AMD Bulldozer FX-8120 Benchmarks

[Idontcare]

y'all guys just mad that you didn't wait for bulldozer and bought into sandy bridge. Who cares if it performs terribly, i'm going to have twice as many cores as you fools

lol

 

[inf64]

This can't be real, no way!

Welcome to the discussion.We have been saying this in like 4 different topics...

 

[SickBeast]

I've given up on waiting for BD to come out. Someone please send me a PM when the official reviews are up.

I will be on hiatus until that happens.

 

[3DVagabond]

Your first clue that benches are not real is that the "reliable" sites aren't touting them. People that we know have reliable sources (We'll use Anand as an example, but there are others. You all can insert your favorite. ) are not confirming these leaks. People who are typically under NDA or have reliable sources that they don't want to give info that would flesh them out.

Typically, when someone is under NDA they can't reveal any info they get from the source of the NDA. Once it's out though, they are able to report generally known info.

 

[BlueBlazer]

Your first clue that benches are not real is that the "reliable" sites aren't touting them. People that we know have reliable sources (We'll use Anand as an example, but there are others. You all can insert your favorite. ) are not confirming these leaks. People who are typically under NDA or have reliable sources that they don't want to give info that would flesh them out.

IMHO actually these are pretty much real. I did a background check on the poster before I post (to make sure its not another fraudster). The second one especially is even more reliable. The original here >> AMD Bulldozer FX-8120 獨家曝光！ (still exist) was in the Chinese VR-Zone site, under Product Testing section. This was also in the English VR-Zone site here >> Early Benchmarks of AMD Bulldozer FX-8120 Zambezi B2 Stepping but the article was pulled out just a few hours (still can be found under news search engine but the link is dead). Cached snippet quotes (from Hardware.info)....

Our team in Taiwan have obtained a sample of the latest B2 revision (rumored retail stepping) of the enthusiast targeted AMD "Bulldozer" FX-8120 (8-core 3.1GHz codenamed "Zambezi" chip) and ran a few benchmarks on it to see how it fared in multithreaded scenarios.

Just like the Sisoftware leak, this one got pulled out either due to NDA violation or complaint (speculations). :hmm:

Typically, when someone is under NDA they can't reveal any info they get from the source of the NDA. Once it's out though, they are able to report generally known info.

Whether these people are under NDA (like the earlier B0 stepping leaks) is unknown.

 

[3DVagabond]

Things can be pulled because they are deemed to be fake. We don't know why they were pulled?

I'm not talking about the leakers being under NDA. I'm figuring they aren't. My point is that someone who is under NDA and couldn't leak the info themselves could report it after it became general knowledge.

Just to illustrate. Let's make believe I was under NDA and had a web news outlet. A reviewer, for example. I can't report anything I've been told from official sources or that I might have tested myself. Once someone else leaked it though, I could then report their leak. If I knew it wasn't true, though, why bother to help spread Fud? How many legit news sites have we seen post any of these leaked benches?

 

[Bill Brasky]

I'm regularly disappointed in the benchmarking abilities of those in Asia with pre-released hardware. Is not possible to run something meaningful with the hardware?

I've always wondered this. Why does the leaked hardware always seem to wind up in the hands of an incompetent boob?

 

[anikhtos]

I think buldozer name is apropriate.
Amd hoped that it will demolized intel.
But unfortunetly buldozer will demolish amd!!!!
i do not know if amd can survive after a fiasco of bulldozer.
As it seems more and more that buldozer will fail hard.
Amd for sale soon???

 

[trollolo]

I think buldozer name is apropriate.
Amd hoped that it will demolized intel.
But unfortunetly buldozer will demolish amd!!!!
i do not know if amd can survive after a fiasco of bulldozer.
As it seems more and more that buldozer will fail hard.
Amd for sale soon???

not before they have a firesale of their bulldozer chip. i'd buy one for 50$

 

[Spikesoldier]

we'll see. but those numbers dont look good at all.

 

[Idontcare]

I've always wondered this. Why does the leaked hardware always seem to wind up in the hands of an incompetent boob?

I think its the other way around, hardware gets seeded to lots and lots of people, but its only the incompetent boobs who'd dare risk their very job and professional future by leaking a few useless/meaningless results based on ES hardware and platform.

 

[Idontcare]

Things can be pulled because they are deemed to be fake. We don't know why they were pulled?

I'm not talking about the leakers being under NDA. I'm figuring they aren't. My point is that someone who is under NDA and couldn't leak the info themselves could report it after it became general knowledge.

Just to illustrate. Let's make believe I was under NDA and had a web news outlet. A reviewer, for example. I can't report anything I've been told from official sources or that I might have tested myself. Once someone else leaked it though, I could then report their leak. If I knew it wasn't true, though, why bother to help spread Fud? How many legit news sites have we seen post any of these leaked benches?

I've seen this happen firsthand.

If you know who is who, and you know to watch their posts, then you'll recognize they intentionally wait for material to become "public domain" before they speak to it. Prior to it becoming public domain they will avoid speaking about the subject itself, for good reason.

There's a few members of this forum who are "in the know", but they can't speak to the topic until something on the topic becomes public domain.

But even then, NDA's prevent you from confirming anything that violates the NDA.

If person A leaks benches, and person B is under NDA but knows the results of the benches are valid, person B is not allowed (per NDA restrictions) to comment on person A's leaked benches in any way that would either confirm or disprove them.

At most, person B can dance around the topic by liberally pointing out that the leaked benches are on ES platforms and hardware and as such the results could be real but still not representative of retail stepping products...

(sound familiar?)

The challenge for the non-insider in the industry is to figure out who to pay attention to and who to wholely ignore, because the posts from both types of individuals will read seemingly similar...but the people who are in the know have to keep their responses appropriately veiled.

