Why I will NEVER go back to crossfire or sli!

[goldsauce]

Just a couple of hours ago I was comparing what happens if I disable crossfire in the catalyst control panel and I found that my god damn games run much smoother and better fps when crossfire is DISABLED! I was shocked at first and now I am furious, because 80% of all of my games work better with crossfire disabled, so now I have taken out my other 7850 and it will be up on ebay tomorrow, let me just say this to anyone looking at getting crossfire, don't get your hopes up! Sure maybe 7850s in crossfire run better on BENCHMARKS but only on the damn synthetic benchmarks you get crazy 100% scaling but once you go in a damn game it turns into rubbish.

 

[Fulgurant]

No, its your uneducated and ignorant opinion.
Crossfire/SLI is not inherently flawed and wont lower your performance, if you meet the minimum standard required. You didn't research what you would need to successfully hold a Crossfire/SLI array and you got burned. Dont turn others way from a legitimate option because you had a bad experience stemming from your own mistakes.

Agreed on all counts, but I think this thread does demonstrate that SLI/Crossfire isn't worth the trouble most of the time. At the extreme high end? Sure, because after a point there's no other option to increase performance.

In the mid-high-end range? Maybe, if two cards of a given (combined) spec are significantly cheaper than the single-card analogue -- and if (as you say) you plan out your configuration carefully.

For every other situation? A single card is almost always the better option -- and perversely enough, that's true in part because a single card leaves you space to add a second later, provided your configuration can support it.

 

[zdbc13]

Very good point, fulgurant. I see sli/crossfire as the domain of high end, radical builders who are willing to pay big bucks for a system and get the most out of it. I don't think low to mid range systems are a good fit for sli/crossfire. Some very good information all through this post. Thanks to all for your ideas and discussion. And thanks for keeping it clean with minimal name calling!!!!! Have a good one...

 

[iam2thecrowe]

Another anecdotal experience that is in complete contrast w/ reality. Upstairs I can hear games being played on the following:

Son No. 2 - Single 580
Son No. 3 - Twin 560 Ti's

Two 560 Ti's (Asus DCII TOP) for $400 outperform the single 580 (EVGA FTW) by 40% .... a mark which is echoed by dozens of published reviews.

Simple Rules of SLI / CF

1. No sense going that way if the rest of your system will bottleneck it.
2. Microstutter is NOT on issue with mid-range ==> high range cards ($200 and up).
3. No more than 2 cards in CF.
4. Power Supply must be beefy enough and must supply stable voltages (1% variation) if you want the highest overclocks.
5. SLI / CF options with mid range cards consistently have a lower cost per frame that the top tier single card solution.

actually microstutter IS an issue with high end cards, its just that normally the FPS is so high you dont notice it, or your using vsync which is like a frame rate cap which abolishes microstutter, so long as you dont dip below that cap. Also, Nvidia cards dont suffer such poor frame latency so microstutter is less noticeable. Also, a 3rd card in crossfire actually reduces microstutter significantly, toms had an article about this a while ago. I guess AMD decided when a third card waws present they needed some algorithm to sync the frames better from 3 cards, but you also dont see as much performance boost from the 3rd card.

I think in the OP's case:
1. CPU bottleneck in many games, the i3 aint enough to push those cards and keep the minimum frame rate up to avoid stuttering.
2. a possible wrong/corrupt installation, you should never see a decrease in FPS even if you are cpu bottlenecked just because crossfire is enabled. Maybe a motherboard issue also.
3. Some games do run worse with crossfire, this is not avoidable it is up to the game devs to fix or AMD to fix the driver.
4. if you want crossfire you need to be a tweaker, experiment with settings, and use aftermarket tools to help. If you want things to be straight forward with no issue, get a single card.

A while ago i upgraded my single 6850 to crossfire 6850's, it was only a month before i sold them due to microstutter/performance issues. A few games ran absolutely fine, i always saw an FPS increase over a single card, but some games the stuttering was all too obvious, even after playing with radeon pro fps caps and different crossfire methods. In some games changing the crossfire method in radeon pro removed the stuttering all together. I also found the 1gb vram was not enough to run games at the settings i wanted to. So i wouldn't say crossfire isnt for everyone, but i would say it isnt for most people that want trouble free gaming. Its best left for those that already have the fastest single card an want extra performance for more monitors or something. In the end i gave up and replaced them with a single gtx660, went with nvidia due to some articles on techreport about them having lower frame latency and smoother gameplay than the AMD equivelant, weather its true or not i didnt take the risk. Now i enjoy smooth trouble free gaming.

