Not getting the performance out of SLi that I would expect, could use some help.

[Overl0rd]

What you could do (just as a test) is disabling SLI and see whether the GPU utilisation of a single 970 increases.

[david cassar]

Haha, yeah it's definitely enabled.

 

 

Temps are fine, 58/64 under load.

 

 

Can you provide actual evidence of this other than just your post?  I would like to read about it.

@Faceman and @zappian have benchmarks which show this.

[kendirect]

What you could do (just as a test) is disabling SLI and see whether the GPU utilisation of a single 970 increases.

 

Same FPS, 98% load.

 

http://i.imgur.com/D0RWtYx.png

 

See I thought you wanted SLi to both be splitting the load and thus they would both be at around 50%, but I'm also apparently dumb.  

[Overl0rd]

See I thought you wanted SLi to both be splitting the load and thus they would both be at around 50%, but I'm also apparently dumb.  

Haha no, ideally both GPU's are at 100%. 

[kendirect]

@Faceman and @zappian have benchmarks which show this.

 

Well, I guess I'll just have to wait a bit and save up for a new MOBO / CPU.  The games I play constantly are running fine at least, and I'll just have to do work arounds on games that don't do so well.  Maybe I'll have enough saved for Skylake later this year.

[Overl0rd]

Same FPS, 98% load.

 

http://i.imgur.com/D0RWtYx.png

 

I presume you have disabled SLI here. 

So now the first 970 is fully utilized. Can you post another screenshot of this game, but now with SLI enabled?

[kendirect]

Haha no, ideally both GPU's are at 100%. 

 

So help me out here (cause I would like to learn) what is it that causes my CPU to affect it at running SLi vs Single?  As clearly the single is running nearly 100%, what causes the dip when I enable SLi?

[kendirect]

I presume you have disabled SLI here. 

So now the first 970 is fully utilized. Can you post another screenshot of this game, but now with SLI enabled?

 

http://i.imgur.com/vv5hDSf.png

 

but also

 

http://i.imgur.com/vVejtNA.png & http://i.imgur.com/Lt9Blxc.png

 

but then again while in combat and a bit more graphically intense...

 

http://i.imgur.com/YGTs2PK.png

[Overl0rd]

So help me out here (cause I would like to learn) what is it that causes my CPU to affect it at running SLi vs Single?  As clearly the single is running nearly 100%, what causes the dip when I enable SLi?

Your CPU is most likely not fast enough to process what needs to processed by the game in order for both cards to be fully utilized. Since they now have to "wait" for the CPU, the GPU utilization decreases.

 

What I presume is that a single 970 is not fast enough for the 8350 to become a bottleneck, however 2 cards are. 

Since with a single card: almost 100%, with 2 cards; both around 60-70%.

[Overl0rd]

http://i.imgur.com/vv5hDSf.png

 

but also

 

http://i.imgur.com/vVejtNA.png & http://i.imgur.com/Lt9Blxc.png

 

but then again while in combat and a bit more graphically intense...

 

http://i.imgur.com/YGTs2PK.png

In these games you can clearly see that most of CPU cores are almost at 100%. 

[kendirect]

In these games you can clearly see that most of CPU cores are almost at 100%. 

 

So ideally you want CPU less and GPU @ 100% is what I'm getting

 

Going to test single vs SLi in BF4 now - probably the same though.

[Overl0rd]

So ideally you want CPU less and GPU @ 100% is what I'm getting

Yup, that's right. In BF4 my 2600K stays around 50-60% whereas both 570's are pegged at 100%.

[kendirect]

The loads in BF4 were somewhat confusing from this line of discussion lol.  I guess showing that the 8350 is a much bigger bottleneck than any of the benchmarks said at all.

 

Single: http://i.imgur.com/INpfs7Y.png

SLi: http://i.imgur.com/bgzP06N.png

 

Single was only even at about 60% whilst SLi only hit around 50%.  SLi however, had much higher CPU loads.

[creativename]

http://www.tweaktown.com/tweakipedia/55/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-2560x1440/index.html

 

I'm just going to say I trust a reputed website more than hear-say if you're not willing to show me better factual information with proven benchmarks.

I'm not going to argue that you're wrong because I honestly do not have the deep technical knowledge needed, but I've seen multiple sites showing that the 8350 should not heavily bottleneck the 9xx series.

 I'll provide links soon.

the 8350 is known to perform decently in games that are able to use multiple threads. In those games, 8350 will look like it performs as well as an i5. Battlefield 4 should be able to take advantage of all 8 cores, your FPS is way too low... Unless you're displaying the minimum FPS, then your numbers would look normal. Are those average fps? Or are those minimum FPS. And you can't just gather this information by looking at frame counters in the corner, and guessing what the average is. Get a software to count for you, that way its more accurate. FRAPS can do this, I'm sure there are other programs that can do it as well.

 

In games that do require single threads, 8350 will perform bad, but some of the games you're playing actually can use multiple threads. 

[kendirect]

 I'll provide links soon.

the 8350 is known to perform decently in games that are able to use multiple threads. In those games, 8350 will look like it performs as well as an i5. Battlefield 4 should be able to take advantage of all 8 cores, your FPS is way too low... Unless you're displaying the minimum FPS, then your numbers would look normal. Are those average fps? Or are those minimum FPS. And you can't just gather this information by looking at frame counters in the corner, and guessing what the average is. Get a software to count for you, that way its more accurate. FRAPS can do this, I'm sure there are other programs that can do it as well.

