GF106 die shot and 512sp GF100

[WeiT.235]

Entry level Fermi stuff is coming...
15.2x15.7=238.6 mm^2
no reference board, good OC capability, performance between HD5750 and HD5770




source:
[EXP]
English version:
http://en.expreview.com/2010/08/04/g...s-up/8903.html

The anticipated full-spec GF100 benchmarked.




source:
[EXP]

[TRANCEFORMER]

u beat me posting this

[-Boris-]

GF106 looks like expected. A GF104 cut in half.

The clocks at GF100 SP512 seems to be the same as SP480.

[rado992]

Anyone notice that the "Fermi 512" card reports running @ PCI-E v.1 x16 and not v2.0? Wonder how much of an impact that could have on scores if it was true.

[Open1Your1Eyes0]

Anyone notice that the "Fermi 512" card reports running @ PCI-E v.1 x16 and not v2.0? Wonder how much of an impact that could have on scores if it was true.

This was probably taken a short time after the benchmark during it's "idling" phase. It bumps down to x16 v1.1 in order to cool down.

[ChinStrap]

The anticipated full-spec GF100 benchmarked.
http://www.expreview.com/img/news/20...GTX480_512.jpg

very repost

http://www.xtremesystems.org/forums/...d.php?t=255025

[Mechromancer]

I hope they somehow magically fixed the fab process to allow this thing to not draw enormous amounts of power or get hot enough to boil water. This thing would make for an excellent F@H monster if so. Oh well, we have the full-shader GF104 on the way to save the day I suppose.

[cegras]

Is the GF106 even an upgrade over the 48xx or GTX2xx?

Is the GF104 even a cost effective upgrade?

[WeiT.235]

very repost

http://www.xtremesystems.org/forums/...d.php?t=255025

This post in EXP haven't said it was the 512 sp Fermi until recently...

[Olivon]

edit : Oops ... Wrong thread, sorry

So, back to subject, 512sp GPU-Z shot is fake or not buddies ?

edit2 : ryboto got me

[ryboto]

Old slide ...



10% Int more
35% FP more
(Thanks Hardware.fr and Marc for datas)

Is it still relevant or not John ?

Wrong thread! quicker than the edit, I was.

[Boissez]

GF106 looks like expected. A GF104 cut in half.

The clocks at GF100 SP512 seems to be the same as SP480.

@ 240 mm^2 it's more like a half GF100.
Has there been any words on on whether it'll be 256 or 192 SP?

[-Sweeper_]

@ 240 mm^2 it's more like a half GF100.
Has there been any words on on whether it'll be 256 or 192 SP?

well if it has the same SM structure as GF104 (48 SPs), then I guess 240 SPs or 288 SPs for the ''full chip''

a GTS450 with 240 SPs might perform well...

radeon HD5770 is not that much slower than radeon HD5830, I doubt a 192 SPs GF106 would be competitive, even at high clocks..

and dont forget the 1/3 die size reduction from GF104 to GF106, theres definitely more than half a GF104 in there...

[Johnny87au]

not bad bench results, Anyone know which cards will be using these?

[zanzabar]

i expect better from them, how hard is it to post a legal 3dm vantage score. i dont care who its from they should know how to run a benchmark without breaking the damn rules for physX

[Nintendork]

GF106 die size: 238mm^2 (too big for a price battle)
HD5770 die size: 170mm^2
HD5670 die size: 104mm^2

GF106 doesn't seem to be competitive at least for nvidia in order to make money. I suppose they want to sell it for less than what 5770 cost and battle for the market share instead of making profit.

[slaveondope]

2x8pin connectors

[~CS~]

That PCB is not GTX485,that's a non reference GTX480 design from some Chinese company.Have come up here in the "News" section as well.

[Oliverda]

GF108

[Caparroz]

I don't doubt there's a gf106 in the making but those die shots look ps'ed to me. The lighting on the part name looks a bit weird... If they're real that's a gigantic chip for a mid to low-end part.

[spursindonesia]

So, GF 100 @ around 530 mm^2, GF 104 @ 366 mm^2, GF 106 @ 238 mm^2, GF 108 @ around 130 mm^2, no wonder nVidia has to play the capacity domination game at TSMC to make sure AMD won't engage them in a price war, since the supply is still constrained.

[saaya]

So, GF 100 @ around 530 mm^2, GF 104 @ 366 mm^2, GF 106 @ 238 mm^2, GF 108 @ around 130 mm^2, no wonder nVidia has to play the capacity domination game at TSMC to make sure AMD won't engage them in a price war, since the supply is still constrained.

well the latest rumors is that nvidia is reserving LESS capacity at tsmc for the coming months... odd huh?

[KainXS]

I don't think that wei hang GTX480 is the 512sp version of the card, see the gpu-z shot, the texture fill rate is wrong for 512sp(it should be 44.9Gtexel/s) and the bios was used in the pre release GTX480's, it the bios with all the fan problems if you remember, it wouldn't more than likely be used in a new card, the die is also blurred, and we all should know that the card wei hang showed had the GF100-375-A3 core which is binned as a 480sp part, I think expreview is lying.

[zalbard]

512sp Fermi has a very nice, clean PCB layout. I like it.
Not a bad performance result, too. Then again, it's just Vantage, so meh...

[WeiT.235]

http://vga.zol.com.cn/190/1903529.html