R9 290 Crossfire Lack Luster Performance

[BoC]

Hey guys,

 

So to start here are my specs:

 

Mobo: M5A99FX Pro R2.0

Cpu: Fx8350 @4.7

Mem: 32Gb Gskill Rip Jaws @1600

Gpu: 2 x HIS R9 290 Reference

Psu: Coolermaster V1000

 

All is under water with a 120, 140, 240, & a 280 radiator.

 

So here is my deal. I'm not expecting these cards to be the end all to be all but I was expecting better benchmarks than I am getting especially from what I have seen from others. The highest score I have gotten in Valley Benchmark is around 3300 to 3400 @ about 82fps on ultra 1920x1080. While its not shabby I don't think its where these cards should be performing (I was thinking in the low 4000s score wise). Is the CPU bottle necking them that badly or is something else at work here?

 

Thanks in advance guys. 

[dragoon20005]

yes your CPU is the main culprit

 

 

run on just single GPU mode and run the test again

[vlkn´]

The 2nd pcie slot is running 4x and the first 16x  @ pcie 2.0

Maybe thats the reason, maybe.

[Glenwing]

Can you monitor the GPU usage with a utility like OpenHardwareMonitor and see if they're both at 100% the whole time?

[Guest Philosophy]

Probably your CPU, I am running 2x290s and I get awesome performance. 

[Faceman]

Its your CPU.  You are even bottlenecking a single 290 in some games/applications.  An i5 or better will fix this.

[-rascal-]

The 2nd pcie slot is running 4x and the first 16x  @ pcie 2.0

Maybe thats the reason, maybe.

 

The 990FX chipset can run two PCI-E X16 slots a full X16 speed. It has a total of 42 PCI-E lanes.

 

 

ASUS M5A99FX PRO R2.0

...

Expansion Slots

2 x PCIe 2.0 x16 (dual x16)

2 x PCIe 2.0 x16 (x4 mode, black)

1 x PCIe 2.0 x1

1 x PCI

 

 

Monitor the temperatures of the two cards. Since you are using reference design R9-290's, they could be throttling down the clock frequencies.

[dragoon20005]

Monitor the temperatures of the two cards. Since you are using reference design R9-290's, they could be throttling down the clock frequencies.

yea could be this too

 

but again the FX 8350 is holding back the dual GPUs

[xAcid9]

The 990FX chipset can run two PCI-E X16 slots a full X16 speed. It has a total of 42 PCI-E lanes.

 

 

 

Monitor the temperatures of the two cards. Since you are using reference design R9-290's, they could be throttling down the clock frequencies.

He meant the 2nd and 4th PCIe slot run at x4, only 1st and 3rd run at 16x when on dual card configurations. Correct me if i'm wrong.  D:

 

Both OP's 290s was watercooled.

 

All is under water with a 120, 140, 240, & a 280 radiator.

[Cheddle]

Hey guys,

 

So to start here are my specs:

 

Mobo: M5A99FX Pro R2.0

Cpu: Fx8350 @4.7

Mem: 32Gb Gskill Rip Jaws @1600

Gpu: 2 x HIS R9 290 Reference

Psu: Coolermaster V1000

 

All is under water with a 120, 140, 240, & a 280 radiator.

 

So here is my deal. I'm not expecting these cards to be the end all to be all but I was expecting better benchmarks than I am getting especially from what I have seen from others. The highest score I have gotten in Valley Benchmark is around 3300 to 3400 @ about 82fps on ultra 1920x1080. While its not shabby I don't think its where these cards should be performing (I was thinking in the low 4000s score wise). Is the CPU bottle necking them that badly or is something else at work here?

 

Thanks in advance guys. 

 

I would definitely expect more - I got 99 fps in valley with a z87, i5-4690 and 2 r9 290's clocked at 1ghz each.

 

Can you run GPUz to confirm that both cards are in fact running PCIE 2.0x16?

 

Your CPU seems good enough? though I am not well versed in AMD CPU's - the best AMD cpu I have is an Athlon 64 :-)

 

http://www.cpubenchmark.net/compare.php?cmp[]=1780&cmp[]=2236

 

The ONLY thing I can think of is there is something bottlenecking the cards communicating between each other.  not sure if its chipset or cpu...

