Asus Dual 6800Ultra

Thats one big MF, now all we need is 1GB GDR3 on board.
And no you wont be able to use two of these in an SLI mobo as the two chips are connected via SLI on the PCB itself hence all 16 lanes would be in use allready (look at the pcb traces between the pci-e interface and the two 6800U cores, you can clearly see the 16 traces shared between them). Also the mobo will have to be SLI ready (similar setup to the gigabyte SLI combo). Asus please correct me if iam wrong.
 
kleox64 said:
Thats one big MF, now all we need is 1GB GDR3 on board.
And no you wont be able to use two of these in an SLI mobo as the two chips are connected via SLI on the PCB itself hence all 16 lanes would be in use allready (look at the pcb traces between the pci-e interface and the two 6800U cores, you can clearly see the 16 traces shared between them). Also the mobo will have to be SLI ready (similar setup to the gigabyte SLI combo). Asus please correct me if iam wrong.

No it doesn't. All the motherboard needs is an x16 PCI-E slot and that's all. Theres no magic to SLi motherboards. It can be done with any chipset with the appropriate number of PCI-E lanes. If the motherboard had to be "SLi ready" then the Gigabyte i915P SLi board wouldn't have been possible. Nor would the IWill DN800-SLi. Both of which are using Intel PCI-Express chipsets.

In fact due to the fact that the only announced P4 SLi board, the Gigabyte one can't actually be found anywhere this card would be the obvious choice for higher end gamers looking for SLi performance on their Pentium 4 machines.
 
Wow.
I am impressed.

Now I want a picture with the heatsinks on it.
I bet someone could photoshop some heatsinks on there. ;)
 
OK, but for someone to enable SLI (dual video cards or single dual GPU) wouldnt you need an SLI compatible mobo? Iam not exactly sure but ill read up on the gigabyte board.
 
kleox64 said:
OK, but for someone to enable SLI (dual video cards or single dual GPU) wouldnt you need an SLI compatible mobo? Iam not exactly sure but ill read up on the gigabyte board.

Not necessarily. If you have the slots it can be made to work. With a single card SLI solution, it's simply done at the driver level.
 
as i remember the intel gigabyte sli board doesnt split the 16 lanes equally.
 
ThomasE66 said:
Not necessarily. If you have the slots it can be made to work. With a single card SLI solution, it's simply done at the driver level.

ok got it.

Lets see what asus say about it.
 
I hate to see how much thats going to cost, Asus = $$$$$ and if there is a chance of two in SLI :eek: there goes the piggy bank. I can see it now, my second job standing beside old Roy as a Walmart greeter.
 
Blazemore said:
I hate to see how much thats going to cost, Asus = $$$$$ and if there is a chance of two in SLI :eek: there goes the piggy bank. I can see it now, my second job standing beside old Roy as a Walmart greeter.

not unless theres 32 pci-e lanes on the mobo.
also you cant as theres no SLI connecter DUH, something that can be fixed with some wire and a soldering iron;)
 
Am I the only one thinking that looks bigger than usual in terms of height? Also wouldn't there be traces running between the processors?
 
Very interesting. Hopefully someone other than Asus will make that card. Asus' cards are always much more than others, and their confusing names are borderline false advertisment.

Would like to see that in a case, it probably wont fit in some mid size towers.
 
mashie said:
It seems to be nearly 50% taller than a normal card yes.

Remember that vidocards have 12 to 16 layers of traces in the PCB, you won't see any except for the top layer. Mind the sticker "x-ray ok" ;)

True, but my thinking was surface traces would be somewhat easier to do (intuitive feeling/the simplest cards use surface traces, why would they go out of their way to make it difficult, please feel free to prove me wrong), and given the lack of them between the cores, why would they make work harder for themselves by making lower level traces.
 
mashie said:
If it was possible everyone would just use one layer of traces. With a 256bit memory bus for example you have 256 traces to deal with. And to jump between layers with traces cost more than staying on the same "dept". Say GPU 1 has to go down to layer 5 in order to get past the other signal traces, to move the signals up to the top layer and then back down to layer 5 for GPU 2 will cost a lot more than staying on layer 5 all the way. Also it might be preferred to have the signalling traces between two ground layers for less interference. As for manufacturing cost it isn't really a difference between making layer 1 or layer 7 since the PCBs are made as 16 0.1mm boards epoxied together. Plated holes has to be added for layer to layer connectivity, hence the added cost for layer jumping. ;)

There is NO difference between a surface trace or inner layer trace other than the impedance it can support. Almost all BGA directly go from ball to via. Once you are on the via it doesn't matter what layer you are on...you just route to the appropraite via.

