Best 3090 model to watercool?

[TheNaitsyrk]

Waiting to buy two to SLI. 

 

Is Founders Edition decent this gen? Assuming 3080 is similar to 3090s models?

[emosun]

who is your question for? nobody here has one.......

[SavageNeo]

1 minute ago, TheNaitsyrk said:

Waiting to buy two to SLI. 

cant sli 3080's. onyl can sli 3090's. so better with single 3090 when you can buy them. 

 

founders is decent

[emosun]

2 minutes ago, SavageNeo said:

cant sli 3080's.

they aren't from the post

[jaslion]

Sli is dead now. Nvidia basically said we will stop supporting it. So just get a single 3090. The 3090 isn't out yet so we do not know which card works with waterblocks.

[Guest]

Founders have traditionally been the best ones to cool support wise since not everyone makes 3rd party PCB blocks and especially not right away.

 

That said, there weren't even 3rd party Titan class cards and the cards aren't out yet, so you will not have a useful answer for quite some time.

[SupaKomputa]

To watercool, i think you should buy the AIB model, as the FE has a different board layout may not be compatible with after market waterblock.

I haven't seen anyone attempting to add waterblock on the FE.

As for the difference between 80 and 90, i guess there will be no difference in board design apart from the SLI bridge.

[Guest]

2 minutes ago, jaslion said:

Sli is dead now.

SLI has been dead a while.

He's gonna do it anyway.

As am I. lol

[jaslion]

Just now, Den-Fi said:

SLI has been dead a while.

He's gonna do it anyway.

As am I. lol

It's been dead for a long time I'm aware but this is an official statement finally that it is dead.

[DildorTheDecent]

12 minutes ago, TheNaitsyrk said:

 

Reference or FE board. Doesn't matter what AIB you want to buy since the power and temp crippling is worse on Ampere than it was on Turing.

[emosun]

1 minute ago, Den-Fi said:

As am I. lol

 

[Guest]

Just now, jaslion said:

It's been dead for a long time I'm aware but this is an official statement finally that it is dead.

The statement says SLI profiles are going away and that they are focusing on working with devs to bake support directly into games. (Of course it's clever speak)

That said, I'm entirely certain he just wants two big ole GPU because reasons like anyone else who still does SLI lol.

[Guest]

2 minutes ago, emosun said:

 

 

I actually use multi GPU for reasons beyond "SLI" but sure.

That's a funny meme.

[Moonzy]

27 minutes ago, TheNaitsyrk said:

Is Founders Edition decent this gen?

think the FE vram was reported running hot on the cooler

but if the waterblock can cool it well, no prob

the power delivery was said to be good for FE, but i havent seen any in-depth PCB analysis* yet so im not too sure myself.

 

*  there are probably some out there, probably techpowerup or something 

 

edit: for 3080, anyway

3090 could be different

[SupaKomputa]

12 minutes ago, Den-Fi said:

I actually use multi GPU for reasons beyond "SLI" but sure.

That's a funny meme.

Just curious, outside gaming, if a program can use SLI and all the combined resource, say 24gbx2 vram?

[Guest]

2 minutes ago, SupaKomputa said:

Just curious, outside gaming, if a program can use SLI and all the combined resource, say 24gbx2 vram?

No, that's a full fat NVLink feature, but most multi GPU workloads don't run over SLI. They simply address the cards individually.

[SupaKomputa]

2 minutes ago, Den-Fi said:

No, that's a full fat NVLink feature, but most multi GPU workloads don't run over SLI. They simply address the cards individually.

but to run NVLink you need to plug the bridge on the sli enabled card isn't it?

[jaslion]

6 minutes ago, SupaKomputa said:

Just curious, outside gaming, if a program can use SLI and all the combined resource, say 24gbx2 vram?

Either nvlink is used (not the half assed version on the 3090) or the normal way is to just use the gpu's as their own entity. Some renderers give 1 frame to one card the next to the other, different tasks or just different sections of the frame.

[Guest]

Just now, SupaKomputa said:

but to run NVLink you need to plug the bridge on the sli enabled card isn't it?

What I mean by full-fat NVLink is that it's a Quadro only feature to be able to use the memory of both cards.

You can't do this with the SLI over NVLink that GeForce cards use. It's a segmented implementation.

[SupaKomputa]

1 minute ago, Den-Fi said:

What I mean by full-fat NVLink is that it's a Quadro only feature to be able to use the memory of both cards.

You can't do this with the SLI over NVLink that GeForce cards use. It's a segmented implementation.

ah i see, no wonder i never seen one. haha.

[Bad5ector]

Be careful FE and Reference designs are different boards. The first round of waterblocks from EK I think were for Reference design (AIB cards) and not FE cards.

 

Edit: Post #69.... giggity

[TheNaitsyrk]

I just want it for older games, and maybe rendering, benchmarking too.

 

That's it really.

[Guest]

32 minutes ago, TheNaitsyrk said:

I just want it for older games, and maybe rendering, benchmarking too.

 

That's it really.

Main PC:

CPU: Octodeca Core 10980XE @ 5.0Ghz 1.305V

 

Idk how, but I guessed that was the reason lol.

Though even if you had none, it's your money and your business lol.

[Tristerin]

Everyone always hating on the @TheNaitsyrk builds/desires - keep on letting me view that crazy setup, and high end hardware, I live vicariously through your posts as I cant afford that shit lol

[The Blackhat]

2 hours ago, SupaKomputa said:

To watercool, i think you should buy the AIB model, as the FE has a different board layout may not be compatible with after market waterblock.

I haven't seen anyone attempting to add waterblock on the FE.

As for the difference between 80 and 90, i guess there will be no difference in board design apart from the SLI bridge.

The FE cards get the most waterblocks, period. Always have, so your notion that aftermarket blocks are mostly for AIB cards is plain wrong. Though this gen FE doesn't get the reference design, so who knows what waterblock makers are doing this time around.