Bulldozer Benchmark leaked

my guess is this person isn't "new" but didn't want to put up his real alias incase these pix turned out to be BS.
 
Dont have stock speeds, but got a 4.0 Ghz speed on 11.5- 6.93

http://hardforum.com/showthread.php?t=1519096

I did testing when i first got my thuban, coulda sworn I had stock speeds in there as well....if you want i can drop the speed later today and run some Cinebench for ya(drop me a PM if your actually interested)

CaptNumbNutz said:
Anyone know the Cinebench 11.5 scores for an AMD 1100T or 1090T at stock, turbo and even OC speeds for comparison? I'm seeing around 5.7's to 5.9's at stock speeds...Something seems off.

Over at XtremeSystems, they have a chart listing some scores for the Thubans:
1090T @3.2ghz = 5.64
1055T @3.4ghz = 5.95
Click to expand...
 
Ugh... I wish they would just release this thing already. Though I'm not due to retire my C2D Conroe just yet. Not that I probably need it, but it'd be pretty sweet going from 2 to 8 cores. :)
 
tangoseal said:
A whole bunch of Bullshit coming out of China or wherever,

LAMO once again let Intel leak some bullshit benches starting their corporate smear campaign.

What the hell makes you think that AMD would even want to leak some crap like this.

This same shit happens almost every new release. Man you people suck if you really believe this.

And those benches look like turbo is disabled on the AMD but not the Intel. Gee... that is really freaking fair.
Click to expand...

Were you around in the months/weeks before "barcelona" was released? Let me tell you or refresh your memory as to what happened.

AMD was very quiet about the actual performance. Only thing we heard about it was "native quad core" the AMD fans jumped on this (myself included admittedly) because that's all they had and well, it sounded good.

Benchmarks and power consumption were leaked. It did not look good. AMD fans claimed BS and they'll wait for real reviews/benchmarks.

Reviews/Benchmarks of production chips were finally available, they more or less mimicked what the leaked reviews showed. Rational AMD fans like myself were disappointed and looked towards Intel. Die hard AMD fans said that it's just a matter of time before barcelona overtakes Core 2 Quad because quad cores were relatively new and most apps weren't properly taking advantage of them... Several years later, they're still waiting for those "optimizations" apparently.

Sure, TDP numbers may change but overall performance is going to be roughly what you see here. You're only going to disappoint yourself if you believe the "official" performance figures are going to be much different.
 
I think it's worth pointing out at this point that this is like the third set of leaked benchmarkd we have seen and I know every time somthing is leaked here the cool thing to do is to cry fake as loudly as possible but all the leaked benches have pretty much shown the same thing namely bulldozer being a hughe step up from phenom II but still only equaling or beating SB in bench makrs that are heavily multi threaded and falling behind by 20+ % in all others.

were lloking at multuple sets of leaked benches that all pretty much show the same thing I suppose it's possible that its all a huge conspiracy but that sure seems unlikely. Now couple that with the constant delays and AMD's refusal to show us anything other than !!!Dirt 3!!! performance and it paints a pretty poor picture for bulldozer's ability to compete with sandy bridge performance wise.

I want AMD to succeed and I really want to believe that BD will be every bit as good as or better than SB but for me to actually believe that would require an act of blind faith on my part that is just not realistic.

Lets be real guys if AMD really had a chip on their hands that could crush SB at this point as close as it is to supposedly releasing they would be publishing bench mark results themselves and handing out early review samples to get people hyped for release not hiding and showing us games that would run just as well on an Athalon X2.
 
kinjo said:
I think it's worth pointing out at this point that this is like the third set of leaked benchmarkd we have seen and I know every time somthing is leaked here the cool thing to do is to cry fake as loudly as possible but all the leaked benches have pretty much shown the same thing namely bulldozer being a hughe step up from phenom II but still only equaling or beating SB in bench makrs that are heavily multi threaded and falling behind by 20+ % in all others.

were lloking at multuple sets of leaked benches that all pretty much show the same thing I suppose it's possible that its all a huge conspiracy but that sure seems unlikely. Now couple that with the constant delays and AMD's refusal to show us anything other than !!!Dirt 3!!! performance and it paints a pretty poor picture for bulldozer's ability to compete with sandy bridge performance wise.

I want AMD to succeed and I really want to believe that BD will be every bit as good as or better than SB but for me to actually believe that would require an act of blind faith on my part that is just not realistic.

Lets be real guys if AMD really had a chip on their hands that could crush SB at this point as close as it is to supposedly releasing they would be publishing bench mark results themselves and handing out early review samples to get people hyped for release not hiding and showing us games that would run just as well on an Athalon X2.
Click to expand...

