AMD Bulldozer Finally!

LeJosh · 24 Sep 2011 at 16:40

Cooper said:
Ok, can someone tell me how an 8 core CPU that can only score 5.95 @ 3.6Ghz in Cinebench R11.5 is impressive???

My 4 core unlocked Phenom II @3.6 Ghz gets 4.35.

100% more cores but only a 40% speedup....that does not seem very impressive to me.

It's actualy still 4 cores but they have 2 modules inside each core making a theoretical 8 cores. I think, I'm a bit hazy with the ins and outs.

Trunks9486 · 24 Sep 2011 at 16:41

Cooper said:
Ok, can someone tell me how an 8 core CPU that can only score 5.95 @ 3.6Ghz in Cinebench R11.5 is impressive???

My 4 core unlocked Phenom II @3.6 Ghz gets 4.35.

100% more cores but only a 40% speedup....that does not seem very impressive to me.

Its not really 8 traditional cores. This for all intents and purposes is the next X4 design. They are just marketing it as 8 core for the marketing benefit.

Still, its not a total disaster i guess. I am curious as to the max stable overclocks and the actual IPC comparison.

Jokester · 24 Sep 2011 at 16:41

Think of it as AMDs equivalent of Hyperthreading.

LeJosh · 24 Sep 2011 at 16:42

Except the threads aren't logical, they're all physical aren't they?

Trunks9486 · 24 Sep 2011 at 16:44

LeJosh said:
Except the threads aren't logical, they're all physical aren't they?

Yes. In a well threaded application the gains are supposingly higher compared to HT.

gurusan · 24 Sep 2011 at 16:47

I meant useful benchmarks....as in not at 7680x1600

FoxEye · 24 Sep 2011 at 16:47

I still think my next CPU is going to be an Intel. Guess it all depends on price. The fact that I'm going to need a new mobo whichever route I take makes it easier to jump ship this time.

gareth170 · 24 Sep 2011 at 16:50

they have 4 modules with 2 cores in each module, total of 8 cores.

Cooper · 24 Sep 2011 at 16:50

Trunks9486 said:
Its not really 8 traditional cores. This for all intents and purposes is the next X4 design. They are just marketing it as 8 core for the marketing benefit.

So in that case, the 6 core bulldozer that i was interested in, is really only a true Tri-core?... :confused:

..

..hmm, the more i look at these Bulldozers, the more the 2500K appeals...

Trunks9486 · 24 Sep 2011 at 16:52

Cooper said:
So in that case, the 6 core bulldozer that i was interested in, is really only a true Tri-core?.....

..hmm, the more i look at these Bulldozers, the more the 2500K appeals...

Exactly. I am pretty sure that they will eventually introduce an 8 Module chip resulting in 16 "cores".

The 2500 and 2600 are both cracking chips, especially overclocked.

gareth170 · 24 Sep 2011 at 16:53

Cooper said:
So in that case, the 6 core bulldozer that i was interested in, is really only a true Tri-core?.....

..hmm, the more i look at these Bulldozers, the more the 2500K appeals...

6 core bulldozer would have 3 modules with 2 cores in each module, total of 6 cores.

Cooper · 24 Sep 2011 at 17:05

gareth170 said:
6 core bulldozer would have 3 modules with 2 cores in each module, total of 6 cores.

..then why is an 8 'core' bulldozer only as fast as a 6 'core' Phenom II..in Cinebench 11.5..?

Somethings not right here...and i hope for AMD's sake its another set of fake benchmarks...

VortX · 24 Sep 2011 at 17:08

What about this?

Doesn't look completely convincing... 4.8GHz with 8 cores the score is 7.8. With my 2600K with 4 cores (8 threads obviously) at 4.6GHz I get 9.04. Or is Cinebench Intel biased?

I know I said I didn't expect Bulldozer to be "that" much better, if at all, but I would expect the scores to be a lot closer, and at 4.8GHz it "should" be overtaking a 2600K...?

EDIT: Hang on... I have just noticed...