[Sintezza]

Scaling on 1080p with 2 highend gpu´s like that will not be that greate in the first place.

Sure an intel i5-4690k will scale a bit better.

But it will also not anywhere near maxout those 2 gpu´s.

At 1080p with a dual highend gpu setup, simply every cpu bottlenecks, even a 5960X overclocked.

Dual highend gpu setups for just 1080p is basicly a waste of f*cking money.

 

If you wanne test gpu performance it self, you could do firestrike.

[Fate]

Scaling on 1080p with 2 highend gpu´s like that will not be that greate in the first place.

Sure an intel i5-4690k will scale a bit better.

But it will also not anywhere near maxout those 2 gpu´s.

 

At 1080p with a dual highend gpu setup, simply every cpu bottlenecks, even a 5960X overclocked.

 

If you wanne test gpu performance it self, you could do firestrike.

You don't need a 5960x overclocked for two 290's lmao...

A 4690k will likely alleviate the bottleneck. 

[Sintezza]

You don't need a 5960x overclocked for two 290's lmao...

A 4690k will likely alleviate the bottleneck. 

 

At 1080p, even the 5960X will no where near scale 100%.

Especialy not on DX11

 

A 4690K will also bottleneck.

Dont talk about intel, if they are some kind of mythical gods because they arent.

Scaling at 1440p will be allot better, sure thats because you get GPU bound then.

[Fate]

At 1080p, even the 5960X will no where near scale 100%.

Especialy not on DX11

 

A 4690K will also bottleneck.

Benchmarks? Proof? Source? Something???

[Sintezza]

Benchmarks? Proof? Source? Something???

 

Sure check how badly the scaling is at 1080p.

Some games gain an extra 30Fps but thats still a terrible performance gain.

No where near 100% scaling.

 

 

At 1440p the scaling is allot better, sure but again, you getting GPU bound.

You see that on 1440p the scaling is allmost 80% / 100% in allot of games.

 

At 4k the scaling will be 100% in allmost everything, no matter which cpu.

[Faceman]

Holy crap Sintezza.  Let it go.  You are the worst warmongerer for the FX processors.  They suck.  Get over it. It has been proven time and time and time again that the FX processors are holding back high end GPUs, at 1080p, and 1440p.  Not Ultra 4k.

[Sintezza]

Holy crap Sintezza.  Let it go.  You are the worst warmongerer for the FX processors.  They suck.  Get over it. It has been proven time and time and time again that the FX processors are holding back high end GPUs, at 1080p, and 1440p.  Not Ultra 4k.

 

This video isnt about FX dude, this is about 5960X on 4.2GHz, just looks how badly this cpu bottlenecks at 1080p. this video just points out that a dual highend gpu setup will get bolltenecked by every cpu badly at 1080p.

 

Sure if op wants better scaling, then an 4690k will help some, but dont expect nowhere near 80% scaling.

So this basicly means that also a 4690K bottlenecks as fuck.

Thats why i´m saying dual highend gpu´s for just 1080p is basicly a waste of money, unless you are happy with 20fps gains....

 

Just face it, or ignore it, i dont care at all

[Briggsy]

Hey guys,

 

So to start here are my specs:

 

Mobo: M5A99FX Pro R2.0

Cpu: Fx8350 @4.7

Mem: 32Gb Gskill Rip Jaws @1600

Gpu: 2 x HIS R9 290 Reference

Psu: Coolermaster V1000

 

All is under water with a 120, 140, 240, & a 280 radiator.

 

So here is my deal. I'm not expecting these cards to be the end all to be all but I was expecting better benchmarks than I am getting especially from what I have seen from others. The highest score I have gotten in Valley Benchmark is around 3300 to 3400 @ about 82fps on ultra 1920x1080. While its not shabby I don't think its where these cards should be performing (I was thinking in the low 4000s score wise). Is the CPU bottle necking them that badly or is something else at work here?

 

Thanks in advance guys. 

 

something is off. 3300 is what you would be scoring in valley with an fx8350 and R9 280's in crossfire. there's no way its just the cpu as a gtx770 SLI with an fx8350 can close in on a 4000 score in valley.

 

are you sure both cards are in 16x slots?

[Briggsy]

I just did a couple runs in valley with my 290's at stock speeds, and Even without doing any performance optimizations, and with my i5 3570k underclocked to 3Ghz I was still getting 3500 score.

