• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Possible bottleneck (how to confirm?) - 3770k & 2 x 7990s CF

Arr0way

Arr0way

Arr0way

Associate
Joined
12 Nov 2013
Posts
77
Location
...
Hi Guys,

So I know the general opinion / possible fact is that a 3770k will bottleneck with 4 cards (2 7990s in CF), but I wanted to try it anyway.

If so you reckon an Intel 4930K 3.40GHz VGA2011 & a Asus Rampage IV Formula Intel X79 (Socket 2011) will fix?

How can I find out if the CPU is bottleneck is it as simple as checking the usage or is this a bus / throughput issue?
 
Both the CPU's have equal IPC performance so in games that use 4 cores or less there won't be any difference.

It is as simple as checking GPU use while gaming or benchmarking.

If we lived in a perfect world where CPU limitations did not exist then all available GPU cores would run at 100% in every game.

That is not the case though so you're best off seeing what use you're actually getting.

Running 4 GPU cores itself will reduce your minimum frame rate as it places a higher load on the CPU and thus reduces the amunt of CPU cycles it can dedicate to processing frames for the GPU to process.
 
Last edited:
Although you've got four GPUs, it's not the same as four cards as you can still run at 8x8x.

Scaling is more of an issue than the CPU. Adding more GPUs doesn't mean higher CPU usage in itself.
 
I will check my CPU / GPU usage tonight.

Looking on 3dmark site (and the leader board on this forum) I see the best results are coming from the 2011 socket 3930 or 4930 chips.
 
In many games there'll be no difference. e.g. in Civ 5 (CPU demanding), 4 7970s at 1440p is giving 108fps for the 3770K and 107fps for the 3930K. In Sleeping Dogs (GPU demanding) it's 100fps and 101fps.
 
Well, I used to run 2 x 7970 on an i5 then swapped to an i7 3770k.

I did see an improvement in games like BF3 (bf4 was not out at the time), but BF multiplayer is a CPU heavy game that uses all threads.

I only did a quick run of 3dmark this morning I will look into it more tonight... 3Dmark did say my CPU was not up to scratch or something similar in the results.
 
It varies. Any game that uses 1-4 threads only is not going to show an improvement from an i5 to i7. Demanding games that use 4-8 threads will - e.g. BF3, BF4, Crysis 3, etc. Those games can gain from hex cores too for obvious reasons.

For multiple GPUs the PCI-E bandwidth of Ivy and Haswell will definitely have an effect with 3-4 cards.
 
Arr0way said:
Well, I used to run 2 x 7970 on an i5 then swapped to an i7 3770k.

I did see an improvement in games like BF3 (bf4 was not out at the time), but BF multiplayer is a CPU heavy game that uses all threads.

I only did a quick run of 3dmark this morning I will look into it more tonight... 3Dmark did say my CPU was not up to scratch or something similar in the results.
Click to expand...

Read the Anandtech review already posted they are a very reliable source of information.

http://www.anandtech.com/show/7189/c...september-2013

Anything above 4.5Ghz removes the bottleneck in BF4 for crossfire.

So can you tell me what is your fps in BF4 mins, max and average?
 
dmferrett said:
Read the Anandtech review already posted they are a very reliable source of information.

http://www.anandtech.com/show/7189/c...september-2013

Anything above 4.5Ghz removes the bottleneck in BF4 for crossfire.

So can you tell me what is your fps in BF4 mins, max and average?
Click to expand...

I will check them for you tonight, any preferred map, number of player etc?

I suspect you are right regarding the CPU, but it does not explain the much higher 3DMark results for the X79 chipsets?

I will also try clocking to 5ghz (normally run at 4.6) and seeing it makes a difference.
 
OP hexcore i7s will score high in all the futuremark benches as it is down to the physics tests (this has nothing to do with GPUs) in these benches. What you need to compare is the graphics score when comparing scores with any futuremark bench.

I also don't think a 3770k @4.6 will have any problems with a pair of 7990s.

To check for a bottleneck use the Heaven 4 bench with extreme settings. The scaling is very good in this bench and you should be looking for nearly double the score going from one to two HD 7990s.

Don't go out and buy a 4930k (some are poor overclockers), if you get a bad one it will be even worse than using a 3770k.
 
dmferrett said:
Read the Anandtech review already posted they are a very reliable source of information.

http://www.anandtech.com/show/7189/c...september-2013

Anything above 4.5Ghz removes the bottleneck in BF4 for crossfire.

So can you tell me what is your fps in BF4 mins, max and average?
Click to expand...

almighty15 said:
There's a BIG difference between crossfire with 2 GPU and 4 GPU cores.
Click to expand...

I would not take anandtech seriously

What have Sleeping Dogs, Dirt3 and Metro2033 all got in common ?

Answer, they are all totally useless on high end 4 GPU setups and totally inappropriate for testing in this context.
 
Kaapstad said:
I would not take anandtech seriously

What have Sleeping Dogs, Dirt3 and Metro2033 all got in common ?

Answer, they are all totally useless on high end 4 GPU setups and totally inappropriate for testing in this context.
Click to expand...

Why useless? They scale pretty well, and Sleeping Dogs especially is very demanding on the GPUs.

You can easily show bottlenecks in games like Crysis 3 on the 3770K compared to a 3930K on high end GPUs because it uses >4 threads and is CPU demanding too.

e.g.

XgNCpfC.png
 
Last edited:
teppic said:
Why useless? They scale pretty well, and Sleeping Dogs especially is very demanding on the GPUs.

You can easily show bottlenecks in games like Crysis 3 on the 3770K compared to a 3930K on high end GPUs because it uses >4 threads and is CPU demanding too.
Click to expand...

On sleeping dogs you hit a wall around 170 fps which makes it useless for testing GTX 690s let alone 4 Titans.

Dirt3 is not very demanding and once you get above 200fps you get all sorts of weird numbers.

Metro2033 has negative scaling on 4 GPUs, 2 Titans are faster than 4 lol.

Games that are good for testing bottlenecks are Crysis 3, Tomb Raider, BF3 and BF4 as they scale well, the only problem with BF3 and BF4 is they are capped at 200fps.
 
dmferrett said:
Acoording to Anand Tech a 3770k will not be a bottleneck with 4 cards.

They even suggest that an i5 4670k isnt a bottleneck with 3 cards providing you have high enough pciE lanes.

People on here say a i7 for sli/crossfire, I say that is BS.

http://www.anandtech.com/show/7189/choosing-a-gaming-cpu-september-2013
Click to expand...

It's not BS at all. Multiple people tested BF3 on an i5/i7 with 2 GPUs and their findings are all largely in line.

It's not a massive difference but GPU usage did increase a lot with an i7 over an i5.

Is it worth it when most games don't scale this way? Probably not but to call it BS is ironically BS itself.
 
To be honest 3 is the sweet spot so you would maybe be better with a 7990 and a 7970 as from what I have read 3 seems to be a much smoother experience than 4. As I said though that is just what I have read and Kapp is the right man to see for crazy multi GPU set ups :D
 
Kaapstad said:
I would take that graph with a pinch of salt as well, who runs Crysis 3 at the settings they used.

Can you get a monitor with a native resolution of 1280 x 720 these days lol.
Click to expand...

That's just to rule out the GPU, it shows CPU limitations. The benchmarks at higher resolutions are reflected but with one GPU you'd be GPU limited in Crysis 3 anyway.
 
You must log in or register to reply here.
Back
Top Bottom