How is a Radeon 7770 faster than a 6770?

[internetswag]

I'm trying to figure this out. GPU generations have improved performance over their previous generations due to an increase in the number of stream processor units or CUDA cores. As well as other contributing factors such as ROP's, Texture mapping units etc.

Let's take an example, the Radeon 5670 has 400 Stream Processor Units/Cores. it's successor the Radeon 6670 has 460 SPU's. This causes an increase in graphical performance.



However, when we look at the Radeon 7770, it has 640 SPU's. Compare this to the 800 SPU's the Radeon 6770 has and I'm stumped. Yet reviews indicate that the 7770 is very superior to a Radeon 6770. I don't see how this is possible.

You might now argue that the reason for the increase in GPU power is because the 640 SPU's are running at 1GHZ. Which, is pretty smart to consider. But when you take into account the fact that the 6770 ran at 900 mhz, I highly doubt the reason for the graphical superiority of the 7770 lies in a tiny 100 mhz core clock increase.

Because by that logic a 6770 @ 1ghz should be way faster than a 7770 @ 1ghz. I guarantee you that won't happen.


So all in all, I'm trying to figure out why the 7770 is faster than the 6770 - teach me. If you think I'm coming across as brash, read the first sentence of this thread.

 

[esrever]

the cores are simply more efficient since its a new architecture.

 

[internetswag]

the cores are simply more efficient since its a new architecture.

!!!

I didn't know that was possible. I thought you only had core speed influencing core efficiency.

I wonder how they do it though, since this makes it harder to compare old graphic cards to new ones. I guess benchmarks will always prove as testers.

Thank you, esrever.

 

[esrever]

the 5870 has more cores than the the 6970 yet the 6970 was faster, this isn't really new. Im not sure exactly what they do.

Its like how sandy bridge i5 is faster than a c2q even tho they both have 4 cores even if they ran at the same ghz.

 

[shadow_city]

The 7770 uses 28 nm so it can squish in more transistors which results in better performance. The 6770 and 5770 use 40 nm transistors.

 

[sewalk]

The Netburst architecture of the Pentium4 is a textbook example of why raw clock speed is a really poor way to compare different architectures, even in products from the same manufacturer. The PentiumM and Core processors ran at markedly lower clock rates than the chips they replaced and yet performed better and ran cooler. With multi-core processors, branch prediction performance is critical.

 

[internetswag]

Best answer selected by InternetSwag.

 

[Mousemonkey]

This topic has been closed by Mousemonkey