AMD chief say BD will offer only 35% not 50% more performance than previous gen

[ieatdonuts]

I am going to have to agree with bridito. There's really no guarantee that there is higher IPC vs current architectures. The rumours that have floated out there have more often been negative than positive - negative in that Bulldozer may be nothing more than just 2 more cores and higher clocks.

 

[ed29a]

So in other words for the average consumer a 6 or an 8-core CPU is a waste of $ if it's not faster in 2-4 threaded apps. Agreed.

But they hell don't need a 2500K. As I mentioned, a simple low end dual core does the job.

Intel had another architecture alongside the Netburst - Pentium M. Intel didn't fall asleep, they just miscalculated the ability for their Netburst architecture to scale with higher clock speeds.

Monopoly + a tech company led by marketing instead of engineers. No miscalculation.

Regardless, you aren't giving AMD engineers any credit. The Athlon 64 had 64-bit support, onboard memory controller, among other innovations in the consumer market.

This thread is not about giving AMD credit for past successes. But just to re-assure you:

[Disclaimer] My gaming rig is a quad core Phenom II and my linux workstation is a hexa core Thuban. Other than laptops, I never bought intel desktops.

What do consoles have anything to do with this discussion?

You were giving me meaningless statistics about strategy PC games being 34some% of total PC games sales. And I pointed out to you that PC game sales are a joke compared to consoles. 34% of not much is still ... not much! So there isn't any market demand to support all those PC gamers out there. Just to reassure you:

[Disclaimer] I never had a console in my life, and I game a lot.

You still think consumers don't care about gaming on their PCs/Macs?

http://www.apple.com/macbookpro/

First page of MacBook Pros:

"Game-changing graphics. AMD Radeon graphics processors on the 15- and 17-inch models are up to 3x faster....".

Apple even provided marketing charts for HL2 and Portal. Mac is not even a "gaming" platform so to speak and even Apple got Steam support.

Oh wow. You complain about AMD core marketing yet you give me a marketing line from Apple. The company that can market anything to anyone, the company that has invented the reality distortion field. I see Steve Jobs did a good job on you! By the way, you realize that HL2 and Portal are old, and I mean old games? It came out in 2004. Seven years ago.

Let that sink in a second. Seven years ago.

I have a question though, if Apple puts high power quality components in their PCs, where are the macs that run the i7 2600K? Or the 2500K at least. Where are the macs that run the 990X? Where are the macs that run the 580 GTX or 6970? Why do they cram slow mobile graphic cards in their macs? I challenge anyone to run in 2560 x 1440 (27 inch iMac) resolution any modern game with a 6970M card with high details/settings. Seriously ... Apple and gaming?

The average consumer will walk into the BestBuy store and say, I want a snappy computer for every day tasks, good enough for casual games, doesn't run too hot and consumes little power. What should I get? The answer isn't going to be a 6- or 8-core BD. Those CPUs target guys who know what IPC is what performance/watt is and what a core means.

Oh boy, you never been inside Best Buy? Seriously, the salesman have barely more knowledge than clients. They recite a script. They don't know jack about computers but they are DAMN good salesmen.
Client: I am want a fast PC and I want to play games on it.
Salesman: Here is a random HP/Acer PC with a mediocre video card.
Client: Is it good?
Salesman: Oh yes, it has gigahurtz and memory and lot of hard drive!!!!!
Client: I'll take it.

Again, you deluding yourself by thinking that consumers know/care anything about hardware especially details like power consumption and heat production. Besides, the salesmen will guide them through their high priced gaming machines from Acer or HP that are mostly AMD processor driven anyways (read: costs less).

The reason AMD has been able to sell their CPUs all these years is because they continue to compete in the <$140 price range. Their Athlon II X4 CPUs were excellent budget CPUs.

Agreed but I would move the price to about 160$. You can't beat the value of a 150-160$ 1090T Thuban.

But AMD can't just continue to give us more cores without improving the performance / watt of those cores - that's all I am saying.

