AMD tapes out its Bulldozer CPU architecture

[CDdude55]

[Techtu]

Back on track though...



Any price point heard of for these new line-ups?

[wolf]

I want to see stock clocked chips at 4+ ghz, even just 4 cores, I have no need for more than that, just give me sheer speed already.

Having said that I really doubt AMD will be the first to throw a 4ghz stock clocker on the table, at least at the mainstream consumer grade.

[TheMailMan78]

Originally Posted by DrPepper

Watercooling accident if anything.

Well at least he went with honor! My condolences. Here in honor of your friend have a beer on me!

[Super XP]

Originally Posted by CDdude55

If they actually make a CPU that beats my i7 this time and it's at a good price, i shall be switching over.

Bulldozer is meant to complete with Intel's next gen or some may call it Nehalen 2. So obviously if Bulldozer does not beat Core i7 then AMD has major issues. Bring it on SandyBridge, we need competition to help drive prices down into the dirt.

[DrPepper]

Originally Posted by TheMailMan78

Well at least he went with honor! My condolences. Here in honor of your friend have a beer on me!

http://www.grapesandgrainsnyc.com/wp...rfest-girl.jpg

Cheers.

I always try see the funny side of things.

[WhiteLotus]

Originally Posted by TheMailMan78

Well at least he went with honor! My condolences. Here in honor of your friend have a beer on me!

http://www.grapesandgrainsnyc.com/wp...rfest-girl.jpg

plus Hofbrau (the beer) is fucking good drinking.

[CDdude55]

Originally Posted by Super XP

Bulldozer is meant to complete with Intel's next gen or some may call it Nehalen 2. So obviously if Bulldozer does not beat Core i7 then AMD has major issues. Bring it on SandyBridge, we need competition to help drive prices down into the dirt.

I agree.

[WarEagleAU]

Wow, about time, been speculating on this with a tPU member or two and some personal friends. Cant wait to see what the prices and yields are. I would love to see just about most software now come out with at least dual cores in mind (if not quad cores). That would benefit greatly. Wonder if Windows 7 is optimized for Dual cores. Anyone know?

[pantherx12]

Bah, waiting for tomorrow for news!

Looking forward to this chip, if its decent price/performance I'm getting an AMD platform next year.

Well unless their graphics turn out crappy next year.

And vice versa <_< or both.

[cadaveca]

Performance: We release benchmarks at launch, so don’t expect too much detail there anytime soon. From a performance standpoint, if you compare our 16-core Interlagos to our current 12-core AMD Opteron™ 6100 Series processors (code named “Magny Cours”) we estimate that customers will see up to 50% more performance from 33% more cores. This means we expect the per core performance to go in the right direction — up. That is all I will say until launch.

http://blogs.amd.com/work/2010/08/02/what-is-bulldozer/


For those unaware, there is TONNES of info on the AMD website about these upcoming processors. When I said speculation is pointless, I really meant it.

We can extrapolate on the info I posted here form AMD, and suggest that each new Bulldozer core will offer about 4% more performance, clock-for-clock. Hardly what people are looking for, and hardly an i7 killer. Again, I said it's two years too late, and I stand by that comment.

[Steevo]

we estimate that customers will see up to 50% more performance from 33% more cores.

So for a example we will say the 12 core system can execute 1000 standard complex threads in one second. 1000/12= 83.33 per core. So now we have a standardized core work number.

So they say that the new processor will do 1500 standard complex threads in once second. 1500/16=93.75 threads per core per second, or a 12.5 clock efficiency increase per core.

[cadaveca]

Originally Posted by Steevo

we estimate that customers will see up to 50% more performance from 33% more cores.

So for a example we will say the 12 core system can execute 1000 standard complex threads in one second. 1000/12= 83.33 per core. So now we have a standardized core work number.

So they say that the new processor will do 1500 standard complex threads in once second. 1500/16=93.75 threads per core per second, or a 12.5 clock efficiency increase per core.

Flawed logic. Try your math again. Your numbers are not correct. Uh, let's start with 93.75-83.33....reversing the math give 151+%.



I look at it this way... we added 33% more cores, but got 50% more performance. that leaves 17% extra performance, spread across 16 cores... I see your logic as interpreting the increase from only those extra 4 cores....

[largon]

that leaves 17% extra performance, spread across 16 cores

[rhetorical question]
If a quad core processor "A" is 17% faster than quad core processor "B", then how much faster is processor "A" per core?
[/rhetorical question]

That's right. 17% increase per core would be really nice if you ask me.

[cadaveca]

Originally Posted by largon

[rhetorical question]
If a quad core processor "A" is 17% faster than quad core processor "B", then how much faster is processor "A" per core?
[/rhetorical question]

That's right. 17% increase per core would be really nice if you ask me.

No doubt. I am expecting AT MOST 4& per clock...mind you, this core is so very different, that may be completely off-base.

So, if we look at Steevo's numbers, he says 12.5, but really, he gets "12" and not "1.2" by using a 1000-base number for the calculation instead of 100-base.

Anyway, it's more like 1.03% increase per core.

[largon]

I reeeaaallly don't see why you're expecting that 4%.

[cadaveca]

Hoping. That is all. If that's the case though, I'd be swapping out my server farm. I REALLY want to do this, so I gotta hope.