All seemed well...except the Cinebench scores were exactly what a six core Thuban 1100T would get at stock speed. That part is not "a huge step up from Phenom II". The FX-8150 has 8 cores. Cinebench also scales surprisingly well with more cores. The OP was also a 2 day noobie that joined just to post this thread...something seemed off. The rest of it seemed ok.
 
it is amazing how AMD once again dissapoints its people.

this shit wont be worth it.

next build -> Intel again!
 
what amazes me is that everyone is jumping over every piece of crap posted on the whole interweb, of which 100% has to be complete BS. Only those privvy to actual hardware know how it performs and they are not allowed to say. only when [H] has a review can we comment properly and the appropriate fanboi's claim the glory. in the meantime may i suggest that everyone takes a deep breath and chill out;)
 
You can be sure it is not faked.
Click to expand...

You can be SURE of this guys....I just got my chip from newegg and my numbers are the same.

...wait...the chip isnt out yet. Guess we can't be sure of anything yet.
 
CaptNumbNutz said:
All seemed well...except the Cinebench scores were exactly what a six core Thuban 1100T would get at stock speed. That part is not "a huge step up from Phenom II". The FX-8150 has 8 cores. Cinebench also scales surprisingly well with more cores. The OP was also a 2 day noobie that joined just to post this thread...something seemed off. The rest of it seemed ok.
Click to expand...

Once again:

Thuban core!= Bulldozer core

Cinebench is a floating-point heavy Benchmark. Thuban contains more FP units than Bulldozer. Now call your facts. FX-8150 Bulldozer has 4 modules, you have two Integer cores in every module but you don't have two FP units in it. It's not a traditional Octacore where all units are exclusive for every core. That's why some of the insiders calling it Quadcore not Octacore. We will see how it scales. And with turbo enabled FX-8150 is faster in Cinebench than 1100T with less FP units. Not huge but easily faster at least.
 
Paran said:
Once again:

Thuban core!= Bulldozer core

Cinebench is a floating-point heavy Benchmark. Thuban contains more FP units than Bulldozer. Now call your facts. FX-8150 Bulldozer has 4 modules, you have two Integer cores in every module but you don't have two FP units in it. It's not a traditional Octacore where all units are exclusive for every core. That's why some of the insiders calling it Quadcore not Octacore. We will see how it scales. And with turbo enabled FX-8150 is faster in Cinebench than 1100T with less FP units. Not huge but easily faster at least.
Click to expand...

True, Thuban has more FPU cores. 6x 128 bit, Bulldozer has 4x 256 bit, ergo it can process 8x 128 bit code. So Bulldozer can process 8x 128 bit interger and floating point operations per clock, and Thuban can process 6x 128 bit interger and floating point operations per clock.
 
Paran said:
Once again:

Thuban core!= Bulldozer core

Cinebench is a floating-point heavy Benchmark. Thuban contains more FP units than Bulldozer. Now call your facts. FX-8150 Bulldozer has 4 modules, you have two Integer cores in every module but you don't have two FP units in it. It's not a traditional Octacore where all units are exclusive for every core. That's why some of the insiders calling it Quadcore not Octacore. We will see how it scales. And with turbo enabled FX-8150 is faster in Cinebench than 1100T with less FP units. Not huge but easily faster at least.
Click to expand...

Not less FP units, it can act as two 128-bit FP units or one 256-bit FP unit unless I misunderstand how it is supposed to work.

Bulldozer -
"Two symmetrical 128-bit FMAC (fused multiply–add capability) floating-point pipelines per module that can be unified into one large 256-bit-wide unit if one of integer cores dispatch AVX instruction and two symmetrical x87/MMX/SSE capable FPPs for backward compatibility with SSE2 non-optimized software"

Phenom II -
"■128-bit floating point unit (FPU)
■High performance (128bit internal data path) floating point unit per core. ◦Benefit: Larger data paths and quicker floating point calculations for better performance."
 
Paran said:
Once again:

Thuban core!= Bulldozer core

Cinebench is a floating-point heavy Benchmark. Thuban contains more FP units than Bulldozer. Now call your facts. FX-8150 Bulldozer has 4 modules, you have two Integer cores in every module but you don't have two FP units in it. It's not a traditional Octacore where all units are exclusive for every core. That's why some of the insiders calling it Quadcore not Octacore. We will see how it scales. And with turbo enabled FX-8150 is faster in Cinebench than 1100T with less FP units. Not huge but easily faster at least.
Click to expand...

Now we know you are a troll. Thanks for removing all doubt!
 
