nVidia X80, X80ti and X80 Titan rumoured Specs

[Paragon_X]

According to videocardz the specs of the aforementioned GPUs have hit the web. I truly believe that on both sides this new generation will have the biggest leap in performance in the recent years. Also Judging on the Titan having HBM2 it means it will come later this year maybe Q4.



It also shows an new naming scheme for the upcoming GPUS naming them Geforce X


Source

[TrigrH]

lets add "nvidia confidential" to the end of our excel table so it seems more legit....

 

if they seriously call it Geforce X im going polaris,

[19_blackie_73]

Holy **** - so many cuda cores????!!!

[Sakkura]

1 minute ago, 19_blackie_73 said:

Holy **** - so many cuda cores????!!!

It's just doubled vs. Maxwell, which is nothing special for nearly 2 nodes of fab process upgrade.

[19_blackie_73]

Just now, Sakkura said:

It's just doubled

ehm ... I didn't expect that many

[ChrisxIxCross]

4 minutes ago, TrigrH said:

lets add "nvidia confidential" to the end of our excel table so it seems more legit....

 

if they seriously call it Geforce X im going polaris,

It's just called Geforce X because they don't know what naming scheme Nvidia will be using with the new gpus

[SeanAngelo]

i was confused at first with those series, i thought they were gonna release all of them. silly me 

 

MOAR CUDAAAAAAAAAAAAAAAAAAAAAAA

[Sakkura]

1 minute ago, 19_blackie_73 said:

ehm ... I didn't expect that many

In theory, if 28nm and 16nm were completely accurate representations of transistor dimensions on each node, then density would be able to scale up by a factor of just over 3x. Doubling is significantly less than that, which is probably a combination of the inaccuracy of "16nm", thermal limitations, and maybe a slightly smaller chip.

 

That's all assuming those specs are accurate anyway. But they're certainly plausible.

[VagabondWraith]

Give me a Titan NOW!!! 

 

If this comes out q4/q1 2017 that falls inline perfectly for my upgrade path.

[Master Disaster]

Hang on, let me just check my maths...

 

8,192 - 5,632 = 2560

2,560 is roughly half of 5,632

 

So the x80 is capable of roughly 1.5 times the Tflop output of the 980Ti?

 

Holy fucking shit!!!!

 

Edited March 17, 2016 by Master Disaster
Fixed typo

[TrigrH]

why would the cut down titan not use HBM, such bs

 

[Master Disaster]

1 minute ago, TrigrH said:

why would the cut down titan not use HBM, such bs

 

Because HBM 2 yields have been very low and theres just not enough of it in existence.

[Sakkura]

2 minutes ago, Master Disaster said:

Hang on, let me just check my maths...

 

8,192 - 5,632 = 2560

2,560 is roughly half of 5,632

 

So the x80 is capable of roughly have the Tflop output of the 980Ti?

 

Holy fucking shit!!!!

 

It says the X80 would be capable of 8.192 TFLOPS where the 980 Ti does 5.632 TFLOPS. Shit just got holy hand grenaded.

1 minute ago, TrigrH said:

why would the cut down titan not use HBM, such bs

 

Price? GDDR5 is probably a lot cheaper, and in better supply.

[TrigrH]

Just now, Master Disaster said:

Because HBM 2 yields have been very low and theres just not enough of it in existence.

yes but I don't think they can use a different memory system for the same die? look at FURY X, Fury, and the nano.

[Real_PhillBert]

That Titan actually looks worth buying this time around. DO WANT.

[Master Disaster]

Just now, TrigrH said:

yes but I don't think they can use a different memory system for the same die? look at FURY X, Fury, and the nano.

There is anohter, IMO much more important factor.

 

It is in Nvidias interest (and AMDs for that matter) to keep the status quo and not jump to a new die and new memory all in one go. If they can just shrink the die and reuse DDR memory then they can charge us for that, then in the future they can jump to HBM and charge us again for that instead of misisng a price point by offering both in one upgrade.

[jasonvp]

Sign me up for two Titans please.  Just make sure they don't have those God-forsaken DVI ports on them, please!

[Dabombinable]

4 minutes ago, TrigrH said:

yes but I don't think they can use a different memory system for the same die? look at FURY X, Fury, and the nano.

It'd be the same base architecture. As it is Fiji is just Tonga blown up and with HBM.

[CommandMan7]

So what i'm getting from this is that only their $1000 flagship will even come close to the performance of the Radeon Pro Duo, but even then fall significantly short? IMHO, if you're going to be spending $1000 or more on a graphics card you may as well go for the better performance of the Radeon Pro Duo for the higher cost.

[TidaLWaveZ]

Nvidia Geforce GTX X -> Nvidia Geforce GTX XX -> Nvidia Geforce GTX XXX -> Nvidia Geforce GTX XL

[Master Disaster]

6 minutes ago, Sakkura said:

Price? GDDR5 is probably a lot cheaper, and in better supply.

I think I heard that GDDR5X was in fairly short supply and production wasn't expected to increase significantly until the summer so although it will certainly be cheaper I dont think it'll be as cheap as we all think it will/should be.

[Dabombinable]

Just now, CommandMan7 said:

So what i'm getting from this is that only their $1000 flagship will even come close to the performance of the Radeon Pro Duo, but even then fall significantly short? IMHO, if you're going to be spending $1000 or more on a graphics card you may as well go for the better performance of the Radeon Pro Duo for the higher cost.

Note that single GPU solutions have less issues than multiple GPU solutions-even if they use the PCIe slots to communicate instead of the old (and retarded) bridge.

[Starelementpoke]

28 minutes ago, Paragon_X said:

According to videocardz the specs of the aforementioned GPUs have hit the web. I truly believe that on both sides this new generation will have the biggest leap in performance in the recent years. Also Judging on the Titan having HBM2 it means it will come later this year maybe Q4.



It also shows an new naming scheme for the upcoming GPUS naming them Geforce X


Source

Correct me if I´m wrong, but doesn´t the Fury X have 8.6 TFLOPS?

[TrigrH]

Just now, Dabombinable said:

Note that single GPU solutions have less issues than multiple GPU solutions-even if they use the PCIe slots to communicate instead of the old (and retarded) bridge.

let alone the 4gb of vram

[Master Disaster]

1 minute ago, TrigrH said:

let alone the 4gb of vram

Whats worrying here is the X80 only having 6GB, so does that mean the X70 will only have 4GB again? If so then that's really bad, the GTX970 is really starting to show how much 4GB of Vram is holding it back.

 

I was hoping for an 8GB X80 and a 6GB X70