Before we even have details about HD6000M...
For the first time in the world: the first details about AMD's 28nm GPU! - Donanimhaber.com
For the first time in the world: 28nm, AMD Mobility Radeon HD 7000 series GPUs will be used - Donanimhaber.com
Interesting Bits:
Wimbledon - Ultra High-End; 2GB GDDR5; 256-bit memory bus; L3 Cahce; 65W TDP; MXM 3.0; 25% faster than Blackcomb; production in Q2 2012.
Heathrow - High-End; 1.5GB to 2GB GDDR5; 128-bit or 198-bit memory bus; 33W-45W TDP; MXM 3.0; 30% faster than Chelsea; production Q4 2011 (available probably CES 2012).
Chelsea - Mainstream; 1GB GDDR5 or 2GB GDDR3; 128-bit memory bus; 20-30W TDP; MXM 3.0; 30% faster than Whislter; ES June 2011; production Q4 2011 (available probably CES 2012).
Thames - Entry-Level; 1GB to 2GB GDDR3 or GDDR5; 128-bit memory bus; 15-20W TDP; MXM 3.0; twice as fast as Seymour; production Q4 2011 (available probably CES 2012).
-
Nice, I'll replace this g73jh in 2012.
-
And I care about 2012 why?
-
25% faster seems very disappointing. I would think that at least 40% would be necessary to have the fastest card. And I was hoping for a late 2011 release (not that it matters to much to me, since I won't be buying until summer 2012). -
Time for me to start speculating:
Even with the die shrink, I'm not too sure if those TDP numbers look high enough to include up to 2GB of GDDR5. If they're GPU only then total TDP for Wimbledon could be a good bit over 75W.
Gives me new hope for Blackcomb being over 75W too.
-
Heck I can't even figure out what's coming in Q1 2011.
But thanks for info. -
Looks like I'll be skipping the HD 6000 series for my desktop, assuming the desktop Radeons will also be switching to 28nm.
-
The key word is "MXM 3.0"
-
So looks like Ivory Bridge times..... well im happy i bought right now, its good enough to reach Q2 2012.
-
-
-
Along with AMD Trinity APU. :yes: -
They should actually bring whistler and the other new material that has already been announced before throwing some numbers at us about stuff that is so far ahead. For the moment all I've seen about 6xxxM is some crappy rebrands with a few new functions enabled.
-
Crazy. 5000 series just seems so antiquated all of a sudden.
-
gee thanks for the news... your making me feel bad about my 5870M... Hopefully it lasts 4 years...otherwise i'm going to be playing games on medium..
Panther214 -
Hi, thanks for posting it, but production Q2 2012 should be Q4 2011. They want to sell notebooks with these mobile parts into Q2 2012. Quite a difference in the IT world.
I wonder how much shaders they will put into these babbies. 5800M wasn't really a big deal (4800M already had 800 shaders)
I hope they can double it again, like they did from 3800M to 4800M -
3870m to 4870m was more then doubeled 320 to 800
-
-
That's strange, normally they introduce their fastests parts first (?)
-
Not always. Mobility HD3000 series followed the same pattern.
Probably has a lot to do with brand new architecture, on a brand new process, using brand new materials(HKMG). AMD might be rolling out the smaller chips on one of the lower process HKMG and then have to wait until the higher process HKMG are ready later.
That's just a guess though. -
Tsunade_Hime such bacon. wow
Hope TSMC fully tests 28nm before mainstream use.
-
If AMD uses TSMC. There's GloFo 28nm too.
-
-
Last few years? lol. It's been like this for a long, long time. Heck, when nVidia launched the 8800gt on 65nm, TSMC also had supply issues. More likely? TSMC is just reported more, since the products comming out of it's fabs have a short product cycle and heavy demand that doesn't sustain itself in the long run of each process node. That being said, MS sourced the xbox360's fabs to 4 different companies to help ensure it didn't run into heavy issues with supply (since it was launching a mere 3 weeks or so after the final IC revision can out of the fabs).
-
AMD will probably use both TSMC and GloFo not one or the other.
TSMC will likely still handle the next gen Bobcat APU families, Krishna and Whichita...while GloFo handles Bulldozer APU like Trinity. GPU arch would have to be designed differently for each fabs' 28nm process but they could still be split between GloFo and TSMC. e.g. GloFo gets Heathrow, TSMC gets Wimbledon. -
dont expect 28nm gpus to pop up this year at least for laptop gpus its still a year away
-
-
swheeett hd7000 28nm woohoo!
-
Awesome, why not go with a foundry that has a history of problems and completely skipped a manufacturing process node instead...oh well...
