Pics of gt300

[jaredpace]

they might be using very oddly mounted power plugs, and those 2 could be for the side one, and the other 8 are shared by the top and side?

Possible. However, I only see 10 solder points, 6 + 8 = 14.

[Manicdan]

Possible. However, I only see 10 solder points, 6 + 8 = 14.

yeah thats why im wondering if they are being shared, i dont know how much that matters on the board itself

[jaredpace]

thanks for the pix, eastcoasthandle.

It just looks like they sawed the end of the pcb off:

[Eastcoasthandle]

If you noticed in that pic above, their is a small outline for what I believe to be the 4-pin fan connector (it could be for something else). Which is underneath the 6pin connector. However, there is no soldering points.

[Helloworld_98]

this card must use loads of energy, this is the tesla version which has a lower power usage than the GTX series and it still uses between 225w and 300w, so unless dual 8 pin and a 6 pin is an option for nvidia, we probably won't see a dual GF100 card.

[Manicdan]

is it possible this card is just a mockup for show? build the case but use an old PCB cut to fit? nothing about these connectors make sense. if they were optional, like a 8pin on top and side, and you choose which to use, that would be cool. but to say you have to have room up top AND on the side, its gonna look retarded ugly in a clean case.

EDIT: there a ****ing screw hole under the 6pin connector, that cant be right

[jaredpace]

EDIT: there a ****ing screw hole under the 6pin connector, that cant be right

add to the fact that you cannot exhaust air out of the thing, and I'd say you're on the right track



oh, and how are you supposed to put your sli connectors on?

[largon]

where? you mean the gpu shot? thats not a die shot! die = silicon, what they show is a lid labeled U90 (...)

Err?
The page does have a die shot. This:

[Chickenfeed]

this card must use loads of energy, this is the tesla version which has a lower power usage than the GTX series and it still uses between 225w and 300w, so unless dual 8 pin and a 6 pin is an option for nvidia, we probably won't see a dual GF100 card.

This is what I've been assuming for sometime now. If we see one it will be on 32/28nm and at reduced clocks ( if not disabled units as well ) For Nvidia's sake hopefully the 380 can compete with the 5870x2 otherwise they are in trouble ( as they don't look to have any mid / low end GT300 cards in the design pipeline in any reasonable amount of time ; eg the real profit earners ) I sense GT200 rebrands...

[saaya]

does anybody else have a dejavu thing with that pcb?
i cant help but think ive seen it before...

they definately cut off the back part and made it shorter, like literally cut it off with a saw LOL... they even cut some solder pads in half as you can see in the better high res pics...

its not any gtx260/75/85 pcb variation... its not the old or new 295 pcb, its not the 9800gx2 or 7950gx2 pcb... its not anything g92 either from what i can tell...
the odd thing is that it has the solder spots for 8pin to the top and then apparently another 6 pin to the top further to the back, but only 2 pins are visible, the rest of the pcb has been sawed off...

previous tesla cards have this odd one power connector up and one to the side thing as well, BUT they have the 8 pin to the side and the 6pin to the top... well, this card has as well, plastic connector wise, but the actual pcb solder spots are contradicting that completely

the heatsink shround will probably be used for the tesla card once its finished, they cant make it much longer cause theres just not more space inside their tesla blade servers and i think they dont want to redesign those... but the pcb is completely diferent from what the final cards will have...

what really sticks out when looking at this card compared to the previous generations of nvidia cards, is that the gpu seems to be centered above the pciE interface... thats odd, cause then either the nvio chip needs to be moved, and where to? or the memory chips need to be moved... that would mean there are only memory chips on 2 sides of the gpu... which sounds more likely.

conclusion:
very lame launch event from nvidia... very very lame... as in setting a new record in paper launches...
they should have at least shown a dummy card that looks more like the actual retail cards... or maybe they dont even know it yet cause they are still that far away from actually mass producing them

its beyond irony that its nvidia that criticized intel for their lrb pr campaign all the time, rightfuly!, and now they beat intel at lame pr campaigning by having a paperlaunch based on hot air and bold words, oh and a video of some floating green spots and a frankenstein sample card based on some old sample parts they fished out of the rnd labs trashcan

