AMD FX 8150 Revisited

CPU by leeghoofd @ 2011-11-01

After the hectic week of testing this brand new 8 core CPU from Advanced Micro Design, it's time to go a bit deeper. In this review we are gonna retest the Bulldozer CPU versus it's main rivals. Being AMD's own Thuban 1090T, Intels 2600K and the almighty Gulftown 990X CPU. Mainly because our test suite had to be slightly updated to give the new Zambezi architecture a shot to maybe show it's true potential. But most important to show some people the real deal. I've myself read through a few articles on AMD's latest flagship. To be honest some reviews made me wonder if they were really done or just a copy paste of the marketing slides. It was also kind of funny to see some renown websites include completely GPU bottlenecked game benchmarks. Kinda hard to tell the importance of the CPU part don't you think ? Even if they call it real world scenarios, it still made my eyebrows frown as they hardly used any game tests in older reviews. Why now include them ? So without boring you too much with my frustrations, let's get it on...

  • prev
  • next

GTX480 SLI Gaming...

In the game tests on page 2 we saw that we were closing in on the GPU wall. Raw CPU speed didn't help much more as the GPU was fully stuffed with polygons. Time to lift that bottleneck and include a 2nd GTX 480 GPU in the tests. Let's see if the AMD platform is still able to match the competitors and it's previous processors. The CPUs remain overclocked to deliver enough speed to the dual GPUs. Sadly I don't have access to a bigger than 24" monitor. So no extreme high resolution gaming, 1920 x 1080 was the native gaming resolution tested. This will probably will be another discussion point.  Yet it can clearly show the strenghts and/or weaknesses of a platform.

 

 

Futuremarks 3DMark 06 hardly scales by adding a 2nd GPU with the AMD lineup. The Intels score an impressive 9-10K extra, while the Thuban scales a bit better then the FX CPU. Both AMD CPUs were tested on the Crosshair V with 9911 bios. Take note that the 1090T will be even more efficient on the older ASUS 890FX boards. Not a good showing of the FX in 3DMark 06. Something newer now with a Vantage run.

 

 

The Vantage score was a real surprise to me. The Thuban still outperforms the FX "8" core CPU , the latter running at 600Mhz higher clock speeds. Interesting enough is the good CPU score set by the FX 8150, yet the GPU score is way below the 1090T. At first I thought I had a few bugged runs (results are the average out of three). So I reinstalled the 1090T and redid the Nvidia drivers. Sadly for the FX the same outcome. Is this an indication for games to stay way from the Bulldozers when using multiple GPU setups ? Let's run a few more tests...

 

 

Mafia II still runs better on the older overclocked technology. At stock clocks this game was pretty evenly matched accross the different stock CPUs. Though with the added CPU clockspeed and the extra GPU we see a healthy increase in FPS ( from 61 to +100 ), but the 1090T and Intels are pulling away from the FX.

 

 

Lost Planet II shares the same experience. Close to doubling the framerates from stock setups with a single graphics card. The SandyBridge 2600K is in a class of it's own here. Crysis II tested on hardcore mode next...

 

 

Crysis II benefitting from the extra GPU big time. All platforms giving tremendous power to the GPUs. Yet again bulldozer trailing the rest....

 

 

 

Far Cry 2 known to be a great scaling engine with extra CPU and GPU speeds. Close game again between the two AMD CPUs, the Intels remain untouchable. Racing F1 cars on Monte Carlo shedding a different light on the FX ?

 

 

+

 

F1 2011, does not benefit in our test version from the 2nd GPU. (EDIT : This is apparently is driver issue related as testing with a newer driver gives better results on all platforms ) with the newer driver So results remain pretty evenly matched.Time to give Dices latest First Person shooter some extra GPU power by adding a 2nd GTX480. While with a single GPU the average FPS were rather close, this time it shows a totally different outcome. We crancked up the game detail to the preset Ultra setting, yet due to monitor limitations, the resolution stays put at 1920 x 1080. Even though the AMDS manage fluid framerates, the raw power of the Intel design gives far more minimum FPS. Naturally if a bigger or dual monitor setup is used things will get more GPU bound again.