Again, you assume that the average consumers cares (or knows) about those things and most importantly needs all that power. A low end dual core from either companies is more than enough for them. I recently built my mother a Zacate E-350. She is very happy. Facebook + Skype + Browsing doesn't need more than that.

 

[LOL_Wut_Axel]

I am going to have to agree with bridito. There's really no guarantee that there is higher IPC vs current architectures. The rumours that have floated out there have more often been negative than positive - negative in that Bulldozer may be nothing more than just 2 more cores and higher clocks.

Even if there's no IPC increase or even a 10&#37; IPC decrease it's still miles ahead of Atom in performance. So no, his argument makes zero sense. He [bridito] when faced with facts just tries to substitute them for his personal opinion and when you point it out he'll say you're the one that's wrong because of ridiculous would-be arguments he made up, so please stop feeding the troll. It'd be pretty nice if we could focus on the technical aspects of Bulldozer instead of derailing the thread.

 

[GaiaHunter]

Did you not read what I am saying? If BD = Phenom II in IPC, then the 80&#37; statement applies.

Assuming 2A (Bulldozer) = 2B (Phenom II), AMD said A (module design) is only 80% as efficient as B design. Therefore --> 0.8 (2A).

How are you not following this? In other words, if BD has equal IPC to Phenom II, then out of the gate, its 2 cores are only 80% as fast as 2 equivalent independent cores. In other words, assuming no change in IPC over Phenom II, 2 bulldozer cores will only be 80% as fast at the same clock speeds.

He claimed that BD will be 1.35x faster than Phenom II. In other words, he claims that 0.8 (2A) / 2B = 1.35.

The only way for this to occur at the same clock speeds is if 2A is 1.69x faster than 2B (i.e., 2 core Bulldozer is 69% faster than a 2 core Phenom II).

Now think about how ridiculous that statement is.

Why should BD IPC be equal to Phenom II?

Completely different architecture, different instructions, different resources, different everything.

How fast is Phenom II processing workloads that are optimized using instructions like SSE4.2 compared to Nehalem?

It is a completely different architecture but apparently the basis point, instead of being 0 is Phenom II.

That is why people thought it was impossible for Intel to turn Netburst into Core 2. And in fact it was - what Intel did was a new architecture not based on Netburst, just like Bulldozer isn't based on Deneb/Agena.

See Agena -> Deneb is somewhat predictable, since Deneb is a revision/evolution of Agena architecture.

Wow, you just love twisting information, don't you?

Read the paragraph as it is.

Insufficient clock-speeds of Bulldozer are probably the reason why AMD now claims that the 16-core offering will be 35% faster than 12-core solution (which is natural, given 33% higher core count) and not 50%, as it initially expected. It is also noteworthy that Bulldozer's per-core performance is not projected to be much higher compared to existing microprocessors.

In the above paragraph, the comparison is clearly being made between a new 16-core AMD micro architecture and their modern (i.e., current) 12-core micro architecture. The reference to "per-core performance" is in regard to previous statement (i.e., their own existing microprocessors, and has nothing to do with SB).



The statement explains that AMD made a compromise between performance and die space. They accepted 80% performance of 2 full fledged cores for an increase of die space of just 12%. It sounds like a great tradeoff, but it also means a 6-core BD with same IPC as Phenom 2 is by their own definition can only be 80% as fast as a 6-core Phenom II at the same clocks. It's no wonder AMD is doing everything they can to launch BD at much higher clock speeds than Phenom II ships at.

The paragraph states "existing microprocessor". It doesn't say "existing AMD microprocessors", "current AMD microprocessors" or "AMD current offerings", while it is being very clear that the 35% is compared to X product.. That is your brain interpreting based on your opinions.

Additionally a performance reduction based on lower clocks doesn't say anything about IPC.

A 4.5GHz SB has higher performance than a 3.5GHz SB. Which one has higher IPC?

Lastly, as it was explained before, the 50% claim was stated when the top Operaton run at 2.3GHz. Now the top Operaton runs at 2.5GHz.

1.5x(2.3/2.5)=1.38

No IPC decrease, only clocks.

Lastly, as I keep reading, there is no need for 8 cores for desktop, right?

