ExtremeTech: AMD Bulldozer FX pricing revealed: a lot cheaper than Sandy Bridge

[classy]

They might as well let the true numbers out of the bag now. I am sure in the next week or so numbers will be leaked out anyway.

 

[Arzachel]

Unfortunately, all signs point to Zambezi being less than stellar; I suspect that clock for clock, a single BD core will be slower than current K10.5 stuff, but you'll get more cores. It will also be interesting to see how Turbo Core plays out; if it's as cautious as some of the other chips, the eight core chips will only run at the base (or base + 1) frequency for anything more than dual-core workloads.

Identical IPC to Phenom II

I have to call bullshit on these claims/guesses.

JFAMD has confirmed several times that BD IPC INCREASES. Increases doesn't mean it gets worse. It also doesn't mean it stays the same. It can only become worse than projected due to bugs, and those will hopefuly get ironed out before release.

Why the hell is it even under debate, its one of the few things we can be sure about. The question is, is it closer to Llano or Nehalem.

 

[RussianSensation]

Why the hell is it even under debate, its one of the few things we can be sure about. The question is, is it closer to Llano or Nehalem.

"we can be sure about"

You mean you can be sure about that?

There may be an actual IPC increase per 1 core, but the BD module design penalizes dual-threaded scaling by 10-20%, unless this statement is wrong:

"The firm claims a Bulldozer module can achieve 80% of the performance of two complete cores of the same capability."

So if 2 BD cores (in 1 module) have a 25% faster IPC than 2 Phenom II cores, that means 0.8 (module penalty) * 1.25x faster IPC = 1.0x 2 Phenom II cores.

What you end up with is an IPC increase per core is offset by sharing of core resources in a module setup. So while "per core" IPC increases, in practice (when running 6-8 threads), it's may not better in IPC over an 8 core Phenom II/

JF-AMD's statement may be 100% true under these scenarios:

You run 2 threads on a 4-core BD (so you don't share resources, so you get full 25% IPC increase per thread).
You run 3 threads on a 6-core BD
You run 4 threads on an 8-core BD

 

[AtenRa]

Ok guys, just think of this

If BD has the same IPC as Phenom and a 6 Core BD will have lower performance due to 80% core scale than 6-Core Thuban, then AMD would actually just shrink Thuban cores and they would produce an 8 Core Thuban that would be faster than 8 Core BD.
They would save time, resources and money from the R&D and they could probably have already released that CPU a quarter at least earlier.

Now, who actually believes BD has the same IPC as Thuban ???

 

[Arzachel]

"we can be sure about"

You mean you can be sure about that?

There may be an actual IPC increase per 1 core, but the BD module design penalizes dual-threaded scaling by 10-20%, unless this statement is wrong:

"The firm claims a Bulldozer module can achieve 80% of the performance of two complete cores of the same capability."

So if 2 BD cores (in 1 module) have a 25% faster IPC than 2 Phenom II cores, that means 0.8 (module penalty) * 1.25x faster IPC = 1.0x 2 Phenom II cores.

What you get is IPC increase is offset by sharing of core resources. So in theory, "per core" IPC increases, but in practice (2-4 threads), it's no better unless you are only running 1 thread.

No.

The 1.8x figure was for two hypotethical "full" BD cores, that wouldn't share parts, NOT for Phenom II. That kind of BD doesn't exist, so the 1.8x figure is useless beyond showing how modules make sense from a perf/m^2 standpoint.

 

[RussianSensation]

Ok guys, just think of this

If BD has the same IPC as Phenom and a 6 Core BD will have lower performance due to 80% core scale than 6-Core Thuban, then AMD would actually just shrink Thuban cores and they would produce an 8 Core Thuban that would be faster than 8 Core BD.
They would save time, resources and money from the R&D and they could probably have already released that CPU a quarter at least earlier.

Now, who actually believes BD has the same IPC as Thuban ???

1. Pentium 4 Williamette had worse IPC than PIII. But Pentium 4 Northwood was cheaper to produce, and eventually performance scaled with more clock speed, addition of HT, improved FSB/memory bandwidth, etc.

2. Pentium E Prescott had worse IPC than Pentium 4 "C" Northwood, but scaled beyond Northwood and was actually faster past 3.2-3.4ghz.