More layers make it EASIER to route but also more expensive. It is actually technically harder to make a board w/ less layers than one with more as circuit complexity rises. If cost and board manufacturing were zero I would take a board with a billion layer just so I could brain dead route it as 1 route per layer.

The last board I had routed was 20 layers because we needed board done in 1 week with 2 guys working stacked 12 hour shifts. The board was expensive but it got done. I think the expidite cost was $8000 per board *cringe*. Could we have done it in less layers...yes. Would it have been done in a week...hell no.

-tReP
 
Taken from my post at XBit:
MAValpha said:
Color me skeptical.

The PCB looks far too simple, starting with the apparent lack of power circuitry surrounding two PEG 6-pin power connectors. While I realize that KISS reigns supreme, I think this takes it a bit too far. I'm sure there is interest in such a product, but I think this is more the result of wishful thinking and a late-night PhotoShop session.
I might have believed that they borrowed this from NV48's more simplified power supply, but NV48 is too recent a product for them to have had time for this. Also, I would have expected two of these circuits given that a single AGP 6800Ultra requires Molexes on two separate chains.
It also looks like someone just copy/pasted two close-ups of a normal 6800PCB, since the heatsink tracings would suggest a normal NV40/45 heatsink shifted a few inches against a second heatsink, with no interaction between them.

Off-topic, I think it's interesting that Asus' 2x6600GT seems to mirror Gigabyte's 3D1 PCB, shown at Tom's. Did they collaborate? If so, why was Gigabyte's offering available so much sooner?
 
i don't really think the card is authentic......it doesn't have any molex connectors on it, and why would the shim have the name of the card on it? It makes no sense.....
 
calchala said:
i don't really think the card is authentic......it doesn't have any molex connectors on it...
PCI Express cards don't use Molex, they use six-pin connectors (3x2). The card supposedly has two of them, the black boxes along the upper edge of the card.
However, see my previous post.
 
justacow said:


seems like noone has posted this here yet, so i will :)

source

Where and when can I buy one...lol... nice card man, nice find... :D One more thing though How much?... 1000$ RIGHT
 
And hopefully this mofo supports the 30" Apple Display at 2560x1600 (beating dead horse)
 
mentok1982 said:
Wow.
I am impressed.

Now I want a picture with the heatsinks on it.
I bet someone could photoshop some heatsinks on there. ;)
:p

2 regular-sized heatsinks won't fit on it. I'd hope that Asus puts some kind of Arctic Cooler-type external exhaust HSF on it. Just for scale, look how big (huge) the card is. I used the lines on the circuit board to align the stock 6800GT heatsinks mounting holes (off by a little due to the fish-eye distortion of the original):

bigdual6800u.jpg


Notice that the card uses 2 x 6 pin power connectors. If you have a PSU that includes one you'll still need to use a 12v molex to 6 pin adapter. :eek:
 
^ thats hilarious^

If the card is really that big then how would anyone fit it into their case? It looks liknd of cutesy with those puny HSF over it.

That being said, SLI on a single card is really interesting. You wouldn't need overpriced and bloated mobos to get the benefit of dual gfx cards...
 
Guys did it occur to anyone that this could be an engineering/reference board hence the large size?
 
calchala said:
i don't really think the card is authentic......it doesn't have any molex connectors on it, and why would the shim have the name of the card on it? It makes no sense.....
#1 it uses two 6 pin power connectors for pcie, AGP uses the molex connectors
#2 the shim has the name of the gpu not the card and pcie cuz it has a bridge chip on it
 
If they are going to make a card that big why don't they just stick four DVI ports on the thing. Then if it were possible to have two of these cards in a system you could have 8 monitors (in non SLI mode, unless nVidia updates their drivers).
 