 

is the 8350 really that bad that it runs worse than a 3Ghz i5? I've heard that a 4Ghz 8350 is equal to a 2ghz i5 for gaming... but I didn't believe it until now.... 

[vlkn´]

video + badass music=mlg benchmark

[Sintezza]

I just did a couple runs in valley with my 290's at stock speeds, and Even without doing any performance optimizations, and with my i5 3570k underclocked to 3Ghz I was still getting 3500 score.

 

is the 8350 really that bad that it runs worse than a 3Ghz i5? I've heard that a 4Ghz 8350 is equal to a 2ghz i5 for gaming... but I didn't believe it until now.... 

 

intel scales better on 1080p, because they have better single threaded performance.

But still also an intel cpu will bottleneck 2x290X or what ever highend gpu setup at 1080p.

Because you simpley dont get 100% scaling, only in GPU bound games.

 

Thats why i have said tons of times, that the term "bottlenecking" is miss-used by allot of people.

Because no matter intel or amd, every cpu will become a bottleneck at a certain point.

But intel cpu´s are simply better, because they have a better per core performance, so they will also scale better.

But still there is a bottleneck even with intel on 1080p with dual highend gpu´s.

[Najuno]

intel scales better on 1080p, because they have better single threaded performance.

But still also an intel cpu will bottleneck 2x290X or what ever highend gpu setup at 1080p.

Because you simpley dont get 100% scaling, only in GPU bound games.

 

Thats why i have said tons of times, that the term "bottlenecking" is miss-used by allot of people.

Because no matter intel or amd, every cpu will become a bottleneck at a certain point.

But intel cpu´s are simply better, because they have a better per core performance, so they will also scale better.

But still there is a bottleneck even with intel on 1080p with dual highend gpu´s.

Intel CPUs score better in every higher resolution... if you are not convinced just go to 3D Mark's HoF and look in the Top100 if you can find any FX CPU there. Or in our benchmarking threads Valley and 3D Mark, there's nobody with FX in the top notches. My i7 5930K OCed to 4.7GHz eats every FX CPU alive... I could deactivate 2 cores and HT and would still be faster.

 

To run a SLI/CF with highend cards in 1080p is kind of pointless anyhow unless you want to downsample or show off in front of your buddys... either way highend cards scale very good in 1440p and above. When I run my GTX980s in 2way SLI it's just the sweet spot for 1440p. With the 3way SLI activated I get up to 85FPS in 4K Very High settings. There's a big difference between a 2way and 3way config for 4K, but a shame anyhow right now that you need to invest that much money in order to make full use of a 4K 60Hz monitor.

[KeltonDSMer]

Uhh, Valley barely loads the CPU. Of all synthetic GPU benches, that is the one that should be impacted the least by a less powerful CPU.

 

I am not arguing saying 83xx is equal to a haswell 4 core in gaming or anything like that, but Valley specifically should not be hugely affected by using that CPU compared to a high end Intel.

[Briggsy]

intel scales better on 1080p, because they have better single threaded performance.

But still also an intel cpu will bottleneck 2x290X or what ever highend gpu setup at 1080p.

Because you simpley dont get 100% scaling, only in GPU bound games.

 

Thats why i have said tons of times, that the term "bottlenecking" is miss-used by allot of people.

Because no matter intel or amd, every cpu will become a bottleneck at a certain point.

But intel cpu´s are simply better, because they have a better per core performance, so they will also scale better.

But still there is a bottleneck even with intel on 1080p with dual highend gpu´s.

 

Oh I've argued the same case before too in different threads. The very existence of the CPU in the system causes a performance decrease in the first place. If a graphics card had its own dedicated CPU built into the card that only performed tasks related to the GPU and shared the GPU GDDR5 Memory, we wouldn't have to deal with as much cpu overhead. too bad APU's are weaksauce at this point.

[BoC]

OK wow thanks for all the replies everyone.I'm on my phone let me try and answer some of the question the best I can at the moment.

Yes both cards are in the x16 slots

Cards are stock clcks

Temps don't go over 45c for the cards

Even with catalyst enabling xfire cpuz and hwm only show specs and on one card

Both cards perform fine on there own.

And a 1440 monitor is in the pipeline