Clear answer on CFX performance.

Hey guys,

I need a clear answer on something here.

Official statement from AMD says:

"What happens when graphics cards with varying memory speeds and engine clocks are paired together?
​The overall performance will be adjusted to match the memory speeds and engine clocks of the lower clocked card."

Source: http://support.amd.com/en-us/kb-articles/Pages/AMDCrossFireFAQ.aspx

But right now, i have a two guys claiming, that they obtain better benching results when overclocking all 4 cards (quadfire) individually, in pratice meaning that AMD cards do not run from the clocking of the slowest card.

Is this correct or are they somehow mistaken?

You mean they overclocked the slower cards to match the faster ones on GPU/Memory clocks? Of course in that situation the resulting crossfire configuration would improve.

No.

They claim to be running 4 cards, at different speeds, and getting better benchings with it.

All 4 cards at different speeds.

First card, HD 7970 @ 1180MHz
Second card, HD 7950 @ 1000MHz
Third card, HD 7970 @ 1125MHz
Forth card, HD 7950 @ 980MHz

Claiming this gave better results, he has no results where all cards are clocked identical.

Pierre3400 said:

No.

They claim to be running 4 cards, at different speeds, and getting better benchings with it.

All 4 cards at different speeds.

First card, HD 7970 @ 1180MHz
Second card, HD 7950 @ 1000MHz
Third card, HD 7970 @ 1125MHz
Forth card, HD 7950 @ 980MHz

Claiming this gave better results, he has no results where all cards are clocked identical.

Click to expand...

Clocked identical at what speed though? I'd have to see some data because that set up looks screwy to me... Not to mention those are some pretty heavy overclocks. Also, the scaling is different in quadfire than crossfire so the results may not be as significantly impacted by one card.

King107s said:

Clocked identical at what speed though? I'd have to see some data because that set up looks screwy to me... Not to mention those are some pretty heavy overclocks. Also, the scaling is different in quadfire than crossfire so the results may not be as significantly impacted by one card.

Click to expand...

I am sorry its not in English, but here you go: http://hwt.dk/Test/38433/Stortest-AMD-CrossFire/Overclocking

I am calling this test 100% useless, since there is no testing done with identical clocks vs, his indiviual clocking that he claims works, and works better, than running all cards at the same speed as the slowest card.

As additional info, now they claim that the test was made before AMD changed the drivers to revert all cards to the clock. Which i am pretty sure have been so, for at least 2 years.