Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

[tweakboy]

I want to see side side performance results until I believe this...:'(

 

[HW2050Plus]

I wonder if they're going to reach a 2 cycle L1 or remain at 3 cycles and have extra headroom for clock speed? It would most certainly seem that they traded size for the extra associativity and snooping vs shared resources probably played a part in the small L1 as well.

L1D even increased from 3 cycles on Phenom to 4 cycles on Bulldozer. But with all questions you have to consider other very important paramters e.g. the bandwidth and that has increased afaik. And the 1 cycle more for L1D is quite good if you consider the high speed design. Because latency-wise that means that the L1D is not slower (33% more cycles on ~30% less time).

A big part of improvement over Phenom might be the L2 cache bandwith. I assume it was doubled.

Btw I am wondering when the first taiwan bench leaks appear. As so many samples are out now there should be something coming, right? That would be appreciated since the current "leaks" are more fake than anything.

 

[Mopetar]

Because latency-wise that means that the L1D is not slower (33% more cycles on ~30% less time).

If you must wait, wait faster.

 

[ShadowVVL]

The more I read the more I want one and the more impatient I get.
I only need 4 cores at the most, but even so that damn bulldozer isnt pushing me away but reeling me in.

 

[Mopetar]

The more I read the more I want one and the more impatient I get.
I only need 4 cores at the most, but even so that damn bulldozer isnt pushing me away but reeling me in.

I'm fairly certain there are going to be two-module/four-core versions of the chip.

 

[drizek]

I'm fairly certain there are going to be two-module/four-core versions of the chip.

Ya, but where's the fun in that?

 

[ShadowVVL]

there most likely will be a 4 core chip. but I have been thinking about building a future proof system to carry me to 2014. something like 8 cores 8 gigs of ram and 2 ssds in raid which would be great for what ever software and game developers can throw at us for the next few years.

the new hardware seems to get me vary excited and upgrade happy.
but I might just stick with my C2Q and upgrade when needed.

 

[Dark Shroud]

I hope so, because I want a two-module/four-core CPU for my family PC as well my planned Windows Home Server build.

 

[Idontcare]

Ya, but where's the fun in that?

Laptops?

 

[HW2050Plus]

Ya, but where's the fun in that?

Those are for the low end market that previously used 2 core chips.

What AMD will initially ready:

Consumer: 4C, 6C, 8C in Q2/2011
Server: 8C, 12C (MCM), 16C (MCM) in Q3/2011

Then dependend on the results and what Intel does we might see later on (next year) more cores (e.g. Consumer 12 Core, Server 24 Core) and higher clocks. Though Intel and AMD announced 20 cores for 2012 I am sure that AMD will top this in 2012 to at least 24 Cores. 32 Cores would be possible, but then the clocks would be quite low beause of the TDP limit, therefore I do not think we will see that.

@Idontcare
Possible for Laptops, but I believe that the APUs will make big share not only for business PC but for laptops in special. Therefore the laptop market will be likly see a widespread use of Llano APUs.

 

[podspi]

32 Cores would be possible, but then the clocks would be quite low beause of the TDP limit, therefore I do not think we will see that.

...
Therefore the laptop market will be likly see a widespread use of Llano APUs.

That would actually be ideal. A ton of cores for multithreaded applications, and when you're doing something that requires singlethread performance, shut most of the cores down and crank the clockspeeds. I'm pretty sure this is what BD is going to do, really all we have left to see is how well it works in practice


I agree with you on Llano, I think for laptops the market for it will be huge. I've never heard anybody complain about the performance of an AMD-based laptop, but I have heard people complain about the battery life. If they can get good battery life AND able to fit in mainstream gaming, they're going to have a hug win, benchmarks be damned.

 

[Mopetar]

If they can get good battery life AND able to fit in mainstream gaming, they're going to have a hug win, benchmarks be damned.

The cores for Llano are based on the K10 design so I imagine the power characteristics will be somewhat similar. They may have made a few modifications to get better idle power draw results and the process shrink should allow for some further reductions, but at the end of the day I don't know if it will be good enough to match what we're seeing from Intel right now.

