Return of AMD FX: My OC'd AMD FX 8150 with OC'd 6990 Review - First Results Up!

[polyzp]

Ok sorry you are right, by IPC i mean actual performance per core, regardless of the actual OC ghz, just around the ceiling of the cards performance.

ie.

i7 920 @ 3.9 Ghz = roughly near its limit
FX 8150 @ 4.8 = roughly near its limit

Now all i am saying is that based on this one H264 benchmark, when one core is utilized bulldozer actually manages to beat the i7 920 according to these posted results from the official website of the benchmark.


----


PCMARK 7 benchmarks!
Round 11 : PCMARK 7




RESULTS:


Pre-Patch VS. Post-Patch



Before Patch installation

After Patch installation

Comparison

We can see that PCMARK 7 is very happy with the Windows 7 FX Patch. The only performance decrease is the system storage score which is probably due to the use of my SSD. Garbage Collection seems to be doing its job however. The most notable increase in performance is in the computation Score, where the patch shows a +16.6% increase in performance. An honourable mention to the entertainment score as well, which noticed a +4.4% increase in performance.


WinRar / Cinebench 11.5 Revisited with Patch!




Cinebench 11.5 - with Patch


RESULTS:

When compared to without the patch we score +0.25% (from 7.90) higher in the CPU test, and +4.3% (from 72.95) in OpenGL score (6990 @ 990/1500 Mhz). The single core score does not show any increase in performance.


WinRar - with Patch


RESULTS:

We can see here that at stock 3.6 Ghz, the FX 8150 manages to benefit from the patch by +3.4% when compared to without, and running at 4.8 Ghz performance increases by +3.9%. Opposite of what the initial preliminary patch released by Microsoft showed, where WinRar performance managed to decrease.



7-Zip Benchmarks Revisited **Updated with Patch results**

7-Zip Benchmarks - With Patch

We remember FX being a beast in 7-zip, how will it fair with the patch?


RESULTS:

Over 100% more performance than i5 2500k @ 3.7 Ghz Turbo


As we can see here, FX manages to marginally benefit from the patch in Decompression only. Compression shows little to no improvement. 7-zip really shows Bulldozer's strength.

DIRT 3 Benchmarks! FX King?

ROUND 12 : DIRT 3 Benchmarks

RESULTS:

source: Tomshardware

As you can see DIRT 3 really takes advantage of FX architecture. The most notable comparison is with the 6990 @ stock settings 830/1250 Mhz. The AMD FX 8150 @ 4.8 Ghz manages to squeeze out 131.4 AVG FPS and 118.2 MIN FPS, while the intel i5 2500k @ 4.0 Ghz manages to only get 104.3 AVG FPS and 97.0 MIN FPS. Thats 26%/22% MORE FPS. I was even shocked to see this! Good Job AMD!

Also to be noted is the patch's modest improvement in FPS of 2.0%/3.6% for MIN/AVG FPS.

http://AMDFX.blogspot.com

 

[Don Karnage]

i7 920 @ 3.9 Ghz = roughly near its limit
FX 8150 @ 4.8 = roughly near its limit

I can't seem to find my 4.2Ghz screengrab. 4.2 was quite easy with a UD3 and good cooling

 

[RavenSEAL]

Very nice, i'm hope that they really improve the architecture with Piledriver by increasing the IPC and lowering the power usage.

Hopefully they keep AM3+ so my upgrade path doesn't get derailed

 

[polyzp]

By roughly i mean within 300 mhz or so error ^^

-----------



ROUND 13: TrueCrypt 7.1 Benchmark

RESULTS:


CPU: AMD FX 8150 @ 4.8 Ghz (SHOWN IN RED)

OS: Windows 7 x64 SP1

source: Pugetsystems

Intel vs. AMD


In the TrueCrypt 7.1 benchmark we can see that the AMD FX 8150 @ 4.8 Ghz beats an i7 990x @ 3.6 Ghz Turbo in all tests, and just trails the i7 3930k. Note that this is using Windows 7 x64 SP1. Comparing to an i7 2600k @ 3.6 Ghz Turbo, across all tests FX wins by an average of over +54%. Bulldozer's architecture is seemingly taken advantage of with this specific benchmark, but now we will look at Ubuntu Linux Performance.




CPU: AMD FX 8150 @ 3.9 Ghz Turbo

OS: Ubuntu 11.10

source: PCimpact


Here we can see that the AMD FX 8150 performs much much better at stock settings when compared to with Windows 7. Linux seems to be taking much more advantage of Bulldozer's architecture, and this just comes to show that optimization for Windows is not near completion, and shows us just what could be in store for Piledriver when it comes out.

AMD FX 8150 @ 3.9 Ghz Turbo



In Linux, FX @ 3.9 Ghz Turbo , FX manages to even significantly beat overclocked (at 4.8 Ghz) performance on Windows 7, and comes much closer to performing on-par with a 3960x. It would be interesting to see overclocked performance in Linux, as I suspect its drastic.

Round 14 : Sandra SiSoftware Benchmark Results!

