Imagine 10 Nvidia Geforce GTX 1080s!

[App4that]

2 minutes ago, Dabombinable said:

Blower coolers are awesome.........

 

But those ZOTAC aren't, their open coolers. Trying to figure out how they get enough air.

[Dabombinable]

1 minute ago, App4that said:

But those ZOTAC aren't, their open coolers. Trying to figure out how they get enough air.

Remember the loud "turbines" in the front of the case? that's where the airflow comes from.

[App4that]

Just now, Dabombinable said:

Remember the loud "turbines" in the front of the case? that's where the airflow comes from.

Damn, if they move that much air no wonder the noise didn't bug him. That's crazy.

[handymanshandle]

1 hour ago, Yoo Song Won said:

Aren't they cheaper than Titan Xs?

Yeah they are but 7 or 8 of them is still quite a bit of money, for like 8 of them that'd be like $5280 if they were $660/ea

 

Anyways to your OP, I wouldn't want 10 1080s I'd rather maybe two or three R9 490Xs or 480Xs either one would be fine.

I'm just not that interested in Pascal because I'm hoping on Polaris being good. At least good enough for 1440p 60FPS high to ultra gaming. Basically I just want the 480/480X to perform about as well as the 390/390X and be cheaper than the 390X. Though the 390 doesn't seem to be as expensive as I thought it to be. It's like maybe a little more expensive than the 380X. Regardless I'm hoping to get 390X performance I could get away with a GTX 1070 or R9 480X.

[Dabombinable]

1 minute ago, App4that said:

Damn, if they move that much air no wonder the noise didn't bug him. That's crazy.

He actually replaced the originals with quieter and more bearable-but still loud fans. Now the fans in use are about as loud as my 20GB Seagate U5 and 20GB Samsung Spinpoint HDD. Which my brother has to deal with if he wants a gaming rig with a Gameport for my Mum's old joypad and joystick.

[Jaggsta]

1080 is only 25% faster than 980Ti in 3dmark benchmarks not really impressive. Once GTX 1080Ti or Titan comes out next year with HBM2 memory and more powerful GP100 series chip you will see double the speed of 980Ti or close to double.

[TTTufayl]

3 hours ago, Yoo Song Won said:

Do you know that 10 PCI slot motherboard Linus was making with the 980tis. Imagine that but with 1080s instead of 980tis!

 

Take a look at the benchmarks: http://www.tweaktown.com/news/52097/leaked-geforce-gtx-1080-benchmarks-show-dominating-titan-fury/index.html?utm_campaign=trueAnthem:+Trending+Content&utm_content=573715d804d3016fbff89ddd&utm_medium=trueAnthem&utm_source=facebook

nvidea will only support the 2 way sli so 4 way ur on ur own and 10 way mate you need to think through your life

[App4that]

1 hour ago, TTTufayl said:

nvidea will only support the 2 way sli so 4 way ur on ur own and 10 way mate you need to think through your life

Incorrect, the people who know are losing their minds because they want to smash that rumor, but the NDA keeps them from doing so. There will be up to 4 way SLI with Pascal. The reason for the 3 bridges  

 

And if you look these are not in SLI anyway.

[Big-G]

4 hours ago, Godlygamer23 said:

Technically, it'd be pretty powerful, but let me know how you'd use them all at once. If you had the ability to put them in SLI, good luck on your scaling and latency. 

DX12 mGPU? AFAIK doesn't need SLI or CFX...

 

4 hours ago, Atmos said:

I'd rather imagine 10x polaris 11 running in cfx in dx12 for that crazy performance.

Pascal was never really the new card to watch for, the real surprise is gonna be polaris and its performance in dx12.

 

All pascal brought was a die shrink and minor architecture changes. Still no support for a-synch, still no hbm 2, still no real new technology. Just more performance, and lower tdp like any die shrink will bring.

U huh?.... Yeah.... Mkay...

 

2 hours ago, App4that said:

How the shit do they keep them cool with as close as they are?

Blower style coolers, they the secret man...

 

2 hours ago, App4that said:

But those ZOTAC aren't, their open coolers. Trying to figure out how they get enough air.

True, those puppies will thermal throttle like a biatch... but the reference style one's will be fine.

 

 

[Godlygamer23]

29 minutes ago, GidonsClaw said:

DX12 mGPU? AFAIK doesn't need SLI or CFX..

You're making at least two assumptions there: Games that use DX12 will also support that feature, and the games will support that many cards in the first place. Remember, Multi-Adapter is a developer-enabled feature. They can choose to not use it at all if they wish. Just because the feature is available, doesn't mean it will be widely used. 

