Same clock speeds on same GPU but different performance?

[EpIcPoNaGe]

I've always known that you can't compare clock speeds of different products, but when comparing the same product such as a GPU, why is there sometimes a rather large deviation in performance? 

[sgloux3470]

Stability isn't always the same.  One card's memory may be spitting out invisible errors at higher clockspeed.  Their other components could be slightly different marginally affecting performance.  

[SkilledRebuilds]

I've always known that you can't compare clock speeds of different products, but when comparing the same product such as a GPU, why is there sometimes a rather large deviation in performance? 

Same GPU with Same CPU/MEM/Mobo as well?

If not using the EXACT EXACT spec across both machines, this is where the disparity shows up.

 

If using the same exact config, software may introduce random overhead leading to a difference that is more than expected.

 

If the GPU uses a boost clock, one card may boost higher at the same settings than another... due to a slightly better binned core (co-incidental or not)

[EpIcPoNaGe]

Stability isn't always the same.  One card's memory may be spitting out invisible errors at higher clockspeed.  Their other components could be slightly different marginally affecting performance.  

 

I see. That does make sense.

 

Same GPU with Same CPU/MEM/Mobo as well?

If not using the EXACT EXACT spec across both machines, this is where the disparity shows up.

 

If using the same exact config, software may introduce random overhead leading to a difference that is more than expected.

 

If the GPU uses a boost clock, one card may boost higher at the same settings than another... due to a slightly better binned core (co-incidental or not)

 

Hm.... I didn't think software would make that much of a difference. 

 

Other components make that big of a difference when comparing a gpu dependent benchmark such as Valley? 

[sgloux3470]

I see. That does make sense.

 

 

Hm.... I didn't think software would make that much of a difference. 

 

Other components make that big of a difference when comparing a gpu dependent benchmark such as Valley? 

 

Valley isn't *completely* GPU dependent.  It bottlenecks for me with a second 980 due to my CPU for example.

[EpIcPoNaGe]

Valley isn't *completely* GPU dependent.  It bottlenecks for me with a second 980 due to my CPU for example.

 

Its mostly GPU dependent though and it doesn't use much CPU anyway.

[bob345]

It all depends on how well the chip was etched. Chips that are toward the center of the wafer when the chips are being made tend to perform better and are binned for use in higher end cards. Some chips will just perform better than others.

[EpIcPoNaGe]

It all depends on how well the chip was etched. Chips that are toward the center of the wafer when the chips are being made tend to perform better and are binned for use in higher end cards. Some chips will just perform better than others.

 

Huh, that's cool! 

[Speaker1264]

It all depends on how well the chip was etched. Chips that are toward the center of the wafer when the chips are being made tend to perform better and are binned for use in higher end cards. Some chips will just perform better than others.

 

They typically have less leakage, meaning they are more efficient with the power, using less of it, and creating less heat.  However, this does not necessarily mean that it will overclock any better than other chips.

 

Also, the same architecture clocked at the same speed should always perform the same in terms of FPS, but there are other factors involved.