Paran said:
Once again:

Thuban core!= Bulldozer core

Cinebench is a floating-point heavy Benchmark. Thuban contains more FP units than Bulldozer. Now call your facts. FX-8150 Bulldozer has 4 modules, you have two Integer cores in every module but you don't have two FP units in it. It's not a traditional Octacore where all units are exclusive for every core. That's why some of the insiders calling it Quadcore not Octacore. We will see how it scales. And with turbo enabled FX-8150 is faster in Cinebench than 1100T with less FP units. Not huge but easily faster at least.
Click to expand...
No shit a Thuban core is not a Bulldozer core. I am very well aware of that. All I'm saying is it doesn't seem right that an 8 core/8 thread cpu from a newer generation performs exactly the same as a 6 core/6 thread cpu from the older generation on a benchmark that is known to scale incredibly well with more threads and cores. Combine all that with the fact that you just joined 2 days ago and all your posts are in this thread, "Bullshit" seems like the appropriate word to describe all this. I'm seeing 1100T's across the net scoring in the 5.9 range on cinebench 11.5. This FX-8150 that you showed scored a 5.93. That is the score that was posted before you took the pics down in a fit of rage. Honestly, I think you should have left the pics and link up. Then when the product was released, you would have been proven right. But since you pulled them, it seems even more fishy. If these scores are indeed proven correct, I will gladly admit I am wrong and back down.

Untitledone said:
True, Thuban has more FPU cores. 6x 128 bit, Bulldozer has 4x 256 bit, ergo it can process 8x 128 bit code. So Bulldozer can process 8x 128 bit interger and floating point operations per clock, and Thuban can process 6x 128 bit interger and floating point operations per clock.
Click to expand...
Exactly... Unless the FX-8150 took a major step back in performance, something seems really off. Even if the performance of each core in an FX-8150 is exactly the same as a core in a Phenom II or Thuban, there are more of them.

Is Cinebench 11.5 using 256bit FPU code? That might have made the difference...
 
Last edited:
Don't get too excited over rumors, wait until they've actually been shipped out and only then will we get real benchmarks. That been said, don't expect too much since this chip has been delayed over and over, meaning performance is lacking (due to design and/or fabrication quality).
 
cyclone3d said:
Not less FP units, it can act as two 128-bit FP units or one 256-bit FP unit unless I misunderstand how it is supposed to work.

Bulldozer -
"Two symmetrical 128-bit FMAC (fused multiply–add capability) floating-point pipelines per module that can be unified into one large 256-bit-wide unit if one of integer cores dispatch AVX instruction and two symmetrical x87/MMX/SSE capable FPPs for backward compatibility with SSE2 non-optimized software"

Phenom II -
"■128-bit floating point unit (FPU)
■High performance (128bit internal data path) floating point unit per core. ◦Benefit: Larger data paths and quicker floating point calculations for better performance."
Click to expand...

I still think that if everything scalled perfectly noone at amd would bother with all this modules stuff but they would call it 8 cores from the start.
 
cyclone3d said:
Not less FP units, it can act as two 128-bit FP units or one 256-bit FP unit unless I misunderstand how it is supposed to work.

Bulldozer -
"Two symmetrical 128-bit FMAC (fused multiply–add capability) floating-point pipelines per module that can be unified into one large 256-bit-wide unit if one of integer cores dispatch AVX instruction and two symmetrical x87/MMX/SSE capable FPPs for backward compatibility with SSE2 non-optimized software"

Phenom II -
"■128-bit floating point unit (FPU)
■High performance (128bit internal data path) floating point unit per core. ◦Benefit: Larger data paths and quicker floating point calculations for better performance."
Click to expand...

Which is why those "real" benchmarks are very confusing.
 
cyclone3d said:
Phenom II -
"■128-bit floating point unit (FPU)
■High performance (128bit internal data path) floating point unit per core. ◦Benefit: Larger data paths and quicker floating point calculations for better performance."
Click to expand...
PII can do one 128 bit Add and one 128 bit Multiply per cycle, so the theoretical FLOPS of a PII core is equal to the BD module. BD is more flexible in that both of its 128 FPUs can do Adds or Multiplies and can be increase by 2X with the right algorithm and rewritten software that takes advantage of the FMA.
 
Bulldozer is a myth that is never going to come out.

And if it does I sincerely hope it smashes intel in the face and makes them run for the R and D hills.

Why because I want socket 4800 or whatever the hell is next to be absolute badass so I can crunch videos even faster haha.
 
SLee said:
PII can do one 128 bit Add and one 128 bit Multiply per cycle, so the theoretical FLOPS of a PII core is equal to the BD module. BD is more flexible in that both of its 128 FPUs can do Adds or Multiplies and can be increase by 2X with the right algorithm and rewritten software that takes advantage of the FMA.
Click to expand...