180% scaling is when resources are being shared. If the resources aren't being contested the performance is 100% per core.

 

[waffleironhead]

I don't know how to make myself any more clearer.

Assuming you converted 2 Phenom II cores into a 2 Phenom "module" design, you would instantly lose 20% performance according to AMD. Which means your 2 core BD module design would have 0.8x of the performance of your original 2-core design (independent 2 cores).

I thought this was explained differently, buy maybe I misunderstood.

My interpretation.
BD module
Thread1 =100%
Thread2(if needed) =60%

So the 80% performance figure is only important if you are using the second thread of a module, because a user will still be at the same or higher performance of k10.5 when using half the module.

This looks like to me, that if someone wants a full 4 cores of performance, then they need to buy a 4 module bulldozer, the extra performance that the second thread of the module throws in is gravy.

The 80% performance level is kinda funky to me as I've never seen anyone say the nehalem with hyperthreading has only 65% the performance per thread than C2Q. Which would be the same analogy since the extra thread of nehalem only adds around 30% performance iirc.

 

[podspi]

IIRC, there was a thread (it may not have been on AT, could have been on SA or another forum) where JFAMD clarified the 80% statement.

The second core(thread, whatever you want to call it) provides 80% of the throughput an additional core (with dedicated everything) would. This translates to each individual thread running at around 90% of its standalone throughput.

Please, if anyone remembers where that thread is, and/or if I remembered it wrong


Wouldn't be surprised to see Nehalem level performance in strictly ST (one thread per module) workloads. That translates to less than Nehalem (per-thread) performance in multithreaded workloads, but comparable performance (per clock!) to SB-E hex-core.

IMHO, the clocks will make it or break it. Here is hoping somebody does an in-depth look at IPC in BD. If only because as some others here have said, IPC does matter if you are overclocking.

 

[LOL_Wut_Axel]

I thought this was explained differently, buy maybe I misunderstood.

My interpretation.
BD module
Thread1 =100&#37;
Thread2(if needed) =60%

So the 80% performance figure is only important if you are using the second thread of a module, because a user will still be at the same or higher performance of k10.5 when using half the module.

This looks like to me, that if someone wants a full 4 cores of performance, then they need to buy a 4 module bulldozer, the extra performance that the second thread of the module throws in is gravy.

The 80% performance level is kinda funky to me as I've never seen anyone say the nehalem with hyperthreading has only 65% the performance per thread than C2Q. Which would be the same analogy since the extra thread of nehalem only adds around 30% performance iirc.

You lose 20% performance, not 40%. So, two cores in a Bulldozer module: 80% performance of two normal cores. One core: 100% performance of a normal core.

As for HyperThreading, it's not a real core like the cores in a Bulldozer module. With HyperThreading you add 20% performance to the core in multi-threaded applications; that's it. I hope that clears it up.

EDIT:

To the above poster: that makes sense, but at the same time I'm now confused. Do you have a link to where he clarified?

 

[waffleironhead]

You lose 20% performance, not 40%. So, two cores in a Bulldozer module: 80% performance of two normal cores. One core: 100% performance of a normal core.

As for HyperThreading, it's not a real core like the cores in a Bulldozer module. With HyperThreading you add 20% performance to the core in multi-threaded applications; that's it. I hope that clears it up.

EDIT:

To the above poster: that makes sense, but at the same time I'm now confused. Do you have a link to where he clarified?

I dont think you are understanding what I am saying.

Thread 1 at 100% thread 2 at 60% (100+60)/2 =this magic 80% number that everyone is throwing around. This doesnt mean that BD suddenly only has 80% the ipc of k10.5 nor does it need to be clocked higher.

Nehalem hyper threading only adds 20% on the second thread? damn then nehalem is only 60% the performance of c2q per thread when using hyperthreading.

 

[LOL_Wut_Axel]

I dont think you are understanding what I am saying.

Thread 1 at 100&#37; thread 2 at 60% (100+60)/2 =this magic 80% number that everyone is throwing around. This doesnt mean that BD suddenly only has 80% the ipc of k10.5 nor does it need to be clocked higher.

