SLI Performance Scaling

[nicklmg]

NCIX: http://bit.ly/1sw4ugZ
Amazon: http://geni.us/33eG

 

NVIDIA doesn't have much solid information on their site about how performance scales in SLI... so we decided to run some tests. At 4K.

 

[Clyne]

Thanks Nick for the upload!

[Deletive]

Woah 4 way sli actually scales at 4k... is new.

Thanks Nick for the vid  

[Tech_Dreamer]

Whoever is in charge of creating the Thumbnail for Linus's recent videos , you deserve a raise  (or atleast the last piece of cookie from the jar Linus hides under his desk)   :lol:

[Phllips]

I cock blocked the "first people" again

[SkilledRebuilds]

As a LOT of cheaper motherboards for common consumers have x16 / x4 Linus should do a AMD performance scaling guide.

Crossfire

16x / 4x - Cheaper Boards = Limited Performance?

VS

16x / 8x - More dollars = More Performance?

 

Triple/Quad Fire

16x / 8x / 8x / 8x - The best bet?

VS

16x / 8x / 8x / 4x - No ill effects perceived compared to above^.?

 

As long as there are common 16x / 4x boards I think people should know more about the 4x slot and 'perceived loss' if any.

 

Remember, most of us THINK WE KNOW that it's not saturated, but when dealing with certain configs, it may be the best to make sure before you say so.

[Speedbird]

Thanks! You should try Crossfire too.

[qwertywarrior]

WHY does all your videos look too bright

its annoying and kills all the color and contrast

 

its like im watching a vid and took the control and turned up the brightness

[Narussistic]

WHY does all your videos look too bright

its annoying and kills all the color and contrast

 

its like im watching a vid and took the control and turned up the brightness

 

Yeah. And talk about the quality. This is what I see.

[megaant4]

You put the CPU of the test bench as the Core i7 4930K... I think it's meant to be the i7 5930K?....

[Eyal]

Yes, I would take two graphics cards and a nice watercooling loop over three graphics cards any day.

[MAXXHEW]

You guys are so lucky to be able to run those type of tests... most of us only dream of a set up like that. Keep up the good work LTT.

[Celmor]

Putting a Linus on a box, Linus still smaller than Luke.

[ImBleu]

 

> MFW i7-4930k + X99

 

comic sans because why not

[KamiKatze]

Am i the only one who is astonished by the power consumption? I would've liked full throttle torture test power consumptions but i guess gameplay power consumption will do. 970 SLI off my 550W PSU, here we come!

[Paul567]

Is it just me or does anyone else want to see standard deviation instead of minimum frame rates in these videos?

 

Using a minimum frame rate highlights a single result independant from the rest of the samples.

Standard deviation shows how much these samples deviate from the mean and will give you a better real world example of how much your frame rate will actually vary while gaming.

[oreozzy]

I for one would rather hear about whether 2 lower end cards are as good as or worse than one high end card.

Example: 1x GTA 980 ($800 AUD avg) vs 2x GTA 770 ($400 AUD avg).

In terms of power consumption and temps then obviously just one card, but what about from a purely performance pov?

Edit: my prices are from here:http://www.umart.com.au/umart1/pro/Products_list.phtml?id=10&bid=4&id2=247

[aahrg]

You said you were using an ax1500i but in the video there's an antec 1000w psu

 

EDIT: I realize now that it was put there because you were going to talk about the low power consumption of the cards

[jezza39]

I for one would rather hear about whether 2 lower end cards are as good as or worse than one high end card.

Example: 1x GTA 980 ($800 AUD avg) vs 2x GTA 770 ($400 AUD avg).

In terms of power consumption and temps then obviously just one card, but what about from a purely performance pov?

Edit: my prices are from here:http://www.umart.com.au/umart1/pro/Products_list.phtml?id=10&bid=4&id2=247

 

In a fully optimised game, two lower end cards will generally outperform a similarly priced higher end card:

http://www.guru3d.com/articles_pages/geforce_gtx_760_sli_review,16.html

 

But in games where SLI isn't optimised, you get sweet fa from the second card and the performance tanks.

 

O/T: PCCG > Umart

[aahrg]

I for one would rather hear about whether 2 lower end cards are as good as or worse than one high end card.

Example: 1x GTA 980 ($800 AUD avg) vs 2x GTA 770 ($400 AUD avg).

In terms of power consumption and temps then obviously just one card, but what about from a purely performance pov?

Edit: my prices are from here:http://www.umart.com.au/umart1/pro/Products_list.phtml?id=10&bid=4&id2=247

Go with the single high end card. You might get slightly less performance in SOME games than with the dual 770s, but some games don't have SLI support, or horrible scaling and i'd take the slightly lower average framerate over the headache of tweaking config files just to make it playable.

[sherwind2010]

Can I run GTX 980 SLI in my computer??

 

2600K CPU

8GB 2.4GHz RAM

Asus P8Z68-V PRO/GEN3 mobo

850W XFX PSU 

[Jumper118]

surely the minimum fps being so low is due to something else slowing the cards down with 3 and 4 way sli, because the minimums were about the same. i bet if they had the cpu at 5ghz those minimum frame rates would go up.

[deadbc77]

Thanks! You should try Crossfire too.

He only does Nvidia Cards now.(just kidding)......but kinda seems that way

[cthomas1489]

Linus, out of curiosity, is the test bench using a 4930k or did you mean 5930k? You mentioned a X99 motherboard.

[cthomas1489]

Can I run GTX 980 SLI in my computer??

 

2600K CPU

8GB 2.4GHz RAM

Asus P8Z68-V PRO/GEN3 mobo

850W XFX PSU 

 

You would more than likely need a decent OC to keep the bottleneck minimal but it'll probably still bottleneck slightly. Now, will the bottleneck be noticeable in real world gaming? I'm not sure but you could always buy the cards and upgrade your cpu/mb at a later date.