CPU 1: Intel i7 2600k @ 4.3 Ghz
CPU 2: Intel i7 2600k @ 4.6 Ghz, DDR3 @ 2133 CL11
CPU 3: AMD FX 8150 @ 4.8 Ghz, Gskill DDR3 @ 2183 CL10

RESULTS:

Here we see that the FX Patch brings a decent boost in performance, averaging +5.17% across all 12 tests. The largest performance increase comes in the .NET Arithmetic - Dhrystone test, where we see a +24.6% difference. This is the most significant increase in performance I have yet to see for the FX patch.

Comparing my AMD FX 8150 @ 4.8 Ghz to an intel 2600k @ 4.6 Ghz for 10 of the 12 tests, and a 2600k @ 4.3 Ghz for 2 of the 12 tests, we see an average performance difference of -0.07% , implying that AMD is still not so behind in this notoriously Intel favoured benchmark. The reason that two of the tests were not carried out @ 4.6 Ghz in the .NET Arithmetic scores , but instead @ 4.3 Ghz , is because scores @ 4.6 Ghz were not included in the internal comparison benchmarks listed. Of course this will play into the averaged difference, so I suspect that the 2600k @ 4.8 Ghz should beat the FX 8150 @ 4.8 Ghz by an average of 5% give or take. I might re visit this later on with my own testing to confirm this.

We can really see that intel and AMD have different strengths and weaknesses across the 12 tests in this benchmark.

Also notice how memory scores for G.Skill Ram @ 2183 Mhz CL10 only trail the intel's score with Ram @ 2133 Mhz CL11 by -4.8%/-4.6%.

Alien Vs. Predator Revisited **With Patch**
​

Results:



source: Tomshardware

​

We can see here that the FX Windows 7 Patch brings about +1% in performance when compared to without it. This pushes AVP performance with a stock 6990 above that of with an intel 980x @ 4.0 Ghz by +4.3% without AA, and by +23.5% with 4xMSAA. I used Catalyst 12.1b for the pre-Patch scores, and 12.1 Final Build for post-Patch scores.


For Comparison's sake we may also examine results from HEXUS.net with Two 7950's @ 900/1250 in Crossfire and catalyst 12.1 Final Build (which is also what I used in my updated Patch FPS). This test is with a Stock i5 2500k @ 3.6 Ghz Turbo. It should be noted that this test is with 4xMSAA but also with 16xAF as well, which the above test with 4xMSAA lacks - in accordance to Tom's review. So take these results with a grain of salt! (Although AF affects fps minimally in this game, while AA affects it significantly).

​

source: HEXUS

​

As we can see, Two 7950's OCd @ 900/1250 in crossfire only score a measly 114.0 FPS with an i5 2500k @ 3.6 Ghz Turbo. Despite the fact that this test has AF enabled, the other settings and identical. My Patched FX 8150 @ 4.8 Ghz with 6990 @ 990/1500 scores 124.5 FPS , which is +9.2% higher FPS than the intel rig with 2 x 7950 OC @ 900/1250 in Crossfire. We can also see that FX with a 6990 @ 990/1500 scores +37.1% more fps than the i5 2500k @ 3.6 ghz with Two GTX 580's in SLI OC @ 797/1594.



​

http://AMDFX.blogspot.com

 

[polyzp]

First gaming results will be up tomorrow! AVP!! Check into my blog or this post to find out

 

[BallaTheFeared]

AVP is a gpu test, try to pick some stuff that isn't totally gpu bound too.

 

[polyzp]

What are some other games with built in benchmarks i could use?

 

[Sunburn74]

Those two things are what drove me to stay away from bulldozer this time around. I recently sold my X4 Phenom II and bought an X6 to tie me over. If I hadn't just invested so much (err, well it was a sale/budget build actually) into an AMD system I would have been going with intel all the way. Oh well, this rig still does everything I need it to do.

Have you ever heard the phrase "throwing good money after bad"?

 

[ctk1981]

All a matter of opinion. The machine does what I need and that's the important thing. Some of the tasks I do daily benefit from the additional cores the x6 offers. If I had gone intel, it would have been a quad core with ht which puts the price in a whole different category. So while I'm somewhat regretful, my wallet is a little happier.

 

[Netopia]

polyzp

Thanks for all the work you put into this and WELCOME TO THE FORUMS!

Joe

 

[polyzp]

Passmark benches up ^^

 

[BallaTheFeared]

The most noteable comparison is between my stock 6990 @ 880/1250 paired with my AMD FX 8150 @ 4.8 Ghz, and their stock 6990 with the exact same clocks paired with a 980x @ 3.6 Ghz Turbo. The difference in score is mostly due to a difference in Graphics Score, as the 980x generally destroys the FX in physics and combined results even at stock.


This tells us alot about where FX bottlenecks or not, and the answer seems to be NO, atleast when comparing to a 980x @ 3.6 Ghz. The difference in score is roughly 5%, where the FX is slightly favoured.

Your conclusion seems to be born out of a misunderstand of what the results mean.

What it means is in the only test that places any load on the cpu (all eight cores) and the gpu at the same time (combined) the bulldozer cpu is bottlenecking just as much as last gen Phenom II's were, despite its additional core advantage and clock speed advantage.

