Nvidia Announces GV100 Volta Based GPU, Built on TSMC 12nm FinFET with a Die Size of 815 mm^2

[DocSwag]

4 minutes ago, Princess Cadence said:

Makes me happy that I ignored the GTX1080ti in favour to keep saving money for the TITAN X (Volta).

Well you're probably gonna be waiting until next year April-June

[Princess Luna]

Just now, DocSwag said:

Well you're probably gonna be waiting until next year April-June

Not an issue, planning on making a huge jump, go from the 1080p ultra wide to 1440p ultrawide, I'll probably only have the money save up for both the monitor and GPU by then

 

Until then, I'm more than satisfied with my Maxwell Titan X still rocking ultras on 2560x1080p [:

[DocSwag]

Just now, Princess Cadence said:

Not an issue, planning on making a huge jump, go from the 1080p ultra wide to 1440p ultrawide, I'll probably only have the money save up for both the monitor and GPU by then

 

Until then, I'm more than satisfied with my Maxwell Titan X still rocking ultras on 2560x1080p [:

Outta curiosity, does your Titan x max out all games at that resolution@80hz or no?

[Carclis]

9 minutes ago, DocSwag said:

No wonder it's $15k...

 

The GV102 GPU that's gonna be for consumers is also gonna be insane... Judging by GV100 and GP100 vs GP102 it's gonna be around 600 mm^2...

Yeah...If it has 5120 cuda cores I think it's gonna crush the 1080ti performance.

[Princess Luna]

4 minutes ago, DocSwag said:

Outta curiosity, does your Titan x max out all games at that resolution@80hz or no?

I can't max every single thing on the utmost demanding ones, however with a little bit of tweaking it works good enough, I also tend to play third person games such as The Witcher 3 on my 40inch 1920x1080p60hz TV with gamepad, then I can max all out but first person shooters I only game on the monitor with keyboard and mice; here's my personal rig performance on Deus Ex Mankind Divided:

 

(unlisted video, no desire for self promotion).

[IceCold008]

This is good news for people like myself if its true. Which have bought a 10 series GPU. I dont expect to see volta until mid 2018 and a volta ti card until sometime late 2019. Gives me time to get good use out of my 1080ti card before or if i need to upgrade. So people saying they are unhappy or Bored with the 1080ti cards suck it up the card its a beast and everytime you buy a new GPU you no its going to get outdated at one point. Thats the nature of the beast you have to play there games just dont get miserable when it dont go your own way

[Dylanc1500]

This is pretty awesome. It is capable of running FP32 and FP64 instructions simultaneously due to having separate cores for each. Each SM having 16 FP32 cores, 8 FP64 cores in addition to 16 INT32 and two of the new Tensor cores. All capable of being simultaneously used. Those SMs are packed.

[DocSwag]

10 minutes ago, Dylanc1500 said:

This is pretty awesome. It is capable of running FP32 and FP64 instructions simultaneously due to having separate cores for each. Each SM having 16 FP32 cores, 8 FP64 cores in addition to 16 INT32 and two of the new Tensor cores. All capable of being simultaneously used. Those SMs are packed.

GPUs have always been able to do fp32 and fp64 at the same time, not new.

[Senzelian]

Cool, another one of those graphics thingies.

[Crossbred]

Don't you guys remember? Vega came out q2 2016. At least that's what they said, right?

Also, I love that nvidia can get hbm2 cards out before amd despite the fact that amd was the original promoter of it.

Also, VEGA -> EVGA. *x-files theme*

[Dylanc1500]

28 minutes ago, DocSwag said:

GPUs have always been able to do fp32 and fp64 at the same time, not new.

Sorry, I meant INT32 not FP64. For some reason my phone autocorrected to that. 

 

Also those Tensor cores are huge. This explains the die size.

[leadeater]

24 minutes ago, Crossbred said:

Don't you guys remember? Vega came out q2 2016. At least that's what they said, right?

Also, I love that nvidia can get hbm2 cards out before amd despite the fact that amd was the original promoter of it.

