GTX470 renders faster than GTX760

[rutger79]

Hello,

What am I missing? I changed my older gtx470 for the gtx760. I expected the gtx760 to be much faster considering the amount of cuda cores.
specs gtx 470;
Operating System: Windows 10 Pro, 64-bit
DirectX version: 12.0
GPU processor: GeForce GTX 470
Driver version: 361.91
Direct3D API version: 11.2
Direct3D feature level: 11_0
CUDA Cores: 448
Core clock: 700 MHz
Shader clock: 1400 MHz
Memory data rate: 3348 MHz
Memory interface: 320-bit
Memory bandwidth: 133.92 GB/s
Total available graphics memory: 9457 MB
Dedicated video memory: 1280 MB GDDR5
System video memory: 0 MB
Shared system memory: 8177 MB
Video BIOS version: 70.00.21.00.03

GTX760;
Operating System: Windows 10 Pro, 64-bit
DirectX version: 12.0
GPU processor: GeForce GTX 760
Driver version: 361.91
Direct3D API version: 12
Direct3D feature level: 11_0
CUDA Cores: 1152
Core clock: 1019 MHz
Memory data rate: 6008 MHz
Memory interface: 256-bit
Memory bandwidth: 192.26 GB/s
Total available graphics memory: 12273 MB
Dedicated video memory: 4096 MB GDDR5
System video memory: 0 MB
Shared system memory: 8177 MB

Hope someone can help me. The only difference i could find was the shader clock.

[abreukers]

hi, the biggest difference here is the architecture of those cards. the gtx470 is fermi while the gtx760 is kepler. you'll find that kepler cuda cores are just roughly half as fast as their fermi predecessors in terms of rendering, but they have other advantages in other terms. take a look the octanebench scores here: https://render.otoy.com/octanebench/results.php

the Kepler cards have many more cores than Fermi but these cores are designed differently. the kepler cores run at a much lower power requirement which implies they have less performance per core but they have better performance per watt. (this info based on a writeup by nvidia in 2012, during the advent of the kepler generation cards).

[prehabitat]

Hello,

What am I missing? I changed my older gtx470 for the gtx760. I expected the gtx760 to be much faster considering the amount of cuda cores.

the 470 is a Fermi - same as the 5XX series. The 760 is Kepler.
There are some formulas and other stuff floating around; but generally a Fermi core>Kepler core in terms of octane speed.

Take a look at https://render.otoy.com/octanebench/res ... ingleGPU=1
The 'old' gtx580 (available in 3gb variants) beats everything up to a GTX770, and is only 20% slower than the GTX970.

Marketing and politics aside; the only difference is power consumption/waste and price.
Fermi uses/wastes a lot of power (as heat), whereas the later architectures (Kepler, Maxwell, etc) waste less power as heat (ie run cooler for the same work).
Fermi is very very old; generally only available used - generally without warranty.

[rutger79]

Thank you both for the reply. It was very useful! I wasn't aware of the Kepler/Fermi architecture.
Although my old gtx470 is a bit faster I'll stick with the newer gtx760 for better allround pc performance and of course the battle against global warming;)