Nehalem hyper threading only adds 20% on the second thread? damn then nehalem is only 60% the performance of c2q per thread when using hyperthreading.

Well, there's now another poster saying the second core is 80%, in which case it'd be 90% the same performance overall. Let's see if a statement from John can be found on this.

As for HyperThreading, that's not how it works. HyperThreading works by taking a single core and making more efficient use of its pipeline, thus delivering higher performance. HyperThreading adds the equivalent of a virtual core to a real core. HyperThreading does not add another real core, but merely makes better use of that single real core. The module concept, in comparison, adds another real core that shares resources with the other one.

 

[waffleironhead]

Well, there's now another poster saying the second core is 80%, in which case it'd be 90% the same performance overall. Let's see if a statement from John can be found on this.

As for HyperThreading, that's not how it works. HyperThreading works by taking a single core and making more efficient use of its pipeline, thus delivering higher performance. HyperThreading adds the equivalent of a virtual core to a real core. HyperThreading does not add another real core, but merely makes better use of that single real core. The module concept, in comparison, adds another real core that shares resources with the other one.

That just confirms my understanding that the 80% number being throw around is performance per thread normalized across the module. AMD is already admitting that the weak side of the module is not going to be at 100% just like intel admits that HT does not equal a full second core.

So anyone taking this 80% performance number as a detriment just needs to compare to intels 65% per thread with HT on. In my mind its the same thing.

Unless I'm mistaken and there isnt a strong side/weak side of the module and indeed running two threads thru a module drops each side of the module to 80% of normal single thread performance.

 

[LOL_Wut_Axel]

That just confirms my understanding that the 80&#37; number being throw around is performance per thread normalized across the module. AMD is already admitting that the weak side of the module is not going to be at 100% just like intel admits that HT does not equal a full second core.

So anyone taking this 80% performance number as a detriment just needs to compare to intels 65% per thread with HT on. In my mind its the same thing.

Unless I'm mistaken and there isnt a strong side/weak side of the module and indeed running two threads thru a module drops each side of the module to 80% of normal single thread performance.

The module concept has its advantages and its weaknesses. One of its advantages is that it still allows for great single-threaded performance and allows for a smaller die in comparison to having two cores each with their independent resources. That means that at the same die size more cores can be added, which also means higher multi-threaded performance.

The flip side of the coin is that if you're using the second core of the module as well it'll have to share resources with the first one and, as a consequence, both will lose performance in comparison to before. Each core will have the same resources, and therefore the same (reduced) performance. In comparison to two fully independent cores, performance for each one should go down by 10%, more or less. Whereas with a fully independent design you'd have 200% performance for both cores, with the shared one you'd have 180% performance. This brings me to another point:

AMD could have designed Bulldozer and made some tweaks so that in mildly-threaded applications like gaming the CPU schedules the task to only one core in each module, hence meaning it'd perform 100% like a normal core. At that stage, though, you also have to take into account how Windows assigns each core. If they can pull something like this off, it's a win-win. In single-threaded applications you have the single core with all of its resources intact and hence you have max performance. In mildly multi-threaded applications you assign the load to only one core in each module, hence you also have no performance deficit when running up to four cores/threads. In multi-threaded apps you have all cores working and given it has so many, it'd excel in that area even with the 10-20% performance deficit for each core.

If the clock speeds and IPC is high enough, this idea would fit perfectly.

 

[RussianSensation]

Lastly, as I keep reading, there is no need for 8 cores for desktop, right?

If most of the programs you run benefit from 6-8 threads, then under such a scenario, you absolutely would want an 8-core processor, assuming your budget suffices.

However, if your typical load is not like that, then an 8-core processor may no longer be the fastest in applications which only support 3-4 threads. In such case, you'd probably want a much faster and more power efficient 4-core processor.

