Saturday, June 6th 2020

NVIDIA GeForce RTX 3080 Pictured?

Here are what could be the very first pictures of a reference NVIDIA GeForce RTX 3080 "Ampere" graphics card revealing an unusual board design, which is the biggest departure in NVIDIA's design schemes since the original GeForce TITAN. It features a dual-fan aluminium fin-stack cooler, except that one of its fans is located on the obverse side, and the other on the reverse side of the card. The PCB of the card appears to extend only two-thirds the length of the card, ending in an inward cutout, beyond which there's only an extension of the cooling solution. The cooler shroud, rather than being a solid covering of the heatsink, is made of aluminium heatsink ridges. All in all, a very unusual design, which NVIDIA could implement on its top-tier SKUs, such as the RTX 3080, RTX 3080 Ti, and in a cosmetic form on lower SKUs. We get the feeling that "Cyberpunk 2077" has influenced this design.
Sources: ChipHell Forums, HXL (Twitter), VideoCardz
Add your own comment

225 Comments on NVIDIA GeForce RTX 3080 Pictured?

#201
EarthDog
Chrispy_
Third party cards were usually cooler and quieter, all whilst having cheaper-to-produce, simpler cooler designs on them and using lower-binned, hotter-running silicon.
It's like, how it has always been...........AIB cards are cooler and quieter. Shocking :p :)! I think all people are saying is that they are better and less noisy than the blower style coolers. :)

Outside of that, I don't about lower-binned silcon...as far as I could see, most AIB cards overclocked father than the FE....
www.techpowerup.com/review/asus-geforce-rtx-2080-super-strix-oc/32.html
www.techpowerup.com/review/asus-geforce-rtx-2080-ti-matrix/32.html

The first two I looked at ^^. Clearly not all are that way, but this small sample shows card partner cards close or higher than the FEs.
Posted on Reply
#202
Vario
This cools really well but I don't think we'd ever see any manufacturer sell a card looking like this
Posted on Reply
#203
Chrispy_
EarthDog
I don't about lower-binned silcon...as far as I could see, most AIB cards overclocked father than the FE....
Binning on Nvidia cards is really easy to see, the die is actually stamped with the bin. Take a 2070 for example.

TU-106-400A-A1 is the FE, top bin. It was sold to some vendors at a premium for their top-tier factory overclocks, too.
TU-106-400-A1 is the standard 2070, that most 3rd party boards near the MSRP used.

The reason the FE didn't overclock as well was down to Nvidia strictly enforcing their own recommended power limit. The A-die let them get away with smaller, sexier coolers because they were higher-yield, more efficient silicon. 3rd party vendors had two options - buy the cheaper chip and throw the savings at beefier cooling to handle the higher voltages required for the non-A die, or buy the A-dies at a premium, go bonkers on the VRMs and cooling, and sell it as a top-tier card with a hefty overclock.

I believe that yields have improved since Turing's launch and there are now no "A dies" because all of them are considered good enough quality.
Posted on Reply
#204
EarthDog
Chrispy_
Binning on Nvidia cards is really easy to see, the die is actually stamped with the bin. Take a 2070 for example.

TU-106-400A-A1 is the FE, top bin. It was sold to some vendors at a premium for their top-tier factory overclocks, too.
TU-106-400-A1 is the standard 2070, that most 3rd party boards near the MSRP used.

The reason the FE didn't overclock as well was down to Nvidia strictly enforcing their own recommended power limit. The A-die let them get away with smaller, sexier coolers because they were higher-yield, more efficient silicon. 3rd party vendors had two options - buy the cheaper chip and throw the savings at beefier cooling to handle the higher voltages required for the non-A die, or buy the A-dies at a premium, go bonkers on the VRMs and cooling, and sell it as a top-tier card with a hefty overclock.

I believe that yields have improved since Turing's launch and there are now no "A dies" because all of them are considered good enough quality.
i recall that, yes... but I dont remember it paying off in actual clocks. The differences seemed negligible.
Posted on Reply
#205
Chrispy_
EarthDog
i recall that, yes... but I dont remember it paying off in actual clocks. The differences seemed negligible.
Well vendors would do their own binning too. So the A die were likely used for the Kingpin and FTW variants, but EVGA would have bulk-purchased loads of non-A silicon and would likely sub-bin that themselves to put the better silicon in the SSC Ultra and the worse silicon in the stock SC models.

I'm sure all vendors offering a range of factory overclocked models of increasingly aggressive clocks would have sub-binned the non-A dies too, so the fact that some non-A dies went into agressively-clocked cards is no surprise.