Also, VEGA -> EVGA. *x-files theme*

Neither has them out yet and AMD did actually announce HBM2 cards first, AMD MI25. Now it's a race for first to actually ship HBM2.

[DocSwag]

11 minutes ago, leadeater said:

Neither has them out yet and AMD did actually announce HBM2 cards first, AMD MI25. Now it's a race for first to actually ship HBM2.

P100 was announced before that though and it uses HBM2 

[leadeater]

1 minute ago, DocSwag said:

P100 was announced before that though and it uses HBM2 

Damn it you're right, I'll just leave in a pit of shame.

[Dylanc1500]

3 minutes ago, DocSwag said:

P100 was announced before that though and it uses HBM2 

Which the Quadro now uses.

[Guest]

17 hours ago, DoctorWho1975 said:

That's because AMD has/had a shitload of shit going on with Ryzen 7/5/3, the 5XX launch and Vega. You don't want announcements/releases trampling on each other, you'll get your Vega soon enough. 

AMD CPU and graphics divisions are "seperated". Radeon is handeled by Radeon Technologies Group, which has a seperate Markting team.

[DocSwag]

51 minutes ago, Dylanc1500 said:

Which the Quadro now uses.

No... Quadros are all based on GP102. Only GP100 based GPU out there AFAIK is the P100

[Dylanc1500]

5 minutes ago, DocSwag said:

No... Quadros are all based on GP102. Only GP100 based GPU out there AFAIK is the P100

No they released the Quadro GP100. It dropped SLI and uses a NVLINK connector.

[DocSwag]

2 minutes ago, Dylanc1500 said:

No they released the Quadro GP100. It dropped SLI and uses a NVLINK connector.

oh damn my bad didn't realize they did.

 

It also costs a metric crapton lol

[Lord Nicoll]

21 hours ago, The Benjamins said:

so the Flops IPC has not improved over pascal.

 

1080ti               V100

3584 CUDA vs 5120 CUDA

10.6 TFLOP vs 15 TFLOPS

42.8% more CUDA cores and it is 42% more FLOPS.

 

(note: boost clocks at very close)

I expect the boost clocks of the GeForce cards to be higher, since GPU compute cards don't need raw clock speed as much, power and thermals mean more to businesses and such.

[DocSwag]

1 minute ago, Lord Nicoll said:

I expect the boost clocks of the GeForce cards to be higher, since GPU compute cards don't need raw clock speed as much, power and thermals mean more to businesses and such.

1080ti has about 7% or so higher clocks, so if you factor that in I think there is a bit of a difference, albeit one that's not that huge.

 

EDIT:

 

Wait just realized that makes no sense.

 

TFlops is always cores*2*clock speed

 

It's not gonna improve. @The Benjamins

[Dylanc1500]

3 minutes ago, DocSwag said:

oh damn my bad didn't realize they did.

 

It also costs a metric crapton lol

Lol yep, but hey, it's only 8.9k a card. What I hope is that with Volta (consumer) they introduce NVLINK connectivity and drop SLI.

[Trixanity]

40 minutes ago, FTL said:

AMD CPU and graphics divisions are "seperated". Radeon is handeled by Radeon Technologies Group, which has a seperate Markting team.

Separate but still intertwined. They'll mess up the marketing push if they launch everything at once. That's why they delayed the RX500 series a week so Ryzen 5 could have the spotlight all to itself.

[DoctorWho1975]

4 hours ago, FTL said:

AMD CPU and graphics divisions are "seperated". Radeon is handeled by Radeon Technologies Group, which has a seperate Markting team.

 

Separate marketing team but why would you stomp all over another division with your release? You are nuts to think that these releases haven't been coordinated to take most advantage of a news cycle.

[VerticalDiscussions]

9 hours ago, Valentyn said:

-snip-

Omg lmao, its true xD!!!

 

Heres how hes acting towards other users in those comments:

Spoiler

1- I don't think you realise I just buttressed your snark with indisputable facts. Chill out.

2- Eh, my blunt charm's gotten me pretty far up to now.

And here is what someone had to say about him:

As far as getting universally hated, yes.