AMD Bulldozer / FX-8150 Desktop Performance Review @ [H]ardOCP

zero2dash said:
Don't get me wrong ~ I totally wanted AMD to come out, guns blazing, and kick Intel square in the pants. I've wanted that to happen for at least 2 product cycles now.....so far, AMD's failed us (in that regard). I'm happy with my X4 640 (board+cpu combo @ MC for $150 I believe; great deal), I've been happy with my other AMD systems in the past (Tbird 1.4 baby :D). It's just sad to see that Intel is continuing to clobber them though. :(
Click to expand...

I had a 1.2 Tbird years ago as well and I loved it (had a thermaltake orb on it, looked sweet!).

That said, I agree I am slightly disappointed by the performance not going above and beyond Intel's offerings, but at this point, edging past them is good enough for me. If I was really concerned with getting the best performance EVAR!!!!1 then I would have gotten Intel's $1k cpu. I know that with this chip and it's flaws (if this review is a truly accurate depiction) I can still get damn close, and for that I'll be happy. And I get to support the company that has never steered me wrong or left me with a dead computer, performance be damned.
 
bcorcoran said:
I was referencing 8120 vs 8150, but your point stands. My gripe with complaining about the 8120's performance is that there IS an 8150 sitting right there which is rated for higher clocks. I don't see any logic in why the reviewer chose to focus on the 8120 and used that to come to his conclusion, other than that it was cheaper and *could be* (likely that it is) from the same die.

AMD doesn't just toss a coin to decide which chips get the 8120 or 8150 badge. If they did, they'd be in for some real bullshit involving many many lawyers. Without seeing any numbers I can all but guarantee you that the 8150 will overclock better than the 8120. Just like the Opteron 165 in my sig overclocked way better than the consumer-level equivalent at the time. It's not a direct relationship in that sense, but it's similar.

Fact of the matter is, and I know this isn't in the spirit of overclockers in general, but if you want the best performance and your goal in a review is to review the PROCESSOR and not the overclocking performance, then at LEAST focus on each version individually rather than focus mostly on the #2. There's no merit to saying "well, enthusiasts will probably buy this chip, so I'll focus on that". That's just doing a disservice to your readers.
Click to expand...

I will disagree with you in that when we spoke to Bulldozer we spoke the performance of the 8 core part when overclocked, not to simply the 8120 or 8150 model. If you go back and look, the 8150 and the 8120 offer the same performance when overclocked.

The O in HardOCP stands for "Overclockers.'"

You can disagree with how we presented Bulldozer to our readers, but it hardly changes the performance. I spelled out to our readers what they would be buying in the 8150. It is my opinion that the 8120 is where the value is and I made that clear. And I hardly suggested that a coin toss was involved in putting the model number on the silicon. Below is my quote on the conclusion page that negates your arguments.

"There is a silver lining to this. AMD's unlocked FX-8120 is to be priced at around $200 in retail. You noticed we did not spend a lot of time pontificating about the FX-8150? The reason is because, from an enthusiast standpoint, there is simply no reason to buy an 8150. The 8120 and the 8150 are the same silicon. Yes, AMD is doing some silicon culling looking for which ones will take less vCore to hit the 4.2GHz Turbo specifications, and of course those parts are going to get the 8150 badge. For $200 you can purchase an AMD FX 8 core processor that will overclock to the mid-4GHz speeds fairly easily."

HardOCP caters to enthusiasts, always have, always will.

And this is why we have the HardForum. It gives you a place to tell us we are wrong. I appreciate you delivering your criticism in an adult manner. That is always appreciated.
 
Stoly said:
Although I don't think it hurts conclusions, I gotta question the systems setup, more so the 4ghz "stock clocks" but also the overclocking results.

why no real stock clocks? why settle for 4ghz when none of the cpus tested run at that speed at stock? Its like mild overclocking vs real overclocking tests.

And then the OC results, I think they mean very little as it doesn't represent in any way the real world overclocking potential of either AMD nor intel cpus.

4.8 is hardly the top speed for a 2500k/2600k, and we don't know the real potential of AMD cpus yet.

I think the Gaming performance review did a much better job, that's the one that I'll take into account when deciding what will be my next upgrade.
Click to expand...

