• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

3090 sli

criccio said:
I cannot believe people bought 2x 3090's for gaming. On what planet did that purchase make sense.
Click to expand...

Also how did someone get two when most of us can’t even get one? Unless they paid $2200 plus tax each on eBay. Makes you even more dumb. Lol
 
arestavo said:
It takes time to implement it in any API. With DX9 and DX11, Nvidia assisted game developers directly. With DX12 mGPU, Nvidia straight up said that game developers are on their own - and that's why we have so very, very, very few DX12/Vulkan games that support mGPU.

Time = money. Game devs (for the most part) don't care enough about the 1% or less of the gamer population to warrant the cost to them.
Click to expand...

Nvidia and AMD work with major developers regardless.

DX9/11 it kind of just worked if you took some steps to help the driver. There's not too much you can do to directly implement it and there's not really tools to control the process in detail - more like "don't do this because you'll either cause corruption or trash the performance gain." From what I've read, it's pretty very annoying to debug issues because of how hands off it can be.

DX12 and Vulkan, it's basically just distilled to "yeah you can just do whatever the fuck you want with however many GPUs you have, it's on you to figure it out." This mentality exists throughout the API, the driver can't/won't even force certain things like changing the pre-rendered frames setting. It does nothing from the control panel.
 
"mUh sLi iS bAd bRo" :rolleyes:

0Mr9Diz.jpg

HmZfuB9.jpg
 
arestavo said:
Be sure to let us know how it games! Would love to see some screen caps of your FPS and GPU utilization in all the games that support 3090 SLI so I can show you my single 3090 FTW3 Ultra numbers.

(For those not in the know, that was me pointing out what was already pointed out earlier in this thread)
Click to expand...

Will do! Just set it up yesterday and got some baseline numbers and slight OC. Still tweaking that of course. The cooling on these cards is phenomenal. The Titan RTX SLI setup I had would make the top card shoot up to 87C within a few minutes but the 3090 gets to about 68C at its hottest!

mnewxcv said:
What motherboard is that?
Click to expand...

It's the EVGA SR-3 Dark (using the W-3175X CPU @ 4.50Ghz. :)
 
Lucky75 said:
Thank you, but why to get a big screen 4K Oled monitor when I need to reduce a lot of things ? Before that summer I have 2x Titan X Pascal and performance was abosilutelly the same like RTX 3090 so where is big improvement ?
Click to expand...
I'm with you here. I don't understand the logic why people are happy to see SLI and Crossfire die. Yes, a single powerful GPU will always be the best solution, I don't think anyone would argue otherwise. The problem is when even the most powerful single GPU isn't powerful enough to meet whatever your needs or goals are. Having SLI in a working state at least gives people the option to increase their GPU power when you already have the best of the best. I know we're talking about the .1% of gamers but it simply comes down to an option that is no longer on the table.
 
FlawleZ said:
I'm with you here. I don't understand the logic why people are happy to see SLI and Crossfire die. Yes, a single powerful GPU will always be the best solution, I don't think anyone would argue otherwise. The problem is when even the most powerful single GPU isn't powerful enough to meet whatever your needs or goals are. Having SLI in a working state at least gives people the option to increase their GPU power when you already have the best of the best. I know we're talking about the .1% of gamers but it simply comes down to an option that is no longer on the table.
Click to expand...
I don't think anyone was happy to see SLI being abandoned by more and more developers, requiring more and more end-user configuration to get mediocre scaling for twice the money.

People are understandably happy to see Nvidia finally acknowledge that reality, and stop spending any effort on something virtually no devs implement and 0.001% of gamers utilize.

If it were as simple as "add another card, get ~175% performance" then it would likely be more popular, but it's almost never been universally that simple.
 
VanGoghComplex said:
I don't think anyone was happy to see SLI being abandoned by more and more developers, requiring more and more end-user configuration to get mediocre scaling for twice the money.

People are understandably happy to see Nvidia finally acknowledge that reality, and stop spending any effort on something virtually no devs implement and 0.001% of gamers utilize.

If it were as simple as "add another card, get ~175% performance" then it would likely be more popular, but it's almost never been universally that simple.
Click to expand...
in some sense I don't understand why it doesn't work at a driver or hardware level, it shouldn't be up to devs. They should simply pool all resources and that's it, pooled compute, pooled vram, and it's done. Obviously they can't seem to figure it out, but I don't really understand why. The resources are there.
 
mnewxcv said:
in some sense I don't understand why it doesn't work at a driver or hardware level, it shouldn't be up to devs. They should simply pool all resources and that's it, pooled compute, pooled vram, and it's done. Obviously they can't seem to figure it out, but I don't really understand why. The resources are there.
Click to expand...
Because a lot of developpers are lazy/incoompetent not to mention time constraint (hence all the game ready drivers which are mostly there to remedy a lot of the previous mentioned issues)

Why spend a lot of time optimising your game for SLI which hardly anyone uses when you have more then enough work getting it to run on the dozens of single card solutions.

Why buy 2 cards when SLI does not work decently and on the other hand if noone buys it who is going to bother supporting it. Nvidia does not care, they can't make cards fast enoughto keep them in stock.
 
mnewxcv said:
in some sense I don't understand why it doesn't work at a driver or hardware level, it shouldn't be up to devs. They should simply pool all resources and that's it, pooled compute, pooled vram, and it's done. Obviously they can't seem to figure it out, but I don't really understand why. The resources are there.
Click to expand...
I'm not really an expert on this. The last time I did multi-GPU it was with Crossfire, years ago, meaning by and large I needed to make all my own profiles for games. Just stepping through that showed me a little bit of what goes into it. There are a dozen different methods for divvying up graphical workload between two (or more) GPUs. Some are better for different games than others. Nvidia could probably throw a single boilerplate profile in, but there's no guarantee that I would be helpful. In my experience in many cases, though the frame rate was higher than single card, it was visually worse because the frame pacing caused it to look like a lower framerate than it was.
 
mnewxcv said:
in some sense I don't understand why it doesn't work at a driver or hardware level, it shouldn't be up to devs. They should simply pool all resources and that's it, pooled compute, pooled vram, and it's done. Obviously they can't seem to figure it out, but I don't really understand why. The resources are there.
Click to expand...
I think a lot of it has a lot to do with how loosely coupled multiple GPUs are compared to the compute units within the GPU. Not enough bandwidth and too high latency on the interconnects for the hardware/driver to split the load in a manner that's totally opaque to apps, particularly latency-sensitive ones.
 
I admit that's all true. It's not just about getting both gpus to do work, it's about the gaming experience as a whole. They could probably get a 2x performance metric at the cost of latency.
 
VanGoghComplex said:
Two cards doesn't magically double the power if the game isn't programmed to utilize two cards. Even then, even in the best scenarios, it doesn't scale 1:1 and it introduces other problems.

There's a reason SLI is dead. SLI sucked.
Click to expand...
I was a pretty big fan of my 8800GT SLI setup. Worked great! ...most of the time ha ha
 
I think I saw a 30 percent improvement with my nvidia 6800s in sli. It was the first gen of sli more or less and it let me run better image quality at the same resolution in battlefield 2 so I was happy.
I still have crossfire im my current rig but that is only because the second card had no other use. I dont game as much as I used to.

Sucks getting old and slow.
 
You must log in or register to reply here.
Back
Top