apples to apples comparison (same gpu with higher bus, slower speed mems and the other with lower bus, higher speed mems).
The most close one but also innacurate, 4890 with mem underclock vs HD5770.
There's also the question of more complex PCBs to handle the extra traces for a bigger bus width. Also more memory chips to put on + QA. Granted, there's that tradeoff against higher speed GDDR5 which is more costly, but that requires actually knowing prices to know which is which. Nvidia did pretty much state the GDDR5 controller was far more complex than they thought it was, part of why Fermi had many delays
Depends on the architecture too. I remember these same arguments rehashed every gen since the 4800 was released - "oh it's 256-bit, it'll be eaten alive" etc.Quote:
Originally Posted by DilTech
AMD's high end cards simply haven't been memory bandwidth limited in the same way some of Nvidia's cards have
So the Chinese rumor from about 4 months ago about Cayman having less SPs than Cypress was right all along... Just like the Chinese rumor early on about RV770 having 800SPs... Good times.
As neliz explained, the switch is an advance overclocking switch for those with highend air/water or dice/ln2 so they don't have to do a hard mod to get a decent increase to voltage, though some no doubtly will still hardmod to get even more.
Saw something early today that makes me think someone with the card already has it under LN2, can't wait to hear the numbers.
Supposedly that French vendor cannot get direct supply from AIBs so he has to go a round-about way of getting inventory, i.e. not having them at launch and high prices. Like I mentioned at the end of Nov, most AIBs have already sent out first shipments to priority retailers and they should have received them last week. Volume is good. Depending on sales, there might be a small hiccup in Feburary due to Chinese New Year (and another reason).
Some seem to think Cayman will have dynamic voltage for the GDDR5, idle/2d/3d. I don't think it will but that would be nice.
Die size is supposedly 75% of GF110, based on GF110 being 530mm2. (Edited due to confirmation)
Someone else mentioned it being less than 20% larger than Cypress, meaning just under 400mm2. As I said before, the most specific number I have heard is 385mm2, which is what I am leaning towards.
TDP is similar to Cypress according to neliz.
Let say it cost a little less and is 10% performance under 580gtx at 150w less tdp.
Now, a 6990 under such tdp, going to totally rock.
single card performance, is though to do when the chips get big, small efficient and fast, the crossfire performance with the 6850 has been phenomenal.
there been more rumors, and I read Gibbo as a super great card for the price/performance ratio, which 90%+ people buy.
150W lower TDP than gtx 580!!! lets be reasonable here and not let the rationality go out the window. There's a reason why there's is a 8 pin connector and a 6 pin connector on the 6970. The 6870 still consumes 150 watts and the gtx 580 at the most consume 300 watts and most of the time a bit under 250. I can't imagine this card using 100watts to 150 when it has a 8 pin connector and a 6 pin connector. The 6950 has to consume more than 150 watts considering the 6870 consumes that.Quote:
Originally Posted by flopper
Forgot to add to the above post, Antilles doesn't use same bin as Pro.
The 8 pin can be for the clitoris thing.Quote:
Originally Posted by tajoh111
I think AMD definitely aimed too low, given that a 420mm^2 Cayman XT probably takes the crown from the 580 pretty easily. Problem with that line of thinking though is that we don't know how much yields/power consumption would change going that direction. After all, the last (and only) 400mm^2 card they tried was the R600, and that was an utter disaster. It's always easier to second guess
That said, it all comes down to performance, price, and availability again. The 6950 at < $350 might be a killer deal, much like the 5850 was
FWIW, I saw this from a different perspective as well, which myself being an engineer was pretty cool to think about:
3870 was 65% of 9800GTX
4870 was 75% of GTX 285
5870 was 85% of GTX 480
6970 is ?? % of GTX 580
If a < 400mm^2 < 200W card can perform at 90%+ of a GTX 580, that's a pretty impressive engineering feat and it nearly makes needing a dual GPU more of a e-peen "look what we can do" thing, rather than the necessity it was for the 3800's and 4800's
Of course this is irrelevant until we get the actual performance and actual #'s reviewed
150W statement being absurd aside, we actually don't know if it uses a 6+8 now... some people were saying it's 6+6 and since the card is coming in at under 200W, its possibleQuote:
Originally Posted by tajoh111
In fact, if you look at that leaked slide with the switch, it's blurry but it doesn't look like an 8pin
http://h-5.abload.de/img/switch_pdfom24.png
Please call it the voltage unlock switch... otherwise, as has been said before, nobody will be able to find it.Quote:
Originally Posted by Nintendork
Hmm its speculation but maybe with voltage unlock the card passes 300w ? It clocks like mad at those volts but AMD couldnt release it like that so they give the option to keep TDP under 300w ?
Cant wait seems it wont go faster than 580 ..... also maybe delay was to finalise clocks from 800-880mhz ? maybe at 800 it would have been too close to 570 even and AMD adjusted for that. 570 and 580 are pretty close anyways if this is going to be inbetween them its probably gonna cost ~420$ ?
250w not 300w.Quote:
Originally Posted by Zloyd
Can't change clocks that close to release, depending on where you are in the process it takes at least a few months to change clocks.Quote:
Originally Posted by Zloyd
I believe XT is still at the $449 MSRP, which is what I heard awhile ago.
Nah... they'll find it cause they've seen it on pictures, but they won't know how to operate it.Quote:
Originally Posted by Death^Dread
Actually 6870 does not consume 150W. the actual figures are lot lower than the rated TDP of that chip.Quote:
Originally Posted by tajoh111
TDP for 580gtx is 350w.Quote:
Originally Posted by tajoh111
cayman likely is around 200w.
its 150w difference.
300w difference for 6990, now, that is why amd went value and small mm/W design.
they beat nvidia with efficiency and offers better price/performance ratio and now for 3 straight designs they just hammer nvidia.
amd sells with high profit, if Nvidia didnt have their professional market they be gone like 3dfx.