• 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.

What is source of gpu power?

Joined
Aug 25, 2015
Messages
77 (0.02/day)
Location
Turkey
System Name Leviathan
Processor i5 4690
Motherboard H81M-P33
Cooling COOLER MASTER HYPER TX3 EVO
Memory 12GB
Video Card(s) ZOTAC GTX 970 AMP! Extreme Core Edition
Storage WD Caviar Blue 1TB 7200Rpm 64mb
Display(s) LG 21"
Case Zalman Z3plus
Power Supply Corsair VS650
Mouse Razer Chroma
Keyboard Razer Chroma
Software Win10 64bit
Hello friends i was checking differences between a gtx 780 and 970. So 780 has 2200~ cuda cores and 970 has 1664~ cuda cores. I bought my 970 a month ago and i was using a gtx 760 which has 1150~ cuda cores. İ have noticed the power difference clearly when i got 970 with 500 cuda core difference so how is that possible to 970 and 780 has same fps when 780 has 600 more cuda cores? What is exact origin of gpu power?
 
It's not just how many cores they have, there are a number of other factors when looking at different architectures, so the efficiency of those cores, ROP count and a bucket full of other things.
 
Just noticed that 780 has 160 billion/sec texture fillrate while 970 has 109 gigatexels/sec texture fill rate. Maybe thats the difference i dont know anymore. İs there a chart that shows the difference between billion/sec and gigatexels/sec?
 
Hello friends i was checking differences between a gtx 780 and 970. So 780 has 2200~ cuda cores and 970 has 1664~ cuda cores. I bought my 970 a month ago and i was using a gtx 760 which has 1150~ cuda cores. İ have noticed the power difference clearly when i got 970 with 500 cuda core difference so how is that possible to 970 and 780 has same fps when 780 has 600 more cuda cores? What is exact origin of gpu power?
Sooooooooooooooooooooo many different variables on that question. There are hardware changes and efficiencies as well as software from generation to generation. You cannot look at different generation CUDA cores in the same way. As Tatty said, there are different ROPs/TMUs, etc that a JUST A PART OF the difference.
 
The origin of GPU power is the same as origin of CPU power.

uarch improvements. You can't just compare the numbers on spec sheet if they are from different generation. It's the same thing with i3-6100 raping P4 of the same speed.
 
The power source is the screams of console peasants.
 
Should be a wall socket somewhere :laugh:
 
Hello friends i was checking differences between a gtx 780 and 970. So 780 has 2200~ cuda cores and 970 has 1664~ cuda cores. I bought my 970 a month ago and i was using a gtx 760 which has 1150~ cuda cores. İ have noticed the power difference clearly when i got 970 with 500 cuda core difference so how is that possible to 970 and 780 has same fps when 780 has 600 more cuda cores? What is exact origin of gpu power?
So you want to know what is the difference between Kepler and Maxwell, I will try to explain that as good as I can (I'am no engineer, just a hardware enthusiast):

Kepler is a universal engine for games and professional work, it's optimized for games as well, but not specially built to suit them.

Maxwell is a strictly for gaming built architecture, that's why it's more efficient and even faster than Kepler. Also it has a texture compression feature, means that it uses it's Ram more efficient than Kepler.

So, to compare the GPUs that you were talking about:

1) you can't compare CUDA cores of different architectures 1:1, you have to have the architectural differences in mind.

That means 2334 cores in GTX 780 are inferior to the 1664 cores in GTX 970. The GTX 970 has simply shaders better suited for gaming. Also because the architecture is built up different it's able to clock a lot higher. Where GTX 780 maybe can achieve up to 1200 or 1250 on Core (without any changes to the graphics card), the GTX 970 easily runs on 1400 to 1500 MHz. This on top of the better architecture (for gaming), leads to the speed boost and even means the GTX 970 is about as fast as the GTX 780 Ti (with 2880 Kepler cores).

2) the ROP count is always important.

The GTX 970 has 56 ROPs, the GTX 780 (or GK110) only 48, which means it has a edge there too, in rendering pixels etc.

3) the Vram.

the GTX 970 has 1 GB or 0.5 GB more Vram, which also has texture compression, so practically spoken has "a lot" more Vram (texture compression is about 30%).

If you want to know even more, consultate a engineer I'd say. :) That's my try in explaining things.
 
So you want to know what is the difference between Kepler and Maxwell, I will try to explain that as good as I can (I'am no engineer, just a hardware enthusiast):

Kepler is a universal engine for games and professional work, it's optimized for games as well, but not specially built to suit them.

Maxwell is a strictly for gaming built architecture, that's why it's more efficient and even faster than Kepler. Also it has a texture compression feature, means that it uses it's Ram more efficient than Kepler.

So, to compare the GPUs that you were talking about:

1) you can't compare CUDA cores of different architectures 1:1, you have to have the architectural differences in mind.

That means 2334 cores in GTX 780 are inferior to the 1664 cores in GTX 970. The GTX 970 has simply shaders better suited for gaming. Also because the architecture is built up different it's able to clock a lot higher. Where GTX 780 maybe can achieve up to 1200 or 1250 on Core (without any changes to the graphics card), the GTX 970 easily runs on 1400 to 1500 MHz. This on top of the better architecture (for gaming), leads to the speed boost and even means the GTX 970 is about as fast as the GTX 780 Ti (with 2880 Kepler cores).

2) the ROP count is always important.

The GTX 970 has 56 ROPs, the GTX 780 (or GK110) only 48, which means it has a edge there too, in rendering pixels etc.

3) the Vram.

the GTX 970 has 1 GB or 0.5 GB more Vram, which also has texture compression, so practically spoken has "a lot" more Vram (texture compression is about 30%).

If you want to know even more, consultate a engineer I'd say. :) That's my try in explaining things.

This was helpful thanks. also is it true that nvidia downgraded kepler on purpose? Lots of ppl complaining about that in nvidia forums.
 
This was helpful thanks. also is it true that nvidia downgraded kepler on purpose? Lots of ppl complaining about that in nvidia forums.
Not one backed it up...
 
Back
Top