[trickson]

Looks like AMD may have some thing to finally bring to the table . Still I do not see myself getting a new system for another 3 years as well the Quad I have should be just fine for many more years to come . It would be nice to see AMD on top again but till I see some real chips tested no one can say for sure what Bulldozer will be able to do .

[cadaveca]

Originally Posted by trickson

Looks like AMD may have some thing to finally bring to the table . Still I do not see myself getting a new system for another 3 years as well the Quad I have should be just fine for many more years to come . It would be nice to see AMD on top again but till I see some real chips tested no one can say for sure what Bulldozer will be able to do .

Yeah, and AMD is last presenter tomorrow @ HotChips. So the info is most likely NOT gonna be known until long after that. We need yeild info, and that is too far off..we won't get real performance until launch, next year some time. Would be nice to see a pre-holiday launch, but I think that's more wishfull thinking on my part.

[Techtu]

Originally Posted by cadaveca

Yeah, and AMD is last presenter tomorrow @ HotChips. So the info is most likely NOT gonna be known until long after that. We need yeild info, and that is too far off..we won't get real performance until launch, next year some time. Would be nice to see a pre-holiday launch, but I think that's more wishfull thinking on my part.

No, not just on your part... seem's like pretty much all of us in this thread are wanting to see some results quicker than AMD can make them.

[Beertintedgoggles]

Originally Posted by cadaveca

Flawed logic. Try your math again. Your numbers are not correct. Uh, let's start with 93.75-83.33....reversing the math give 151+%.



I look at it this way... we added 33% more cores, but got 50% more performance. that leaves 17% extra performance, spread across 16 cores... I see your logic as interpreting the increase from only those extra 4 cores....

I've got to agree with Steevo and his math on this one, well since it's right...

Just clarifying his numbers using the 1000 threads per second example

33% more cores: so 12 cores * 1.333333..... = 16 cores
50% more performance: 1000 * 1.5 = 1500 threads per second

So the old 12 core proc. would do 1000 threads per second while the new 16 core will do 1500 per second (threads of whatever, not relavent here... just some AMD spec'd performance numbers):
that means old did 1000/12=83.33 threads per second per core
new will do 1500/16=93.75 threads per second per core

so that's 93.75 for the new and 83.33 for the old. To find the percentage increase that's 93.75/83.33=12.5% faster per core than the old architecture.

Now I'm not sure if AMD's numbers are skewed and that 50% performance increase was more than likely on a specific test; however, going by those numbers they posted the math shows the new cores should be 12.5% faster per clock.

[trickson]

I just hope the AMD fanboys wont get all upsad if they are not what they expect . Nut I sure hope AMD finally gets it right . My hope is they are going to at the very least mach the core i7's . that is a tall order seeing as AMD has been trying for years now .

[cadaveca]

Originally Posted by Beertintedgoggles

I've got to agree with Steevo and his math on this one, well since it's right...

Just clarifying his numbers using the 1000 threads per second example

33% more cores: so 12 cores * 1.333333..... = 16 cores
50% more performance: 1000 * 1.5 = 1500 threads per second

So the old 12 core proc. would do 1000 threads per second while the new 16 core will do 1500 per second (threads of whatever, not relavent here... just some AMD spec'd performance numbers):
that means old did 1000/12=83.33 threads per second per core
new will do 1500/16=93.75 threads per second per core

so that's 93.75 for the new and 83.33 for the old. To find the percentage increase that's 93.75/83.33=12.5% faster per core than the old architecture.

Now I'm not sure if AMD's numbers are skewed and that 50% performance increase was more than likely on a specific test; however, going by those numbers they posted the math shows the new cores should be 12.5% faster per clock.

Again, flawed equation. If each core was 12.5% faster, there'd be more than 50% increase. 12.5%, times 16 cores= what? 200%? and 12 cores is 150%?


Please. FIX. THAT. math.

[CDdude55]

Originally Posted by trickson

I just hope the AMD fanboys wont get all upsad if they are not what they expect . Nut I sure hope AMD finally gets it right . My hope is they are going to at the very least mach the core i7's . that is a tall order seeing as AMD has been trying for years now .

I agree on that, there past two lineups have yet to truly past an i7. I'm hoping we will see a good resurgence with this architecture. As a person who doesn't cum my pants over a particular company(aka a fanboy), i truly won't be heartbroken if these chips don't succeed.

[Beertintedgoggles]

Originally Posted by cadaveca

Again, flawed equation. If each core was 12.5% faster, there'd be more than 50% increase. 12.5%, times 16 cores= what? 200%? and 12 cores is 150%?


Please. FIX. THAT. math.

Um, no there wouldn't. Simple math. Very simple in fact.

You have a 50% increase with 33% more resources.

You want to do it the easiest way:
the old cpu had 1.00 (100% resources), new has 1.33 (33% more or 133%)
old cpu did 1.00 work (100%), new does 1.50 (50% increase)

That's now 1.5/1.33 = 1.125 ish. So still your 12.5% increase per core. AMD didn't use number of cores do denote how much faster they are, just 33% more cores make a 50% increase. The 12 and 16 cores I'm assuming were just pulled out since they make that 33% more a very convenient whole number. But still, if you think about it as 50% more work is done with only 33% added resources you can see that they should more than 4% faster. They should be 1.5/1.33 faster.