A lot of your SSE code is bogged down by data movement anyway so the practical result is almost like having 2 SSE units per core, example

Code: 
movaps xmm0, [eax+ecx]          // 4 data movement instructinos (load)
movaps xmm1, [eax+ecx+16]       // the other thread can use the 2 SSE pipelines while
movaps xmm2, [eax+ecx+32]       // we wait for the memory controller to get our data
movaps xmm3, [eax+ecx+48]
mulps xmm0, [edx+ecx]           // 4 multiply instructions
mulps xmm1, [edx+ecx+16]        // now its our turn
mulps xmm2, [edx+ecx+32]
mulps xmm3, [edx+ecx+48]
movaps [edi+ecx], xmm0          // 4 data movement instructions (store)
movaps [edi+ecx+16], xmm1       // store the results and let the other thread use
movaps [edi+ecx+32], xmm2       // the SSE pipelines again.
movaps [edi+ecx+48], xmm3



I just hope the instruction decoder can keep up.
 
I also wanted to add to my last post by noting that the other problem AMD has is that Bulldozer has been in development for so long that even if it comes out and performs every bit as well as SB Intel is literally sitting on multiple new cpu's that are even better than what they have now. The more I look at it the harder it is to believe that AMD will ever be able to catch up.
 
kinjo said:
I also wanted to add to my last post by noting that the other problem AMD has is that Bulldozer has been in development for so long that even if it comes out and performs every bit as well as SB Intel is literally sitting on multiple new cpu's that are even better than what they have now. The more I look at it the harder it is to believe that AMD will ever be able to catch up.
Click to expand...

New CPU' or new architecture/completely new designs?
 
cyclone3d said:
Not less FP units, it can act as two 128-bit FP units or one 256-bit FP unit unless I misunderstand how it is supposed to work.
Click to expand...

Yes (4 FPUs but twice as wide to my understanding). So it's 8 Integer cores and 4 FPUs (which should act as 2x128 in Cinebench multithread). Thuban has 128Bit FADD and 128 Bit FMAC on each core. Eight exclusive cores with its own units should scale better. Depends on the workload. What I wanted to say is hat Cinebench may not Bulldozers best discipline because its FP heavy. You can't count it as 8 traditional core design until its clear how it scales with Bulldozer. Don't expect 95-100% scaling.

Poisoner said:
Now we know you are a troll. Thanks for removing all doubt!
Click to expand...

Thanks for confirmation that it's better to hide Bulldozer Benchmarks, some Fanboys goes crazy. Was a mistake to open a thread with BD Benchmarks here. Gonna ask a Mod to close this thread.
 
Last edited:
Paran said:
Yes (4 FPUs but twice as wide to my understanding). So it's 8 Integer cores and 4 FPUs (which should act as 2x128 in Cinebench multithread). Thuban has 128Bit FADD and 128 Bit FMAC on each core. Eight exclusive cores with its own units should scale better. Depends on the workload. What I wanted to say is hat Cinebench may not Bulldozers best discipline because its FP heavy. You can't count it as 8 traditional core design until its clear how it scales with Bulldozer. Don't expect 95-100% scaling.



Thanks for confirmation that it's better to hide Bulldozer Benchmarks, some Fanboys goes crazy. Was a mistake to open a thread with BD Benchmarks here. Gonna ask a Mod to close this thread.
Click to expand...

Dude no offense stop being a pussy you posted leaked bench marks of course some one is going to cry fake. You could submit them with a letter sidned in blood from the president of AMD vouching for their authenticity and someone here would still cry fake. Welcome to the internet man 90% trolls and douche bags and 10% great conversation with intelligent people it's just like real life except with a giant open forum for retards to express their opinion publicly.... Wait on second thought real life has Fox news, so yeah its just like real life. Im not saying it's fair or right but that the way it is if you want to contribute you have to deal with trolls and insane fanboys.
 
I want Bulldozer benchmarked while it's compiling Qt (all of it), or some other big C++ project. As long as it uses an SSD, the CPU should be fully loaded. C++ scales nearly perfectly.
 
Paran said:
Thanks for confirmation that it's better to hide Bulldozer Benchmarks, some Fanboys goes crazy. Was a mistake to open a thread with BD Benchmarks here. Gonna ask a Mod to close this thread.
Click to expand...

dude, grow the hell up.
 
iconiK said:
I want Bulldozer benchmarked while it's compiling Qt (all of it), or some other big C++ project. As long as it uses an SSD, the CPU should be fully loaded. C++ scales nearly perfectly.
Click to expand...

++. I really need to see compiling benchmarks....
 
Being a programmer who regularly compiles millions of lines of C++ code I too would like to see that.
 
Really odd that they did not test against Thuban or Deneb. They did however test with a wide range of Intel chips.
 
They must have patched the BIOS for the 760 board themselves right? I can't believe one with stock BIOS would function beyond failsafe clocks with BD.
 
You must log in or register to reply here.
Back
Top