Many games nowadays are more GPU-heavy, thus reading a gaming review to see a CPU's performance doesn't seem like a great idea. The performance of a chip can depend on the specific game. Testing with a wide range CPU benchmarks, both synthetic and not, is a way to avoid the specificity. I think we'll move to a future with more cores, and it seems AMD and Intel both agree that it's inevitable, but even in the threaded applications where it should have shined the bulldozer didn't really pull ahead and was often bested by the thuban.

There's a billion tech sites out there that have tested the chips at stock speeds and reached the same conclusions: they underperform and are too expensive for their suggested retail price.
 
Rizen said:
So my question is, weren't these initially targeted to the server market first? Those power consumption numbers are horrendous, I suspect that will really hurt their server market...
Click to expand...

umm the power numbers where overlocked the review i read at tom's has the power usage at stock clocks and they are like 30watts more than a i5-2600k at load.


http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-22.html

i'm not personally sure what to make of that since ars didn't list non oc power consumption.

I posted last night... I'm not necessarily saying i'm thrilled with what I'm seeing ..but it is hardly the disaster (performance wise) some make out.. It is also a future looking design in that it shines in heavy multithreaded loads.

Personally i think BD is getting sort of sandbagged ... maybe AMD's own fault for hype and delay but it seems people's expectations where somewhat unrealistic for what BD would do..
 
Last edited:
It's getting sandbagged because it's slower in some instances then an architecture they've been running with for years, which costs less and runs cooler.

To top it off they pulled out the once fabled FX label that represented top performance. What has been shipped is anything but.
 
atarione said:
Personally i think BD is getting sort of sandbagged ... maybe AMD's own fault for hype and delay but it seems people's expectations where somewhat unrealistic for what BD would do..
Click to expand...

I would be the first person to defend it if there was any shred of anything good in this product launch. The fact of the matter is, AMD managed to squander ~2 billion transistors in this design, on their hot new 32nm fab process and it barely competes with CPUs half its size, even on well multithreaded benchmarks.

The result is, you get an expensive chip, that uses tons of power and it still under delivers in single thread benchmarks compared to Thuban or Deneb (almost 3 year old architecture).

There is not a single thing to commend AMD on when it comes to Bulldozer. I guess there is one thing. All the CPUs in the lineup will be multiplier unlocked. So they will probably be good to mess around with and impress your friends with LN2 overclocks.
 
whoa...and i thought i rushed into buying a 2500k this summer. Best DECISION EVER TO GO INTEL this time around...sad I was really hoping for AMD to pull a rabbit out of the bag and beat the living shit out of SB.
 
i would like to point out the FX is faster than thuban, or phenom II is all areas except single thread performance.

I don't know if the [H] team has even tested the FX cpu on the windows 8 beta, but it does show some interesting performance gains. That being said why did you only test with 1600 ddr3 ram? Not that it makes a huge difference but might as well show it in its best light.

I can understand the negative comments about Bulldozer. I honestly don't feel as if it is a fail from my own personal findings. I have only had my fx 8150 for a week, and there is certainly a lot of tweaking and such to do. I have managed to get to 5.1ghz across all 8 cores @ 1.5 vcore.

This being a gaming site, I am still surprised to see people "sandbagging" this processor, its overall gaming results are pretty good.
 
:eek: I was excited to see the benchmark and thought amd couldn t possibly be as bad as the rumours.... this feeling...... :( It's like the first phenom ..... that made me go core 2 duo. :eek:

and i never regretted it! :D I guess amd 's more focused on getting the fusion parts right....
what a shame :eek: :rolleyes:
 
ebduncan said:
i would like to point out the FX is faster than thuban, or phenom II is all areas except single thread performance.

I don't know if the [H] team has even tested the FX cpu on the windows 8 beta, but it does show some interesting performance gains. That being said why did you only test with 1600 ddr3 ram? Not that it makes a huge difference but might as well show it in its best light.

I can understand the negative comments about Bulldozer. I honestly don't feel as if it is a fail from my own personal findings. I have only had my fx 8150 for a week, and there is certainly a lot of tweaking and such to do. I have managed to get to 5.1ghz across all 8 cores @ 1.5 vcore.

