Question 'Ampere'/Next-gen gaming uarch speculation thread

Page 77 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Ottonomous

Senior member
May 15, 2014
559
292
136
How much is the Samsung 7nm EUV process expected to provide in terms of gains?
How will the RTX components be scaled/developed?
Any major architectural enhancements expected?
Will VRAM be bumped to 16/12/12 for the top three?
Will there be further fragmentation in the lineup? (Keeping turing at cheaper prices, while offering 'beefed up RTX' options at the top?)
Will the top card be capable of >4K60, at least 90?
Would Nvidia ever consider an HBM implementation in the gaming lineup?
Will Nvidia introduce new proprietary technologies again?

Sorry if imprudent/uncalled for, just interested in the forum member's thoughts.
 

Karnak

Senior member
Jan 5, 2017
399
767
136
Well, since even GDDR6X is now official I think all Igor said so far was true. Will probably be the same for his rumoured up to 350W TDP. Guessing that's due to Samsung's bad 8nm LPP in terms of higher clock speeds.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
That's the back side of the PCB. Looks like 22 GB on that example, double-sided memory coming to GeForce, and GDDR6X 19-21 GT/s only coming in 1 GB packages.

They have front and back sides of the PCB. There is very clearly no places for memory on the backside of the board. There is 11 modules on the front.
 

Bouowmx

Golden Member
Nov 13, 2016
1,147
551
146
They have front and back sides of the PCB. There is very clearly no places for memory on the backside of the board. There is 11 modules on the front.
NVIDIA-GeForce-RTX-PCB.jpg

Because front-side of graphics card has the visible screws to the heatsink that is still attached, no visible VRM components, and just the 8-pin connector solder points, not the connector itself. It's the back side.
 

Bouowmx

Golden Member
Nov 13, 2016
1,147
551
146
Ampere to have 2X FP32 units per SM compared to Turing, this accelerates RT performance greatly in combination with RT cores improvements.
A weird rumor flying under the radar.
I thought doubled FP32 would mean:
7 GPC * 6 TPC * 2 SM * 128 cores = 10752 cores

kopite7kimi thinks it's a bit more nuanced:
SM would have 128 FP32 + 64 INT32 ALUs, compared to Turing 64 FP32 + 64 INT32.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
View attachment 27970

Because front-side of graphics card has the visible screws to the heatsink that is still attached, no visible VRM components, and just the 8-pin connector solder points, not the connector itself. It's the back side.

I missed the heatsink behind the card on that image. I took it for a mirrored front side since it has a GPU sitting there. And the pads for the vregs and power connectors and such are there.

This image very clearly has no memory on the backside, so maybe we are seeing two different PCBs?

EDIT: Looking again, I am quite sure those first two images are of the front side, the heatsink is there behind it, but not attached. The inner screws attach to the backplate.
1597413204315.png
 

Bouowmx

Golden Member
Nov 13, 2016
1,147
551
146
I missed the heatsink behind the card on that image. I took it for a mirrored front side since it has a GPU sitting there. And the pads for the vregs and power connectors and such are there.

This image very clearly has no memory on the backside, so maybe we are seeing two different PCBs?

EDIT: Looking again, I am quite sure those first two images are of the front side, the heatsink is there behind it, but not attached. The inner screws attach to the backplate.
View attachment 27971
I think this pic was included as a reference to the current NVIDIA RTX 2080 Ti. There are no solder points for memory toward the I/O bracket, like the current 2080 Ti.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
I think this pic was included as a reference to the current NVIDIA RTX 2080 Ti. There are no solder points for memory toward the I/O bracket, like the current 2080 Ti.

No 2080Ti's have three 8pin connectors on them that I know of? Maybe some super high end OC model?

But even then, those first two images clearly show the GPU chip, that wont be visible on the back.
 

Bouowmx

Golden Member
Nov 13, 2016
1,147
551
146
Wow, 384bit 21Gbps would be what, over 60% more than the 2080 ti? Is the 3090's memory bandwidth overkill, or is it really going to be that strong of a card?
Overkill? maybe. But at least GDDR6X helps turn back the race of core speed vs memory speed a bit.

If calculating the ratio of FLOPS (at rated boost frequency) to bandwidth:
GeForce GTX 780 Ti: 0.063
1080 Ti: 0.042
2080 Ti: 0.046
3090 (1.7 GHz): 0.056
 

beginner99

Diamond Member
Jun 2, 2009
5,233
1,610
136
12GB in a high-end part is piss-poor given we've had 11GB in consumer space for over 3 years. Especially considering this thing will almost certainly be horrifically overpriced.

