Clock Speed VS. Raw Processing Cores, which is more valuable?

[trollolo]

What the title says. Right now EVGA's 460 and 550 are pretty much the same price, but when i look at the specs. the 460 has 336 proc-cores coupled with low clock speeds, while the 550 only has 192 proc-cores, but its clock speeds are higher.

which one is more important? do they balance out? and last but not least, which one will run cooler?

 

[Seero]

What the title says. Right now EVGA's 460 and 550 are pretty much the same price, but when i look at the specs. the 460 has 336 proc-cores coupled with low clock speeds, while the 550 only has 192 proc-cores, but its clock speeds are higher.

which one is more important? do they balance out? and last but not least, which one will run cooler?

The answer you will like to know:
Reviews shows that 460 is faster, cooler, and quieter then 550Ti.

 

[RussianSensation]

In simple terms, GTX460 is faster than a GTX550 Ti.

=========Read below if you want a more detailed explanation====================

Even if you do look at the GPUs in terms of stream processors, that's still not enough. You need to look at memory bandwidth, texture unit (TMU) performance and Raster Operator Units (ROPs).

GTX550Ti:

Shader operations (SPs x GPU clock speed) = 192 x 900mhz = 172,800 MOperations/sec
Pixel fill-rate (ROPs x GPU clock) = 24 ROPs x 900mhz = 21600 MPixels/sec
Texture fill-rate (TMUs x GPU clock) = 32 TMUs x 900mhz = 28800 MTexels/sec
Memory Bandwidth (Bus Type x Memory speed) = 192-bit x 4100mhz / 8 bits (per 1 Byte) / 1000 = 98.4GB/sec

vs.

GTX460 1GB

Shader operations (SPs x GPU clock) = 336 x 675mhz = 226,800 MOperations/sec
Pixel fill-rate (ROPs x GPU clock) = 32 ROPs x 675mhz = 21600 MPixels/sec
Texture fill-rate (TMUs x GPU clock) = 56 TMUs x 675mhz = 37800 MTexels/sec
Memory Bandwidth (Bus Type x Memory speed) = 256-bit x 3600mhz = 115.2GB/sec

So basically the GTX460 is >= in every performance metric imaginable.

Without doing these calculations, you can always refer to GPUReview.com. Be mindful that it's not useful to compare NV to AMD based on these metrics on paper. It's relevant within each brand though.

OR an even better method is to check benchmarks like this or this or this.

It's more difficult to answer your question which one "runs cooler" since that depends on the type of aftermarket cooler the card in question and the specific factory vs. stock frequency parameters of the GPUs:

Obviously an MSI Cyclone/Hawk GTX460 or GTX460 Windforce will run cooler than a reference card.

Depending on the price, a good alternative to the GTX460 1GB is the HD6850 or for another $30, HD6870 is even faster than both.

GTX460 can overclock extremely well though (i.e., > 900mhz). For $120, it's a good value assuming you get the 1GB version and your CPU is fairly modern (i.e., not something slow like an E6400!)

And of course, the obligatory song:

http://www.youtube.com/watch?v=t6FUR_nhGX8

 

[tviceman]

I would also like to add that it depends on both the number of cores (or rather, the difference in the number of cores you are comparing) and the speed of the cores (again, the difference). Generally speaking, having more cores tends to be more efficient than having less cores running at higher frequencies. At their reference speeds, the gtx560ti or more efficient than a gtx570. BUT, if you increase the core speed of the gtx560ti to match the gtx570 in performance, the gtx570 is suddenly more efficient.

Case in point, a reference gtx570 operates 480 cores @ 732mhz. 480 X 732 = 351,360 core mhz operations per second. A reference gtx560ti operates 384 cores @ 822mhz, but in order for a gtx560ti to roughly equal a gtx570 in speeds, it needs to run at about 1000mhz core frequency. So, 384 x 1,000 = 384,000 core mhz operations per second.

 

[trollolo]

Even if you do look at the GPUs in terms of stream processors, that's still not enough. You need to look at memory bandwidth, texture unit (TMU) performance and Raster Operator Units (ROPs).

>mfw that post

thanks for the help tho, time to do some more lernin