[Big-G]

1 minute ago, Godlygamer23 said:

You're making at least two assumptions there: Games that use DX12 will also support that feature, and the games will support that many cards in the first place. Remember, Multi-Adapter is a developer-enabled feature. They can choose to not use it at all if they wish. Just because the feature is available, doesn't mean it will be widely used. 

True, but it would be a way of allowing for these flights of fantasy

[Godlygamer23]

1 minute ago, GidonsClaw said:

True, but it would be a way of allowing for these flights of fantasy

I have to admit though, it'd be interesting to see what kind of performance would be theoretically achievable with that kind of setup. 

[Commander Llama]

Helluva folding machine, that's for sure.

[Scott17818]

5 hours ago, 0ld_Chicken said:

I believe they are limited to 2way sli, so... folding??

edit: seems to not be true!

as stated before it's not scalable... 10 1080s is not going to be 10x better than 1 1080. after 2 or 3 way sli the returns will not be worth it.

[BuckGup]

5 hours ago, 0ld_Chicken said:

I believe they are limited to 2way sli, so... folding??

edit: seems to not be true!

Like there is 10 way SLI

[BuckGup]

4 minutes ago, Scott17818 said:

as stated before it's not scalable... 10 1080s is not going to be 10x better than 1 1080. after 2 or 3 way sli the returns will not be worth it.

For SLI. But there isn't such a thing called 10 way SLI so the cards are being used for rendering and they are being used to 100% of their ability 

[0ld_Chicken]

1 hour ago, Scott17818 said:

as stated before it's not scalable... 10 1080s is not going to be 10x better than 1 1080. after 2 or 3 way sli the returns will not be worth it.

 

1 hour ago, SamStrecker said:

Like there is 10 way SLI

Never said there was lol, I said I'd sell 7 or 8 of them. They could be used for VM gaming though like 7gamers 1cpu, though that isn't sli obviously

[BuckGup]

20 hours ago, 0ld_Chicken said:

 

Never said there was lol, I said I'd sell 7 or 8 of them. They could be used for VM gaming though like 7gamers 1cpu, though that isn't sli obviously

oOOooOoOoo

[MadyTehWolfie]

On 5/15/2016 at 3:01 PM, 0ld_Chicken said:

I believe they are limited to 2way sli, so... folding??

edit: seems to not be true!

Nope Jayztwocents already got three for skankworks...skunkworks.../s /bad joke

[Scott17818]

I'm patiently waiting for my order.. lol.. hoping to get a founders edition 1080. saving my pennies.

[App4that]

12 minutes ago, Scott17818 said:

I'm patiently waiting for my order.. lol.. hoping to get a founders edition 1080. saving my pennies.

Why would you get the reference when a better performing EVGA or MSI will be cheaper, by 100 bucks?

[App4that]

3 minutes ago, rrubberr said:

It seems like the whole point of this would be compute, but Nvidia cards are absolute worthless trash at compute... so... what you're looking at is about a single Radeon Pro Duo.

The 1080 has over 11Tflop at 2.5GHz, which it will hit in AIO form. The Pro Duo, a dual GPU card has 16Tflop and costs over twice as much.

 

So what was that about Nvidia and compute again?

[App4that]

4 minutes ago, rrubberr said:

I explained very eloquently that that 11 teraflops was utter poo, so try again. Mind you that compute also requires double precision, which Nvidia lost after Fermi. Also, on a side note, Nvidia limits OpenCL performance at the driver level, slowing them down even further. It's nice to see an NVsheep, though.

Ah, time traveler. Now I see the problem. Could you do me a solid and tell me from the past that Moore machine won't be on the final please?

[Dabombinable]

6 minutes ago, rrubberr said:

I explained very eloquently that that 11 teraflops was utter poo, so try again. Mind you that compute also requires double precision, which Nvidia lost after Fermi. Also, on a side note, Nvidia limits OpenCL performance at the driver level, slowing them down even further. It's nice to see an NVsheep, though.

I expect that kind of response from a console peasant, not some one dealing with hardware.

[App4that]

1 minute ago, rrubberr said:

What does that mean? I don't usually have to ask for clarification on forum posts, but you didn't refute anything, and just marched right on with the NV warship.

The 1080 costs 600us and at 2.5Ghz which it's confirmed it hits is capable of over 11 Tflops, the Pro Duo at 1500us with two GPUs hits 16T flops.

 

Two 1080s is 1200us, 300us less than the Pro Duo. Two 1080s would give you 22 Tflop. 22>16 1200us<1500us

 

Get it yet?