Phenom II X4 and X6 processors didn't stand a chance in the last 3 years (hence their only option was to compete on price) because with Core i5/i7 and SB you didn't have to choose between having the fastest processor for 2-4 threaded apps or for 6-8 threaded apps (thanks to amazing IPC, hyper-threading support, and high overclocking potential). With Phenom II X6, you got significantly inferior performance in apps that used 2-4 threads, and it only became competitive in programs that actually used 6 threads.

The problem is today by far the majority of consumer programs use at most 4 threads. That likely won't change in the next 2-3 years either. As such, for BD to be truly successful, it won't be enough to simply be superior in 6-8 threaded apps, as that only captures a niche target market. If BD is to make a huge splash, it would have to be competitive in 2-4 threaded apps . For 2-4 threaded apps, you want high clock speeds, high IPC, high overclocking headroom, and low power consumption, not more cores. The challenge is how do you design an 8-core processor like that for $300?

And if you did have an 8-core processor that was competitive with Intel's fastest 4 core processor in 2-4 threaded apps and also easily beat it in 6-8 threaded apps, then why wouldn't you launch your 4-core version of the said processor for $300 and sell the 8-core for $500-600? Yet AMD is going to launch their 8-core processor for "only" $300? It makes no sense to me unless BD is worse in 2-4 threaded apps but is better in 6-8 threaded apps. Otherwise, it just sounds too good to be true.

Finally, if the FX-8100 series BD was such a beast in performance per clock/watt (i.e., an 8 Core Nehalem as some project), then why hasn't AMD launched an 8-core 3.0ghz-3.3ghz version of such a processor 8-12 months ago? Surely such a processor would have beaten the mighty Core i7-980/990X, which Intel sold for $999!! Why did AMD need to re-spin BD to get it to 3.6-4.2ghz? Under this conservative scenario, a 3.2-3.3 ghz 8-core i5 "Nehalem" for $300 from AMD would have sold extremely well.

Could it be that BD is nowhere near Nehalem in IPC at 3.3ghz? That's probably why it needs 4.2ghz Turbo and 8 cores to compete with a 4-core 2600k 3.4ghz? :hmm:

 

[Joseph F]

Please report whoever is holding that gun to your head to the authorities.

You have a good point. However, I am quite tired of seeing the CPU forum filled with BD speculation threads.

 

[intangir]

btw with that "same clock speeds" looks like you don't know what IPC means by any rate. IPC doesn't change with clock speed.

*cough* http://forums.anandtech.com/showpost.php?p=32089813&postcount=101 *cough*

In actuality, all data shows that IPC does vary with clock speed. It is supposed to represent clock-normalized performance, and as such it functions as a useful tool with which to compare microarchitectures. But you should be aware that it is only an abstraction, and reality cannot be so simply modelled.

It's much like how electrical resistance is given a constant value R. It's defined as the ratio between the voltage and the current. But this is actually only constant at a given temperature, and not all materials follow this relationship. Even so-called ohmic resistors that are supposed to conform to this model may not follow it exactly, especially when dealing with semiconductor circuits with sub-micron sized components.

 

[intangir]

In civ 5 27 to 52 is pretty darn close to 100%, well within the margin of error.

Yes, there is a margin of error. However, you would think that if the actual scaling number were 100%, there would be results both above and below that value. But there aren't. All such measurements end up being below 100%, which is a strong suggestion that the actual value is below 100%.

 

[tulx]

Lastly, as I keep reading, there is no need for 8 cores for desktop, right?

Isn't the idea behind the 8 core FX (Zambezi) processors that each 2 cores form a module, thus, in kind-of a reversed Hyper Threading, resulting in 4 logical cores?

8 cores (threads?) for a desktop PC is really ahead of its time, yes. A lot of games don't even use 4. Most good ones do, though.

 

[podspi]

It wasn't on AT, he said it on SA:

http://semiaccurate.com/forums/showpost.php?p=114359&postcount=1107

The important bit:

3. The 80% number keeps getting thrown around. Nobody understands it. The number that we have said is that 2 bulldozer cores in a module would be 180% of the throughput for a single bulldozer core. 90% + 90% = 180%. So the "overhead", so to speak, of the architecture is ~10% per core.

This is a pretty important distinction. It isn't "80% of two cores" it is 180% of one core.