Let's get real here: if this thing isn't a fake, who would buy this piece of shit? Who wants a video card over half the size of a motherboard in their case messing up airflow and just plain being ugly.

Waste of silicon.
 
zeebs said:
Let's get real here: if this thing isn't a fake, who would buy this piece of shit? Who wants a video card over half the size of a motherboard in their case messing up airflow and just plain being ugly.

Waste of silicon.

If it's real, then I do if you can do two of them in an SLI configuration :)
 
Sir-Fragalot said:
No it doesn't. All the motherboard needs is an x16 PCI-E slot and that's all. Theres no magic to SLi motherboards. It can be done with any chipset with the appropriate number of PCI-E lanes. If the motherboard had to be "SLi ready" then the Gigabyte i915P SLi board wouldn't have been possible. Nor would the IWill DN800-SLi. Both of which are using Intel PCI-Express chipsets.

In fact due to the fact that the only announced P4 SLi board, the Gigabyte one can't actually be found anywhere this card would be the obvious choice for higher end gamers looking for SLi performance on their Pentium 4 machines.

WHat it is is basically 2 AGP 8x cards, on one PCB, if you look closely, you can see half of thos little lines going to the connectors going to one chip, while the other half goes to the other

I would be more excited if this wasn't asus. It's probably like 6800NUs with the MHZ pumped up.
 
zeebs said:
Let's get real here: if this thing isn't a fake,
Cebit starts today. If it is fake, we'll know soon since Asus is supposed to be introduced at Cebit.
 
ThomasE66 said:
Not necessarily. If you have the slots it can be made to work. With a single card SLI solution, it's simply done at the driver level.

I thought this too, but not quite: http://www.hardocp.com/article.html?art=NzEyLDI=

"We had to try another motherboard -- a motherboard that does not natively support SLI at all. The ABIT AA8 Duramax is an Intel 925x-based motherboard with a PCI-Express x16 slot. We installed the Gigabyte 3D1 on this motherboard to see what would happen.

Sadly, it wouldn’t even POST with the video card installed. No initialization at all. Well, at least we tried"

==>Lazn
 
Lazn_Work said:
I thought this too, but not quite: http://www.hardocp.com/article.html?art=NzEyLDI=

"We had to try another motherboard -- a motherboard that does not natively support SLI at all. The ABIT AA8 Duramax is an Intel 925x-based motherboard with a PCI-Express x16 slot. We installed the Gigabyte 3D1 on this motherboard to see what would happen.

Sadly, it wouldn’t even POST with the video card installed. No initialization at all. Well, at least we tried"

==>Lazn

Damn, that goes contrary to what I've been told and read to date :(

Ah well, I don't plan on buying one soon anyways. I'm happy with my dual 6600 GT's. I'm waiting to see what the next gen GPU's offer before I upgrade again. It's bad enough that I find myself buying a new card or cards every six months :)
 
I think some1's wet dream spilled onto the internet.

Even if that is legit it's frickin HUGE!!
 
I wonder if you can pencil mod the DFI Ultra-D and then use that thing since you wouldn't need the bridge connector :).
 
BTW, the reason they didn't include an SLI connector was because this board CANNOT be run in SLI mode with any board currently available.
In order for four NV4x's to run on one x16-capable chipset, each processor would need 4 dedicated PCIe lanes. It is fairly obvious from the PCB that each GPU has 8 lanes running to it, dividing the x16 slot roughly in half. In a traditional SLI motherboard, the first half of the x16 slot is turned into an x8 while the second half is disabled. This means that the second GPU on this board would get no data on an x8 slot.

The only two ways I can see to work around this are:
1) base a board on a chipset with a total of 32 or more PCIe lanes, and have them split between 2 x16 slots.
2) use it in conjunction with Tyan's new motherboard, which has two nForce4 chipsets onboard. If Gigabyte's 3D1 is any indication, the Asus card would only work with Asus boards- more specifically, one of their A8N-SLI variants.
 
skinegibbs said:
That kinda sucks though...its just a 6600 lol

Your missing the point.

It may not be the fastest card out there but
  • Its two cores on the same card.
  • It not vapourware its in the shops NOW
  • If its working with 6600's the it will also work with other cores
 
Back
Top