8120 after 1 week

[Daveburt714]

Ok folks, I've had this 8120 for a little over a week now.
I'm not sure I have it dialed in for the best perfomance yet, but it is stable...
I decided to run my normal benchmark suite. They may be a little outdated and I didn't run multiple tests to make sure I got the best results, (the fact that I've had a few beers probably doesn't help), but these are all legit...

CB 11.5 w/stable clocks:


HWBot Heaven Extreme Bench. This was done with a 6870 oc'd to 1000/1095:


CB 10 64bit scores:



3DMark06 scores (kinda nice to see >2.6 Fps on Red Valley.)





Last but not least, WinRAR and WPrime scores (the 32m Wprime (seems a little low), but at least it shows some amount of stability ):


There's been alot of talk about how bad FX's performance is, but in reality, it's a nice chip, and it's alot of fun to OC!

In all honesty, if your short on cash and you already have an X6 it's probably not worth the upgrade... But all this talk of it being "EPIC FAIL" is BS...

[The Stilt]

Whats the maximum Prime95 Large FFT you can do with all cores enabled?

[Baam]

There's been alot of talk about how bad FX's performance is, but in reality, it's a nice chip, and it's alot of fun to OC!

In all honesty, if your short on cash and you already have an X6 it's probably not worth the upgrade... But all this talk of it being "EPIC FAIL" is BS...

I agree, tons of little tweaks and some probably untapped.
Have you tried to undervolt it and see what the max overclock is at lower voltages?

[BeepBeep2]

Dave, the CPU is an okay multi-thread CPU...however without resource sharing disabled it is a fail at < 4 threads. Power consumption on a 32nm process is also concerning, those are the kind of numbers we were seeing by GloFo's 45nm with a CPU that in many ways outperforms this one.

Good results though.

[FlanK3r]

n1 Dave, yours chip is in my eye :-D (any chance to change your chip for Gulftown or SB-E)?)

[charged3800z24]

Looks Great Dave. I agree it is a nice chip and has loads of tweeking fun. I'll be doing OCing test as soon as I get through My comparison testing. Takes a lot of time lol.

[dess]

Dave, you may want to experiment with 4CU/4C "mode" for games with 3-4 threads, and 2CU/4C mode (or perhaps 2CU/2C is better) for games with 2 threads (and maybe 1CU/2C or 1CU/1C for games with only one thread), because disabling CMT there is a decent performance boost! (Somewhat like disabling HT on Intel CPU's.) There is also more OC headroom from doing it or disabling whole CU's (it depends on the game which variation is best in the given case).

There is a fine topic about it here.

[Particle]

I have to agree. I bought an 8120 last week too and am actually pleasantly surprised by its performance now that I have it in hand. A lot of the launch day reviews were incompetently performed (perhaps understandably due to time) and cast an unrealistically pessimistic light on the chip. It's not the new top dog of CPUs, but it wasn't supposed to be either. It's a darn good piece of silicon in its own right.

That's not to say that I'm not still planning on going SB-E next month if it looks good. Regardless of how good the FX is, the application I spend myself using the most (and makes me stare at a wall wasting time) is optimized for Intel CPUs and isn't at all threaded. It completes in roughly half the time on their architecture.

[SEA]

Dave, question.
Did you turn off the c'n'q? Did you turn windows power mode to max performance?
I ask because i found this:
freq-graph-2.png
It shows processor frequency in cinebench r11.5 in single thread mode. So I suspect that c'n'q has significant performance hit.
Can you run tests for cool'nquiet on and off for single- and multi-core bench (not exactly CB R11.5, maybe other things would discover that) pls

P.S.
I found it here:
http://www.3dnews.ru/cpu/618486/index2.htm

[Zeus]

Seems like you have a lot of fun with your new cpu and it looks like you have a very good clocker too, although we haven't seen many other as a reference.

I'm curious what your temps are like with Prime 95 putting full load on all cores, is your watercooling up against it? How does it scale with let's say 1.45V?

The only FX cpu i have found available here in The Netherlands is a FX6100, no FX81x0 yet.

[Manicdan]

daveburt, you going to set up a turbo setting for single threaded stuff?

[Olivon]

I have to agree. I bought an 8120 last week too and am actually pleasantly surprised by its performance now that I have it in hand. A lot of the launch day reviews were incompetently performed (perhaps understandably due to time) and cast an unrealistically pessimistic light on the chip. It's not the new top dog of CPUs, but it wasn't supposed to be either. It's a darn good piece of silicon in its own right.

That's not to say that I'm not still planning on going SB-E next month if it looks good. Regardless of how good the FX is, the application I spend myself using the most (and makes me stare at a wall wasting time) is optimized for Intel CPUs and isn't at all threaded. It completes in roughly half the time on their architecture.

I'm sorry but Bulldozer FX is only good for AMD die hard fans ...
Globally, this CPU is a deception and brings almost nothing compared to X6 processor (as daveburt said fairly in the first post)
Worse price/perf/efficiency than its predecessor isn't good and you don't need to blame reviewers for that ...

[Dumo]

Awesome runs Dave

[keithlm]

A lot of the launch day reviews were incompetently performed (perhaps understandably due to time) and cast an unrealistically pessimistic light on the chip.

