[R300King!]

Originally Posted by flippin_waffles

Interesting: I just did a search for WolfLSI and this is what i've come up with.

http://wolflsi.deviantart.com/
http://www.geocities.com/monsterlsi/

There's more, but I think that demonstrates what i'm thinking.

I don't intend this to be evidence that these images are BS, just an observation, and may very well be a coincidence. It may not even be the same person. And even if it is the same person, it still may be irrelevant. I'm rooting for the red team btw, so i hope these are indeed real shot's of her backside!

It looks like the same person, I wouldn't be surprised at all. But that does not mean anything really. Only means he likes to draw. ...or he could be the photochop expert. LOL It's probably just one of his game characters.

[flippin_waffles]

Originally Posted by R300King!

It looks like the same person, I wouldn't be surprised at all. But that does not mean anything really. Only means he likes to draw. ...or he could be the photochop expert. LOL It's probably just one of his game characters.

Yeah, that's definitely possible. It certainly looks real!

[Razor1]

Originally Posted by R300King!

It looks like the same person, I wouldn't be surprised at all. But that does not mean anything really. Only means he likes to draw. ...or he could be the photochop expert. LOL It's probably just one of his game characters.

Yeah why would anyone go to that lenght of time to fake something, I already looked for errors in the photo, can't find any or anything obvious at least

[R300King!]

I got a pin count of 1985 on this one. Daamit this pin counting stuff is hard.

[R300King!]

Seriously though, I've compared the only two known R600 dies shots and only see one real difference in them, circled in red. Seems they are from a different layer/batch(what da ya call it?), possibly from a different revision of the chip altogether. Can't see the date on the 1st chip.

[flippin_waffles]

Originally Posted by R300King!

Seriously though, I've compared the only two known R600 dies shots and only see one real difference in them, circled in red. Seems they are from a different layer/batch(what da ya call it?), possibly from a different revision of the chip altogether. Can't see the date on the 1st chip.

True, there are a couple other minor differences as well. The spacing and position of most of those are different.
I'd also like to see a pic of the whole thing with the heatspreader on. At least i assume that is what is hanging out at the top of the pictures of her backside?

[Razor1]

actually something doesn't look right with that bottom chip, look at the lower right hand edge of the silicon, its glossy, yet there is a shadow being casted to right side of that very edge.

[Kaotik]

Originally Posted by Razor1

actually something doesn't look right with that bottom chip, look at the lower right hand edge of the silicon, its glossy, yet there is a shadow being casted to right side of that very edge.

It's no shadow, it's whatever thing it's called that's around the chip itself, "protecting" it or something, it's all around the silicon, makes the green a bit darker all around the silicon close to it.
The same thing can be seen in the upper chip, too, or are you suggesting the chip is casting that shaped shadow?

[Razor1]

Originally Posted by Kaotik

It's no shadow, it's whatever thing it's called that's around the chip itself, "protecting" it or something, it's all around the silicon, makes the green a bit darker all around the silicon close to it.
The same thing can be seen in the upper chip, too, or are you suggesting the chip is casting that shaped shadow?

hmm it looks like a shadow because if the light is coming from the upper left hand side of the chip the green is stronger on lower right and side of the chip by wolf. And the sides that the light is hitting from its glossy. The upper left hand side is more glossy and the upper right hand side is less glossy. Just as we would expect if the light was hitting from the upper left hand side.

Also the wolf pics the only thing I could see have we ever seen a chip that has "sample" written on it? I'm not sure just asking.

Yep there were chips with sample on em, I think the wolf one is real and the other chip, maybe some fudging going on.

[MulciberXP]

looks to me like someone smudged the right facing corner of the chip in the 2nd picture. seems too smudged to just be compression artifacts.

[Mintmaster]

Anyone else find the last few pages of posts rather funny?

Man are you guys desparate for R600 discussion material

[tEd]

Originally Posted by Mintmaster

Anyone else find the last few pages of posts rather funny?

Man are you guys desparate for R600 discussion material

disturbing fits better

[Ailuros]

Originally Posted by Mintmaster

Anyone else find the last few pages of posts rather funny?

Man are you guys desparate for R600 discussion material

Pin count masturbation is kewl

[Skrying]

Originally Posted by Mintmaster

Anyone else find the last few pages of posts rather funny?

Man are you guys desparate for R600 discussion material

I think its very telling of how we are here at B3D. Lol, I know of no other forum where you can get pages out of just counting pins!

[BRiT]

Originally Posted by Skrying

I think its very telling of how we are here at B3D. Lol, I know of no other forum where you can get pages out of just counting pins!

It wouldn't have gone on for so long had people not counted incorrectly.

[Twinkie]

Heres a few tibits for the desperate crowd

http://www.hardspell.com/news/showco...?news_id=31823

These guys did quite a few measurements based on a mockup model of the R600 die.

Now this picture that revealed the true size of the R600 core. The LGA775 processor and the first with some friends, R600 size of the package to Congress circle, But Die many relatively small size, and according to this, we have to analyze the size of R600.

