• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

How to calculate GDDR6 speed from GPU-Z?

Joined
Aug 4, 2016
Messages
340 (0.11/day)
Processor 8700k OC @ 5GHz (1.35 V)
Motherboard ASUS Prime Z390-A
Cooling Corsair H100i v2
Memory 16GB DDR4 OC @ 3000MHz
Video Card(s) EVGA RTX 2070 XC Ultra
Storage 4 x 250GB PNY SSD & WD BLACK 1TB HDD
Display(s) Samsung UHD 4K / BenQ RL2455HM
Case NZXT H500i
Power Supply Corsair 750W
Software Windows 10 Pro
So apparently my new RTX 2070 has 14000MHZ on the GDDR6, however GPU-Z reads at around 2000. I remember with GDDR5 you just multiply that by 4 since it's "quad pumped", but how do I calculate the GDDR6 speed?
 
The same way.


Your card is 1750 MHz at stock IIRC. Its DOUBLE DATA RATE and Quad pumped. Just how you would have done this with GDDR5. 2x the value, then 4x.
 
Last edited:
The same way.


Your card is 1750 MHz at stock IIRC. Its DOUBLE DATA RATE and Quad pumped. Just how you would have done this with GDDR5. 2x the value, then 4x.

Yeah i dont thing gddr6 is octal pumped as gddr5 was quad pumped
 
GDDR5X and GDDR6 are both 8x the actual clock.

So GPU-Z will report 1,750MHz, which is the actual clock speed of the VRAM. You multiply that by 8 to get 14,000MHz effective clock speed.

Its DOUBLE DATA RATE and Quad pumped. Just how you would have done this with GDDR5. 2x the value, then 4x.

That isn't what double data rate means, well sort of. Double data rate, or DDR, means you multiply the actual clock by 2. Really GDDR5 should have been called QDR not DDR, as it was no longer Double Data Rate, it was Quad Data Rate. GDDR5X and GDDR6 are both Octo Data Rate. But for some reason they are still calling them DDR, I'm guessing just because that is what people know.
 
Last edited:
GDDR5X and GDDR6 are both 8x the actual clock.

its not 8x the actual clock.
its B x DDR x ( R+F)

where R+F is the data operations on the rise and fall of the clocks, GDDR6 does 2 ops on the rise and 2 ops on the fall (hence quad pumped)

The math in the below screenshot starts with the 1750 base and is 1750 * 2 * (2+2) for 14GT/s
127704
 
Last edited:
its not 8x the actual clock.
its B x DDR x ( R+F)

where R+F is the data operations on the rise and fall of the clocks, GDDR6 does 2 ops on the rise and 2 ops on the fall (hence quad pumped)

The math in the below screenshot starts with the 1750 base and is 1750 * 2 * (2+2) for 14GT/s
View attachment 127704

1750*2*4= 14000
1750*8=14000, breaking math down is simpler than having to pemdas everything

So GDDR5 is quad pumped and GDDR6 is octo pumped
 
You're misusing what "pumping" actually means to come to this.


Pumping is another word meaning data transfer or requests can occur on the leading and trailing edge of the wave form and or how many operations occur within one cycle.
 
Have y'all been successful with pushing memory past 14000 on RTX cards? My 2060 Super doesn't seem to tolerate memory oc as high as my old 1060. I guess the new gddr6 is already running nearly maxed out, whereas the old mature gddr5 had some headroom.
 
Back
Top