Actually in many reviews if you look at the results you find that in many cases BD is very competitive; but when you read the review you find that it is tainted by personal opinion. It is probable that if BD had been released with identical results to SB in all cases it would have been declared to be uber fail by many reviewers.

[Particle]

I'm sorry but Bulldozer FX is only good for AMD die hard fans

You're entitled to your opinion of course. Just remember that opinions are opinions, not facts. My opinion was more similar to yours before I actually tried one for myself on the bench. Now I'm of the opinion that your view is inaccurate.

[Dumo]

I'm sorry but Bulldozer FX is only good for AMD die hard fans ...
Globally, this CPU is a deception and brings almost nothing compared to X6 processor (as daveburt said fairly in the first post)
Worse price/perf/efficiency than its predecessor isn't good and you don't need to blame reviewers for that ...

Do you have BD on your bench table Olivion?

Personally, I'm not even read the reviews, but I'm starting to get to know this chip better because I own and benched it myself. I own a few high clocking 2600Ks, but I just refused to compare it with FX-8150...nothing good will come out of it.

Theres a lot of thread that focus on what you're talking about, just leave this thread as it is......Peeps posting about fun stuff while benching BD

Please post here when you already tested BD yourself

[Particle]

On the topic of fun benchmarking:

Last night I received and installed a Corsair H100 on the 8120 on my bench machine. It really helped get thermals under control. Before, if I'd try to do Cinebench at 4.8 GHz or so using 1.43V, temperatures under load would get healthily into the 80-90C bracket. As you might imagine, that's neither good nor stability inducing. Now, temps get into the 50-60C range under heavy load. The 8120 is CB 11.5 benchable at 4.94 GHz. I get 8.1 which is quite a bit better than I ever got with my 1100T overclocked as far as it could go for a CB run (4.1 GHz). Curiously, HTT does seem to have a slight impact. 253 x 19.5 (4.94 GHz) with slower RAM settings (DDR3-1680 about) scores about 0.1 higher than 200 x 24.5 (4.94 GHz since the board overclocks HTT about 1.5 MHz when set to 200) with DDR3-1866 (same timings).

[Manicdan]

Curiously, HTT does seem to have a slight impact. 253 x 19.5 (4.94 GHz) with slower RAM settings (DDR3-1680 about) scores about 0.1 higher than 200 x 24.5 (4.94 GHz since the board overclocks HTT about 1.5 MHz when set to 200) with DDR3-1866 (same timings).

what were your NB and HT speeds? sin saw best results and very noticable results when they were at 10x HTT, so 2530mhz on both might be a sweetspot for you

[Particle]

At 10x, the chip can't complete POST. I had to run one multiplier below that for CPU NB.

[Manicdan]

how about 247 x 20 or 241 x 20.5 to get a NB that can sync without a crash?

[Dumo]

Enabling C6 will help to keep load/heat low when idle in windows

[eRazorzEDGE]

Nice chip dave, I'm hella envious. My 8150 can't do CB 11.5 at above 4.8GHz, now matter what voltage. I'm really disappointed by it's performance in games as well. It's a good percentage below what my X6 1090T could produce.

@Particle, you said that upping your HTT from the base of 200 to 253 alone helped out your score. I'm just breaking down the math from what you said in the last two posts so I have a grasp on it exactly, no offense meant here.

On the core you have 200x24.5 = 4900 or 201.5x24.5 = 4936.75 (with the amount the boards tend to bump the HT clock at its base frequency)
Same HT clock as above and on the IMC and HT Link you have 201.5x11 = 2216.5 (assuming both IMC and HT Link multi's are set the same).

Now with the HT clock upped, you have 253x19.5 = 4933.5
Same HT clock as above and on the IMC and HT link you have 253x9 = 2277 (again, assuming both IMC and HT Link multi's are set the same and by going with "one multiplier below (10)" statement in previous post)

So with the math, it's not the HT clock by itself, but what else the HT clock affects. And you get almost a 3% increase in IMC and HT Link frequency which is probably a more likely reason for the better scores. I'm going to test at home with my 8150 on settings that achieve the exact same final clocks of all 4 main frequencies (cpu, imc, ht link, ram) just to see if there is an increase or not, for sure.

[Particle]

You might also enjoy testing what impact HT frequency alone makes (HTT * Mult) versus the HTT bump. I didn't test it myself, but others have said that a higher HT speed alone doesn't make a difference, just the reference clock.

[FlanK3r]

I'm sorry but Bulldozer FX is only good for AMD die hard fans ...
Globally, this CPU is a deception and brings almost nothing compared to X6 processor (as daveburt said fairly in the first post)
Worse price/perf/efficiency than its predecessor isn't good and you don't need to blame reviewers for that ...

Again, Im dont agree....I read most of reviews! Whats the main diference? Practice performance...Rendering is better than Thuban, big improvements I see in Adobe and practice compression and decompression. Example for me, every day I "working" with archives (movies etc). Some videocodecs, not all, but most of them. NExt Microsoft office. Everything is of every day use.

But I dont tell, FX is superchip etc but it is not also fail.

[dess]

Nice chip dave, I'm hella envious. My 8150 can't do CB 11.5 at above 4.8GHz, now matter what voltage. I'm really disappointed by it's performance in games as well. It's a good percentage below what my X6 1090T could produce.

If you want better performance, you could try 4CU/4C mode with games.