Does GTX 1660Ti 8G exist?

[chisaki]

I searched the vbios database and found the 1660Ti 8G vbios.

Is it possible for TU116 to have 8G RAM?

 

[Udyr]

Warning: You are viewing an unverified BIOS file.
This upload has not been verified by us in any way (like we do for the entries listed under the 'AMD', 'ATI' and 'NVIDIA' sections).
Please exercise caution when flashing it to your graphics card, and always have a backup.

Here's your answer.

Officially, it doesn't exist. Maybe someone modded one to have 8GB.

 

[chisaki]

According to the SKU 20, it might exist. The ES GTX 1660TI FE is 699-1G160-0021-700, which has SKU 21.

I think the Engineering Sample exists.

 

[eidairaman1]

Whats your reasoning for this? Bear in mind there are bogus cards out there

 

[chisaki]

For reference when search for Engineering Sample cards.

 

[arbiter]

given the gpu it 192bit memory bus means likely doesn't unless there was engineering sample that they were testing diff memory bus width likely 128bit.

 

[AusWolf]

given the gpu it 192bit memory bus means likely doesn't unless there was engineering sample that they were testing diff memory bus width likely 128bit.

Even that's unlikely. Why would they cut 1/3 of the memory bus for an engineering sample?

Given that the 192-bit bus only lets you have 6 memory chips, that's either 3 or 6 GB, I find it logical to assume that 8 GB cards don't exist.

 

[Onasi]

@AusWolf
I mean, TECHNICALLY it can be done with a wacky non-symmetrical memory config with chips having different parts of the bus in terms of bandwith dedicated to them, but the question would be - why would NV do that and on a budget SKU no less?

So no, OP, there is no 8G 1660Ti in the wild. The BIOS is probably a modded fluke.

 

[:D:D]

Given that the 192-bit bus only lets you have 6 memory chips, that's either 3 or 6 GB, I find it logical to assume that 8 GB cards don't exist.

Shouldn't that be the other way around, i.e. "6 memory chips only allows a 192-bit bus". IOW 8 chips would give a 256-bit bus and not only increase VRAM but bandwidth by 33% too.

 

[chisaki]

I think there are two ways.

1. 8x1G RAM chips, 256 bit like TU106 (2060s/2070).
2. 4x2G RAM chips, 128 bit like TU116 with doubled capacity RAM chip

 

[AusWolf]

Shouldn't that be the other way around, i.e. "6 memory chips only allows a 192-bit bus". IOW 8 chips would give a 256-bit bus and not only increase VRAM but bandwidth by 33% too.

No. Chip makers design their GPUs with a certain memory controller in them. That determines the memory configuration that you can use.

 

[chisaki]

In my opinion, anything may exist in engineering samples. They create 4G TU116 ES, if doubled the capacity, then we achieve an 8G 1660ti (or other name, using TU116 core).

No. Chip makers design their GPUs with a certain memory controller in them. That determines the memory configuration that you can use.

 

[AusWolf]

In my opinion, anything may exist in engineering samples. They create 4G TU116 ES, if doubled the capacity, then we achieve an 8G 1660ti (or other name, using TU116 core).

View attachment 363364View attachment 363365

Sure, you can use fewer memory chips, but then you're also not using part of the memory controller. Pretty pointless for an engineering sample, imo.

 

[eidairaman1]

In my opinion, anything may exist in engineering samples. They create 4G TU116 ES, if doubled the capacity, then we achieve an 8G 1660ti (or other name, using TU116 core).

View attachment 363364View attachment 363365

Why are the driver version and date blurred out?

 

[Onasi]

Why are the driver version and date blurred out?

The shader count, ROPs and TMUs are also wrong for any 1660 variant. Whatever this is (if it’s real), it’s no 1660/S/Ti I have ever seen.