 

In games that do require single threads, 8350 will perform bad, but some of the games you're playing actually can use multiple threads. 

 

We're a little bit farther now .  I use FRAPS for my FPS counting with MinMaxAvg on.  Up a few posts you'll see BF4 and DA:I and some more information on what's going on with my system.

[creativename]

We're a little bit farther now .  I use FRAPS for my FPS counting with MinMaxAvg on.  Up a few posts you'll see BF4 and DA:I and some more information on what's going on with my system.

I clearly didn't skip every single post and comment ._.

[SteveGrabowski]

All the games you play require very good single threaded performance. Your 8350 is bottlenecking bad. Especially in games like dying light , planetside and arma. Get an I5 Asap. Also at 1080p why did you get 970 sli? A single 970 would have been enough. Anyways the 8350 is known to bottleneck a single 970 let alone two.

 

Dying Light's performance has actually improved pretty dramatically with newer patches, to the point it runs pretty evenly over four cores now. It doesn't touch virtual cores 5-8 on my CPU though.

[Godlygamer23]

Would you recommend it? Or is it more of a game-to-game basis?

http://i.imgur.com/q8TfCpr.png

That's what I'm looking at after about 5 minutes of gameplay.

Generally, I use it unless I run into issues. For example, NFS Rivals is locked to 30FPS. Normal V-Sync is desired there because you'll get constant tearing with Adaptive V-Sync. It's not a flaw of the function. It's working exactly how it's supposed to.

[SteveGrabowski]

http://www.tweaktown.com/tweakipedia/55/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-2560x1440/index.html

 

I'm just going to say I trust a reputed website more than hear-say if you're not willing to show me better factual information with proven benchmarks.

I'm not going to argue that you're wrong because I honestly do not have the deep technical knowledge needed, but I've seen multiple sites showing that the 8350 should not heavily bottleneck the 9xx series.

 

Honestly, it seems like 80% of the time someone comes on one of these forums and says his GTX 970 isn't performing up to expectations, it turns out he has an FX-8350. The other 20% of the time he has an FX-6300. The FX series is 2012 technology that wasn't even that high end back then. To not bottleneck an SLI 970 1080p setup too much you're going to need an i7-4790k or better.

[david cassar]

Dying Light's performance has actually improved pretty dramatically with newer patches, to the point it runs pretty evenly over four cores now. It doesn't touch virtual cores 5-8 on my CPU though.

Didn't know about that haha. Was assuming that performance is the same at launch. Thanks for the information

[Sintezza]

its a cpu bottleneck.

 

But keep in mind, that every cpu will bottleneck a dual GTX970 at 1080p in cpu demending games

There is not any cpu which will get a nearly 100% scaling.

Even a 4790K will not be capable of doing that.

 

So yeah, there will allways be a cpu bottleneck at a certain point.

No matter which cpu you trow at it.

 

Offcourse a GTX970 Sli will scale better on an intel haswell i7 in cpu demending games @1080p.

This is because the Haswell i5´s / i7´s have simply faster cores.

But the cpu bottleneck will still be there.

[kendirect]

Yup, that's right. In BF4 my 2600K stays around 50-60% whereas both 570's are pegged at 100%.

 

So all in all what would you say the final verdict is?  Disable SLi for now and wait until I can get an Intel CPU (probably going to wait for the Skylake release at this point,) or leave SLi enabled and re-enable it for games that I know the 8350 can handle it on?

[Overl0rd]

So all in all what would you say the final verdict is?  Disable SLi for now and wait until I can get an Intel CPU (probably going to wait for the Skylake release at this point,) or leave SLi enabled and re-enable it for games that I know the 8350 can handle it on?

I'd leave SLI on for now since you'll probably get a bit better performance than a single card, despite the CPU bottleneck.

For the time being I'd also try to overlclock the 8350 further since that'll make the bottleneck a bit less worse.

 

Hopefully my advice and tips have been helpful to you You can also PM me if you have any more questions

[Suika]

http://www.tweaktown.com/tweakipedia/55/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-2560x1440/index.html

 

I'm just going to say I trust a reputed website more than hear-say if you're not willing to show me better factual information with proven benchmarks.

I'm not going to argue that you're wrong because I honestly do not have the deep technical knowledge needed, but I've seen multiple sites showing that the 8350 should not heavily bottleneck the 9xx series.

Those multiple websites did not benchmark correctly, because the FX-8350 can easily bottleneck the GTX 770.

 

The only reason the website you linked shows the FX-8350 not bottlenecking is that it's running on 1440p on undemanding titles. Even the Bioshock: Infinite benchmark I provided shows rather minimum differences. And honestly, that article is only fuel for AMD fanboys, I'd hardly say it's credible.

 

If you're doing some high-budget gaming on high resolutions, don't get a budget CPU, it will bite you in the ass, as it has already.

[kendirect]

Those multiple websites did not benchmark correctly, because the FX-8350 can easily bottleneck the GTX 770.

 

The only reason the website you linked shows the FX-8350 not bottlenecking is that it's running on 1440p on undemanding titles. Even the Bioshock: Infinite benchmark I provided shows rather minimum differences. And honestly, that article is only fuel for AMD fanboys, I'd hardly say it's credible.

 

If you're doing some high-budget gaming on high resolutions, don't get a budget CPU, it will bite you in the ass, as it has already.

 

Just an FYI almost all the benchmarks you just sent me are on 290s.  I also said 970 not 770.

 

Thanks for the info though, I wish you had posted more on the 970s so I could look at them.