gtx 1180 leaks and infofmation from wccftech

[Guest savagepain]

https://wccftech.com/nvidia-geforce-gtx-1180-specs-performance-price-release-date-prelimenary/

Quote

                           Specifications

According to the latest leaks, rumors and information provided to wccftech the GeForce GTX 1180 is powered by a 104 class GPU, codenamed GT104. The GPU measures around ~400mm², features 3584 CUDA cores, a 256-bit GDDR6 memory interface and 8 to 16 gigabytes of 16Gbps GDDR6 memory. The graphics card is expected to have a core clock of around ~1.6GHz and a boost clock of around ~1.8GHz. The TDP of this graphics card is unconfirmed to date, but is expected to be somewhere between 170-200W.

Quote

Performance

Peak FP32 compute performance is expected to be around 13 TFLOPS, depending on how high of a clock speed the graphics card will be able to hit and how often. This puts it slightly ahead of the existing GeForce GTX Titan Xp and GTX 1080 Ti. So you can expect around GTX Titan Xp performance or slightly better, but at significantly lower power consumption.

So tldr they are saying it will be a bit more powerful than a titan xp

Quote

Pricing and the release date are the most fluid parts of a product launch, so this could definitely change. But, you can expect the GeForce GTX 1180 to be released in the summer. All leaks and rumors point to a debut around the month of July.

So about july we can expect the new series of cards

[ItsMitch]

wccftrash so take it with a dumpster of salt 

[iRileyx]

Don't get your hopes up, it's wccftech claiming this lol.

 

[AlTech]

@savagepain Gotta love my infofmation. *Information

 

How can it have that many cores and still be 170w-200w? They're not using the Nvidia 12nm Fake FinFet to do that and GDDR6 is not that much more efficient. So something else is going on .

[TVwazhere]

@savagepain bravo for being forward with the WCCFTech source in the title. Keeps us grounded

[TheRandomness]

1.6GHz base and 1.8GHz boost according to the rumour? That seems a little low, does it not?

[Taf the Ghost]

It's a slightly shrunk (improved node) Titan Xp, with generational improvements and somwhat faster memory.

 

WCCFTech's "leak" is likely correct because it's right in the range of what's been expected. (It's not a full Node change at TSMC; that comes after Turing.) 

 

I'm going to guess it's $699 USD for AIB and $849 for the FE.

 

Only interesting question is what improvements come from the generational improvements.

[Master Disaster]

So GT104 confirms Turing then (if true of course)

[AlTech]

Just now, Taf the Ghost said:

It's a slightly shrunk (improved node) Titan Xp, with generational improvements and somwhat faster memory.

Which new node though?
 

Nvidia's custom 12nm process is a fake process based off of TSMC's 16nm and the only real change is that it allows larger die sizes. Efficiency is basically the same.

[Taf the Ghost]

1 minute ago, TheRandomness said:

1.6GHz base and 1.8GHz boost according to the rumour? That seems a little low, does it not?

It's about 100 Mhz base & OC clock increase at the Reference design. What people get on their OC is different than what is Reference.

[Taf the Ghost]

1 minute ago, AluminiumTech said:

Which new node though?
 

Nvidia's custom 12nm process is a fake process based off of TSMC's 16nm and the only real change is that it allows larger die sizes. Efficiency is basically the same.

I thought Pascal was on the 16nm while this was going to be on Nvidia's customized node at TSMC? They'd have a much larger shrink if this was on 7nm.

[Razor01]

5 minutes ago, AluminiumTech said:

Which new node though?
 

Nvidia's custom 12nm process is a fake process based off of TSMC's 16nm and the only real change is that it allows larger die sizes. Efficiency is basically the same.

 

 

Dude 12nm node has less layers than the 16nm, uses less voltage.  Its more efficient but not by much that is true.  Volta's efficiency comes from its architecture not form the node.

 

Its not a die size thing, nV used 12nm TSMC, to cut down costs and a little power consumption, not much.  Its a refined 16nm node, I wouldn't even call it a half node, its just tweaked.

 

 

[AlTech]

2 minutes ago, Taf the Ghost said:

I thought Pascal was on the 16nm while this was going to be on Nvidia's customized node at TSMC? They'd have a much larger shrink if this was on 7nm.

Pascal is based on 16nm.

 

Volta is based on their fake 12nm which is actually 16nm FinFet but with some very tiny modifications that allow larger dies.

[Taf the Ghost]

29 minutes ago, Razor01 said:

 

 

Dude 12nm node has less layers than the 16nm, uses less voltage.  Its more efficient but not by much that is true.  Volta's efficiency comes from its architecture not form the node.

 

Its not a die size thing, that nV used 12nm TSMC, they used it to cut down costs and a little power consumption, not much.  Its a refined 16nm node, I wouldn't even call it a half node, its just tweaked.

 

 

 

28 minutes ago, AluminiumTech said:

Pascal is based on 16nm.

 

Volta is based on their fake 12nm which is actually 16nm FinFet but with some very tiny modifications that allow larger dies.

Having trouble finding clear information on TSMC's 12nm FinFet node because they announced a "12nm" FD-SOI in the mean time.

 

https://www.fool.com/investing/2017/01/18/taiwan-semiconductor-mfg-co-ltd-confirms-12nm-tech.aspx

 

So it's 16FFC+ which they'll call 12FFC. 

 

The important bit is that Volta & Turing means Nvidia has fully split their design branches. Until this generation, they really were building the Big GPU, then cutting down the designs. The introduction of Tensor Cores means the designs are going to start getting pretty separated, which also means they can remove the things that don't effect gaming as much.