This being a gaming site, I am still surprised to see people "sandbagging" this processor, its overall gaming results are pretty good.
Click to expand...

Core i3 isn't bad for gaming either and its a hundred bucks? please see that link AntiSocialMunky just posted.
 
ebduncan said:
i would like to point out the FX is faster than thuban, or phenom II is all areas except single thread performance.

I don't know if the [H] team has even tested the FX cpu on the windows 8 beta, but it does show some interesting performance gains. That being said why did you only test with 1600 ddr3 ram? Not that it makes a huge difference but might as well show it in its best light.

I can understand the negative comments about Bulldozer. I honestly don't feel as if it is a fail from my own personal findings. I have only had my fx 8150 for a week, and there is certainly a lot of tweaking and such to do. I have managed to get to 5.1ghz across all 8 cores @ 1.5 vcore.

This being a gaming site, I am still surprised to see people "sandbagging" this processor, its overall gaming results are pretty good.
Click to expand...

Because there is hardly any gain going over 1600MHZ...i just saw a website test from 1333MHZ all the way up to 2133MHZ...hardly any change at all. I'll post the link if i can find it.
 
sirmo said:
That really puts it into perspective. What was AMD thinking? Really?
Click to expand...

im impressed...with the a8 lol


been using them for friends/family and the new FX isnt "that" much faster and the a8 even has decent graphics at less than half the cost
 
I still can't believe people are justifying buying these things over a 2500K, it blows my fucking mind.
 
I am one of the suckers with a AM3+ 990FX motherboard with a placeholder Athlon II X2 waiting for Bulldozer (upgraded from FX60 with AGP). From my dated knowledge of processors and parallel computer design, I was a bit concerned when I read about Bulldozer's "module" design, but hey "what did I know". Having done some non-ridiculously parallel programs, I would prefer the simplicity of programming for full cores. But, heh, I assumed I could just turn off/ignore the crap like you can hyperthreading, make it act like a 4 core chip and performance would be good. Surely, AMD engineers performed massive architecture simulations proving the "module" concept. Given the performance results, it looks like Bulldozer architecture is a clusterf*k. I cannot stomach the incredibly bad single thread performance which I think is very important. Not all programs can be ridiculously parallel and I refuse to buy a parallel-only CPU. To think that my current cheap "placeholder" CPU is faster than Bulldozer per clock is just sickening. Either they wasted too much time on the "module" architecture instead of IPC or the "module" architecture itself does not promote single thread performance. And from what I have read on the net that the AMD imbecils were merely trying to "hold the line on IPC" (and failed). This probably means that low single thread performance is by design not a result of non-tuned compiler output code. If there is a "Mr Module" "genius in is his own mind" egomanicac at AMD, then he needs to be fired.

It looks like I will be ripping the new 990FX motherboard out of my case and going to Intel (who I hate). In the future, I will not be buying without reading benchmarks.
 
Guess I'll be sticking with my 1055t. Maybe piledriver will be the time to uprgade
 
AntiSocialMunky said:
http://www.anandtech.com/bench/Product/434?vs=289

Oh, the pain!

:D
Click to expand...

Heh, what's even worse is this one http://www.anandtech.com/bench/Product/434?vs=146 When I bought a 1090T from my buddy last december because I was tired of waiting for bulldozer, I was a little worried that when it was released with awesome mindblowing performance I'd regret it a bit. Seems they go toe to toe with each other, and I think the 1090 may even edge out bulldozer overall. Looking at those numbers I can sadly say that I made the right decision, and my only regret is that I don't regret it.
 
Spongebob, that must have been difficult to write. At least your damage was mainly psychological - and will pass. You'll probably get a good price for your 990FX board from a true believer who wants to support AMD with upgrade(?) dollars. I've a couple acquaintances at another forum who have their brand new AM3+ boards on their benches waiting for the AMD champion's appearing. As of this writing, they are waiting for the blessed processor to appear at their favorite etailer. The term they are offering is "good value." I really want to tell them that term would be better be applied to the 1055T.
 
My sympathy goes out to all the AMD AM3+ board owners that were waiting for this for an upgrade. I suggest you score a 1100T before the shit melts down @ AMD.
 