 

[Tom Burnqest]

I had the same issue when trying CF back in 2009 then I returned my mobo as it was only PCI-E x16 x4 for a PCI-E x16 x16 board and all problems went away. CF/SLI is not bad it's just that people like the OP give it a bad name because they are ignorant to what parts they need beforehand to make it work correctly. The problem here people is between OPs keyboard and chair and not AMD Crossfire.

 

[lowcountrysmoke]

I agree. I have the same mobo and it's only built for one gpu..yes,it "can" crossfire but there is only one 3.0 socket. I only use a single card so it i perfect for me., if you want to crossfire you need a mobo designed for that.

 

[goldsauce]

hahaha finally someone said it.... i had 2x 7870's running in z77 extreme6 and benching 14000+ on 3dmark... But getting drops to 30fps in golf and other maps.... My cpu is i5-3570k @ 4.6 and there is ABSOLUTELY no botteneck. Tried all drivers, Disable CF and what do you know... lowest drop is 50... no stuttering, Smoother gameplay, Just less fps when looking at a wall... ROFL to CrossFire !!!!!!!!! I RMA's both cards and got a 680 for $11... Bench's like half of what i used to... RUNS A MILLION TIMES BETTER IN EVERYWAY. Thanks for this post !!!!! ROFL

I might go for the 7970 once the 8000 series comes out so I can get it for like $250.

 

[goldsauce]

If I end up getting a i5 or i7 and then getting a 7970 will that balance my system better?

 

[bigshootr8]

It all depends on what you get if you get a i5 3470 or above I would say yes you would have a more balanced system.

 

[MatildaPersson]

If I end up getting a i5 or i7 and then getting a 7970 will that balance my system better?

i7's have hyperthreading, which makes no difference for gaming -- zero. That's the only real difference between an i5 and an i7.

The i5 3570 or 3570K is a fair bit cheaper and will do the exact same job for gaming. The only reason to get an i7 is if you do things like video editing, sound engineering and graphic design. Video games and general use will show zero gain going from an i5 to an i7.

 

[bystander]

i7's have hyperthreading, which makes no difference for gaming -- zero. That's the only real difference between an i5 and an i7.

The i5 3570 or 3570K is a fair bit cheaper and will do the exact same job for gaming. The only reason to get an i7 is if you do things like video editing, sound engineering and graphic design. Video games and general use will show zero gain going from an i5 to an i7.

It may not make a difference very often, but hyperthreading does a few games performance by up to 10-15%.

 

[MatildaPersson]

hahaha finally someone said it.... i had 2x 7870's running in z77 extreme6 and benching 14000+ on 3dmark... But getting drops to 30fps in golf and other maps.... My cpu is i5-3570k @ 4.6 and there is ABSOLUTELY no botteneck. Tried all drivers, Disable CF and what do you know... lowest drop is 50... no stuttering, Smoother gameplay, Just less fps when looking at a wall... ROFL to CrossFire !!!!!!!!! I RMA's both cards and got a 680 for $11... Bench's like half of what i used to... RUNS A MILLION TIMES BETTER IN EVERYWAY. Thanks for this post !!!!! ROFL

I hate to tell you, but you're doing something wrong. Crosfire and SLI don't exist for benchmarking; they exist because some people want maximum performance and two, three or four GPUs are better than one. This is a fact. This is also the reason why the GTX 690 and the 7990 are the best performing cards you can buy; both are dual GPU cards, meaning they are SLI and Crossfire -- the former being two GTX 680's and the latter being two 7970's.

Don't blame the technology on personal failures. My two GTX 670's do their job nicely 😉

 

[MatildaPersson]

It may not make a difference very often, but hyperthreading does a few games performance by up to 10-15%.

Can you show any evidence of this? I've yet to see any tests that demonstrate any tangible difference at all.

As an aside, hyperthreading generates a surprising amount of heat. Turning it off will result in a cooler, higher overclock.

 

[bystander]

Can you show any evidence of this? I've yet to see any tests that demonstrate any tangible difference at all.

As an aside, hyperthreading generates a surprising amount of heat. Turning it off will result in a cooler, higher overclock.

