Monday, September 27th 2010
AMD Radeon HD 6700 Series ''Barts'' Specs Sheet Surfaces
Here is the slide we've been waiting for, the specs sheet of AMD's next-generation Radeon HD 6700 series GPUs, based on a new, radically redesigned core, codenamed "Barts". The XT variant denotes Radeon HD 6770, and Pro denotes HD 6750. AMD claims that the HD 6700 series will pack "Twice the Horsepower", over previous generation HD 5700 series. Compared to the "Juniper" die that went into making the Radeon HD 5700 series, Barts features twice the memory bandwidth thanks to its 256-bit wide high-speed memory interface, key components such as the SIMD arrays split into two blocks (like on Cypress), and we're now getting to learn that it uses a more efficient 4-D stream processor design. There are 1280 stream processors available to the HD 6770 (Barts XT), and 1120 stream processors to the HD 6750 (Barts Pro). Both SKUs use the full 256-bit memory bus width.
The most interesting specification here is the shader compute power. Barts XT churns out 2.3 TFLOP/s with 1280 stream processors, GPU clocked at 900 MHz, while the Radeon HD 5870 manages 2.72 TFLOP/s with 1600 stream processors, 850 MHz. So indeed the redesigned SIMD core is working its magic. Z/Stencil performance also shot up more than 100% over the Radeon HD 5700 series. Both the HD 6770 and HD 6750 will be equipped with 5 GT/s memory chips, at least on the reference-design cards, which are technically capable of running at 1250 MHz (5 GHz effective), though are clocked at 1050 MHz (4.20 GHz effective) on HD 6770, and 1000 MHz (4 GHz effective) on HD 6750. Although these design changes will inevitably result in a larger die compared to Juniper, it could still be smaller than Cypress, and hence, more energy-efficient.
Source:
PCinLife
The most interesting specification here is the shader compute power. Barts XT churns out 2.3 TFLOP/s with 1280 stream processors, GPU clocked at 900 MHz, while the Radeon HD 5870 manages 2.72 TFLOP/s with 1600 stream processors, 850 MHz. So indeed the redesigned SIMD core is working its magic. Z/Stencil performance also shot up more than 100% over the Radeon HD 5700 series. Both the HD 6770 and HD 6750 will be equipped with 5 GT/s memory chips, at least on the reference-design cards, which are technically capable of running at 1250 MHz (5 GHz effective), though are clocked at 1050 MHz (4.20 GHz effective) on HD 6770, and 1000 MHz (4 GHz effective) on HD 6750. Although these design changes will inevitably result in a larger die compared to Juniper, it could still be smaller than Cypress, and hence, more energy-efficient.
245 Comments on AMD Radeon HD 6700 Series ''Barts'' Specs Sheet Surfaces
:roll:
:nutkick:
But a tiny crappy heatsink ontop :laugh:
They have gone from a 4 simple + 1 complex arrangement to a 4 medium complexity arrangement. So there's no way to know atm how much die area the new 4D ALU will take compared to the old 5D ALU.
www.semiaccurate.com/2010/09/06/what-amds-northern-islands/
What I'm impressed is that the chip is meant to be a mid-range card with the same number of processors as a 5870 and 256bit mem bandwidth. First thing come to my mind is that the chip could be at least as big as the 5870. But then if ATI wants to sell these stuff cheaply and still make good profit, they need to shrink the size while still using 40nm manufacturing.
So I reckon... Their new Northern island is even smaller than the previous generation of processors. Now thats is something mighty impressive. Semi accurate says something like 80% of the previous gen processor size. Possibly a 334*90% = 301mm^2 ??? still much bigger than the 170mm of juniper though.
AMD's general direction has been to make smaller chip that maintains same performance which result in better energy efficiency and reduced production cost. Bulldozer is a good example where they make 2 CPU into 1.15x size of 1 CPU with minimal compromise in speed.
Overall, I'm really excited about this... and I may be thinking of retiring my loyal 4890 if the 6xxx cards are worthy.
Bottom line is that it's not as easy as saying double the SPs == double die size, or double that part of the chip, whatever. Like I said, GF104 has much more than 66% of the working units in GF100 put together in 66% of transistors, my point being that SPs themselves, don't take a lot of space and hence a comparatively small 768 SP Nvidia chip is feasible. Figure it out how many they can add through parallelism* until they are close or even mid-way to Ati's number of SPs.
*Ati has parallelism in the SPs 5D, 4D. Nvidia is adding parallelism with superscalar SIMDs, but it works the same way, it adds more throughoutput without adding a lot of transistors, at the expense of some inhefficiency. That's been Ati's architecture for 5 yeasrs already and they are continuing with it, except they are going with 4D now because it's been found over and over again that their average ALU utilization was around 3.6-4.5 all the time.
too much nv fanboy in here that want to derail the thread.
please stay on topic
back to the topic i hope bartpro wont be expensive, and i hope this will push developer to push eye candy a lot further,
And i agree, i hope it's cheap.:)
really good discussion going on here IMO.
To be honest i'm just hoping that the 6870 does not cost too much, not enough power in those 6770's :D (yes i know i don't have a clue how well they perform :p)
One thing that is annoying me more though is the lack of spec on cayman so far as that is the chip that interests me the most, what makes it worse is for several generations the top end chip has been basicly double the mid range chip and all these barts specs make me wonder if it will be the same this time or something thats cut down to keep power usage in check, i geuss only time will tell.
so maybe the cayman just have 20 % increase in SP, we can't predict it because its too many possibilities and combination, i hope after the bart we can have some info leaked after all cayman was planing to be released on november its just a month a way from bart
But i think no matter what ati is doing for caymen it will be a nice bump in speed/power and if i'm lucky allow acceptable framerates using eyefinity on a single 6870 (i hope), now to just hope someone hurrys up and starts leaking some specs.
shit why PC games can be ended like this, full of consollities with crap graphic and poorly written games and on top of that a draconian DRM :cry:
That would be cool to have PCIE-E slot filled with Japer and Xenos chips on PCB. There would be no reason to port console games. We would be playing them on PC.
But very true at the same time.
that'd be like chevy announcing the new corvette ZR1 and instead it's the 425hp v8 camero. fast sure, but not faster than the 638 hp corvette the market was expecting.
barts is the 6700 series and anyone who says differently is in charlie's pocket.exactly all that says is that there is a document entitled "HD 6800 Series Lauch Guidelines" in which more details on the nda are listed. it has no bearing on product names, specs, or anything else. It just points you to another document.
I for one am looking forward to the amd 6870 cayman chip :D