The other thing to remember is that binning is normally done by voltage - and one thing us overclockers have learned is that leaky/inefficient/higher-voltage silicon often overclocks better - provided you can cool it. Back when ASIC quality was exposed to GPU-Z, we'd often see the highest-clocking cards with ASIC quality of under 75% whilst the incredible undervolting champions were only found above 85%. So binning for voltage is definitely not the same as binning for clocks.
Posted on Reply
#206
EarthDog
Details. :)
EarthDog
as far as I could see, most AIB cards overclocked father than the FE....
Posted on Reply
#207
Chrispy_
EarthDog
Details. :)
Well you're making the incorrect assumption that Nvidia is binning for clockspeed; It wasn't. It was binning for efficiency and one thing us overclockers have learned is that leaky/inefficient/higher-voltage silicon often overclocks better - provided you can cool it.

Back when ASIC quality was exposed to GPU-Z, we'd often see the highest-clocking cards with ASIC quality of under 75% whilst the incredible undervolting champions were only found above 85%. So binning for voltage is definitely not the same as binning for clocks.
Posted on Reply
#208
EarthDog
Sorry.. of course there is more to it than clocks! I didn't feel I needed to mention that. :)
Posted on Reply
#210
dr emulator (madmax)
from the pics i've seen it looks crazy ( in a good way )
buuuut i would prefer a fan grill to stop parts falling in it, obviously that would impede the airflow
just my 2 cents :D

bwahaha i was thinking of the other cooler :roll::laugh:
Posted on Reply
#211
Animalpak
Sooo uhmm according to the dimensions of the card, water cooling it will be convenient since the waterblock will be the same dimension of the card... Less material = cheap ... NICE !
Posted on Reply
#212
RH92
Ok given this leak this is probably whats going on imo :

Posted on Reply
#213
Chrispy_
Aw shucks, that's a stupid-ass design.
Seriously, having the passively-cooled chunk of fins in the middle walled off like that ruins the whole thing. Not only does it remove those fins from active cooling duty, it also completely stifles the front fan.

This is not a high-TDP design - it's 200W max and the aesthetics really are all that Nvidia care about with this because cooling potential is roughtly halved by that single decision to wall off the central fins.

Dumb AF, avoid!

At least it confirms where the power connectors are going to go at last - and yes, there will be stupid wires to the PCB like on the 2060FE again (sigh...)

Posted on Reply
#214
Valantar
Chrispy_
Aw shucks, that's a stupid-ass design.
Seriously, having the passively-cooled chunk of fins in the middle walled off like that ruins the whole thing. Not only does it remove those fins from active cooling duty, it also completely stifles the front fan.

This is not a high-TDP design - it's 200W max and the aesthetics really are all that Nvidia care about with this because cooling potential is roughtly halved by that single decision to wall off the central fins.

Dumb AF, avoid!

At least it confirms where the power connectors are going to go at last - and yes, there will be stupid wires to the PCB like on the 2060FE again (sigh...)


Are you sure that isn't just a cut-out for the sides of the fan, like around the other fan? No fins on the top or bottom there after all. I'm starting to wonder if the cut-out in the passive fin stack is for an angled 8-pin connector or something.

Other than that, I entirely agree that the walled-off fins are a really dumb move. Seems to be motivated entirely by price, as it lets them use a simple aluminium extrusion for this part of the fin stack, but ... wow, it's so stupid. Wasted potential for sure.
Posted on Reply
#215
Chrispy_
Valantar
Are you sure that isn't just a cut-out for the sides of the fan, like around the other fan? No fins on the top or bottom there after all. I'm starting to wonder if the cut-out in the passive fin stack is for an angled 8-pin connector or something.
we're both wrong.
I'm wrong because if you go back to the first set of images, there's no cutout in the shroud there for power connectors. The thing I circled is probably just the fan blade clearance you mentioned.
You're wrong about the cutout in the passive fin stack, it's far too narrow, but you did make me take a closer look and it yielded results:



To me, those look like thick power delivery wires covered in black heatshrink. This is where the wired PCIe connectors meet the PCB I guess, which means the wires probably run under the fan hub and have a cutout on the end of the shroud, which hasn't been photographed yet and is likely the only place left for the PCIe connectors to be hiding.
Posted on Reply
#216
r.h.p
Vario
This cools really well but I don't think we'd ever see any manufacturer sell a card looking like this

actually this is what my old vega 64 needs ... :)
Posted on Reply
#217
Caring1
RH92
Ok given this leak this is probably whats going on imo :


Those look like injection molded plastic mock ups.
Posted on Reply
#218
BoboOOZ
Vario
This cools really well but I don't think we'd ever see any manufacturer sell a card looking like this

They could, they would just have to include a pci riser and a way to adapt the io part.

IMO they should do this, since modern GPU's are way more power-hungry than modern CPUs. Either this, or come with more water-cooled models.
Posted on Reply
#219
Chrispy_
Caring1
Those look like injection molded plastic mock ups.
Looks 100% like machined metal to me.