 

Still, looks more like 100-200 Reference Clocks and then generation improvements. Faster GDDR6 is probably as important to discussion as anything else. It's still mostly just a slightly faster Titan Xp for the 1080 Ti price.

[Guest savagepain]

1 minute ago, Taf the Ghost said:

 

Having trouble finding clear information on TSMC's 12nm FinFet node because they announced a "12nm" FD-SOI in the mean time.

 

https://www.fool.com/investing/2017/01/18/taiwan-semiconductor-mfg-co-ltd-confirms-12nm-tech.aspx

 

So it's 16FFC+ which they'll call 12FFC. 

 

The important bit is that Volta & Turing means Nvidia has fully split their design branches. Until this generation, they really were building the Big GPU, then cutting down the designs. The introduction of Tensor Cores means the designs are going to start getting pretty separated, which also means they can remove the things that don't effect gaming as much.

 

Still, looks more like 100-200 Reference Clocks and then generation improvements. Faster GDDR6 is probably as important to discussion as anything else. It's still mostly just a slightly faster Titan Xp for the 1080 Ti price.

yeah i see what you mean but still take it with a bottle of salt 

[Taf the Ghost]

Just now, savagepain said:

yeah i see what you mean but still take it with a bottle of salt 

I'd take it with a couple of bottles, but it fits within the information that's already been floating around. More interesting question is when the replacement for the 1060 happens, as that'll sell a lot more.

[Guest savagepain]

1 minute ago, Taf the Ghost said:

I'd take it with a couple of bottles, but it fits within the information that's already been floating around. More interesting question is when the replacement for the 1060 happens, as that'll sell a lot more.

yeah i have a 1060 would love to see the 1160 

[Razor01]

46 minutes ago, AluminiumTech said:

Pascal is based on 16nm.

 

Volta is based on their fake 12nm which is actually 16nm FinFet but with some very tiny modifications that allow larger dies.

 

Right now there is no one that has a full 12nm node.  You can't even look at nodes like this with transistor sizes, gate sizes, and what not all being different sizes now.  Intel is probably the only ones that really have close to all the components the size that the node is set at lol.

[Razor01]

20 minutes ago, Taf the Ghost said:

 

Having trouble finding clear information on TSMC's 12nm FinFet node because they announced a "12nm" FD-SOI in the mean time.

 

https://www.fool.com/investing/2017/01/18/taiwan-semiconductor-mfg-co-ltd-confirms-12nm-tech.aspx

 

So it's 16FFC+ which they'll call 12FFC. 

 

The important bit is that Volta & Turing means Nvidia has fully split their design branches. Until this generation, they really were building the Big GPU, then cutting down the designs. The introduction of Tensor Cores means the designs are going to start getting pretty separated, which also means they can remove the things that don't effect gaming as much.

 

Still, looks more like 100-200 Reference Clocks and then generation improvements. Faster GDDR6 is probably as important to discussion as anything else. It's still mostly just a slightly faster Titan Xp for the 1080 Ti price.

As far as rumors are concerned, that would be a 50% improvement at 30% drop in power consumption.

[Camoxide]

"GT104" 

 

Wouldn't it be GV104? So they're either rebranding the Volta architecture or this is a separate architecture. 

 

They've previously used "GT" for chips that were based on the Tesla architecture. 

[Taf the Ghost]

11 minutes ago, savagepain said:

yeah i have a 1060 would love to see the 1160 

No clue on Nvidia's roll out for the full scale of the 11xx series. Looks to be Q3 for the 1180, which probably means the 1180 & 1170 at launch, then about 4-6 weeks later the 1160. If it ends up being say mid-October for the 1180, then around Black Friday for the 1160. That's been Nvidia's release cadence going back several generations.

[AlTech]

4 minutes ago, Camoxide said:

"GT104" 

 

Wouldn't it be GV104? So they're either rebranding the Volta architecture or this is a separate architecture. 

 

They've previously used "GT" for chips that were based on the Tesla architecture. 

Yeah so this is either inaccurate or Nvidia being silly.

[Taf the Ghost]

1 minute ago, Camoxide said:

"GT104" 

 

Wouldn't it be GV104. So they're either rebranding the Volta architecture or this is a separate architecture. 

 

They've previously used "GT" for chips that were based on the Tesla architecture. 

Maybe. Though it's not Volta. Volta & Turing as definitely related, but their pretty far apart due to the Tensor Cores.

1 minute ago, Razor01 said:

As far as rumors are concerned, that would be a 50% improvement at 30% drop in power consumption.

I wonder if the split design branches let's them cut out a lot more of the design than we realize, as those are full-node improvement numbers, which I don't think their getting from generational improvements alone.

[AlTech]

1 minute ago, Taf the Ghost said:

Maybe. Though it's not Volta. Volta & Turing as definitely related, but their pretty far apart due to the Tensor Cores.

I wonder if the split design branches let's them cut out a lot more of the design than we realize, as those are full-node improvement numbers, which I don't think their getting from generational improvements alone.

We could see Tensor Cores in Nvidia's next gen. We'll just have to wait and see.

[DildorTheDecent]

There's no way nvidia is going to go full speed with GDDR6 right off the bat. Like they (and AMD) did with GDDDR5 they'll probably run it lower. Maybe around 12Gbps to start and go from there.

 

Wonder how well it overclocks too. Wonder if it takes volts and scales. Hopefully Samsung are back making it too. There GDDR5 was damn good with volts.

 

>Inb4 it's just Hynix and Micron.

That would succ.