[Helloworld_98]

For Nvidia's sake hopefully the 380 can compete with the 5870x2 otherwise they are in trouble ( as they don't look to have any mid / low end GT300 cards in the design pipeline in any reasonable amount of time ; eg the real profit earners ) I sense GT200 rebrands...

we'll just have to see how the drivers are, and also this list of things that could put the 58X2 ahead,

1. GF100 is purely a Tesla GPU, unlikely but the lack of hardware tessellation and seeing as this is focused at the GPGPU market, it's possible

2. GF100 uses so much power and emits so much heat, that they have to dial it down a few notches for consumer use.

3. GF100 costs about the same as the 5870 X2, performs a lot better in some games, gets owned when DX10.1 and DX11 come into equation.

4. In the GPGPU segment, larrabee takes the crown.

[saaya]

Err?
The page does have a die shot. This:

ohhhh die shot as in logic/design die shot!
my bad!

yeah but that doesnt mean anything, does it?
you can create this easily by using the design tools + photoshop...
you not only dont need working silicon, you dont even need any actual silicon for it...

This is what I've been assuming for sometime now. If we see one it will be on 32/28nm and at reduced clocks ( if not disabled units as well ) For Nvidia's sake hopefully the 380 can compete with the 5870x2 otherwise they are in trouble ( as they don't look to have any mid / low end GT300 cards in the design pipeline in any reasonable amount of time ; eg the real profit earners ) I sense GT200 rebrands...

they claim they do... but considering that they didnt mention them AT ALL, and didnt even bother to invest 20mins to glue some vga parts together to create a another dummy card... i really dont think they will have gt300 cut down cards very soon... i mean actually cut down silicon, not half broken gt300 trash...

helloworld 98, that makes sense...
i think it was a huge mistake from nvidia to make their gpus that gpgpu heavy...
the smart way is doing it the intel way, intel doesnt release a new cpu that is more gpu like, at the cost of cpu performance and price, they create a diferent chip. thats what nvidia should have done as well...

[Helloworld_98]

does anybody else have a dejavu thing with that pcb?
i cant help but think ive seen it before...

they definately cut off the back part and made it shorter, like literally cut it off with a saw LOL... they even cut some solder pads in half as you can see in the better high res pics...

its not any gtx260/75/85 pcb variation... its not the old or new 295 pcb, its not the 9800gx2 or 7950gx2 pcb... its not anything g92 either from what i can tell...
the odd thing is that it has the solder spots for 8pin to the top and then apparently another 6 pin to the top further to the back, but only 2 pins are visible, the rest of the pcb has been sawed off...

maybe it's half a mars, and then they played around with the power connectors to make it look special.

[jaredpace]

saaya, my guess is it's either 1/2 a 9800gx2, a last gen tesla board, 1/2 a gtx295, 1/2 a mars, or an internal lab sample. I've been kinda surfing around to see if i could find which board it is too

very very lame... as in setting a new record in paper launches...

lol you mean like "faking" a paper launch?

What i found particularly ironic & funny was that with semiaccurate reporting all that "only seven samples came back from A1" bull, there was a #7 written in black marker on top of the week 35 A1 fermi sample's IHS!!! Lol, the irony.

[Manicdan]

it is kinda funny they hold up a mockup, i truly wonder what kind of silicon is under that, if any.

and can it run crysis?

[jaredpace]

I dunno dan, but speaking of crysis... I was reading at [H] a post by trinibwoy - there was a poster talking about geforce/tesla and asking why all Nvidia did was talk about gpgpu and GPUc at GTC. Trinibwoy responds saying, "it's GPU Technology Conference, not Crysis Framerate Conference." LOL!

[highoctane]

EDIT: there a ****ing screw hole under the 6pin connector, that cant be right

That could very well be a spot of solder on copper, its not a screw in the hole and there's not light shining from the other side to illuminate it.

[Eastcoasthandle]

That could very well be a spot of solder on copper, its not a screw in the hole and there's not light shining from the other side to illuminate it.

I believe he is correct. Here take a look at the gtx285. As you can see there is a ring around every hole were a screw is used.

