Source: http://pc.watch.impress.co.jp/docs/c...04_404182.html
http://pc.watch.impress.co.jp/img/pc...404/182/02.jpg
http://pc.watch.impress.co.jp/img/pc...404/182/01.jpg
Printable View
Okay folks, get your gear ready & count those stream processor amount ! :D
nice work - looks like nearly half the die size gone for gpu. please someone qualified compare the size to the rumored 480 shaders and dont forget to factor in 32nm =)
here is a 40nm redwood 5670 die shot to get one started:
http://benchmarkreviews.com/images/r..._Die_wDime.jpg
now the architecture for redwood:
http://media.bestofmicro.com/7/6/236...%20Diagram.jpg
i think i counted 24 boxes of the same size in the GPU area.
anyone want to confirm?
yea, looks like 24.Quote:
Originally Posted by stangracin3
K10 core: 9.69mm^2
Whole chip will be >220mm^2?
For the ay they are arranged there's the possibility of a dual ultra-threaded dispatch processor like barts.
...an earlier slide:
for some days we will have here AMD Financial day 2010, think, we will smarter :)
Two things that are very important to remember is the enhancements mae and its nemeses "Sandy Bridge". As i have said several times before sandy bridge is much stronger in the cpu department but not as strong as the llano in the GPU department. Now there is a thing i left out the ratios, i mean ya the llano's GPU is stronger but by how much or the CPU part of SNB is faster but by how much?
http://pc.watch.impress.co.jp/img/pc...404/182/04.jpg
This was a part of discussion "Because we have no job's assigned thank you Intel" and the outcome was that a Intel GPU can reach the top not with huge number of shaders but with improved efficiency and speed. Intel does not want people to waste die space on stupid GPU's :( there were number of meetings held to decide if the current standard of GPU products were good enough to drive basic operations till Windows XP everything was a-ok but then Vista changed a lot of things it was concluded that for a entry level computer to run OS like Vista with bling they needed more GPU power at first they wanted to include a chip on the mobo itself but it worked out to suck a lot of juice and so it was decided to go on chip "Went down easy with most people" and then on die "Veery hard to digest people did not like that die space was being used for a dam GPU but had to be done"
i think were coming to a point where cpu speed is fast enough, and its a transition to the gpu for anything that "really needed" a faster cpu. no point in waiting for scaling of the cpu, when if you code for scaling the gpu is more than ready (pending the code can be run on a gpu)Quote:
Originally Posted by ajaidev
so if we compare a slew of benchmarks between the 2 chips (of a same tdp and price if possible, like what 35W and 250$ gets you) then i think the perf difference will sway ALOT depending on the benchmark, and its really a matter of which ones the user needs it for.
then im curious if in 3 years or so when its time to upgrade that same laptop, which one feels slower then, the one with a stronger cpu, or stronger gpu? i think that part is really most important considering people are keeping their computers longer than ever, which one will have the longer life.