Or so I am led to believe...

 

[BlueBlazer]

Things can be pulled because they are deemed to be fake. We don't know why they were pulled?

I'm not talking about the leakers being under NDA. I'm figuring they aren't. My point is that someone who is under NDA and couldn't leak the info themselves could report it after it became general knowledge.

Just saw this at XS. Does that answer your question?

someone deleted my thread in VRFORUMS (english version). Saying that they need to clarify with AMD what's wrong with the result. Maybe some L3 prefetching and CnQ microcode issue in BIOS?
see the screen shot.

we'll see. but those numbers dont look good at all.

Actually, if you look at the Cinebench R11.5 scores then it came very close to Nehalem in the multi-threaded scenario (if you consider Bulldozer "cores" as threads). Unfortunately Nehalem was more than 3 years ago and Intel is a "moving target".

 

[Arg Clin]

As many have pointed out - these leaks lack common sense and realism. a Phenom II X6 can beat BD, why even bother? AMD could just have given Ph II a die shrink instead.

At worst they are pure fakes, at best the represent a shapshot of the performance of an unfinished product, which is utterly meaningless, since it won't have any bearing on end user experience.

 

[TurtleCrusher]

Im calling shens louder than ever on this one. As stated before, if Phenom 2 could easily beat it, BD wouldnt have made it to ES production unless is was extremely, extremely cheap to manufacture.

Lets get real here. These are bullshit benchmarks. Its a good joke if ive ever heard one.

 

[classy]

I don't believe for one second AMD would release a cpu that is slower than the ones they already have. These may be benchmarks of true hardware, but clearly without bug fixes. K7 early on didn't look much better than K63 either, but when all was said and done it shined.

 

[formulav8]

I can't see AMD going this far with BD if it couldn't beat their current top chip. Simulations would have long ago helped them to realize the potential of BD. The simulations couldn't be That far off could it?

 

[jhu]

I can't see AMD going this far with BD if it couldn't beat their current top chip. Simulations would have long ago helped them to realize the potential of BD. The simulations couldn't be That far off could it?

Intel has done it twice (Pentium 4 and Itanium). Hopefully AMD doesn't follow the same path.

 

[frostedflakes]

As many have pointed out - these leaks lack common sense and realism. a Phenom II X6 can beat BD, why even bother? AMD could just have given Ph II a die shrink instead.

At worst they are pure fakes, at best the represent a shapshot of the performance of an unfinished product, which is utterly meaningless, since it won't have any bearing on end user experience.

I don't think it's inconceivable that BD could be slower than X6. Remember that P4 Willamette was slower than P3 when it was released, for example. It had the potential to be better, but it took a little while for the process to mature to the point that they could hit the clocks necessary for the architecture to be competitive.

Due to the shared resources approach, all else being equal a 3 module BD will be slower than an X6. And AMD added more pipeline stages with Bulldozer as well I think, which will allow for higher clocks but also have a negative effect on IPC. Of course the increased cache, improved prefetch and decode, and other tweaks will help make up for the IPC loss from the shared cores and longer pipeline. How much these tweaks will offset IPC loss from other architectural decisions seems to be the one million dollar question, though. I have a bad feeling that performance per clock and per core will end up being lower than Phenom II. With all the delays and silicon revisions it seems clear that Bulldozer's success hinges on it being able to hit relatively high clocks. Which isn't necessarily a bad thing and doesn't preclude BD from being successful in the long run if the clocks continue to scale well as their 32nm process matures. But it's kind of disappointing as I'd rather see more of a focus on efficiency and IPC.

 

[inf64]

I can't see AMD going this far with BD if it couldn't beat their current top chip. Simulations would have long ago helped them to realize the potential of BD. The simulations couldn't be That far off could it?

Well simulations are one thing and real silicon is another. Anyway,they knew what they are up against.It was going to be Nehalem's successor.So they knew performance levels of both Nehalem and Westmere(6C/12T). If they designed this thing to be just barely faster than X6 (Thuban/Lisbon) with 8 thread versus 6,then they must have know how will it fair against intel's cores in 2011(read: not too well).

 

[BlueBlazer]

Prices from BLT....

FX-6100 = $188.32
FX-8120 = $221.73
FX-8150 = $266.28

These do seem to be close (and similar) to TankGuys Bulldozer "pre-order" prices....

FX6100 X6 - $204.99
FX8120 x8 - $236.99
FX8150 x8 - $259.99

 

[RobertPters77]

not before they have a firesale of their bulldozer chip. i'd buy one for 50$

Smartest thing I've heard about BD in a while.

P.S. Given up waiting for BD. Happy with my brand new i5-2500k(Best 150$ Clearance Item I bought).

 

[RobertPters77]

Intel has done it twice (Pentium 4 and Itanium). Hopefully AMD doesn't follow the same path.

Yeah, When they thought they were ahead of the competition. Or when they focused on a single metric to rate performance. Amd with it's 'More Cores!' approach is making the same Mistake Intel made 10 years ago with 'More Clocks!'

 

[sequoia464]

Smartest thing I've heard about BD in a while.

P.S. Given up waiting for BD. Happy with my brand new i5-2500k(Best 150$ Clearance Item I bought).

I'm kicking myself right now - I had a $150 i5-2500K from Fry's that I returned. Best price I'm seeing now is over $200.

 

[Idontcare]

I'm kicking myself right now - I had a $150 i5-2500K from Fry's that I returned. Best price I'm seeing now is over $200.

Nah, don't be hard on yourself, with computer tech the best deals are when you don't spend any money and you wait for the next round of releases. 6 months from now you'd be looking at your 2500K wishing you'd have saved the $150 to put towards an IB system.