 

[bridito]

No, I'm not gonna "try to understand". There's nothing to understand, as your would-be argument makes no sense. If you have nothing technical to contribute to the thread apart from flame bait and are gonna keep making up ridiculous would-be arguments for your amusement please get out of the thread. We're here to discuss technical facts.

Sorry if this sounds rude, but you constantly trying to trash or deviate this thread is getting old.

My dearest amigo. Last time I checked there was not a Mod after your name, and with you fresh from at least two vacations I do not believe that you have the authority to tell me which thread I can post in or not. When it comes to trashing this thread, I am not the only participant who has commented on your sustained insistence on discussing topics ad nauseum which are based only on fantasy and unbased speculation. Since this is a technical forum it would then follow that the subjects discussed be based on something other than illusions. The vast majority of your statements are illusory. Period. I will respectfully and with all due consideration continue to counter your broadcasting of incorrect, unsupportable, and thoroughly unbased chimeras until I am informed to cease by the Mods. Muchas gracias y viva Puerto Rico!

 

[3DVagabond]

No offense to JFAMD. I really appreciate his presence and contributions to the forums. He has thrown around a lot of numbers/terms though that seem to often be misinterpreted.

The bottom line is 80%, 90%, 180%... of what exactly? +35%, +50%, of what, exactly? Single core? Single module? Total throughput? Clock for clock? Watt for watt? At base clocks? At turbo speeds?

I do a lot of rendering with 3D apps. I'm really hoping that an 8 core BD for ~$300 blows the doors off of a i7 2600k when the software is capable of maxing out all 8 cores. Will it?

Because of it's advanced turbo will I not have to O/C it to get the most performance? Maybe just a moderate O/C will do rather than having to O/C by 1GHz+ to get top performance like with SB? That should put less wear on the system along with lower power usage. That would be desirable.

Surely AMD can afford some hard numbers by now. Even if Intel could do something to counter a month or so before launch, surely they'd have no problem doing it immediately after launch anyway. All this spy vs. spy is getting to be a bit much. :'(

 

[LOL_Wut_Axel]

My dearest amigo. Last time I checked there was not a Mod after your name, and with you fresh from at least two vacations I do not believe that you have the authority to tell me which thread I can post in or not. When it comes to trashing this thread, I am not the only participant who has commented on your sustained insistence on discussing topics ad nauseum which are based only on fantasy and unbased speculation. Since this is a technical forum it would then follow that the subjects discussed be based on something other than illusions. The vast majority of your statements are illusory. Period. I will respectfully and with all due consideration continue to counter your broadcasting of incorrect, unsupportable, and thoroughly unbased chimeras until I am informed to cease by the Mods. Muchas gracias y viva Puerto Rico!

Nothing better to do on this thread than incessantly trolling, right? If you have nothing to contribute go away.

 

[Rifter]

No, I'm not gonna "try to understand". There's nothing to understand, as your would-be argument makes no sense. If you have nothing technical to contribute to the thread apart from flame bait and are gonna keep making up ridiculous would-be arguments for your amusement please get out of the thread. We're here to discuss technical facts.

Sorry if this sounds rude, but you constantly trying to trash or deviate this thread is getting old.

If you want to stick to technical facts then you should do the same. You are the one trying to push performance numbers based on NOTHING. We have no benchmarks, we have no confirmed clock speeds, we have no IPC numbers. At this point we have nothing fact related to discuss.

You are the one making up numbers, he is simply saying until we have benches anything is possible.

 

[Rifter]

No offense to JFAMD. I really appreciate his presence and contributions to the forums. He has thrown around a lot of numbers/terms though that seem to often be misinterpreted.

The bottom line is 80&#37;, 90%, 180%... of what exactly? +35%, +50%, of what, exactly? Single core? Single module? Total throughput? Clock for clock? Watt for watt? At base clocks? At turbo speeds?

I do a lot of rendering with 3D apps. I'm really hoping that an 8 core BD for ~$300 blows the doors off of a i7 2600k when the software is capable of maxing out all 8 cores. Will it?

