Printable View
Cayman XT , PCB
http://img825.imageshack.us/img825/4420/95215812.jpgQuote:
Within the sample or the sample card, core frequency of 900MHz, the final number of frequencies will be used is unclear, but I think I must be 800 ~ 900MHz or less, external power supply for the 6 +8 pin, PCB much like the HD5870
Note that with 4+1 -> 4 shader transition they gain almost 25 % increase in perf/mm˛, so they can get 25 % more performance with roughly the same transistor budget. The thing is, are they going to take the performance or shrink the chip.
My speculation.
* Antilles: 6990 ~300W = 2*Barts
* Cayman XT: 6970 ~270W+
* Cayman PRO: 6950 ~230W+
* Barts XT: 6870 ~200W+
* Barts PRO: 6850 ~170W+
* Juniper XT: 6770 ~140W+
* Juniper LE: 6750 ~110W+
* Turks: 66xx / 65xx ~80W+/~60W+
* Caicos: 63xx / ~50W+
The consumption is just "thrown in" for my point in naming scheme. If Cayman is big die with 250ish consumption, then there really cant be Antilles = 2*Cayman and in that case also Antilles can be only "little" faster than Cayman. So naming like this would make lot of sense.
Oh, and yes in this scenario it would make sense for Cayman to have 384bit memorybus.
PS. for nVidiots, yes renaming 5770 is bull:banana::banana::banana::banana:, they should not rename it even if it takes that place in lineup.
that is a good question, and i think it greatly affects what we see happen with 28nmQuote:
Originally Posted by Calmatory
they have 40nm done well, the fact we see 1600SP chips for a year now is a good sign they have enough control to go even bigger. but this is the last of the 40nm vs the first of the 28nm. so if we see 400mm2 now, when its shrunk to 28nm will still be very close to 400mm2 (depending on how much is changed, but just thinking about 2x SP growth here)
considering they have a very wide array of SP counts across chips, and all very successful (except 5830), i think they will go bigger. they probably have enough confidence to build a big chip now, and still plan to build a big chip at 28nm when ready, even if the first gpu released may only be 6770 (like the 4770)
huh, looks like 256-bit after all.
some more info:
http://gpudesign.bafree.net/radeon-c...70-pcb-picture
Quote:
Core Clock: 800~900 MHz
Memory Chips: 8 on the front side
Memory bus width: 256-bit
Power Connectors: 8+6 Pin connectors
Power Controller: CHiL CHL82xx Digital PWM
Output ports: 2x DVI | 2x Mini Display Ports | 1x HDMI
i still dont understand why they still use dvi ports. just make the dvi ports hdmi or display port and throw in a hdmi/display port to dvi dongle
This card is starting to shape up. If the performance numbers are true, I think I will get this generation. However, it looks like antilles might not be a dual cayman part but a dual barts part to fit the power limit.
I can imagine sucky crossfire performance that makes dual cayman cards not really faster than dual gtx 480 part, especially with new drivers and new parts. AMD is usually terrible with getting new crossfire cards up to speed.
If i am not misstaken HDMI is limited and is not usable in 2560x1600 60Hz so i beg to dissagre and this is why DVI(-DL) is still needed as native unless ATI not want the cards to be used where it can make good use of its power.Quote:
Originally Posted by phantomferrari
At least i wanna be able to use highrez on dedicated monitor even of it is "only30" " rather than in the "lowrez" compared to 2560x1600 on a TV may so be if it is 50-60" but only 1920x1080 and looks rubbish compared to the higherrez monitor and still show less content than the smaller screen.
What about 120Hz LCD users? :p:Quote:
Originally Posted by phantomferrari
Well as you can see from the picture, it is 256bit, and your power consumption speculation is way off.Quote:
Originally Posted by Mechanical Man
Antilles is a dual Cayman part...Quote:
Originally Posted by tajoh111
Crossfire is vastly improved with the newer cards, whether that is related to performance/drivers/other software I'm not sure.
is the crossfire bottleneck because of the lack of bandwith for gpu to gpu communication via the dongle ???? or is it related more on the type of rendering typicly used by ati ????
well, hdmi is probably the worst connection interface for pc ever. it is inferior to dvi in almost every aspect, imo. there's only one reason why we have hdmi on our graphics cards if you ask me: TVs. however, a hdcp ready dvi interface with a dvi-to-hdmi adapter would work for that purpose as well...Quote:
Originally Posted by phantomferrari
don't know about displayport, but there are a few myths floating around - regarding higher latency than dvi etc. i've yet to find a reliable source that proves these claims though, but rumors like this mostly don't come up without a reason.
i'm very happy with the fact that dvi is still around - but this proves my point somehow.
The only person I've heard the high latency(aka input lag) rumour from is zanzabar. I tried googling. Absolutely nothing. zanzabar, if you're reading this, give us some info.Quote:
Originally Posted by RaZz!
Plus, all the eyefinity setups around the world already proven displayport has the same latency as dvi.
The vast majority of monitors use DVI connections. Very few people connect their desktop PC's to televisions or projectors. Not sure why you would want to force the majority of graphics card buyers to use a dongle unnecessarily.Quote:
Originally Posted by phantomferrari
I believe the higher latency has to do with the content protection hardware. It may only be an issue when HDCP compliant sources (movies) are used.Quote:
Originally Posted by blindbox
Also for anyone who wants a laugh, the RIAA/MPAA are attempting to have FM receivers put in all new Cell Phones so they can get a payout. http://arstechnica.com/tech-policy/n...o-wishlist.ars
Wrong thread :p:Quote:
Originally Posted by STEvil
I've run an eyefinity setup on some occasions. My central monitor is always using DP and my side monitors use DVI. Even when watching external camera footage of the three screens, they've never shown motion out of concert that I've seen. The DP connection itself never feels like there's latency, either, beyond that incurred by stuff like render ahead when using crossfire or vsync being on. In other words, there has never been latency that I've seen that is specific to only one connection interface.
It seems to me I researched the latency thing once and found it was true, surprisingly, but it was merely a matter of semantics due to how infinitesimal the difference was. There are many benefits to DP, however. Nobody is ever convinced of anything on a forum, so let's not get into the whole pro/anti DP thing. People love to argue with strangers about anything when they're anonymous.
Same screens ? Active DP->DVI ?Quote:
Originally Posted by Particle
maybe it's not for all screens that DP don't introduce latency ...
no, but thanks for noticing.Quote:
Originally Posted by blindbox
the issue is HDMI, not display port. For the record I use HDMI and havent had latency issues.. so... :shrug:Quote:
Originally Posted by Particle
The two on DVI are Dell 2007FP monitors and the DP one is a 2408FPW. I'll point out that if what you say is true, that alone establishes that DP doesn't have latency issues. Some monitors may, but that wouldn't be DP's fault. From this I'd say to a generic third party, "Put the blame where it really lays."Quote:
Originally Posted by madcho
I'm not sure what you're talking about with regards to replying to me. I'm not talking about HDMI at the moment as DP latency was what was being discussed. HDMI is a topic for a different conversation.Quote:
Originally Posted by STEvil
id like that too, but that would cost them more than having a native dvi i think...Quote:
Originally Posted by phantomferrari
I only recall latency issues being with HDMI, and zanzabar's previous posts on the subject were also about HDMI... but thats only if i'm remembering correctly I guess. :shrug:Quote:
Originally Posted by Particle
Humm dont like this one bit first off all 2x6 pin power then single Crossfire connector :(
I do hope that Barts Pro has a single 6 pin but i highly doubt it.