[WCCF] AMD Radeon R9 390X Pictured

Page 77 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Feb 19, 2009
10,457
10
76
That is a huge die. Easily >600mm2.

Looks like .vodka's calculations based on HBM size are close.

http://forums.anandtech.com/showpost.php?p=37445180&postcount=1751

@Grooveriding
Bare die with HBM shinies! :D

Edit:
Hawaii (~440mm2) vs GK110 (~550mm2) = Equal performance.
Fiji (>600mm2, probably close to 700mm2) vs GM200 (600mm2) = ?????

There's no way Fiji will be slower than Titan X with such a huge die AND HBM advantage.
 
Last edited:

Azix

Golden Member
Apr 18, 2014
1,438
67
91
I think you guys are maybe mixing up the actual GPU with the rest of the structure. It looks like what we saw before and it may look bigger relative to her.

more speculation.
 
Feb 19, 2009
10,457
10
76
If the dimensions of those HBM chips can be acquired, then it would be easy to get the die size in photoshop. That die is at least 600mm though o_O

Already done by .vodka, i put the link in that post. HBM dimensions are published & known. I would guesstimate ~650mm2 at least.
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,476
136
That is a huge die. Easily >600mm2.

Looks like .vodka's calculations based on HBM size are close.

http://forums.anandtech.com/showpost.php?p=37445180&postcount=1751

@Grooveriding
Bare die with HBM shinies! :D

Edit:
Hawaii (~440mm2) vs GK110 (~550mm2) = Equal performance.
Fiji (>600mm2, probably close to 700mm2) vs GM200 (600mm2) = ?????

There's no way Fiji will be slower than Titan X with such a huge die AND HBM advantage.

The package is massive. The die on its own is definitely closer to 600 sq mm. We need somebody to make some accurate calculations based on the photo of the Fiji package shown at Computex. Its already known that HBM is 7x5 mm. I am sure we will have die size calculations within a day or two.

AMD has gone all out with Fiji die size. Hopefully they have packed in significant architectural improvements to make Fiji the fastest GPU in the world without a doubt. :thumbsup:
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
How so?

My 670 was cheaper than the 7970 or 7950. My 970 was cheaper than the 290 or 290x...

All of those were great cards, but you are WAY oversimplifying the situation here. It really depends on what is important to you and when you purchased.

Your statement is terribly inaccurate and false.

Well I am still using my 7950 looking to ride it out to FinFet node shrink GPUs, if I'd gotten a 670 instead I think I'd have replaced it by now. Only thing that might tempt me is if a real price war breaks out between AMD and Nvidia post AMD lineup reveal.
 

Azix

Golden Member
Apr 18, 2014
1,438
67
91
closer to 500mm^2 is my vote. If I'm wrong one of you will have to eat your shoes.

Trying to measure it with software could be worthwhile. Looks like a better estimate though the picture is blurry.

You guys think its massive based on her hand I am thinking, but shes female and asian. Yet she actually has her middle finger bent and seems to be holding it comfortably. Try putting your hand in that position and remember she is holding the entire package. unless she has massive hands...
 
Last edited:

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Looks like AMD's big die will be much closer to Nvidia's in size this time around. Looks 550+ to me.
 

96Firebird

Diamond Member
Nov 8, 2010
5,749
345
126
I did my best... *shrug*

lO6Hhnl.jpg
 
Last edited:

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
26.45*31.48 would be ~832.

I think the visible top portion of HBM is smaller than the total listed dimensions.
 

Azix

Golden Member
Apr 18, 2014
1,438
67
91
the dark should be the hbm based on the GPU color. I'd use 7mm since that's what AMD said.
 

96Firebird

Diamond Member
Nov 8, 2010
5,749
345
126
Oh, ok, I'll give that a shot in a bit...

Here's what I got with the older die shot:

GS0kw1E.jpg


This is really digging into my late-night-drawing session, that need to be done by the end of the week... D:

I see now that the shiny part is not the HBM.

Edit - Second attempt at the newest shot, at this point I have no idea...

lPvVHid.jpg
 
Last edited:
Feb 19, 2009
10,457
10
76
I think the shiny is HBM because the interposer in other sections don't shine like that, specific 4 modules = 4 shiny areas.
 

destrekor

Lifer
Nov 18, 2005
28,799
359
126
Oh, ok, I'll give that a shot in a bit...

Here's what I got with the older die shot:

GS0kw1E.jpg


This is really digging into my late-night-drawing session, that need to be done by the end of the week... D:

I see now that the shiny part is not the HBM.

Edit - Second attempt at the newest shot, at this point I have no idea...

Eo9N6cE.jpg

You measured one size yet calculated a different size. Looks like you may have made a type if using a 10-key keypad (right side keyboard calculator pad).
If you calculate 26.17*21.58, that = ~565mm^2.
 

MrTeal

Diamond Member
Dec 7, 2003
3,919
2,708
136
I think the shiny is HBM because the interposer in other sections don't shine like that, specific 4 modules = 4 shiny areas.

Look at the proportions of the four shiny areas. The top ones are almost square. The bottom ones are more rectangular, like you'd expect from HBM.
They also have the same reflection as the other piece of silicon in the picture that's on the same plane, the big GPU die.
 

96Firebird

Diamond Member
Nov 8, 2010
5,749
345
126
You measured one size yet calculated a different size. Looks like you may have made a type if using a 10-key keypad (right side keyboard calculator pad).
If you calculate 26.17*21.58, that = ~565mm^2.

Oh yeah, good catch. I'll edit that image now...

Not that it may matter much, we can't even agree on which part is the HBM!
 

DDH

Member
May 30, 2015
168
168
111
pt4sw.jpg



You can see the core is raised up against the interposer at the top. I think those lighter/brown areas are the HBM
 
Last edited:
Status
Not open for further replies.