 

 

In Dirt 3 we drove around the Aspen track several times to see how the different platfroms reacted to the added GPU power. Very high FPS all round and finally a good FPS increase over the 1090T.

 

 

 

 

 

 

  • prev
  • next
Comment from Teemto @ 2011/11/02
Always been a fan of AMD, but Bulldozer can only be described as one thing:
Epic fail.
Comment from cesariuth @ 2011/11/02
Why you test only benchmark??
some space for test soft of the real world?
Comment from jmke @ 2011/11/02
yes why not test games or compression software or rendering performance ?

ow wait... you did
Comment from leeghoofd @ 2011/11/02
First up Cesariuth I don't have resources to real rendering software, secondly it is not my interest at all and I find it already very time consuming enough. The test suite in it's current state will highlight most of the performance issues or not from a CPU/platform.

Like said in the two conclusions, this CPU can be good with specially designed software. But in most apps, and for sure with older applications it will not be up to the task to handle even the competition in it's own ranks...

Let's hope for AMD Piledriver will get a bigger boost then the expected 20% ... Piledriver will probably be what Bulldozer should have been from the start...
Comment from Teemto @ 2011/11/02
I seriously doubt it. Optimizations can help but to bridge the gap with Intel (on Sandy Bridge - not even talking about Sandy Bridge-E) they'd have to have made serious design error in the initial design which somehow has crippled performance and which can then be fixed.

But how realistic is that?

But it's not just the performance, the power consumption of this thing is also not up to current standards.

So either they:

Scrap everything and start working on an FX2 with real cores instead of modules. But I doubt they have the funds to do it.

Or

Try to add a GPU to it and make a higher end APU? But that would've only worked if the FX power consumption had been ok.

Or

Stop with trying to make high end CPU's and focus entirely on their current APU products.

Maybe start looking into some ARM products?
Comment from cesariuth @ 2011/11/02
Thanks leeghoofd for your answer, but I think that BD have more performance in the real test vs SB , or benchmark don’t say much of the real performance of this chip...
Benchmark is always benchmark...but this is my opinion I don’t know if this is the opinion of all.

I hope that Pelidriver optimize more than 20% or close to this.

Thanks!
Comment from leeghoofd @ 2011/11/02
I know a benchmark is a benchmark, but one of the main reasons that the initial article had this follow up was due to the "rigorous" testing by some highly acclaimed websites... How to show a CPU works well : put the GPU on it's knees and see how they all score the same... that's ridiculous in my book... it was clearly in the reviewers guide shown how we should have done it it. But them proposed tests setups just looked to be more chosen to make a mainstream product look great...

Real rendering software doesn't cut my budget mate, I also have got no experience in that department. I know Flanker is doing some tests with BD at ExtremeSystems.org Though as always take everything with a grain of salt... I think you will find him in the AMD section
Comment from Massman @ 2011/11/03
It's funny how it's really hard to find these so-called "real-world" applications. No one is using those, but apparently only those hard-to-find apps are the only ones that are showing the "true power" of Bulldozer.

Fyi, you guys might want to read this: http://www.brightsideofnews.com/news...rformance.aspx
Comment from Teemto @ 2011/11/03
Seems there's others which are not afraid to tell the truth

http://www.atomicmpc.com.au/Review/2...sappoints.aspx
Comment from WebNavi @ 2011/11/29
After the hectic week of testing this brand new 8 core CPU from Advanced Micro Design, it's time to go a bit deeper. In this review we are gonna retest the Bulldozer CPU versus it's main rivals. Being AMD's own Thuban 1090T, Intels 2600K and the almighty Gulftown 990X CPU.

 

reply