B*ttleneck

Horrible CPU bottleneck

4 x TX @1441/1977
5960X @4.5
Valley bench max settings

Compare the scores @1080p and @2160p. In theory you should score 4X higher @1080p.

This is why I don't run this bench @1080p.:D

If you compare the temps of cards at the two resolutions you can see they are hardly working @1080p



1080p
wb6Ip3t.png




2160p
RdtJTIq.jpg
 
I'd very much like to see the CPU utilisation graphed through the benchmark before assuming it's the CPU causing the GPUs to have to wait for data. Run CPUZ and GPUZ in the background while rinning the tests again
 
I'd very much like to see the CPU utilisation graphed through the benchmark before assuming it's the CPU causing the GPUs to have to wait for data. Run CPUZ and GPUZ in the background while rinning the tests again

It's not just Kaap, Many online reviewers *JayzTwoCents as an aexample and even he hates the word bottleneck* who have done multi TX have seen that the CPU is the limiting factor, It's not a conspiracy dude it's just a fact ^_^
 
I'd very much like to see the CPU utilisation graphed through the benchmark before assuming it's the CPU causing the GPUs to have to wait for data. Run CPUZ and GPUZ in the background while rinning the tests again

Valley is a well known CPU bottleneck

If you run a single TX on it you get 4X the score @1080p as you do @2160p as the fps are not high enough to hit the CPU bottleneck.

@2160p 4 TXs are fine with no CPU bottleneck and no reduction in performance, this rules out the drivers as they will work the same at both resolutions. If you run 4 TXs @1080p this is where we hit the CPU bottleneck.
 
My definition of Bottleneck : LINK

In simple terms :
If Your GPU is using 3D mode Clocks in games like BF4 - it ain't bottlenecked :P
Smaller utilisation is proof of not having best CPU/GPU combo (or not optimal match), for particular bench/game.

Hope it helps :)
 
Last edited:
Silly argument this whole "bottleneck" thing, any gpu or cpu will "bottleneck" given the right program or game, the conversation should really be focused on why it's taken this long to get a new API and why the hell are we still using Shader model 5 (2008).
 
Back
Top