Silent_Buddha
Legend
AMD: R9xx Speculation
- Thread starter Lukfi
- Start date
Thats just the rear side right? Like they did with the early shots of their Juniper chips, they flipped the wafer so you couldn't see any important structures?
Most likely
Interesting to see that they've already got one of these out. Obviously there will be a lag time based on how long it takes for GF to ramp up production, but that seems a lot more promising in terms of things being out by the end of this year
The foreground wafer looks like an x86. Maybe Magny Cours?
I don't know what the center one has, but it looks like it has L3 sections and would probably be an x86.
I can't see enough detail to place the third wafer's contents as a CPU or not.
I don't know what the center one has, but it looks like it has L3 sections and would probably be an x86.
I can't see enough detail to place the third wafer's contents as a CPU or not.
Silent_Buddha
Legend
Aye, I may have to reconsider my skepticism that NI would be on GF 28 nm this year.
Regards,
SB
Are they encapsulated in plastic? Or how do they preserve the spectacular shininess wooom, preventing them to oxidize or get fat-fingers? :^) I assume they do not have that much trash-wavers that they can put a new one on every show. Just curious.
Silent_Buddha
Legend
I always figured it was standard practice to keep one wafer to "commemorate" the first groundbreaking wafer run(s)?
Regards,
SB
Regards,
SB
itsmydamnation
Veteran
So is this 28nm GPU a rv870 shrink or a "new" GPU? its going to be really interesting to see how well GFs 28nm process does, if there process has a clock speed advantage and this 28nm GPU shows up in the nxt 6 months, then the pain trains comming!
Well basically half as big assuming perfect scaling (330/40^2*28^2). Some things don't scale perfectly so it would be a bit bigger (plus that might be too small to fit all the i/o required for a 256bit gddr5 memory interface, though maybe it would work if it's rectangular enough). Also, there might be differences between tsmc and gf which affect the size.
You can get an idea how things generally scale with rv770->rv740. The latter was very slightly more than half the size, but it also had some units removed (in fact it had ~14% less transistors).
PSU-failure
Newcomer
If my math is not too broken, it would be ~200mm2.
Reasons...
- Juniper is 171mm2
- Redwood is 104mm2
- both have the same 128-bit MC
- Redwood is half of Juniper (ROP, TMU, SP)
Removing the MC from the die size, 5 SIMD and the associated bits use 67mm2, so 20 use about 270mm2.
Assuming scaling is nearly perfect for logic but almost non-existent for IO, that's 270*0.5+(334-270) = 199mm2.
Also, this doesn't account for a possible suppression of the 40nm double-vias workaround neither does it account for architectural changes.
Last edited by a moderator:
I haven't seen them shrink and rename an architecture recently, so why would they start with Cypress unless it really is quite a departure from the norm and they expect to be really slow at getting the mainstream derivatives out the door.
Unless of course they keep the 58xx range around as previous generation parts whilst also offering NI parts at the same time?
I was just thinking if they do hit 28nm in the first half of the year they will just replace cypress and juniper parts with 28nm cypress. 28nm cypress will be much smaller , can have more shaders and will cost alot less for them to produce. They can simply refresh most of the line up.
28nm cypress can fill the $150-$600 parts. 28nm Juniper can move down to the sub $150 parts. Ati will make alot of profit and these new parts on the newer process will most likely be faster than what nvida will have.
Later in the year around Oct they can release their next gen parts. The 28nm cypress and jniper would simply be them working on the process and preparing it for the new parts. Like they did with the rv770.
Bouncing Zabaglione Bros.
Legend
But if you have to do a redesign for a full process drop, why not add/improve stuff at the same time? Going 40nm -> 28nm was always going to be more than just a shrink. I doubt AMD will stand still in the face of continued innovation from Nvidia and Intel. AMD know you can't get away by just shrinking the same old design and hope that no one else comes up with something new.
Similar threads