Good review.Glad I have gotten Intel already.This is some sad perfomance from AMD.
 
Tiporaro said:
Well, it officially took less than 12 hours to sub in "piledriver" in the years-long mantra of "bulldozer will be the time to upgrade"
Click to expand...

well I'm hardly going to spend $200 for minimal performance gains or well over $700 to switch to Sandy Bridge for minimal performance gains.

If bulldozer was epic I would have bought it even though I don't really need to, in a way it's fine, I'll stick with what I have.
 
I expected more as it shows that in lightly threaded applications and not multithreaded applications, shows the Bulldozer IPC weakness, in multithreaded scenarios, it can goe toe to toe with the 2500K and even match the 2600K in some scenarios, something that Phenom II 1100T only could dream off. I didn't expect bulldozer to be the Sandy Bridge killer, but at least more consistent with its performance. It impress suddenly when it matches the 2600K in MT scenarios and then it drops rock bottom in non MT scenarios, barely matching the old Phenom II X6.

AMD says that there's some performance issues with Windows 7 as it doesn't allign the data cache adequately and cannot differentiate between a core with shared cache and a core with non shared cache, apparently Windows 8 fixed this but oh well, I'm happy that I didn't wait for Buttdozer and bought my Sandy Witch xD
 
I have been looking about and something shocking seems to appear in the reviews, the speed of the cache, The L2 cache is running slower then the memory sub system at 1600MHz how come the L2 cache is running slower then the pyhsical memory and 2x slower then the L3 cache? If you still have you sample and if possible can you run a quick bench with the L2 cache turned off. I think it is a problem with the L2 cache or a TLB bug. Wasn't this ment to be corrected a few months ago?
 
Valshistixol said:
I still can't believe people are justifying buying these things over a 2500K, it blows my fucking mind.
Click to expand...

in some programs its faster than the 2600k while being cheaper.

in some programs its faster than the 2500k while bing more expensive.


The real problem with the chip is its power usage. But its inline with my phenom ii x6 and its faster than that chip
 
I am one of the suckers with a AM3+ 990FX motherboard with a placeholder Athlon II X2 waiting for Bulldozer
Click to expand...

Hey Sponge Bob, I was a victim too as I just ordered the Asus Crosshair V this past weekend. I haven't recieved the board yet, but definitely will return for a refund. I feel terribly sorry for Asus for making a hell of a board only to be mangled by a soup sandwich processor.
 
I'm still gonna wait for the head to head with intels lga2011 high end series before I buy this cpu.
 
eastmen said:
in some programs its faster than the 2600k while being cheaper.

in some programs its faster than the 2500k while bing more expensive.


The real problem with the chip is its power usage. But its inline with my phenom ii x6 and its faster than that chip
Click to expand...

And most of the time the 2500K is faster while being CHEAPER.
 
matteos said:
well I'm hardly going to spend $200 for minimal performance gains or well over $700 to switch to Sandy Bridge for minimal performance gains.

If bulldozer was epic I would have bought it even though I don't really need to, in a way it's fine, I'll stick with what I have.
Click to expand...

Oh, I was hardly saying you were incorrect for not wanting to jump on BD. Although, I do wonder how you come up with $700 for SB.
 
Dazz said:
I have been looking about and something shocking seems to appear in the reviews, the speed of the cache, The L2 cache is running slower then the memory sub system at 1600MHz how come the L2 cache is running slower then the pyhsical memory and 2x slower then the L3 cache? If you still have you sample and if possible can you run a quick bench with the L2 cache turned off. I think it is a problem with the L2 cache or a TLB bug. Wasn't this ment to be corrected a few months ago?
Click to expand...

It can't be compared as memory works differently from caches as the latter has more associativities and some performance implications in the way that it works that makes it more faster than any RAM. Here's the results you requested.

http://www.xbitlabs.com/articles/cpu/display/amd-fx-8150_8.html
 
Valshistixol said:
And most of the time the 2500K is faster while being CHEAPER.
Click to expand...

If you multithread a lot and don't play lot of games and simple lightly threaded software, then 2500K is for you. If you are a poweruser which requires a lot of multithreaded performance and don't care much about gaming performance, then BD is better, if you care about gaming performance and need multithreaded performance, get 2600K.
 