Because of it's advanced turbo will I not have to O/C it to get the most performance? Maybe just a moderate O/C will do rather than having to O/C by 1GHz+ to get top performance like with SB? That should put less wear on the system along with lower power usage. That would be desirable.

Surely AMD can afford some hard numbers by now. Even if Intel could do something to counter a month or so before launch, surely they'd have no problem doing it immediately after launch anyway. All this spy vs. spy is getting to be a bit much. :'(

I agree, but JFAMD probably has to jump through so many hoops to release any kind of statement with AMD PR department that what we end up getting is a statement that could mean anything.

 

[LOL_Wut_Axel]

If you want to stick to technical facts then you should do the same. You are the one trying to push performance numbers based on NOTHING. We have no benchmarks, we have no confirmed clock speeds, we have no IPC numbers. At this point we have nothing fact related to discuss.

You are the one making up numbers, he is simply saying until we have benches anything is possible.

I'm not pushing any performance numbers, so I don't know where you pulled that out from. Every time I've made one it's been in relation to Nehalem, and I've made it clear that it's not factual.

And we do know the design of the architecture and how it can affect many different things, and that's what we're discussing. And no, "anything" regarding performance is not possible. Claiming it could be the same speed as an Atom is ridiculous.

Now, how exactly have you contributed to this thread? All you do is look at my posts and make personal attacks while not giving any facts or technical discussion. We have a term for that: flame bait.

 

[Riek]

No offense to JFAMD. I really appreciate his presence and contributions to the forums. He has thrown around a lot of numbers/terms though that seem to often be misinterpreted.

The bottom line is 80&#37;, 90%, 180%... of what exactly? +35%, +50%, of what, exactly? Single core? Single module? Total throughput? Clock for clock? Watt for watt? At base clocks? At turbo speeds?
The 90% or 180% on module level is the efficiency of the design when multiple threads are run over the common resources.
The up to 35% was the estimated performance advantage of interlagos compared to MC 2.5GHz in HPC applications. The up to 50% performance advantage was the estimated performance advantage of interlagos compared to MC 2.3GHz.

I do a lot of rendering with 3D apps. I'm really hoping that an 8 core BD for ~$300 blows the doors off of a i7 2600k when the software is capable of maxing out all 8 cores. Will it?
I doubt it. While BD has 8cores, they share the fpu. So don't expect a lineair performance increase on the 8cores. I suppose the scaling will be better then HT (because that one also shares L/S/sheduler/..). I believe in those situations (fpu intensive) if HT increase performance, the 8core on BD will also increase performance compared to 4Threads.

Because of it's advanced turbo will I not have to O/C it to get the most performance? Maybe just a moderate O/C will do rather than having to O/C by 1GHz+ to get top performance like with SB? That should put less wear on the system along with lower power usage. That would be desirable.
Overclocking goes past TDP (just like turboboost on SB can). This is not the case for BD. So overclocking can still yield performance increases. That said, it doesn't seem likely that BD (this version) will consume less than SB. I suspect they will do the same thing as they did with llano, using to high voltages to keep the yield range as good as possible. You might be able to lower TDP by lowering the clock by 0.2V like llano though.

Surely AMD can afford some hard numbers by now. Even if Intel could do something to counter a month or so before launch, surely they'd have no problem doing it immediately after launch anyway. All this spy vs. spy is getting to be a bit much. :'(.
They probably could, but i doubt they will. We will also get a rough id when interlagos is released. A bit like Opteron and A64.

Added comments inline for once .-)

 

[Rifter]

I'm not pushing any performance numbers, so I don't know where you pulled that out from. Every time I've made one it's been in relation to Nehalem, and I've made it clear that it's not factual.

And we do know the design of the architecture and how it can affect many different things, and that's what we're discussing. And no, "anything" regarding performance is not possible. Claiming it could be the same speed as an Atom is ridiculous.

Now, how exactly have you contributed to this thread? All you do is look at my posts and make personal attacks while not giving any facts or technical discussion. We have a term for that: flame bait.

what part of there are no facts to give are you not understanding?