Epic CPU battles of history: Athlon FX-55, Pentium 4 EE, FX-8150 and Core i7 3960X

[Monstru]

Two icons of CPU history, 6 years of progress, 2 top of the line CPU's from today. From an overall point of view, or clock/clock, core/core with the ancients, how much do you think AMD and Intel progressed in the last 6 years? 4 CPU's, two different approaches, a test of epic proportions!

 

Epic CPU battles of history: Athlon FX-55, Pentium 4 EE, FX-8150 and Core i7 3960X | LAB501

 

 

[Massman]

Checked out some charts yesterday, very nice!

 

Funny how the performance gap between AMD and Intel went from 2.5% to 50% in 6y time. Dramatic performance in Wprime for Bulldozer; I wonder how it'd stack up against a 4x FX-74 system

[knopflerbruce]

I can test fx74, but at 2.6ghz? Also got a pair of fx70's here, they're 2.6 stock I think.

 

Apart from that wprime is just as useless as superpi for performance measurements:D

 

Are both fx8150 tests done at 2.6GHz, btw?

[matose]

The first test is at 2.6GHz with 1 core active (same as FX-55) and the second one is at default setting (3.6-4.2GHz, 8 cores)

[saint19]

wow, AMD need go back to that old FX55 times.

[Dolk]

I would love to read through this if you have an English translation.

[knopflerbruce]

Would be interesting to compare with Thuban and/or Deneb as well.

[wesjuhdabomb]

Omg.. Big fail for amd with 1c1t 2.6ghz on bulldozer. It is not only slower then Phenom 2 but also slower then FX55 of 6 years old.. cmon..

[Monstru]

I would love to read through this if you have an English translation.

 

Google Translate does a pretty decent job in most cases.

[Christian Ney]

I was going to do the same with all CPUs I have, but it takes a shitload of time

[TaPaKaH]

San Diego is not as slow clock-per-clock as many people think, only around 10% slower than Deneb in single-threaded benchmarks.

[OC Nub]

Thanks! I enjoyed reading this.

 

Google translate worked fine.

[I.M.O.G.]

Cool article, thanks. Good read.

 

Not sure about the first graphs showing single core performance - it seems more logical to me to show single thread performance rather than single core performance. Showing single thread performance cuts out the HT/multicore evolution of strategy for both sides, as that perspective is covered sufficiently in the second graph. By dropping the Bulldozer down to one core, it eliminates an important factor - AMD's strategic decision to include more cores at less cost. In the meantime the first graph includes Intel's strategic decision to include virtualized threading - that is an important factor as well. A more meaningful comparison would handicap both chips equally. Not a big deal, but I think at a glance the graphs could be misleading for those who don't stop to think about what is being shown. In the future for an article including this sort of comparison, it would be cool to see HT disabled for a true single threaded performance comparison. The important data including the outcome of their architecture strategy is shown sufficiently in the default runs and scores.

 

Compliments on the article again, a lot of work and cool results. I just wanted to offer feedback to lend my perspective.

Edited January 3, 2012 by I.M.O.G.

[wesjuhdabomb]

Problem is that AMD position their cpu's as 8 cores and Intel as 4 cores. (Bulldozer / Sandy bridge)

[knopflerbruce]

That shouldn't really matter, it's just marketing "BS":D