Return of AMD FX: My OC'd AMD FX 8150 / 6990 review - Check in Daily for New Results!

[Lightman]

Really good job polyzp!
Very informative thread with lots of data points

[polyzp]

thanks for the feedback guys! more would be nice ^^ , I am also taking all requests for benchmarks. At this rate I will have significantly more benchmarks to evaluate FX as a whole than most review sites. It seems to me that most review sites leave out benchmarks where FX shines, and include more where it doesnt, all coming to the same incomplete conclusion that it isnt very good. People actually think an 1100t is a better buy than a FX 8120 when i have seen countless times where Phenom II is left in the dust by both sandy bridge and FX. Low single core performance is more of a myth than reality. At 4.8 Ghz FX matches a 4.1 Ghz Phenom Core, and at 5 Ghz, matches a 4.25 Ghz phenom Core. So when you overclock both the difference in single core performance begins to minimize. Unless you really push the 1100t above 4.3 Ghz, single core performance will not be better than FX. You might aswell take the extra scaling (Phenom II X6 - 5.92 scaling , FX 8150 -6.66 scaling). Also, with a good enough cooling solution Bulldozer can reach clocks as high as 5.2 Ghz, pushing its single core performance above what Phenom II can achieve. The more you overclock FX , the better the OC benefits you, there doesnt seem to be a bottleneck wall at max OC, while Phenom II is generally known to suffer from this.

[Olivon]

Don't waste time and bandwidth dude. We all know the Truth.

[polyzp]

Don't waste time and bandwidth dude. We all know the Truth.

Most people would disagree and for that reason alone I will continue to waste all the time and bandwidth i like ^^

[BeepBeep2]

thanks for the feedback guys! more would be nice ^^ , I am also taking all requests for benchmarks. At this rate I will have significantly more benchmarks to evaluate FX as a whole than most review sites. It seems to me that most review sites leave out benchmarks where FX shines, and include more where it doesnt, all coming to the same incomplete conclusion that it isnt very good. People actually think an 1100t is a better buy than a FX 8120 when i have seen countless times where Phenom II is left in the dust by both sandy bridge and FX. Low single core performance is more of a myth than reality. At 4.8 Ghz FX matches a 4.1 Ghz Phenom Core, and at 5 Ghz, matches a 4.25 Ghz phenom Core. So when you overclock both the difference in single core performance begins to minimize. Unless you really push the 1100t above 4.3 Ghz single core performance will not be better than FX. You might aswell take the extra scaling (Phenom II X6 - 5.92 scaling , FX 8150 -6.66 scaling). Also, with a good enough cooling solution Bulldozer can reach clocks as high as 5.2 Ghz, pushing its single core performance above what Phenom II can achieve. The more you overclock FX , the better the OC benefits you, there doesnt seem to be a bottleneck wall at max OC, while Phenom II is generally known to suffer from this.

No, not really.

First of all, FX is barely doing 4.6-4.8 GHz stable when Phenom II can easily do 4-4.3 depending on if you have a good or bad chip. FX doesn't do 5GHz stable, we can only hope stepping and process improvements make that possible. Therefore, by your post, low single core performance is a reality. By the way, AMD was already behind intel considerably in single thread performance, Phenom II being "low single core performance" already.

Then you talk about the scaling, yep...BD scales ~6.66x vs Phenom II's 5.9x. However BD has 8 "cores" and Phenom II has 6. BD scales 83% efficient while Phenom II manages 98%. That's worse scaling.

I'm not sure what some people are smoking, but phase change is necessary for 5.2 GHz on Bulldozer for things other than maybe only running Cinebench at dangerous voltages.

[polyzp]

In terms of overclocking ability and ease,

4.6 Ghz is to FX as 4 Ghz is to 1100t
4.8 Ghz is to FX as 4.2 Ghz to 1100t
4.9 Ghz is to FX as 4.3 Ghz is to 1100t
5.0 Ghz is to 4.4 Ghz 1100t

Plus or minus 0.1 Ghz on each for error