[highoctane]

I don't think they use fans or shourds in blade configs, the blades themselves should have a flow thru design for airflow.

the heatsink shround will probably be used for the tesla card once its finished, they cant make it much longer cause theres just not more space inside their tesla blade servers and i think they dont want to redesign those... but the pcb is completely diferent from what the final cards will have...

[highoctane]

Here take a look at the gtx285. As you can see there is a ring around every hole were a screw is used.

I understand that but where is the illumination in the center coming from, thats not a screw or an open hole.

EDIT: Look at the other smaller solder spots with copper showing.

[Tao~]

Umm.... Isnt it small (by nvidia's standards ofcourse )

[flippin_waffles]

Lol wow, how embarassing for nv. Well, I guess they actually have to give a *edited censor bypass - DilTech* about their custormers for it to be embarassing.

[saaya]

maybe it's half a mars, and then they played around with the power connectors to make it look special.

i checked the mars pcb, its not the same... the gpu is in a diferent position...

saaya, my guess is it's either 1/2 a 9800gx2, a last gen tesla board, 1/2 a gtx295, 1/2 a mars, or an internal lab sample. I've been kinda surfing around to see if i could find which board it is too

heheh, yeah it DOES look familiar doesnt it? i could have sworn i have seen those pcb vent slits before... but i cant find anything with it now...
i checked all tesla cards i could find, they look diferent...
gtx295, checked old and new, nope...
i suspect its an internal lab sample... possibly even of a gt300...

lol you mean like "faking" a paper launch?

well for a second i even thought of making a video with some friends and staging our own infast vga launch in the same style and putting it on youtube ^^
but i dont have a cam and i think my buddies here in taiwan wouldnt want to spend time on that hehe

What i found particularly ironic & funny was that with semiaccurate reporting all that "only seven samples came back from A1" bull, there was a #7 written in black marker on top of the week 35 A1 fermi sample's IHS!!! Lol, the irony.

LOL i didnt notice that hahah
well didnt anandtech confirm the 7 working gpu thing?
he did put it in p[erspective though and said its not as bad as it sounds, and i agree with him... anyways, the no7 on it is def funny LOL
and it looks like they tried to wipe it off but you could still see it slightly

it is kinda funny they hold up a mockup, i truly wonder what kind of silicon is under that, if any.

and can it run crysis?

my guess is that its an internal lab sample of a dual gt200 card with a tesla mockup shroud.
so about crysis... i doubt it... especially since they sawed off a part of the pcb!

That could very well be a spot of solder on copper, its not a screw in the hole and there's not light shining from the other side to illuminate it.

as if that would change anything :P
the pcb clearly has solder spots for an 8pin connector on top, and next to it a half cut off 6pin connector, both facing to the top... the card shows a 6pin facing to the top and an 8pin facing to the side

I don't think they use fans or shourds in blade configs, the blades themselves should have a flow thru design for airflow.

they do, you can buy tesla cards as a standalone card to use in any pc or server, those cards have a propper shroud and all, google it

one thing i gotta say about this launch event... the plastic shroud of this frankenstein card DOES look really nice!
i really like it... looks better than the ati cards imo, the plastic the ati shrouds use seems really cheap

[highoctane]

Hey I'm not trying to say the pcb makes sense , just my observation that its not an open hole and that there is no screw in it, looks like a ball of solder to me.

as if that would change anything :P
the pcb clearly has solder spots for an 8pin connector on top, and next to it a half cut off 6pin connector, both facing to the top... the card shows a 6pin facing to the top and an 8pin facing to the side

[ACE76]

I think people here are getting too worked up over this videocard and it's upcoming release....let it go and stop analyzing into a few stupid pictures so much!

I don't think Nvidia will make the same mistakes they made years ago with cards like the FX....this card will almost undoubtedly beat the 5870 in performance but at the same time will cost more. I thought it was hilarious when the Nvidia guy cursed when telling Anandtech that when building huge GPUs, it takes a long f***ing time!

More important than trashing Nvidia and their paper launch, is how much money will they actually lose in the months that ATI will have their 5870 on the market for. If Fermi doesn't destroy the 5870 at launch performance-wise, they will lose this round yet again...and that's sad when you have the superior product.