Is SLI worth it?

[mep916]

In my case, running two Ultras in SLi wasn't really worth it. It cost me an additional $700 to run that type of setup, and, even if performance (in FPS) had doubled (which it did, actually, in some cases), it still seems too high. Two GT's, however, could possibly be worth it. The only way to know if it's worth it to you, I guess, is to try it out.

 

[Geoff]

25.7 FPS Single // 46.7 FPS SLI

34 FPS Single // 46 FPS SLI

96 FPS Single // 149FPS SLI

40 FPS Single // 62.1 FPS SLI


Those are where SLI excels, however in almost all older games such as UT2004, CoD2, even some newer games such as CoD4, having SLI makes only a small improvement.

 

[mep916]

even some newer games such as CoD4, having SLI makes only a small improvement.

Yeah, I agree. I've everything on the highest settings with 4X AA and haven't noticed the difference from my SLi setup.

 

[hermeslyre]

Nope its all pre rendered, with the exception of the meta stuff that is done in benchmarks . Plus every consumer should know benchmarks never reflect real world performance in any thing. Just because you buy a new car that was benched from 0 to 60 in 4.5 seconds doesn't mean you are going to get that performance every time. Depends on what configuration, what gas you use, the weather, the driver, the road, the tires, so on and so forth. Just like everything else benchmarks are used for marketing a product.

I hate to use car analogies for computers but that one actually kind of fits.

You know, several benchmarking use ONLY Real world evaluation tests, see [H]'s recent article Benchmarking the benchmarks you can't just out and say they don't count. Your points on configuration are true, but most anyone who looks at benchmarks keeps this in mind, they aren't stupid. If you're coming from a good site like Anadtech or [H] Enthusiast, you read read the entire review, and check their real world benchmarks, then make an educated opinion based on it all. And it's rarely "this is fake, your guys are just tools of Nvidia and ATI, lolololol." Sounds spammerish.

 

[tlarkin]

You know, several benchmarking use ONLY Real world evaluation tests, see [H]'s recent article Benchmarking the benchmarks you can't just out and say they don't count. Your points on configuration are true, but most anyone who looks at benchmarks keeps this in mind, they aren't stupid. If you're coming from a good site like Anadtech or [H] Enthusiast, you read read the entire review, and check their real world benchmarks, then make an educated opinion based on it all. And it's rarely "this is fake, your guys are just tools of Nvidia and ATI, lolololol." Sounds spammerish.

I am not saying it does nothing, and I will come out and say that most journalists are journalists first and whatever they write about second. Tom's hardware has some very in depth articles that have proved most of my points about SLI, and RAID 0 which I have made on this forum.

My opinion still will be that they are both a waste of money for gaming, and until multi core GPUs come out they will still be a waste of money.

 

[hermeslyre]

I am not saying it does nothing, and I will come out and say that most journalists are journalists first and whatever they write about second. Tom's hardware has some very in depth articles that have proved most of my points about SLI, and RAID 0 which I have made on this forum.

Toms hardware is not what I would call a reliable site, maybe they used to be, but some of their "articles" and "tests" have been crap. Journalism first might apply to them. The journalists that are actual enthusiasts and gamers, actually value the position and respect they've been given, the ones that don't love journalism for journalism but for the chance to write about their passion or hobby.. These ones aren't so corruptible as you would think. Maybe they aren't as numerous as the former, maybe both these idea's are nothing more than uneducated opinion.

My opinion still will be that they are both a waste of money for gaming, and until multi core GPUs come out they will still be a waste of money.

Benchmarks don't cost money, and I've already agreed with you on SLI. I disagreed on

Benchmarks do not reflect real world performance

Because the word benchmark covers far too broad and diverse a subject to summed up so easily.

The performance increase in game is not 50 to 80% more, only in benchmarks.

Because of the above oversight, and because Real world benchmarks DO exist and do show improvements of that nature. Many times it's lower but it doesn't remove this fact. This phenomenon isn't (just) the result of increased bandwidth, It's the result of two GPU's working in tandem to render the screen. Sloppily, I already admitted. Nonetheless.

Nope its all pre rendered

^

 

[tlarkin]

meant SLI= waste of money, and I know benchmarks don't cost you anything to run, but nothing in life is free, those benchmark software apps costs people time and money to make

also every site you go to is hit or miss because some of them are actual tech people with writing skills and other are writers with some tech knowledge, I have seen crap articles and good articles on many sites.

and for the bandwidth comment, this is how I see it, and this is an example I am making up the numbers, but the method of how it works is right

One GPU can render/crunch/throughput 1 gig/second, a second GPU in parallel processing mode increases that bandwidth to 1.7gig/second, so it can throughput more data. Thus that is why you get more FPS in some cases if the game is coded to take advantage of it. Now, if developers (and if windows) would thread things more proper and use better memory management it would be even a greater performance increase. This is all good in theory if your bus can handle it and keep dishing it out. Also, when you start throughputting more data like that and if your system can't handle it or it bottle necks you have larger margin for error.

So, is FPS the definitive answer for performance increase? That is what I am asking because FPS do make guns run smoother, but at some point it becomes pointless to add more FPS.

Do you think that running 150 FPS over 90FPS makes a huge difference? A huge enough a human eye can and mind can comprehend?

 

[hermeslyre]

meant SLI= waste of money, and I know benchmarks don't cost you anything to run, but nothing in life is free, those benchmark software apps costs people time and money to make

Yes, but screw apps. I'm talking FRAPS ON, record frame rate. Do it in different places, average it, whatever the hell they do.

Edit: fraps is an App, guh, but I can save myself! In game console command to show FPS.

also every site you go to is hit or miss because some of them are actual tech people with writing skills and other are writers with some tech knowledge, I have seen crap articles and good articles on many sites.

That's when you play favorites, find a great site, stick with it till it craps itself or otherwise. Each article may not have have the same degree of authenticity and truthiness, but nothing in life is STATIC either.

and for the bandwidth comment, this is how I see it, and this is an example I am making up the numbers, but the method of how it works is right

One GPU can render/crunch/throughput 1 gig/second, a second GPU in parallel processing mode increases that bandwidth to 1.7gig/second, so it can throughput more data. Thus that is why you get more FPS in some cases if the game is coded to take advantage of it. Now, if developers (and if windows) would thread things more proper and use better memory management it would be even a greater performance increase. This is all good in theory if your bus can handle it and keep dishing it out. Also, when you start throughputting more data like that and if your system can't handle it or it bottle necks you have larger margin for error.

Remove the GPU's ability to crunch at a competent rate, while keeping the bandwidth strong, and see where you get. Impossible, I know, I'm just saying. the bandwidth is needed to send the data without bottlenecks. The extra muscle is needed to support the bandwidth. Hand in hand.

So, is FPS the definitive answer for performance increase? That is what I am asking because FPS do make guns run smoother, but at some point it becomes pointless to add more FPS.

Do you think that running 150 FPS over 90FPS makes a huge difference? A huge enough a human eye can and mind can comprehend?

This is relative to a monitors refresh rate. You will never see any frames above your VRR, 60 for most LCD. So anything above 60 is shot, regardless of whatever arguments over human eye restriction. Look at the benchies above. Most of the cards hover around 30-40. Great, IMO, but about in a couple months, when you're getting 15 in a certain game. the added 10 or so would be a lifesaver. This is the difference SLI makes. You can't say that the 2 cards are absolutely useless, just like I can't say the one higher end card is, because they're both rendering more frames than you or I have the ability to see.

P.S. Quote and respond is probably annoying. Don't tell me I know.