Valshistixol said:
And most of the time the 2500K is faster while being CHEAPER.
Click to expand...

i don't see that .

taking out the synthetic benchmarks Kyle did in the article there are 11 benchmarks done .

in 6 of the benchmarks its either between the i7 2600k and i7 2500 k or beats both .

In 5 benchmarks its slower


in x264 pass 2 at stock speeds it gets 37.12 the 2500k @4.8ghz (which is 1ghz faster ) only gets 39.43 . The only chip faster is the i7 990x .


I mean the numbers are in the article .

The main problem with the fx is that it uses a ton of power over the sandy bridge chips.

I have a m5a99x evo board so its much cheaper for me to buy an fx 8120 and clock it to 4.5ghz than it would be to buy a new board and sandy bridge . The extra power used is only when gaming or placing load on the cpu. It will take years for the price diffrence to come into play as a serious consideration.

Long before that happens we will be down to 22nm chips and i will have long since left bulldozer and sandy bridge behind. I upgrade cpus every 18 to 22 months. So I'm looking at 2013 chips


I game at eyefinity levels with a single gpu ( i wont ever have 2 gpus its to much of a headache allwyas consantly needing drivers to play a game) I'm going to be gpu bound more often than i'm cpu bound and i can allways increase gpu settings if i find myself suddenly cpu bound. However look at battlefield or rage or the upcoming games , they are all going to be using more and more cores. So single thread or dual thread performance will matter less and less.

Next year i will replace my 6950@6970 levels with what ever is in the $250 - $300 price range of the 7x00 series and the new games will put me right back into being gpu limited
 
@eastmen I would seriously wait to buy it though. This CPU can't sell well, price drops must be imminent. Unless AMD doesn't even care about what's left of their market share.
 
sirmo said:
@eastmen I would seriously wait to buy it though. This CPU can't sell well, price drops must be imminent. Unless AMD doesn't even care about what's left of their market share.
Click to expand...

i'm not buying it today . I'm going to florida on saturday so i wont be buying anything until nov. I'm thinking the fx 8120 will be $200 by then and I will grab it.

My friend wants my 1090t and is going to pass me $150 for it as he is replacing an athlon II x4.

So for me the fx 8120 at $50 more than the 1090t is a no brainer.


Its really not as bad as people think guys. The power drain is ungodly but the performance is actually in line with sandy bridge. They came a long way from the phenom
 
Truly a sad day for AMD desktop fans. Interlagos makes a fine offering for the server market but Bulldozer does not competitively stand up to Sandy Bridge. It's not even in the same ballpark. I've been running AMD system builds exclusively since the release of the 5x86 150mhz cpu back in 1995, 16 years with AMD! Since 1995 the ONLY AMD cpu I have not personally owned is the first generation Phenom cpu's and sadly Bulldozer will make two. :(

Like so many others, I too ran out and upgraded my motherboard to a 990x Bulldozer ready board, disappointed I did not wait. I despise Intel's vile business practices and out of principle I've done everything I can to support AMD throughout the years. I've built 1000's of AMD systems over the years for family, friends, and customers. For the first time, I cannot recommend a AMD system build over a Intel (Sandy Bridge) setup for gamers and most power users. Intel has a superior price/performance CPU with Sandy Bridge and nothing in AMD's lineup competes.

AMD, you need a "Return of the Jedi" with Piledriver. Please dont make me join the dark side (Intel). Principles or not, you still need to earn my loyalty I wont tolerate a second consecutive disappointment with these kind of numbers against Intel's offerings when Piledriver is released.
 
Tiporaro said:
Oh, I was hardly saying you were incorrect for not wanting to jump on BD. Although, I do wonder how you come up with $700 for SB.
Click to expand...

processor, mobo, possibly RAM and possibly new win 7. A quick estimate

Reading through the reviews it seems like bulldozer has promise for the future but doesn't deliver now, I wonder if it will be anything like the first Phenom vs Phenom II. I also wonder how Windows 8 will change its performance. It's a shame, seeing as how Llano is a game changer that bulldozer is anything but.
 
You must log in or register to reply here.
Back
Top