Everything from the paint scratches showing up as lighter material below, to the minor damage resulting in permanently bent metal where plastic would either spring back or snap off. Injection moulded is extremely unlikely - that requires extremely expensive investment dies to be made, not something that's ever done for prototyping. They'd probably 3D Print stuff for prototyping now, and the stuff we've seen so far is far too smooth to be 3D printed, even on the highest-end printer available.
Posted on Reply
#220
Valantar
Chrispy_
Looks 100% like machined metal to me.

Everything from the paint scratches showing up as lighter material below, to the minor damage resulting in permanently bent metal where plastic would either spring back or snap off. Injection moulded is extremely unlikely - that requires extremely expensive investment dies to be made, not something that's ever done for prototyping. They'd probably 3D Print stuff for prototyping now, and the stuff we've seen so far is far too smooth to be 3D printed, even on the highest-end printer available.
Machined? That sounds wrong, CNC'ing or otherwise machining a heatsink would be both wildly impractical, slow, and terribly expensive. Both end pieces (where the fans sit) are likely more or less standard stamped-out sheet metal crimped together at various points, with the cold plate and heatpipes pressed into the structure. The angled middle pieces look like bog-standard aluminium extrusions cut from a long bar. Nothing even remotely exotic about this, but it definitely isn't plastic either. Plastic would be shinier, have rounder edges, and the fins would definitely be less straight (those dense fin stacks have really thin fins).
Posted on Reply
#221
Chrispy_
Cold-rolling and stamping out those thin vanes around the fans is 'machining' and slicing the almumium extrusions is machining too. You're thinking CNC machined = machined, but it's just one tiny part of the 'machining' umbrella term. Pretty much if anything is cut down to size, has threads tapped into it, or is cut from a cast or extrusion, it's machined. That's why they're called machine shops, even if there's not a 3-axis CNC column drill in it.

For what it's worth, the alloy extrusions in the passive fins at the middle are actually CNC machined - the extrusions come off as a symmetrical slice and then the recesses like that one we can see for where the power wires to into the board are likely CNC machined out of the extrusion.
Posted on Reply
#222
Valantar
Chrispy_
Cold-rolling and stamping out those thin vanes around the fans is 'machining' and slicing the almumium extrusions is machining too. You're thinking CNC machined = machined, but it's just one tiny part of the 'machining' umbrella term. Pretty much if anything is cut down to size, has threads tapped into it, or is cut from a cast or extrusion, it's machined. That's why they're called machine shops, even if there's not a 3-axis CNC column drill in it.

For what it's worth, the alloy extrusions in the passive fins at the middle are actually CNC machined - the extrusions come off as a symmetrical slice and then the recesses like that one we can see for where the power wires to into the board are likely CNC machined out of the extrusion.
I wasn't under the impression that 'machining' was commonly used for mostly or fully automated high volume processes like standard heatsink production (which is after all mostly stamping sheet metal and threading that onto hestpipes), but that the term applied more to more manual processes or ones of a lower volume/higher complexity - definitely not limited to CNC milling, but obviously also including routing, turning, drilling, etc. Though I'm no machinist, so it is of course enitely possible that the colloquial meaning of the word I'm used to is inaccurate.
Posted on Reply
#223
Chrispy_
Splitting hairs and going OT at this point but yeah, anything that removes material from an oversized workpiece to bring it down to the final dimensions is classed as machining. The etymology of the word refers to the machinists themselves who worked on machines. When the word was originally coined, machinists used hand-tools exclusively.

Machining today refers to any subtractive fabrication process, material is removed rather than modified or assembled. There's no bending, deformation, or stretching of material so I guess the term can be applied to absolutely anything that cuts a workpiece down to size.

It's a broad term really but pretty much anything that's milled, cut, drilled, tapped, or turned is machined.
Posted on Reply
#224
Caring1
Chrispy_
Splitting hairs and going OT at this point but yeah, anything that removes material from an oversized workpiece to bring it down to the final dimensions is classed as machining. The etymology of the word refers to the machinists themselves who worked on machines. When the word was originally coined, machinists used hand-tools exclusively.

Machining today refers to any subtractive fabrication process, material is removed rather than modified or assembled. There's no bending, deformation, or stretching of material so I guess the term can be applied to absolutely anything that cuts a workpiece down to size.

It's a broad term really but pretty much anything that's milled, cut, drilled, tapped, or turned is machined.
Unless done by hand, if you really want to split hairs. ;)
Posted on Reply
#225
Chrispy_
Caring1
Unless done by hand, if you really want to split hairs. ;)
I dunno, something turned on a manual lathe is still machined.
Posted on Reply
Add your own comment