• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Theoretrical processing power

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

OtherOne

New Member
Joined
Oct 24, 2013
HI. Wondering about video card processing power.
For example i have a videocard with this specs:
- 32 CUDA Cores.
- Graphics Clock: 650 MHz
- Processor Clock 1600 MHz
- Transistors 314M (Millions i guess)

And a core 2 duo intel processor at 2.93 GHz with a number of 228 million processing die transistors (theoretical processing power of 2 cores x 2.93 GHz = 5.86 GHz)

I know about threading and parallel processing and that a multi core processor cannot be compared directly with a single core because of programming drawbacks, Amdahl's law, memory latency & processor's cache etc, but what i'm asking here is just theoretical processing power.

So my questions are:
1. The cuda cores that can be used for general programming, run at 650 MHz or 1600 MHz?

2. If there are 32 cuda cores running at 650 MHz each for example, it would give a total maximum processing power of 32 cores X 650 MHz each = 20.8 GHz (meaning 3.5 times more powerfull than my intel processor)?

3. What is the big difference between graphics clock and processor clock in my video card ? I know one is for shading (calculating and transforming image data) but what the other one does ? And which is which?

4. If for example you answered that there are 32 cuda cores running at 650 MHz. The other 1600 MHz specification tells that there is 1 another processor besides those 32 processors?

5. If the 32x650=20.8 GHz is true, how can the GPU be 3.5 times more powerfull than the CPU with only 100 million transistors extra?

6. The GTX TITAN GPU has 2688 cuda cores, with these specs:
- Base Clock 837 MHz
- Boost Clock 876 MHz
That would mean it has the theoretical processing power of 2250 GHZ ( something like 383 times more than my CPU) ?


I guess that more questions will be asked...

Thanks

_________________________________
Intel processor specs here: http://ark.intel.com/products/36503/

My video card specs:
geforce9500gt_CPUZscreenshot.gif


GTX TITAN GPU specs here:
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-titan/specifications
 
Mhz/ghz are not a measurement of processing power, only speed. A 4ghz Pentium 4 has far less processing power than even a single core ivybridge based Pentium running at 2.5-3ghz. The short answer is that cpus and gpus have much different architecture so what one is fast at won't be the same as what the other is fast at.
 
Last edited:
Thank you for your answer. I know that MHz relates to speed and that other aspects are to be taken into measurement (processor memory cache and bandwidth for example + many many other aspects that could create a bottleneck before reaching the processor) and i guess the title and questions are a bit wrong from this perspective.

I know that GPU's architecture is designed especially for video rendering, geometry etc and that processors are designed for multipurpose tasks (Nvidia cuda processors are leaning towards general multipurpose programming, not just video processing)

But the speed influences the computing power dramatically if the rest of the environment is not out of order and given the theoretical fact that the processor would be fed with information when it can process it.

So yes, i guess that my questions where about speed. Any insights on that ?

Thanks
 
Last edited:
CPUs work in a way that one core does one job, while another core does a separate job. Because of this, it is beneficial to have higher clocked cores (see: 4GHz+).

GPUs work in the way that you have a cluster of slower cores all working on the same job at the same time.
 
Back