I've run benchmarks in Metro 2033 and consistently showed 4-5 FPS higher with hyperthreading on with average FPS around 40-45 (using the built in benchmark).

You should also be able to Google Resident Evil 5 and see that it also gains about 10% in performance. Planetside 2 benchmarks show very noticeable improvements with the i7, though the benchmarks I saw on this didn't specifically test with and without hyperthreading.
http://forums.videocardz.com/topic/188-planetside-2-performance-test/

 

[bigshootr8]

I feel though after a i5 3570k that there are diminished returns in terms of the amount you pay for the higher priced CPU's

 

[bystander]

I don't disagree with that. Just pointing out there is "some" difference.

 

[MatildaPersson]

I've run benchmarks in Metro 2033 and consistently showed 4-5 FPS higher with hyperthreading on with average FPS around 40-45 (using the built in benchmark).

You should also be able to Google Resident Evil 5 and see that it also gains about 10% in performance. Planetside 2 benchmarks show very noticeable improvements with the i7, though the benchmarks I saw on this didn't specifically test with and without hyperthreading.
http://forums.videocardz.com/topic/188-planetside-2-performance-test/

Eh. I'm not sure about that. Benchmarks are different from actual gameplay.

But as you and bigshootr point out, if there is any consistent difference, it's much too marginal to bother with the i7 variant of the Ivy series. Unless you just happen to have loads of money. Or really, really want one or two frames in the 1% of games that will see a difference lol.

As always, GPU >>> CPU for gaming. Put the money where it counts.

 

[bigshootr8]

Agreed typically you will feel it most with a GPU upgrade over a CPU upgrade granted you have a decent CPU i.e., Intel i5 3570k

 

[bystander]

Eh. I'm not sure about that. Benchmarks are different from actual gameplay.

But as you and bigshootr point out, if there is any consistent difference, it's much too marginal to bother with the i7 variant of the Ivy series. Unless you just happen to have loads of money. Or really, really want one or two frames in the 1% of games that will see a difference lol.

As always, GPU >>> CPU for gaming. Put the money where it counts.

I know that RES5 does gain 10% more FPS in game play, and Planetside 2 does too (they may have benchmarked it, but they did so with normal game play).

My point is that you said it made no difference -- zero. You were being very final, and you were wrong on that.

While I won't argue that it rarely helps, and an i5 is a better value, I do disagree that it makes absolutely no difference.

 

[goldsauce]

You guys saying my CPU was a bottleneck don't understand that I played many games that were not CPU Heavy and I still performed better with 1 7850 and not to mention that my i3 3220 is about equal to a i5 2400 and slightly better than a i5 760 in performance although I am planning on getting a good i5 once I feel my CPU is holding me off by a fair amount.

i5 2400:

http://www.anandtech.com/bench/Product/677?vs=363

i5 760:

http://www.anandtech.com/bench/Product/677?vs=191

 

[manofchalk]

Your PCI-Express bandwidth was the bottleneck, not the CPU.

Imagine you have a road 32 lanes wide and you have 32 cars driving abreast, each in a lane. Say that every second 32 cars can travel down this road.
Suddenly reduce that road to four lanes and you still have the same number of cars to get through, you can only get 4 cars through a second. Obviously everything is going to slow down because the road isnt as wide.

That is what you have done by putting in a 7850 on a PCI-2 4x slot, effectively limited the width of the road that the graphics cards can use to communicate with the CPU. That is why you have horrible performance in Crossfire.

 

[goldsauce]

Your PCI-Express bandwidth was the bottleneck, not the CPU.

Imagine you have a road 32 lanes wide and you have 32 cars driving abreast, each in a lane. Say that every second 32 cars can travel down this road.
Suddenly reduce that road to four lanes and you still have the same number of cars to get through, you can only get 4 cars through a second. Obviously everything is going to slow down because the road isnt as wide.

That is what you have done by putting in a 7850 on a PCI-2 4x slot, effectively limited the width of the road that the graphics cards can use to communicate with the CPU. That is why you have horrible performance in Crossfire.

This seems to be the most logical answer, it's just that I don't understand why people are under estimating i3's so much even though they beat first generation i5s? and for some reason of there is a bottleneck in the system other than a graphics card they will say CPU even if it isnt...

 

[goldsauce]

Best answer selected by goldsauce.