3. BD architecture allows you to double the # of cores with only a 12% die size penalty. That's very impressive for markets that need more cores. You couldn't have done this with Phenom II design. Even if you shifted Phenom II to 8 cores on 32nm, the die size would be greater than an 8 core FX (increasing costs and reducing yields per wafer). So in fact, from a die-space efficiency perspective for core scaling, the modular design of Bulldozer is FAR more appealing for AMD in terms of manufacturing costs. It will also allow them to add more and more cores for servers (16-20) and make them more competitive to users who want multi-threaded CPUs.

4. BD architecture has far more advanced dynamic power consumption vs PHII. This will be extremely beneficial for servers and mobile markets (2 growth segments). If you just shrank Phenom II, you wouldn't do much to improve in this area. BD is expected to have major improvements in power consumption over Phenom II.

5. Even with similar IPC to Phenom II, AMD is giving consumers more cores at lower prices. It's quite possible that next versions of Bulldozer will launch at far higher clock speeds, something Phenom II couldn't achieve.

 

[skipsneeky2]

At this point who cares if these chips can or can't touch a 2500k or 2600k.

Only people giving a rats ass i believe are those waiting to build not people like us sitting on 2500k and 2600k chips.

Be a damn miracle if the 6 core for $155 can touch or surpass the i5 760 and if it can,that's enough justice for a 1156 user to upgrade.

 

[BlueBlazer]

I have to call bullshit on these claims/guesses.

JFAMD has confirmed several times that ....

JF-AMD also confirmed there was no delays back then, only to finally admit it later. Then he says Bulldozer will be launched in Q3, but now we know the launch is going to be in Q4. IMHO having trouble believing everything he says nowadays. :\

 

[RussianSensation]

At this point who cares if these chips can or can't touch a 2500k or 2600k.

It matters from a perspective of AMD's competitiveness over the course of the next 2-3 years. If their brand new architecture isn't very successful, AMD loses more $, Intel becomes lazier, innovation slows, we continue to only have 1 option on the desktop. None of these things are good for the marketplace.

Be a damn miracle if the 6 core for $155 can touch or surpass the i5 760 and if it can,that's enough justice for a 1156 user to upgrade.

I don't think people are going to be upgrading from i5 760 to a 6 core BD. These users have initially chosen the i5 over X6, which means their priorities lie in having faster cores, not more slow cores. That's not to say there isn't a market for a 6-core / 8-core BD, because there certainly is.

 

[bryanW1995]

What doesn't look good?

It's a nice chip for 155/185 bucks that will perform adequately in that price range.

Having 229 FPS in games with a i7 2600k for $300 or 156FPS with a FX-6100 for $155, I know where my money will go...

Wow, you sure seem to have quite a bit of quality inside info.

1. How do you know that it will "perform adequately"? How do you even define that?
2. Don't we already have plenty of cpu's that perform adequately?

The whole point of BD was that it was/is a new architecture that is supposed to significantly help out AMD's bottom line and mindshare. Selling the quads at launch for less than $200 b/c their performance is literally THAT BAD is hardly going to garner much mindshare OR bottom line. I, for one, hope that this post is wrong and BD retails for $400 +. Sadly, however, amd hasn't done much in the past 5 years to exactly inspire confidence.

 

[AtenRa]

So if 2 BD cores (in 1 module) have a 25% faster IPC than 2 Phenom II cores, that means 0.8 (module penalty) * 1.25x faster IPC = 1.0x 2 Phenom II cores.

I believe you have 100% for a single core and 180 for two cores (80%)

now two Phenom Cores will be at 190 (95+95)

Take that 180 and mul 1.25 = 225

So a BD module will have 225 vs 190 of dual Phenom cores

225 is 18% faster than 190

How my maths doing ??

What you end up with is an IPC increase per core is offset by sharing of core resources in a module setup. So while "per core" IPC increases, in practice (when running 6-8 threads), it's may not better in IPC over an 8 core Phenom II/

JF-AMD's statement may be 100% true under these scenarios:

You run 2 threads on a 4-core BD (so you don't share resources, so you get full 25% IPC increase per thread).
You run 3 threads on a 6-core BD
You run 4 threads on an 8-core BD

Not happening

 

[Arzachel]

5. Even with similar IPC to Phenom II, AMD is giving us more cores at lower prices. It's quite possible that next versions of Bulldozer will launch at far higher clock speeds, something Phenom II couldn't achieve.

BD IPC INCREASES!