Why would an AMD slide showing the scores for Bulldozer chip have "Your System" as in Intel build... I call BS!

EDIT 2: Nevermind, it's the official screenshot from http://www.maxon.net/downloads/cinebench.html. It could be genuine I guess, but I would have thought in an AMD presentation they would have their own screenshot.

gareth170 · 24 Sep 2011 at 17:08

Cooper said:
..then why is an 8 'core' bulldozer only as fast as a 6 'core' Phenom II..in Cinebench 11.5..?

Somethings not right here...and i hope for AMD's sake its another set of fake benchmarks...

most likely fake results

im sure the 8 core BD as 4 modules with 2 cores in each module, total of 8 cores.

manveruppd · 24 Sep 2011 at 17:13

Cooper said:
So in that case, the 6 core bulldozer that i was interested in, is really only a true Tri-core?.....

..hmm, the more i look at these Bulldozers, the more the 2500K appeals...

Nope, you get 6 full cores, don't worry. Each "module" (2 cores) shares some components, like an FPU (which isn't used very often nowadays) and a L2 cache (which helps with scheduling). They say that each module can reach 80% of the peak performance of 2 complete cores. Note that that's continuous peak performance. In reality, since all but the most efficiently-multithreaded processing tasks aren't continuous (there will be short bursts of peak processing demand followed by troughs of inactivity) and the workload will be asymmetrically distributed between cores, most of the time you wouldn't be seeing a performance difference between a 4-module BD chip and a "true 8-core" chip.

Looks like this basically:

I reckon the reason they're pushing the word "module" so much as a marketing term is because they still expect each individual BD core to be a bit slower than an SB core in terms of maximum performance (inevitably, since it only has half an FPU for one thing), so they want to get reviewers to think about relative performance in terms of 1 module vs 1 Intel core, in which case of course AMD will be made to look like it's ahead. I'm just speculating of course, we won't know for sure how they'll compare until we see some independent benchmarks.

MichaelM · 24 Sep 2011 at 17:14

VortX said:
What about this?

Doesn't look completely convincing... 4.8GHz with 8 cores the score is 7.8. With my 2600K with 4 cores (8 threads obviously) at 4.6GHz I get 9.04. Or is Cinebench Intel biased?

I know I said I didn't expect Bulldozer to be "that" much better, if at all, but I would expect the scores to be a lot closer, and at 4.8GHz it "should" be overtaking a 2600K...?

EDIT: Hang on... I have just noticed...

Why would an AMD slide showing the scores for Bulldozer chip have "Your System" as in Intel build... I call BS!

Haha, why would it also be OS X? a Mac...?

njm106 · 24 Sep 2011 at 17:14

Cooper said:
Ok, can someone tell me how an 8 core CPU that can only score 5.95 @ 3.6Ghz in Cinebench R11.5 is impressive???

My 4 core unlocked Phenom II @3.6 Ghz gets 4.35.

100% more cores but only a 40% speedup....that does not seem very impressive to me.

Looking at the results for Cinebench it would appear that clock for clock that bulldozer is slower than a Phenom II X6, now if that is true its pretty bad unless it either overclockers a lot or uses a lot less power.

AMD have done really well with there APU's recently but it is looking increasing likely that bulldozer will be a bit of a flop, I really hope these benchmarks are wrong.

VortX · 24 Sep 2011 at 17:16

MichaelM said:
Haha, why would it also be OS X? a Mac...?

See my edit

It's the official screenshot used on the Cinebench website... could be fake still.

clocka · 24 Sep 2011 at 17:19

you can tell those benchmarks are cherry picked, they compare the graphics performance to phenom 2 but don't say that the built-in gpu is used with it at the same time, nor do they show the games that its slower to use the built-in and 3rd party amd gpu.

I'd wait for benchmarks from review websites before you get your hopes up guys.

LeJosh · 24 Sep 2011 at 17:21

The teething stage for architectural overhauls is horrendous.

I'm more looking forward to a refresh, or just the 2nd generation of bulldozer now.

But Ivy bridge may be out...