After the PS Lane, The photographs taken by our friends and simulated photographs zooming to the same size ratio. LGA775 processors from the contrast ratio, after measuring and precise calculations, the core size of the inferred R600 21.2x20.5 mm, the core area of 434.6 square millimeters, in general, the closer to the core of R600 square.

R600 estimated size (high) : 21.2 mm "BR> R600 estimated size (width) : 20.5 mm

According to the current information, ATi R600 will be the next generation DX10 graphics core Sale in January next year. This new ATi will be the same this time the core was laying low. Today was a mysterious R600 core users leaked out.

According to the leaked information, R600 has 64 high-frequency and unified framework Shaders. ATi hope eventually to the higher frequency of 700MHz, R600 will use 80 nanometer design. R600 graphics chip will be the highest in the history of the 12-layer PCB AMD-ATi match. used a number of innovations in the entire R600 chip design The most noteworthy thing is 512-bit Xiancun Distributed Power controller design and components.

『Contrast R480 R600 graphics core』

R480 130 nanometer process used earlier, and the R580 using 90 nanometer technology The use of 80-nanometer technology R600, R600, although a substantial increase in the number of transistors. But do not raise the number of core dimension.

This also gives the R480, R480 and R580 is the core area of 250 square mm, 352 mm and square 434.6 square millimeters (estimated).

Maybe my post was too big with too many pictures..

[stevem]

What's interesting is that only the outer circumference is SBGA, whereas the rest of the chip is essentially standard BGA. If they were serious about packing density & pitch/pads were the limiting factors, then you might have expected the whole pad area to be SBGA. I wonder if the SBGA area corresponds to the MC?

[turtle]

You mean...Like a RING bus?



Anyone care to go back a few pages, count the pins around the outer edge of R520/R580, and then compare it to R600 and see what the difference is?

(My god [or lack there-of]....Did I just encourage the pin count debockel to continue? Crap...)

Edit: Realized there is no defining outer edge by looking at the R5xx pcb....But I still think your guess sounds logical.

[trinibwoy]

Originally Posted by turtle

(My god [or lack there-of]....Did I just encourage the pin count debockel to continue? Crap...)

debacle even lol

[turtle]

Firefox didn't have that one in it's dictionary. Give a guy a break.

So...Whatcha think?

4 ring stops with each with two 64-bit access channels each, or 8 ring stops with similar 2x32-bit access like R5xx (which was 4x [2x32])?

Or could have teh inq been right all those days ago with the 16x32-bit report?

[Jawed]

I think the ring-stop count is equal to the ROP-group count (or TMU-quad count).

i.e. in R5xx, the ROPs are grouped in 4s, as well as the TMUs. So 16 ROPs in total are arranged in 4 groups of four, one group per ring-stop.

So it seems likely that R600 will follow the same pattern, with 8 ring stops (each of 2x32), therefore 8 ROP groups, each one working with one ring-stop. So a total of 32 ROPs and I guess 32 TMUs.

Prolly not enough TMUs for Mintmaster.

It seems pretty unlikely that R600 is a 24 TMU, 24 ROP GPU simply because it's messy to organise an equal split of a 512-bit bus across 6 groups. Though that doesn't exclude there being a 384-bit bus (6 ring-stops), of course - so perhaps that's an upper midrange part?

Jawed

[hoom]

I think Dave gave some clues on that in an earlier exchange with Triumphsiao where he took pains to point out that R580 already had 2*32 on its 4 ringstops when Triumpsiao was talking 8 ringstops or 4 that were double R580 (he was assuming 4*64bit).

32 rops seems silly overkill if its a high clocked (like 800mhz people suggest) chip though since 16 is enough at 650mhz.

[trinibwoy]

At this point I don't think it's a safe bet that 32 ROPs is overkill for 140 GB/s of bandwidth Pure fillrate aside there's still FP blending and AA to worry about.

[turtle]

As for TMUs, if the Raster Ops are still coupled with them isn't it possible it could follow Xenos' scheme and have 64 (32 point + 32 bilinear?), 48, or 32?

I know this was touched upon briefly earlier by Pete, but I don't recall if anyone else thought this was viable or even possible. Perhaps I missed it. I know when getting into a question like this it can be diluted with thoughts of wishful thinking, as we most certainly can't EXPECT a 32rop/64tmu/64 4D (256 shader)/512-bit/1Gb highest-bin GDDR4 creature out of a chip that would only be a little over 500mm on 90nm, but it's interesting to look at both the pin count and what Xenos does for it's size. I wonder how much of that magic can be ported over to R600. Think it's more likely rops will stay coupled to tmus and such the traditional GPU styling, or go more oddball as we see in Xenos?

BTW: Thanks for the catch-up on the info earlier from Dave about the bus setup.

[Shtal]

Originally Posted by trinibwoy

At this point I don't think it's a safe bet that 32 ROPs is overkill for 140 GB/s of bandwidth Pure fillrate aside there's still FP blending and AA to worry about.

Would you explain more on why it will be overkill?