6309 for the 4GHz Phenom II, with lower gpu clocks, while the 8150 scores just 6333 a gain of just 24 points.

There also seems to be something wrong with either your gpus on the Intel systems, or your Intel chips. Perhaps a bit of both, your i5 @ 5.35GHz is slower than mine @ 5.27GHz. Both in Physics and Combined, despite the fact that at these clocks (see below) our gpus are nearly the same.

Of note here as well is that the 2011 rig at 4.75GHz is only a few points faster in the combined, it has six cores working on 12 threads, 3DMark11 should eat those up and produce much higher combined scores than what you're getting.

It may be a problem with CF, though. However it's still clear even with the rather odd results on the Intel side that both the Phenom II x6 @ 4Ghz and the Bulldozer x8 @ 4.8GHz are actually pretty much on par as far as bottlenecking goes despite the fact that 3DMark11 will use all eight cores.

For reference here is my i5-2500k /w my 470's the gpu score is pretty close to what you're getting, however in the combined test despite some minor cpu bottlenecking I'm still pulling a combined score of 9361 which is faster than your i5-2500k, i7-2600k, 990x, and almost as fast as the 3960x.

 

[Don Karnage]

Balla I thought those 470s of yours could do 930 core or am I thinking about someone else.

 

[BallaTheFeared]

They can do 930/940 depending on the game, with another 150 on the vram.

 

[IonusX]

What are some other games with built in benchmarks i could use?

street fighter 4
wh40k: DoWII
civ IV
shogun: total war (i think)

 

[cphelps]

Does anyone else find it strange that the PassMark scores for these processors are so disparate between Integer and FP performance? With 4 floating point cores, they are generally competitive with Sandy Bridge, but with 8 integer cores and pipelines, they are many, many percent behind?

I saw the same thing even on Windows 8 Developer Preview, although I think overall my CPU score went up about 12%. How can 8 integer cores do so poorly, and 4 floating point cores do so well?

Chris

 

[BallaTheFeared]

It has eight 128 bit FP cores, they just combine to make four for 256 bit AVX. It's called something like Flex FP or something I don't really remember anymore I stopped following bulldozer and forgot most of what I knew after it came out.

At least that's what JFAMD told us, dunno if it's still true.

AMD has always been competitive with Intel when it comes to FP and things like Linpacks, my 965 would get around 58 GFLOPs which is pretty close to a 1366 i7 with HT off, and my 1090T would get around 76, which is decent compared to "typically clocked" 980x'es.

Where AMD has struggled is integer performance, and bulldozer is really no exception. Integer is far more important in our usage than FP, which is why Intel does so much better in benchmarks like gaming.

 

[BD231]

Where AMD has struggled is integer performance, and bulldozer is really no exception. Integer is far more important in our usage than FP, which is why Intel does so much better in benchmarks like gaming.

Actually a very, very slow floating point operation that already has the data in cache will be *MANY* times faster than an integer operation where an integer needs to be copied from system memory (which is more often than not the case). This why there has been so much focus on x86 Floating Point performance over the years, and the main reason BD has such a large cache scheme. AMD literally tried to make their integer cores function more like FPU's, but failed.

Case in point there are many many more factors that determine why Intels chips are faster, memory interface, faster cache, shorter pipeline, process maturity, higher IPC. It's a lot more than just integer that's giving Intel the lead right now, and for the record gaming is almost entirely reliant on floating point performance not integer .

 

[BallaTheFeared]

for the record gaming is almost entirely reliant on floating point performance not integer .

Can you link me to some documentation on that? Not saying you're wrong, actually I just want to educate myself :awe:

 

[BD231]

http://www.karbosguide.com/books/pcarchitecture/chapter13.htm

Here's the nitty gritty of it for you. E-Mail, word doc's, spread sheets = Integer (all the boring stuff). FPU = anything FUN you could do with a computer (games, video, video editing, music, ect). As far as video encoding goes BD's massive cache and double fpu's are more the reason its competitive in that area.

 

[BallaTheFeared]

Yeah I saw that, but it's pretty old... Does it still hold true today?

 

[polyzp]

GPU score tests is closest to simulating a game than combined score. Tomorrow Revo benches!

 

[BallaTheFeared]

The games you play have no cpu load? Watch your cpu usage while running the gpu tests, more than likely your 8150 is running at 800MHz.

 

[BD231]

Yeah I saw that, but it's pretty old... Does it still hold true today?

3d accelerators do integer rendering, floating point calculations are handled by the CPU.

 

[BallaTheFeared]

Well that's what I get for listening to JFAMD... It makes sense though, because part of the problem with Skyrim's cpu performance was it used x87 FP code...

From ianpatt (a SKSE developer):
It replaces many calls to functions that would have been inlined if the compiler optimizer settings were set to something higher than "off" with actual inline code that nicely fits in the five bytes the jump instruction was previously taking up. It also overwrites several functions, replacing them with SSE implementations instead of the very painfully slow original x87 implementations.

>The More You Know<