Question regarding CPU's throughout the last 10 years

[eblondin]

Thanks again David!

I appreciate your time and that helps me out a lot!

[David (PassMark)]

I don't see any single big leaps coming.
I think it will be more of a slow grind with many small steps and incremental improvements.

If you are doing 3D render work, then maybe your software can offload some of the processing to the GPU. And a GPU upgrade might give better return on your dollar. But it all depends on the software you use and what hardware it can take advantage of.

[eblondin]

Thanks for the speedy reply David!

That makes a lot of sense and answers my question perfectly.

But it is kind of odd that CPU makers aren't offering different flavors of CPUs at consumer prices to meet the demanding needs for different customers. Like you said, I could care less about power consumption or an on chip GPU.

Do you have any insight into how long it will be before we see the NEXT AFFORDABLE big leap like the i7-920 was ($300-$400 price range)??

Most pricing I'm seeing for the new Xenon's E7's are in the 2 grand range and that is just ridiculous.

Thanks again for your time

-Eric

[David (PassMark)]

The i7-920 was a great CPU when it was released. One might argue head it was ahead of its time.

Moore's law was just for transistor counts and not necessarily for performance. The recent focus has been more on reducing electrical power usage rather than increasing performance. Further, more stuff has been crammed into CPUs, adding functionality rather than improving performance. The big thing that was added was on board GPUs (which you probably don't need or want in your case).

It is also a reflection that they are starting to hit the wall performance wise. Especially for single threads tasks. If you look at single threaded tasks then we see,

CPU, Score (single threaded)
Pentium 4, 700
i7 920, 1158
i7-4770K, 2299

So the big jump between the P4 and i7-920 was mostly due to going from 1 core to 4 cores. To get a similar jump again, you would need to go to 16 cores. This is possible, but it would mean going to something like the new 15 core E7-2890 v2 CPU. Or going to a dual CPU system.

[eblondin]

I was wondering if someone could help me understand what has been going on in the last 5-6 years in regards to CPU's.

A bit of a back story. I work as 3D artist and I am all about CPU power. Generally the faster/more cores my CPU is/has the faster I can render out something (unless it is using the GPU). But I'm not made of money, so I only upgrade every 5-6 years or so. When I do upgrade, I typically buy low end of the high range CPUs. Right around $300-$350.

On 9/16/2003 I bought a p4 3.06Ghz HT > here's a benchmark
http://www.cpubenchmark.net/cpu.php?....00GHz&id=1074
Average CPU Mark - 361

I waited a long time to upgrade, waited for a big jump in CPU architecture to warrant dropping over a grand on a new PC.

That day came on 2/5/2009. I bought a core i7-920 > here's a benchmark http://www.cpubenchmark.net/cpu.php?...2.67GHz&id=834
Average CPU MARK - 5,000

The core i7 920 according to these benchmarks was 12 times faster than my Pentium 4 and you know what, it felt like it.

It's been 5 years since 2009 and the best CPU on www.cpubenchmarks.net for my price range ($300-$350) is the Intel Core i7-4770K
http://www.cpubenchmark.net/cpu.php?....50GHz&id=1919
Average CPU Mark - 10,396

That's only a little over twice as fast as the core i7 920! Or am I missing something?

So here's my question, why did a 5.4 year gap between the p4 3.0HT and the core i7 920 result in a 12x performance but a 5.0 year gap between the core i7 920 and the core i7 3930K only result in a 2x Performance? That is a pretty huge gap there. It doesn't seem like Moore's law is working at all. And it makes me sad panda

Thanks!