View Full Version : GTX480 Dissected!
pbmax
03-24-2010, 06:29 AM
Check it out here: http://www.atomicmpc.com.au/Gallery/170368,nvidia-gtx480-disassembly-guide.aspx/1
http://i.haymarket.net.au/Galleries/20100324113531_IMG_0207%20copy.jpg
MegadetHCl
03-24-2010, 08:43 AM
Gotta love the Delta fan - sorry for the spoiler!
Everything about this card screams ARGH! ARGH! I'M ON FIRRREEEEE! ARGH!
I like how the RAM + FETs are cooled by a separate plate. Changing to watercooling should be a breeze :)
Can the mounting hole spacings be discussed?
Same as GT200?
Very interesting. Thanks for that^^ :up:
NKrader
03-24-2010, 09:54 AM
FETs
my backpacks got jets. :shrug:
[DANGERDAN]
03-24-2010, 10:22 AM
Its good to see everything layed out better than previous cards, the gpu die size is still a little big but its still good depending on how nvidia deal with stock coolers this time.
If they make a crappy cooler again they really are cheap bastereds lol cant wait to see :)
SAL36864
03-24-2010, 10:26 AM
Gotta love the Delta fan - sorry for the spoiler!
It is nice to see it use a Delta.
RPGWiZaRD
03-24-2010, 11:45 AM
Isn't it an IHS we're seeing and not the GPU itself?
I'm really curious about the fan tho, 1.8A, it just can't be true... 20W of power to the fan for a small 70mm fan! Does a fan even get above 1A normally at say 5500rpm? 1.8A sounds like it would be enough to spin it at 10000rpm(+). :p: I personally think it'll spin up to 5500rpm at full load and probably noise level probably around 45dB (saw some Delta 70mm 5500rpm at 47dB rating somewhere). :p:
EDIT: AFB0712SH seems to be the biggest and baddest 70x70x25mm fan Delta manufactures atm, 6300rpm and 0.65A and 50.5dB http://www.delta.com.tw/product/cp/dcfans/download/pdf/AFB/AFB70x70x25mm.pdf ... I know they're not the same type of fans but still, that 1.8A rating really looks fishy.
SAL36864
03-24-2010, 12:33 PM
I would trust the markings on the fan. Delta makes some pretty powerful fans, and many that I have seen do not match anything listed on their site.
zanzabar
03-24-2010, 12:55 PM
I would trust the markings on the fan. Delta makes some pretty powerful fans, and many that I have seen do not match anything listed on their site.
ati and the last gen NV cards used a similar one at 1A and it was ear spiting, i cant imagine what this one could do if u let it loose.
i dislike the pwm, they should switch back to voltera, unless the core gets so hot that the card couldent accommodate it
franklin5252
03-24-2010, 01:34 PM
The card iis laid out Great and i hope to see some benches on it soon.
G.Foyle
03-24-2010, 01:59 PM
Can the mounting hole spacings be discussed?
Same as GT200?
61x61 mm, on both GTX480 and 470 :)
So Tek-9 rev. 4.0 SLIM and upwards should fit :)
Don't like the heatpipes protruding out the top, but looks pretty standard design to me.
The integrated "unisink" is a nice touch- might have to cut around a little to make some GPU waterblocks fit
Frag Maniac
03-24-2010, 09:05 PM
I don't see any pics for Atomic's 480 dissection gallery no matter how I access that page. Did Nvidia make them take the pics down or something?
thebluemeanie1
03-24-2010, 09:08 PM
I don't see any pics for Atomic's 480 dissection gallery no matter how I access that page. Did Nvidia make them take the pics down or something?
Not working for me either.:(
Winchip
03-24-2010, 10:37 PM
New link (http://picasaweb.google.fr/PCi.David.l/1003AtomicMPCGeForceGTX480)
YukonTrooper
03-24-2010, 11:46 PM
Wow, that fan is going to whisk me and my house away to the Land of Oz during my first Fermi gaming session.
texasreefer
03-27-2010, 08:26 AM
so i'm confused, this is suppossed to be a 40nm chipset , right? That thing looks huge compared to ati's chipset...is that a temporary chipset bc of the lack of chips/chipsets due to manufacturing setbacks? I heard that the lack of 5870 and other high end cards were set back due to manufacturer and budgets, and that there is a lack of highend dies. Suppossedly, that's why there is a surplus of 240, 250s and 5770s, etc. That could explain the heat issue...smaller die = less heat.
Frag Maniac
03-27-2010, 04:27 PM
What you're seeing is a much higher percentage of added features than the percentage of die shrinkage. That's why the chip is larger despite the smaller die. Like I've said, Fermi would have been much better served with a 32nm die process. It was just not in the cards for them to have the die shrinkage perfected that far by the deadline they wanted to meet. It could still happen though because power wise it's possible with even the existing 6 and 8 pin power connectors and one 16x Pci-Ex slot to have the full 512 cores working, but then the heat might become insurmountable with anything but water cooling. Look for very exciting things to happen should Nvidia manage a 32nm die.
CryptiK
03-27-2010, 06:06 PM
The mosfet/ram cooler plate is very similar to the one used on the single PCB GTX295. With the power the PWM will have to produce to feed this thing, I wouldn't be leaving the plate on and using a cheap 'gpu only' waterblock. Full cover block would probably be the best way to go, so the PWM and ram runs at a lower temp also.
What you're seeing is a much higher percentage of added features than the percentage of die shrinkage. That's why the chip is larger despite the smaller die. Like I've said, Fermi would have been much better served with a 32nm die process. It was just not in the cards for them to have the die shrinkage perfected that far by the deadline they wanted to meet. It could still happen though because power wise it's possible with even the existing 6 and 8 pin power connectors and one 16x Pci-Ex slot to have the full 512 cores working, but then the heat might become insurmountable with anything but water cooling. Look for very exciting things to happen should Nvidia manage a 32nm die.
32nm was skipped for 28nm right.
Frag Maniac
03-27-2010, 11:18 PM
32nm was skipped for 28nm right.Was? With only 40nm being achieved by them thus far, it's still up in the air as to how much smaller they can go. It should be a major clue that in having to disable 32 of the cores to produce the chip they have now, nothing is set in stone as to how much more they can achieve. I'll believe 28nm if and when they can do it. My guess is it will be some time before they can even go 32nm. That's the smallest Intel has achieved thus far, and they lead the industry in die shrinking. Nvidia may not even be able to produce a viable 32nm chip until after ATI comes out with their next gen.