Also, BlueBlazer, release dates and clockspeeds may differ wildly, especialy if the problems come from the GF side, but the only way IPC for the final product can be lower than planed is previously unknown bottlenecks (not likely), huge shift in the way software is written(not likely) and bugs(hopefuly not).

 

[skipsneeky2]

This release reminds of how amd came out with cheaper quads when intel was still charging a arm and a leg for theirs.

People are still wanting the fastest quad cores possible and i do believe Ivy Bridge might put a axe to performance quads and the hexacore could very well be the next standard.

 

[notty22]

Thats all thats left now, JF AMD SAID. :thumbsdown:

He also made qualifications to past statements, such as, he speaks for server side and not client/desktop.

 

[chihlidog]

What is this BD everyone speaks of?

I'm more convinced that Bigfoot exists than I am that Bulldozer exists. I wonder which one we'll have proof of first?

 

[996GT2]

Guys, you need to realize that what JF-AMD said DOES NOT have to equal reality

 

[Arzachel]

I believe you have 100% for a single core and 180 for two cores (80%)

now two Phenom Cores will be at 190 (95+95)

Take that 180 and mul 1.25 = 225

So a BD module will have 225 vs 190 of dual Phenom cores

225 is 18% faster than 190

How my maths doing ??

Everything is nice and fine, but the 180%/1.8x figure is useless when basing your calculations on Phenom II. That figure is actually useless period, everyone should stop (mis)using it, just causes more confusion.

 

[Ancalagon44]

I have to call bullshit on these claims/guesses.

JFAMD has confirmed several times that BD IPC INCREASES. Increases doesn't mean it gets worse. It also doesn't mean it stays the same. It can only become worse than projected due to bugs, and those will hopefuly get ironed out before release.

Why the hell is it even under debate, its one of the few things we can be sure about. The question is, is it closer to Llano or Nehalem.

He would say that, wouldnt he? And as someone pointed out, he only begrudgingly admitted it was delayed.

 

[Phynaz]

JFAMD has confirmed several times that BD IPC INCREASES.

JFAMD also confirmed several times that BD wasn't delayed.

 

[BlueBlazer]

Also, BlueBlazer, release dates and clockspeeds may differ wildly, especialy if the problems come from the GF side, but the only way IPC for the final product can be lower than planed is previously unknown bottlenecks (not likely), huge shift in the way software is written(not likely) and bugs(hopefuly not).

Unfortunately I have to disagree with that view. From all the sources I've read (in the past months), everything points to the opposite of what JF-AMD says. Even Anandtech Editors hinted at the same conclusion (not long ago). You may notice he keeps insisting about being a server guy, and in servers you can use compiler tuning to optimize (for Bulldozer pipleline architecture) and to enable new features in Bulldozer (such as AES, AVX and FMA4) that older generation Magny-Cours doesn't have, to claim increased performance. However on the desktop, the situation may be different. The programs are pre-compiled (fixed) and "run-as-it-is". You can also check JF-AMD's FAQ especially these lines....

Final OS optimizaitons
Final drivers
An app compiled with the latest flags

And for the driver part? There is one bug I've found so far >> Looks like Linus isn't happy with a fix. :hmm:

 

[skipsneeky2]

What is this BD everyone speaks of?

I'm more convinced that Bigfoot exists than I am that Bulldozer exists. I wonder which one we'll have proof of first?

Bigfoot and BD are much alike,perhaps Bigfoot works security at the Amd Lab,s.

 

[Arzachel]

@Everyone: How can you compare something thats down to mostly GF's 32nm process with something that has been set in stone from the day the design was finalised?

And yes, I trust a representative of the company far more, than people that base their opinions on uncomfirmed leaks. I thought that would seem logical?

 

[Majic 7]

Me, I never trust anyone in marketing. Standard policy for any company.

 

[ZOXXO]

Me, I never trust anyone in marketing. Standard policy for any company.

And that to me seems logical.

 

[Iron Wolf]

What is this BD everyone speaks of?

I'm more convinced that Bigfoot exists than I am that Bulldozer exists. I wonder which one we'll have proof of first?

A lot of people are convinced that Bigfoot is real.

Look, even DNF was eventually released. If the king of vaporware was given an actual release date, we have hopes of seeing other long awaited stuff, like Bulldozer CPUs, or even Diablo 3. These last two might as well be somebody's imagination as far as most of us are concerned, though.

And WTF with the codename? Kind of seems ridiculous, given the leaked performance data.