-
-
-
I don't know how I missed this thread, but the fact that next gen is still MXM 3.0 is a big YAY for me. Hopefully the sager NP8150 will work with it.
-
of course it will if its MXM3.0B... otherwise epic fail..
Panther214 -
Hmm, I am pretty sure that the first 28nm GPUs will start shipping in November/December 2011 for both AMD and Nvidia.
EDIT: I just saw this thread and bookmarked.
-
35 watt +? Not 95 watt +, lol. Could mean anything.
-
Most interesting part of the slide is that making up the HD7000m are 4 new GPU code names, in 5 cut-up configurations, instead of just the 3 new GPU (+/- rebrands) separated by G/DDR type, that we've seen the last few generations.
AMD looks like they will continue rebrands of the HD6000M below the value line, then start the new arch with Thames at 2xSeymour perf...which might come out of 192 or 256 4-way stream processors.
Thames @192SP could put Chelsea @384SP, Heathrow @768SP, and Wimbledon at anywhere from 832SP to 1152SP.
-or-
Thames @256SP could mean Chelsea @512SP, Wimbledon @1024, and Heathrow somewhere between 576SP and 960SP (768SP again sounds believable).
-
I am more concerned about the reports on the amount of GDDR memory. Really, the amount of memory these cards are equipped with should be cut in half, it's just a waste of money and TDP.
2GB for mid-range cards... ridiculous.
Thames @192SP could put Chelsea @384SP, Heathrow @768SP, and Wimbledon at anywhere from 832SP to 1152SP.
Sounds plausible and they will almost surely be 4D architecture, but until we see some leaked specs I think it's hard to have a realistic estimate. -
-
Maybe I took the wrong example, but for instance for the upper-mid range they said the cards will have between 1.5 and 3 GB of vRAM. From my perspective even 1.5 GB of vRAM for such cards is an overkill.
-
It's basically an equivalent of putting 1GB of GDDR3 VRAM onto a mid-range mobile gpu with a 128bit bus.
Useless and a waste of money.
Instead, they could have used those resources to increase the bandwidth on the overall gpu's to the point where they will be able to use the halved VRAM effectively. -
Yeah, I never understood that either. Wasting good money on vRAM that isn't necessary instead of optimizing for performance. Never made sense to me. Kinda like building a six car garage when the most you'll ever be able to afford is two cars.
-
I think it makes some marketing sense. A lot of ppl. that ask me for advice on which lappy to buy always point out the size of the video memory. Generally is something like this: Me: "So what video card does it have?" He/She: "I don't know, but it has 512MB of memory, so I think it must be really good."
I think it would be good if we gathered some info on the amount of vRAM used by logging a few games with GPUz and show how much vRAM they've managed to eat (texture settings and resolution should be included). This is a myth we need to demolish! -
Marketing sense yes (in terms of conning people into thinking that a low or mid-level gpu that has massive VRAM is a 'beast' - heck, even manufacturers tout this falseness in their promotional videos, making it look as if large amounts of VRAM is important), but PRACTICAL sense?
Not in the slightest.
Waste of resources that could have been used better.
If you won't increase the bus of the gpu then in that case, increase the bandwidth by replacing the outdated GDDR3 with GDDR5 so we can have a performance boost.
To the manufacturer, this would effectively cost the same as if they put larger amount of GDDR3 onto the gpu, but NO, we have to suffer through their stupidity.
It's bad enough they don't provide us with modularity in laptops and more diverse selections in hardware at a cheaper price tags.
I hope people wake up and force these idiots into finally moving on. -
I like this. It looks like AMD is FINALLY putting something in that huge gap between the midrange and enthusiast cards they keep leaving open.
-
Megacharge Custom User Title
Can't wait for 28nm. Gonna be some nice AMD chips!
-
So running L4D2 at maximum settings: that is everything on high, AA 4x and AS 8x, resolution 1920x1200 has managed to eat a maximum of 780 MB of vRAM.
Unfortunately I don't have any other game right to see how much vRAM they require. -
*insert panic here*
Next people will start panicking in December as they think that is the end of the world too.
The ancient South Americans just had a longer calendar...
Now as for the card... I will believe it when I see it.
I mean, by now, we should be traveling in space and living on mars right?
-
Looks like ill be heading for Wimbledon early 2012 then.
I think ill buy the new ATI card as well.
AMD's 28nm Mobilty HD7000 Series - Coming 2012
Discussion in 'Gaming (Software and Graphics Cards)' started by Phinagle, Dec 30, 2010.