On the other hand, the on-die GPU will be significantly better than what Intel can offer and may balance things out on the whole. If Llano turns out even remotely as well as Bobcat did, AMD can consider it a success.

 

[drizek]

The onboard GPU is by far the most interesting part of it. It's a 32nm SOI GPU in a 40nm world.

If I was a laptop buyer, I wouldn't consider anything other than Llano.

Also, AMD improved power management with Thuban, right? so they can carry that over to Llano.

 

[Anexate]

I just wanted to direct the attention to the title of this thread:
"Rumour: Bulldozer 50% Faster than Core i7 and Phenom II."

and the last news release from AMD:
""The U.S.-based CPU and chip maker AMD will present at CeBIT 2011 processors, which performance exceeds the current top models of the Phenom-II series by 50%,"

other news:
the new Bulldozer CPUs are not solely AM3+ compatible; a BIOS update makes possible for them to function in an AM3 motherboard.

AM3+ platform will double the connection between North and Southbridge.

New platforms and CPU will probably debut at Computex in June.

 

[JFAMD]

Source?

I am not sure that AMD said that. As far as I know the only performance statement was on server, not client.

I am not doubting that it could be true, I am just doubting that AMD said it.

 

[Anexate]

Source?

I am not sure that AMD said that. As far as I know the only performance statement was on server, not client.

I am not doubting that it could be true, I am just doubting that AMD said it.

You know, you could make great career in politics as PR.
You never give any information (this is understandable) and you go quite a long way to make a pudding out of leaks (to put doubts even in doubts).

 

[Idontcare]

You know, you could make great career in politics as PR.
You never give any information (this is understandable) and you go quite a long way to make a pudding out of leaks (to put doubts even in doubts).

I can't help but notice that you seem to be avoiding the question...

 

[IntelUser2000]

He must be talking about this one: http://www.xbitlabs.com/news/cpu/di...ozer_Microprocessors_at_CeBIT_Trade_Show.html

 

[videoclone]

Awesome.. looks like This year is going to be AMD for the win Hope they make lots of money and grow to go toe to toe with intel.

 

[drizek]

Der US-amerikanische CPU- und Chip-Hersteller AMD wird auf der CeBIT 2011 Prozessoren vorstellen, mit denen die Leistung der derzeitigen Top-Modelle der Phenom-II-Reihe um 50 Prozent übertroffen werden soll.

It is taken from AMDs CeBIT page

love that "US-american" phrase.

Also, doesn't the headline of the xbit article contradict JFAMDs earlier statement that AMD wasn't going to show anything at CeBIT?

AMD to Showcase Bulldozer Microprocessors at CeBIT Trade-Show.

I can confirm that you will get [information on] neither [performance nor speedbins] at CeBIT. I have said several times that we disclose both of those at launch.

 

[IntelUser2000]

I can confirm that you will get [information on] neither [performance nor speedbins] at CeBIT. I have said several times that we disclose both of those at launch.

Note the bold.

 

[drizek]

Not speedbins, but "50% faster" is definitely information on performance, right?

Unless of course the CeBit page was just written by a guy who read the topic headline here and took it as fact...

 

[IntelUser2000]

Then again, they could be just rehashing the info given out eons ago comparing Interlagos to Magny Cours.

 

[jvroig]

Not speedbins, but "50% faster" is definitely information on performance, right?

That's the same old info months ago, blogged by JFAMD himself, which was taken out of context by forum-goers and started all the rumors and misunderstandings about how Bulldozer will be 80% faster than everything on the planet, or something equally ridiculous.

 

[drizek]

. The big question, however, is whether the company will initiate a review program for the new chips shortly after the show, will unveil official performance levels of the chip or will let select journalists to "independently" benchmark the new CPUs on site, a more than controversial move.

So I guess xbit don't take the 50% number as "official", and I guess JFAMDs post means that the answer to xbits question is no.