Can't have people to "AI" on cheapo GeForce hardware. You need to shell out even more for that to get you that needed memory. That certainly plays a role plus the fact 12Gb is more than enough for gaming right now and they already seem to have issues with power use.
 

Bouowmx

Golden Member
Nov 13, 2016
1,147
551
146
No 2080Ti's have three 8pin connectors on them that I know of? Maybe some super high end OC model?

But even then, those first two images clearly show the GPU chip, that wont be visible on the back.
videocardz.net tracks all the partner cards: https://videocardz.net/nvidia-geforce-rtx-2080ti
The RTX 2080 Ti is a COLORFUL NEPTUNE: https://www.expreview.com/68821.html

There is another image, a close-up without mosaic. It's covered with an Intel CPU. i'm still on the position this pic is part of the back side.
Ampere-GPU-covered.jpg
 

jpiniero

Lifer
Oct 1, 2010
15,223
5,768
136
I missed the heatsink behind the card on that image. I took it for a mirrored front side since it has a GPU sitting there. And the pads for the vregs and power connectors and such are there.

This image very clearly has no memory on the backside, so maybe we are seeing two different PCBs?

EDIT: Looking again, I am quite sure those first two images are of the front side, the heatsink is there behind it, but not attached. The inner screws attach to the backplate.

Does seem like the FE has a fan on both sides.
 

Tup3x

Golden Member
Dec 31, 2016
1,086
1,084
136
I missed the heatsink behind the card on that image. I took it for a mirrored front side since it has a GPU sitting there. And the pads for the vregs and power connectors and such are there.

This image very clearly has no memory on the backside, so maybe we are seeing two different PCBs?

EDIT: Looking again, I am quite sure those first two images are of the front side, the heatsink is there behind it, but not attached. The inner screws attach to the backplate.
View attachment 27971
I'm pretty sure that it shows two different cards. The first one has Intel cpu there for some reason. Also check the pcie connector positions.

This is probably going to be Fermi 2.0 (at least heat and power use wise).
 

BFG10K

Lifer
Aug 14, 2000
22,709
2,996
126
Untitled.jpg
 

Saylick

Diamond Member
Sep 10, 2012
3,531
7,858
136
videocardz.net tracks all the partner cards: https://videocardz.net/nvidia-geforce-rtx-2080ti
The RTX 2080 Ti is a COLORFUL NEPTUNE: https://www.expreview.com/68821.html

There is another image, a close-up without mosaic. It's covered with an Intel CPU. i'm still on the position this pic is part of the back side.
View attachment 27973
This is a wild guess but wasn't there speculation of Nvidia using an off-die traversal co-processor to accelerate RT performance? I wonder if the leaker intentionally hid the backside of the GPU because that's where such a co-processor, if it exists, would go, and perhaps the 3x8 pin power isn't necessarily because the GPU is power hungry but because it needs to also feed power to a separate die.
 

MrTeal

Diamond Member
Dec 7, 2003
3,614
1,816
136
videocardz.net tracks all the partner cards: https://videocardz.net/nvidia-geforce-rtx-2080ti
The RTX 2080 Ti is a COLORFUL NEPTUNE: https://www.expreview.com/68821.html

There is another image, a close-up without mosaic. It's covered with an Intel CPU. i'm still on the position this pic is part of the back side.
You know it's the back side just by looking at the the position of the PCIe edge connector even outside of any other parts. The only way that could be the "front" of the PCB is if nvidia flipped it around and mounted the die on the opposite side of every other GPU.

An aside, I love how they add the Intel CPU to cover things, and then blur out the CPU details, for extra obfuscation.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
You know it's the back side just by looking at the the position of the PCIe edge connector even outside of any other parts. The only way that could be the "front" of the PCB is if nvidia flipped it around and mounted the die on the opposite side of every other GPU.

An aside, I love how they add the Intel CPU to cover things, and then blur out the CPU details, for extra obfuscation.

nVidia didn't take the images. Its also semi common for images to get mirrored on cell phones, or to make them not match image searches.

Also, Micron lists the card as having 12GB of ram, and each chip as 8Gb (1GB). Which would make it odd to have the memory on the backside of the board. Unless this was maybe done for cooling, to keep the RAM away from the GPU and VRMs.